ArticlePDF Available

Abstract and Figures

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 are particularly prone to bias about victims’ groups (e.g. Blacks, women) 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 (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 concluded that a privileged group (men and Whites) had a more desirable quality relative to a victims’ group (women and Blacks) than vice versa. Ruling out alternative explanations of Bayesian (or other normative) reasoning, significant order effects in within-subjects designs in Studies 6 and 7 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 more liberal ideology and lower credibility ratings when privileged groups were said to score higher on a socially valuable trait. Although not predicted a priori, meta-analyses also revealed Moderates to be the most balanced in their judgments. These findings indicate nothing about whether this bias is morally justifiable, only that it exists.
Content may be subject to copyright.
Running Head: EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
Equalitarianism: A source of liberal bias
*Bo M. Winegard1
*Cory J. Clark2
Connor R. Hasty3
Roy F. Baumeister3,4
1Marietta College, Marietta, OH
2University of Pennsylvania, Philadelphia, PA
3Florida State University, Tallahassee, FL
4University of Queensland, Brisbane, AU
*First two authors contributed equally
DRAFT VERSION 10.30.20
THIS MANUSCRIPT IS UNDER REVIEW AFTER AN INVITED REVISION.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
2
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 are particularly prone to bias about victims’
groups (e.g. Blacks, women) 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 (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 concluded that a privileged group
(men and Whites) had a more desirable quality relative to a victims’ group (women and Blacks)
than vice versa. Ruling out alternative explanations of Bayesian (or other normative) reasoning,
significant order effects in within-subjects designs in Studies 6 and 7 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 more liberal ideology and lower credibility ratings when privileged groups
were said to score higher on a socially valuable trait. Although not predicted a priori, meta-
analyses also revealed Moderates to be the most balanced in their judgments. These findings
indicate nothing about whether this bias is morally justifiable, only that it exists.
Keywords: political psychology, liberal bias, motivated cognition, egalitarianism, prejudice
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
3
Equalitarianism: A Source of Liberal Bias
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.
Bias: The Dark Matter of Psychology
Bias is an important concept in social and cognitive psychology. Unfortunately, it is
exceedingly difficult to define or measure. As we will discuss in greater detail later, there are, to
our knowledge, no empirical studies of bias that entirely escape reasonable objections (usually
from a Bayesian perspective). However, broadly conceived, bias is fairly straightforward: It is a
preference or commitment that shapes and distorts cognition away from the truth or from
impartiality in a predictable, preference congruent manner (Ditto et al., 2019a; Kahan & Braman,
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
4
2006; Taber & Lodge, 2006). If someone, for example, is a devoted fan of the New York
Yankees (a major-league baseball team) and allows her team preference to influence her opinion
of balls and strikes (smaller strike zone for Yankees batters than for the other team’s batters),
then we would say that she is biased. If, on the other hand, she assessed balls and strikes in a
similar manner across teams, then we would say that she is not biased or that she is impartial.
Bias can infect the cognitive process from beginning to end and anywhere between (e.g.,
Ditto & Lopez, 1992; Frenda, Knowles, Saletan, & Loftus, 2013; Iyengar and Hahn, 2009; Taber
& Lodge, 2006). As many scholars have noted, all reasoning is motivated and most people are
prone to bias (Kunda, 1990). Some reasoning is motivated by a concern for the truth, and
therefore is not prone to bias (although it might still lead to incorrect conclusions); and some is
motivated by extraneous concerns such as tribal identity or esteem needs, and therefore is prone
to bias (Taber & Lodge, 2006). As a general rule, bias increases as the strength of one’s
preferences increases (Skitka, 2010; Taber & Lodge, 2006). The fervid Yankees fan is much
more likely than the casual fan to have a biased strike zone. Furthermore, preferences that form
an important part of one’s identity are more likely to impel bias than preferences that do not
(Haidt, 2012; Tajfel, 1974). For example, most people have a strong preference for sunshine over
clouds, but this preference probably does not cause significant bias because it is not an important
component of people’s identities. On the other hand, a preference for one’s home team or sibling
might very well cause significant bias because it could comprise one crucial part of a person’s
social identity. Morally valenced identity preferences generally cause the strongest biases
(Skitka, 2010; Tetlock, 2003). Last, the clearness of the facts/data affects bias (see Felson, 1981;
Kruger & Dunning, 2011 for similar discussions about ambiguity and bias). Generally speaking,
people are not biased about things that are undeniable and obvious (i.e., that have high clarity).
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
5
The less clear, the more ambiguous, facts/data become, the more biased people can be. Balls and
strikes in baseball are less clear than runs, and are therefore a more fertile source of bias.
On whole, then, bias is a function of clarity, accuracy concerns, and extraneous concerns,
such that extraneous concerns increase bias, and accuracy concerns and clarity decrease bias.
This likely explains why partisan bias is such a potent form of bias. First, clarity is often low.
Experts have studied tax policy for many years, and they still don’t have a clear answer about the
optimal marginal rates. Even something such as anthropogenic global warming that compels near
scientific consensus is difficult to predict and measure and therefore unclear to most people. And
second, extraneous concerns are often high. Many people highly value their moral and political
identities and want to protect them from potential threats (Haidt, 2012; Kahan & Braman, 2006;
Skitka, 2010). Often moral and political commitments become sacred values or values that “a
moral community treats as possessing transcendental significance” and that cannot be sacrificed
for other values, even, perhaps, the pursuit of truth (Tetlock, 2003, p. 320; also, Atran, Axelrod,
& Davis, 2007). The intensity of these extraneous values can easily cloud out accuracy concerns
especially when clarity is low, creating a climate extremely conducive to bias. It is worth noting
that from an evolutionary perspective, tribal biases are almost certainly not irrational (Van Bavel
& Pereira, 2018). Group membership and status are probably more important for survival and
reproduction than is the truth about abstruse or abstract questions (Baumeister, Maranges, &
Vohs, 2017).
Political bias. For many people, political (and/or moral) preferences are powerful and
comprise a narrative (often not conscious) that is important to one’s identity (Haidt, 2012;
Huddy, 2001). Therefore, political commitments are very likely to give rise to bias. Indeed, for
many years now, social scientists have examined political personality types and prejudices, often
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
6
creating scales to capture certain traits that are thought to lead to bias, rigidity, and unpleasant
perhaps even deleterious social consequences (Adorno, Frenkel-Brunswik, Levinson, Sandord,
1950; Altmeyer, 1981; 1996; Jost, Glaser, Kruglanski, Sulloway, 2003; Pratto, Sidanius,
Stallworth, Malle, 1994).
One thing many of these studies and theories share is that they depict political
conservatism as potentially malignant, full of bias, and less explicable than liberalism, which is
often assumed to be “normal” or simply correct and therefore without need of explanation
(Haidt, 2012; Tetlock, 1994). (We will call this, in line with other researchers, the asymmetry
hypothesis, which is the belief that Conservatives are more biased than Liberals; see Ditto et al.,
2019a.) For example, a highly influential paper that has been cited several thousand times in the
literature was entitled “Political conservatism as motivated social cognition” (Jost et al., 2003;
italics added). Although the article briefly acknowledged the possibility that there could be
biased cognition among liberals, its main thrust was to depict conservatives as rigid, fearful, and
biased. Many scales in social science reflect this view. That is, they appear to assume that
liberalism or cosmopolitanism (closely related to liberalism) is correct or preferable to
conservatism and therefore measure traits that deviate from liberalism, describe the traits in
pejorative ways, and label the traits pejorative names (see Crawford & Jussim, 2017 for
discussion of political bias in social psychology).
However, throughout the history of the study of political bias, some researchers have
charged that the asymmetry hypothesis is wrong and possibly irresponsible (Rokeach, 1956;
Taylor, 1960). More recently, many social scientists have contended that the asymmetry
hypothesis might be an unfortunate outgrowth of a liberally biased field (Duarte, Crawford,
Stern, Haidt, Jussim, & Tetlock, 2015; Haidt, 2011; Clark & Winegard, 2020). This argument
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
7
suggests that just as Europeans in the 1300s did not notice “Christian bias” because they simply
accepted Christian doctrine as truth, so social scientists do not notice liberal bias because most of
them assume its principles are correct and require no further explanation. Inbar and Lammers
(2012) and von Hippel and Buss (2017) have quantified political beliefs in social psychology and
have confirmed suspicions that the field is dominated by social liberals. This provides at least
prima facie support to the argument that the field’s liberal bias may have contributed to the
asymmetry hypothesis and to the generally unsavory depiction of political conservatives that
dominates social psychology.
Recently, many scholars have worked to correct politically motivated shortcomings in
social science, finding that liberals are often just as biased as conservatives if one scrutinizes in
the correct places (Crawford, 2012; 2014; Graham et al., 2013). In 2019, Ditto and colleagues
reported a meta-analysis on partisan bias and found strong support for a symmetry hypothesis,
noting that the overall effect size for conservative bias was not significantly greater than for
liberal bias (conservative r = .255; liberal r = .235).
However, there is reason to believe that this meta-analysis may have underestimated the
size of liberal bias because it only included a few studies that measured what we will argue is
one of the most potent sources of liberal bias: perceived victims’ groups. And in fact, the one
included study that had the most obvious relevance to victims’ groups (a study regarding
affirmative action and same-sex marriage) found one of the largest effects of liberal bias (r =
.54), and a reverse bias for Conservatives such that they also demonstrated a preference for
affirmative action and same-sex marriage (r = -.20), just to a lesser degree (Crawford, Jussim,
Cain, & Cohen, 2013). In this article, we want to help rectify this problem by directly examining
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
8
liberal bias as related to perceived victims’ groups and by exploring a possible explanation for
this bias, a set of interrelated beliefs about low status groups that we call equalitarianism.
Liberal Bias and Equalitarianism
The present experimental work expands on work by Winegard and colleagues (Winegard,
Winegard & Geary, 2015; Winegard & Winegard, 2015; 2017; see also Clark & Winegard,
2020), which contends that Liberals are particularly disturbed by extant inequalities among
demographic groups and want to ameliorate all such disparities. Liberals, more than
conservatives, are egalitarian (Jost, Nosek, & Gosling, 2008) and empathize more with others
(e.g., Hasson, Tamir, Brahms, Cohrs, & Halperin, 2018), particularly disadvantaged others
(Lucas & Kteily, 2018; see, also, Jeffries, Hornsey, Sutton, Douglas, & Bain, 2012), than do
Conservatives. Inequalities among demographic groups lead Liberals to empathize with groups
that are relatively low-status or experiencing relatively poor outcomes. Although Liberals’
concern for victims’ groups likely stems from admirable compassion, this can lead to ironic
effects, such as patronizing behaviors (Dupree & Fiske, 2018), and double standards that favor
groups that they perceive as victims. For example, one study found that Liberals were more
likely than Conservatives to amplify the successes of women and Blacks than men and Whites,
whereas Conservatives treated the successes of groups more similarly (Kteily, Rocklage,
McClanahan, & Ho, 2019). Another (as yet unpublished) study found that Liberals were more
censorious of information that portrayed women, Blacks, and Muslims unfavorably than
identical information that portrayed men, Whites, and Christian unfavorably, whereas
Conservatives were more equally censorious of information regardless of which groups were
portrayed unfavorably (Winegard, Clark, & Bunnel, 2019). And similarly, a study found that
people more positively evaluated research on female-favoring sex differences than male-favoring
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
9
sex differences, and that more left-leaning participants were increasingly negative toward male-
favoring sex difference research findings (Stewart-Williams, Thomas, Blackburn, & Chan,
2020).
Liberals (more than Conservatives) appear to believe that women and minorities
comprise a victims’ group category that needs to be protected from oppression, exploitation, and
other social harms. This suggests that Liberals will be biased when evaluating information about
perceived victims’ groups in predictable ways. Most broadly, Liberals will be especially
motivated to reject information that appears to pose potential threats to victims’ groups. Thus we
hypothesized that low status groups are a domain of bias for Liberals such that they will more
negatively evaluate information that portrays low status groups unfavorably than information
that portrays high status groups unfavorably.
Whenever group disparities exist, there are at least two potential explanations. One is that
groups differ for predominantly genetic reasons. Another is that society mistreats certain groups,
stunting their potential and inhibiting their success. For those disturbed by inequality, the latter
explanation is probably more appealing because it suggests that social disparities are caused by
injustice, not by difficult to alter genetic processes. If, for example, women’s underrepresentation
among Fields medalists (an award for achievement in mathematics) is due to genetically caused
differences in interests and ability, then it would require massive (and procedurally unfair)
interventions to equalize the representation. But if this disparity is due to social processes,
stereotypes, and sexism, then equality can be achieved in a meritocratic and unbiased society
one just has to eradicate the sexism. Thus, Liberals may believe that most socially consequential
demographic differences (e.g., in median income, representation in various fields, criminality)
are caused by discrimination and other environmental forces, not by characterological
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
10
differences. Indeed, recent research suggests that Liberals are more inclined to impute motives to
researchers who present results suggesting that intrinsic factors such as genetics, hormones, and
neurochemistry influence outcomes such as intelligence, mating strategy, and violence than to
researchers who provide more extrinsic explanations, such as education, nutrition, socialization
and culture, and parenting and development (Hannikainen, 2018).
We call the commitment that might explain liberal bias about perceived victims’ groups
equalitarianism. Equalitarianism is composed of three interrelated beliefs: (1) demographic
groups do not differ biologically on socially valued traits, (2) society is rife with sexism and
racism and that disparate demographic outcomes are likely caused by oppression and prejudice,
(3) people in society should work together to combat pervasive racism and sexism, and that if
successful, no group differences in life outcomes (e.g., educational attainment, imprisonment,
socioeconomic status) would remain. We hypothesized that Liberals would endorse these beliefs
more than Conservatives and that these beliefs might at least partially explain the predicted bias
against information that portrays low status groups unfavorably. Note that these beliefs
themselves are not necessarily a sort of bias, rather we expect that they will partially explain the
relationship between higher self-reported liberalism and biased responding in upcoming studies.
We contend that Liberals have concerns about protecting victims’ groups (e.g., women, Blacks;
see Study 1a for a list of perceived victims’ groups), which leads Liberals to evince bias when
evaluating information about perceived victims’ groups and about potential demographic
differences on socially valued traits (Bawer, 2012; Winegard & Winegard, 2015). Specifically,
we contend that Liberals will more negatively evaluate information that contends that perceived
victims’ groups are lower on average on socially valued traits than perceived privileged groups
than identical information that supports the opposite conclusion. We support this argument with
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
11
eight studies that show that Liberals were consistently biased against information that portrayed
privileged groups more favorably than victims’ groups (small to medium effects across all
studies). In all studies, this bias was predicted by higher support for equalitarianism.
But is it bias?
Most bias studies, including ours, rely upon the principle of invariance: Decision
irrelevant information (extraneous information) should not affect judgments; therefore, the
degree of a person’s bias is reflected by the degree to which the extraneous information affects
his or her judgments (Ditto et al. 2019a; Kahneman & Tversky, 1984). In psychology, the
standard methods for testing bias involve matching as much information as possible, changing
only the conclusions of a vignette or other supposedly extraneous information (such as the race
or sex of an actor), and then having participants evaluate the matched information rather than the
manipulated information. If participants evaluate identical information differently depending on
the extraneous information, this is considered bias. For example, in a famous study on bias, Lord,
Ross, and Lepper (1979) gave participants identical methodology descriptions of studies testing
the deterrent efficacy of the death penalty. The only information that varied in the conditions was
the conclusion: the study found that the death penalty did or did not deter crime. Then they had
the participants rate the quality of the studies’ methods (which again, were identical).
Participants who supported the death penalty rated the methods as worse when the conclusion
contradicted their prior attitude (death penalty deters crime) than when it buttressed it.
However, it is not clear that this paradigm allows a researcher to isolate bias
unambiguously. In methods matching studies, for example, it might be rational to assess methods
differently depending upon the outcomes of those methods. Imagine, for example, a description
of methods that appeared sound but generated results that showed that eating purple muffins
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
12
allows people to see the future. People cannot see the future (but see Bem, 2011); therefore, one
should be very skeptical of the results; and if one is skeptical of the results, then one should
probably be skeptical of the methods that led to them. We call this “the proof of the recipe is in
the eating” or PRE principle. A recipe might look good or bad on paper, but its final value
depends upon the food it produces. If one follows the recipe and gets bad food, it is not irrational
to update one’s assessment of the recipe. More broadly, the results of a process (methods, recipe,
blueprint) provide information about the soundness of the process, and a good Bayesian should
update his or her priors about the process after getting the results (see Kahan, 2016, for a
discussion of Bayesian reasoning and bias).
The same criticism applies to matched vignettes that change the demographic
characteristics of described individuals. Suppose, for example, that a researcher believes that
liberals are biased against White people. She designs a study that includes a vignette describing a
cop shooting a person who was found to be holding a piece of silverware (not a weapon). The
vignette is altered such that in one condition the cop’s victim is White and in the other he is
Black. She then finds that liberals rate the cop as less wrong when the victim was White than
when he is Black and contends that this is due to liberal bias against White people. One might
object, however, that the demographic characteristics Black and White provide information.
Perhaps one believes, for example, that Blacks are unfairly targeted by police officers more often
than Whites. One might believe, then, that in the White condition the spoon must have looked
quite menacing because otherwise the cop would not have shot; whereas, in the Black condition,
one might just think “yeah, cops wrongly shoot Black people all the time… this is very wrong.”
There are a couple of ways to mitigate the force of the Bayesian (normative rationality)
objection. First, one can choose examples in which base rates go in the opposite direction from
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
13
the predicted bias. Suppose, for example, that a researcher thinks that Conservatives are biased
against women. He could use a vignette in which either a man or a woman sexually propositions
a subordinate in a crass way and ask participants if the (identical) behavior is sexual harassment.
In this way, the demographic information is going, if anything, against the direction of the
hypothesis because most people believe that men are more likely than women to sexually harass
others. Thus, if Conservatives rate the identical behavior as harassment only when performed by
a woman (and not when performed by a man), it would be reasonably compelling evidence that
Conservatives are biased against women in this domain.
Second, one could observe order effects in a within-subjects design (for example, see
Uhlmann, Pizarro, Tannenbaum, & Ditto, 2009, which, incidentally, found that Liberals were
more willing to sacrifice a White man to save 100 others than to sacrifice a Black man to save
100 others, whereas race had no influence on Conservatives’ willingness to sacrifice one life to
save 100, somewhat consistent with our predictions here). That is, one could give both vignettes
to participants and manipulate the order of presentation. If participants believe that their answers
in the two conditions should be the same and therefore anchor their second response to their first,
that suggests that people at least believe it is irrational (and biased) to answer them differently. If
an order effect is observed such that both vignettes are evaluated more favorably when the
preference consistent one is presented first than when the preference inconsistent one is
presented first, this would indicate that participants are biased despite their apparent belief that it
is irrational to treat the two conditions differently. In our experiments, we tried to use both
principles to counter possible objections. We still believe that objections are possible; ultimately,
it might not be possible to demonstrate bias in an experiment without putting participants
through timely experiments that allow researchers to rule out Bayesian updating explanations.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
14
Nevertheless, current methods allow us to glean valuable information about potential bias,
which, when combined with theory, should cause us to update our priors about sources of bias.
Research Overview and General Predictions
Across eight studies, we tested the equalitarian theory of liberal bias. We used a novel
measure of equalitarianism, which had an excellent alpha (.88-.93; see appendix for full scale).
Studies 1a-1b did not test for bias, but rather were equalitarianism validation studies, which
simply tested whether liberalism was associated with perceiving certain groups (e.g., women,
Blacks, Hispanics) as victims, a variety of pro-victims’ group attitudes, and intolerance of
putative real world events in which victims’ groups were harmed (e.g., cop shooting an unarmed
Black person), and whether equalitarianism mediated all of these relationships. The remaining
studies tested our main hypotheses regarding liberal bias. In Studies 2-3, participants read
vignettes, which suggested that either a privileged group (men or Whites) or a victims’ group
(women or Blacks) scored higher on a socially valued trait (intelligence) and evaluated the
credibility of the arguments. Studies 4-5 included conditions in which both groups were said to
be equal. Studies 6-7 were conducted within-subjects to test for order effects to increase
confidence that the obtained results indeed reflect bias. Across all studies, we expected that
Liberals would rate the arguments as less credible when the privileged group was said to be more
intelligent than the victim’s group than vice versa, and that Liberals would rate the arguments
that stated that the privileged group was more intelligent as less credible than Conservatives.
However, we expected that Liberals would rate the argument that stated that both groups are
equal as the most credible. We further predicted that higher equalitarianism would mediate the
influence of more liberal ideology on lower credibility ratings when privileged groups were said
to be more intelligent. Table 13 toward the end summarizes all main results.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
15
Note the present work only made a priori predictions about Liberals because Liberals are
a relatively understudied group (Eitan et al., 2018), and very little work deliberately explores
biases among Liberals. Nonetheless, we will discuss patterns discovered among Moderates and
Conservatives as they are identified. Overall, patterns were less consistent across studies for
Moderates and Conservatives than for Liberals; however, meta-analyses revealed some patterns
for these groups as well, which we elaborate on in results and discussion sections and the general
discussion.
Initiative for Open Science Statement
No participants were excluded from any study. There were no additional undisclosed
manipulations or conditions. No analyses were performed before the corresponding data
collection was complete (except as described in Study 6 of the main text). All datasets and
syntax will be publicly available. The only variables deleted from the datasets were potential
identifiers (ip address, latitude, longitude), some irrelevant and potentially sensitive
demographics (relationship status and sexual orientation [asked only in Studies 1a-3] and income
[asked only in Study 2]), two open-ended responses (“Was anything about this study bizarre,
unusual, or suspicious?” and “Is there anything else you would like to tell us about your
participation today?” asked only in Studies 4-7), and other irrelevant data (e.g., date of
completion, time to complete). No other data were collected that are not described in the paper.
No other studies were conducted testing the hypothesis that liberals would be biased against
information that portrays privileged groups more favorably than victims’ groups on a socially
valued trait. There are no file drawer studies.
Study 1a
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
16
Study 1a tested the hypotheses that stronger liberal ideology would predict stronger
beliefs that certain groups are victims of unfair treatment by society, and that our measure of
equalitarianism would mediate this relationship.
Method
Ethics statement. All studies in the present manuscript were approved by the Florida
State University Human Subjects Committee under protocols HSC #2015.16573 and HSC
#2017.22463. Participants consented to participate by clicking ‘next’ to begin the study after
reading a study information sheet.
Participants. U.S. participants (Mage = 36.93, SD = 12.30; 122 female) were recruited via
Amazon Mechanical Turk (MTurk).
1
We aimed for a fairly large sample size of 200; 202 people
participated. This would allow detection of a small to medium r effect size around .2 (at p < .05
with 80% power; G*Power; Faul, Erdfelder, Buchner, & Lang, 2009; Faul, Erdfelder, Lang, &
Buchner, 2007). We discuss efforts to maintain and increase power across studies in the Methods
section of Study 2.
Procedure. Order of procedures was randomized. Participants were asked to rate how
unfairly various groups of people are treated in society on 100-point sliding scales from Treated
completely unfairly to Treated completely fairly. Four were groups that are generally considered
1
We used U.S. MTurk participants in all studies. Though MTurk is not perfectly representative
of the U.S. as a whole, it is quite diverse with respect to age, sex, race, education, SES, and
ideology (and certainly more diverse than most university subject pools). On average, MTurk
workers are slightly younger, more educated, and more liberal than the US population; Asians
are overrepresented and Hispanics are underrepresented (Paolacci & Chandler, 2014). All of
these were true in our samples as well. Women were also slightly overrepresented. We focused
on the U.S. because our theory builds on research conducted in the U.S. and was partly inspired
by patterns of recent events in the U.S. We do not know whether our results would generalize to
other countries and cultures. We suspect there may be important cultural moderators that we
hope future research will uncover. For example, cultures that are more hierarchical and less
egalitarian may not display the reported effects.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
17
victims’ groups (Black people, Women, Hispanic people, and Muslims); three were groups that
are generally considered privileged groups (White people, Men, and Christians).
2
Participants also completed an equalitarianism measure, which contained 18 items
measuring attitudes about whether 1) all groups are equally endowed with socially desirable
traits (e.g., “All ethnic groups have equal abilities on all tasks [for example, mathematics, sports,
creativity]”), 2) prejudiced attitudes are ubiquitous (e.g., “Racism is everywhere even though
people say they are not racist”), and 3) we can and should strive for a more egalitarian society
(e.g., “We should strive to make all groups equal in society”), rated on 7-point scales from Do
not agree at all to Completely agree,
= .92 (see appendix for full scale). Across all studies
reported in the present paper, the alpha for the equalitarianism measure ranged from .88-.93.
Combined across all studies, a principle components analysis supported a one factor solution.
Four components had Eigenvalues above 1, however there was a dramatic drop off from the first
component (Eigenvalue = 7.72) to the second (Eigenvalue = 1.66), and all items correlated with
the first component the strongest (and all above .48). The only exceptions were the three reverse
scored items, which correlated with multiple components to similar degrees and seemed to
explain the presence of the other components, a common problem with reverse-scored items (see
e.g., Conrad et al., 2004). The three reverse-scored items generally correlated the weakest with
all other items in the measure, however, they were still positively and significantly associated
with all other items.
2
One additional group was included (atheists), but this group does not clearly fit as a victims’ or
privileged group nor did we have a priori predictions about this group. But to satisfy curiosity,
these were the results for atheists: fairness rating (M = 55.16, SD = 29.46); correlation with
liberal ideology, r = -.35, p < .001. Thus, it seems Liberals believe atheists to be victims as well.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
18
The only other procedure was a demographics survey on which participants reported a
variety of demographic variables, including political ideology, which was reported on a 7-point
scale from Very conservative to Very liberal. Combined across all studies, equalitarianism was
correlated with more liberal political ideology, r = .54, p < .001.
Results
Participants were slightly above the midpoint on liberalism (M = 4.44, SD = 1.79) and
equalitarianism (M = 4.69, SD = 1.14), and these were positively correlated, r = .53, p < .001. As
can be seen in Table 1, participants viewed Whites as treated the most fairly, followed in order
by Men, Christians, Women, Hispanics, Blacks, and last, Muslims. As predicted, stronger liberal
ideology was significantly negatively related to fairness ratings for all four victims’ groups:
Muslims, Blacks, Hispanics, and Women. Results were slightly mixed for the privileged groups,
such that stronger liberal ideology was significantly positively related to fairness ratings for
Christians, slightly (but non-significantly) positively related to fairness ratings for Whites, and
unrelated or slightly negatively related to fairness ratings for Men.
Table 1
Fairness ratings by group and their correlation with (liberal) ideology
Group
M
SD
r
p
Whites
78.92
23.03
0.09
0.231
Men
78.34
23.22
-0.05
0.514
Christians
68.12
27.74
0.32
<.001
Women
59.81
22.85
-0.39
<.001
Hispanics
51.65
25.12
-0.42
<.001
Blacks
50.30
26.23
-0.44
<.001
Muslims
41.94
28.57
-0.39
<.001
Fairness ratings for the victims’ groups were reverse-scored and combined with fairness
ratings for the privileged groups to create an unfairness index,
= .77. A bootstrap mediation
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
19
analysis (10,000 resamples; PROCESS model 4 [Hayes, 2013])
3
revealed a significant indirect
effect of ideology on unfairness ratings through equalitarianism, 95% CI [-2.99, -1.22]. As can
be seen in Figure 1, more liberal ideology predicted rating victims’ groups as treated more
unfairly (and privileged groups as treated more fairly), and this was partially mediated by their
higher equalitarian attitudes.
Figure 1. Influence of ideology (higher values = more liberal) on unfairness (higher values =
victims’ groups treated more unfairly/privileged groups treated more fairly), mediated by
equalitarianism (higher values = more equalitarian). In this and all subsequent studies, values in
parenthesis are the total effect of the IV on the DV (i.e., prior to controlling for equalitarianism).
Discussion
As predicted, Liberals viewed perceived victims’ groups as treated more unfairly than
Conservatives, and this effect was partially mediated by scores on a measure of equalitarianism.
Study 1b
Study 1b examined the influence of political ideology and equalitarianism on evaluations
of news events and public opinions involving victims’ groups. Participants evaluated two
3
In this study and all upcoming studies, this is how we tested simple mediations.
b = .33, se = .04, p <.0001
Unfairness
Equalitarianism
Ideology
b = 5.98, se = .98, p <.0001
b = 2.13, se = .62, p =.0007
(b = 4.13, se = .58, p <.0001)
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
20
ostensible news events, one involving a cop shooting an unarmed Black man, and one involving
a university using a performance exam on which men outperform women. We expected that
Liberals would evaluate the cop and the exam more unfavorably, and that these would be at least
partially accounted for by their higher equalitarianism scores. Participants also reported their
agreement with a variety of statements relevant to victims’ groups. We expected that more
liberal ideology would predict more pro-victims’ groups and more anti-privileged groups
attitudes, and that these would also be at least partially accounted for by their higher
equalitarianism scores.
Method
Participants. U.S. participants (Mage = 34.68, SD = 11.14; 100 female) were recruited via
MTurk. Given the strength of the relationships in Study 1a (the correlation between more liberal
ideology and overall unfairness ratings was r = .453), we aimed for a slightly smaller sample size
of 150; 151 people participated. This would allow us to detect an r effect size of around .23 (Faul
et al., 2007, 2009). Participants were slightly above the midpoint on liberalism (M = 4.30, SD =
1.77) and equalitarianism (M = 4.78, SD = 1.01), and these were positively correlated, r = .54, p
< .001.
Procedure. Order of procedures was randomized. Participants completed the same
measure of equalitarianism,
= .89, and reported political ideology as in Study 1a. Participants
were also asked to read two ostensible news reports from The New York Times and The Boston
Globe (order of presentation was counterbalanced). One story was about a police officer killing
an unarmed Black man:
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
21
On the night of August 19th, Joe Smith, a New York City policeman encountered Darren
Johnson, an African American, on a playground. Officer Smith had received a call about
an armed robber in the area. Officer Smith confronted Darren Johnson and told him to
put his hands up. Darren Johnson then lifted a shiny object into the air and pointed at
Officer Smith. Officer Smith fired five shots at Darren Johnson, killing him instantly.
After the shooting, police discovered that the shiny object was a ballpoint pen.
The other story was about the introduction of a performance exam, on which men
outperform women:
Washington State University is facing controversy after introducing the Graduate
Performance Test (GPT). The GPT predicts college performance quite well, so
Washington State began to administer it to incoming freshman. However, men perform
much better than women on it. Some activists believe that the test is sexist and have
called on administrators to stop using it. However, others have noted that men perform
better in college at Washington State University, so the test is fair and predictive of
performance.
Immediately following the cop story, participants responded to four questions (“How
justified was the officer’s shooting?” [reversed], “How wrong was the person who was shot?”
[reversed], “Should the officer be punished?”, and “Should the family of the person who was
shot receive money?”) on 7-point scales from Not at all to Very much so, which were combined
into an index of belief that the cop was wrong,
= .80. Immediately following the test story,
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
22
participants responded to four questions (“How justified was the school in using the Graduate
Performance Test?” [reversed], “How right were activists in trying to get rid of the test?”, “Is the
test fair?” [reversed], and “Is the test sexist?”) on 7-point scales from Not at all to Very much so,
which were combined into an index of belief that the test is unfair,
= .88.
Participants rated their agreement with several statements relevant to victims’ groups
(Most police departments are racist, Islam is a religion of peace, Men are physically stronger
than women, Men are better at mathematics than women, The government should spy on
Muslims, Jokes about race are offensive, A woman’s proper role in society is in the kitchen, and
Women are smarter than men) on 7-point scales from 1= Not at all to 7= Very much so.
4
Results
As expected, more liberalism predicted stronger beliefs that the cop was wrong, r = .45, p
< .001, and stronger beliefs that the test is unfair, r = .24, p = .003. Moreover, and consistent
with predictions, stronger equalitarian beliefs partially mediated the influence of liberal ideology
on beliefs that the cop was wrong, 95% CI [.04, .23] and that the test is unfair, 95% CI [.06, .25].
One of these mediations is mapped in Figure 2 below, in which more liberal ideology predicted
stronger beliefs that the cop was wrong to shoot the Black man, and this was partially accounted
for by Liberals’ stronger equalitarian attitudes.
4
For purposes of upholding the cover story that the study was about political attitudes, two
additional statements were included (I think gays should be able to marry, and Abortion should
be legal). We had no a priori predictions regarding these items, but to satisfy curiosity, these
were the agreement rating and correlation with liberal ideology results for the former: M = 5.44,
SD = 2.13, r = .59, p < .001, and the latter: M = 4.80, SD = 2.36, r = .53, p < .001.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
23
Figure 2. Influence of ideology (higher values = more liberal) on beliefs that the cop was wrong,
mediated by equalitarianism.
We next examined the relationships between ideology and agreement with the statements
regarding victims’ groups and privileged groups. As can be seen in Table 2, more liberal
ideology was significantly positively related to beliefs that most police departments are racist
and that Islam is a religion of peace and significantly negatively related to beliefs that men are
physically stronger than women, that men are better at math than women, that the government
should spy on Muslims, and that a woman’s place in society is in the kitchen. There was also a
small (but not significant) negative relationship between liberalism and beliefs that jokes about
race are offensive. There was no significant relationship between ideology and beliefs that
women are smarter than men (later studies will suggest that both Conservatives and Liberals
prefer this conclusion). Higher equalitarian attitudes significantly mediated all relationships
except the relationship between ideology and beliefs that the government should spy on Muslims
and the non-significant relationship between ideology and beliefs that women are smarter than
men. Thus, other than these two exceptions, results were consistent with predictions that more
liberal ideology predicts more pro-victims’ groups attitudes and more anti-privileged groups
attitudes, and that these relationships are partially explained by stronger equalitarian beliefs.
b = .31, se = .04, p <.0001
Cop Wrong
Equalitarianism
Ideology
b = .39, se = .13, p =.0026
b = .27, se = .07, p =.0003
(b = .39, se = .06, p <.0001)
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
24
Table 2
Agreement with victims' groups statements, their correlation with (liberal) ideology,
and mediation of that relationship by equalitarianism in Study 1b
Mediation
Group
M
SD
r
p
95% CI
Most police departments are racist.
3.27
1.82
.42
<.001
.13, .32
Islam is a religion of peace.
3.98
1.83
.53
<.001
.06, .25
Men are physically stronger than women.
4.93
1.77
-.31
<.001
-.30, -.08
Men are better at mathematics than women.
2.50
1.57
-.20
.017
-.35, -.12
The government should spy on Muslims.
2.70
1.89
-.56
<.001
-.18, .05
Jokes about race are offensive.
5.08
1.80
-.13
.102
.09, .34
A woman's proper role in society is in the kitchen.
1.67
1.25
-.24
.003
-.21, -.06
Women are smarter than men.
3.39
1.54
.03
.742
-.05, .13
Note. 1 = not at all agree; 7 = agree very much so
Discussion
As predicted, more liberalism predicted greater opposition to using a test that favored
men and more unfavorable judgments of a police officer who shot an unarmed black person; and
these relationships were partially mediated by higher equalitarianism. More liberalism also
predicted more positive victims’ group attitudes and more negative privileged group attitudes,
and equalitarianism generally mediated these relationships. Studies 1a and 1b showed that
liberalism and equalitarianism were related to group attitudes in the expected ways. Studies 2
through 7 moved on to test the hypotheses that liberalism and equalitarianism predict biases
against information that portray privileged groups more favorably than victims’ groups.
Study 2
So far, liberal ideology predicts 1) beliefs that victims’ groups are treated more unfairly
by society, 2) that a cop shooting an unarmed black man was more wrong, 3) that it is more
unacceptable to use performance exams on which men outperform women, and 4) more
favorable attitudes toward victims’ groups/less favorable attitudes toward privileged groups.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
25
Liberals’ higher equalitarian attitudes at least partially accounted for nearly all these outcomes.
These validation studies suggest that Liberals’ have greater concern for victims groups. Thus,
this concern could be a potential source of liberal bias. In the remaining studies, we expanded
our investigation to test whether Liberals’ stronger equalitarian attitudes were related to bias
against biological group equality, especially when those threats indicated that privileged groups
score higher than victims’ groups on a socially valued trait.
Study 2 tested the prediction that Liberals would be less likely to trust, support, and
accept an exam on which men outperform women than an exam on which women outperform
men. As we noted in the introduction, Liberals desire to protect perceived victims’ groups;
therefore, they likely are more sensitive to potential threats to those groups. When a threat is
detected, we predict that many Liberals will shift from standard to motivated reasoning and will
evince bias. In this Study, we hypothesized that the test on which men perform better will be
perceived as a threat to many Liberals, therefore causing bias against it (i.e., causing them to
assess it differently from the same test when women are said to do better [no threat]).
Using standard methods to detect bias, we had participants read one of two vignettes
about a university’s use of a performance exam, and randomly assigned them to read either that
men outperform women or that women outperform men (on average). Participants then evaluated
whether it is acceptable to use the test. We predicted that liberal participants would be biased
such that they would rate the exam as less acceptable when men outperform women than when
women outperform men. We also expected that Liberals would rate the exam more unacceptable
than Conservatives when men outperform women. These results would indicate that 1) Liberals
evaluate information in a biased manner when that information could portray victims’ groups or
privileged groups in a more or less favorable light, and 2) Liberals (relative to Conservatives) are
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
26
particularly motivated to disparage information that appears to favor a privileged group over a
victims’ groups.
Method
Participants. U.S. participants (Mage = 36.80, SD = 12.75; 113 female) were recruited via
MTurk. We aimed for 100 participants per condition (200 total); 205 participated. In this study,
sample size was derived from the researchers’ personal experience conducting similar work. In
order to maintain sufficient power, we increased the number of participants per cell with each
increasingly complex experimental design, (100 per condition in studies with two groups
[Studies 2 and 3], 150 per condition in studies with three groups [Studies 4 and 5], and 200 per
condition in studies with four groups [i.e., 2 x 2 designs; Studies 6 and 7]). Participants were
slightly above the midpoint on liberalism (M = 4.30, SD = 1.66) and equalitarianism (M = 4.68,
SD = 1.02),
5
and these were positively correlated, r = .42, p < .001.
Procedure. As in Studies 1a and 1b, order of procedures was randomized. Equalitarian
attitudes,
= .90, and political ideology were measured with the same procedures as in Studies
1a and 1b. Participants also read a short vignette about a college entrance exam (below), and
were randomly assigned to read that either men outperform women or women outperform men:
In the past decade, the College Entrance Exam (CEE) has been given to high school
students. It has been shown to have remarkable accuracy at predicting academic
performance in college.
However, universities have been debating whether to use the exam or
5
In this study only, a slightly modified version of the scale was used, which replaced item 4 with
“Many people are biased against people, and such biases threaten society” and item 18 with
“With the right policies, we will increase equality in society”.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
27
not because women/(men), on average, score much higher than men/(women) on the
exam, leading to the acceptance of more women/(men) to college than men/(women).
Following this vignette, participants responded to three questions (“How much do you
think the test should be used?”, “How fair do you think the test is?”, and “How sexist do you
think the test is?” [reverse-scored]) on 7-point scales from 1= Not at all to 7= Very much so,
which were combined into an index of test acceptability,
= .85.
Results
For this and all subsequent studies, interactions could be computed with either the
continuous measure of ideology or by categorizing participants as Liberals, Moderates, and
Conservatives based on conceptual cut points. The former strategy retains all available
information, but the latter is easier to comprehend particularly as the designs get more
complicated in later studies. For these reasons, and for the sake of open reporting, we report the
results both ways. Note that across all studies, both analysis strategies yield similar
interpretations of the data, though in some cases, the continuous analyses have slightly larger
overall effect sizes or smaller p-values, especially for the relevant interaction effects.
Continuous. We regressed test acceptability ratings on the Sex condition, ideology
(centered), and the interaction, controlling for sex.
6
As can be seen in Table 3, there was a
significant main effect of Sex condition on test acceptability such that the test was considered
less acceptable if men outperform women than if women outperform men. There was also a main
effect of ideology such that liberalism predicted lower test acceptability. Somewhat consistent
6
Removing sex as a control does not affect the statistical significance of any effects.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
28
with predictions, there was a small, trending (but not statistically significant) interaction between
the condition and ideology.
Consistent with predictions, simple slopes one standard deviation above and below the
mean of political ideology revealed that liberal participants found the test significantly less
acceptable when men outperform women than vice versa (b = 1.02), t = 3.54, p = .001. In
contrast, conservative participants (one standard deviation below the mean) found the test
equally acceptable regardless of whether women outperform men or men outperform women (b
= .40), t = 1.38, p = .171, though they were still trending in the same direction as Liberals.
Examining the interaction another way, in the condition in which women outperform
men, there was virtually no effect of ideology on test acceptability (b = .01), t = 0.12, p = .902.
Both Liberals and Conservatives found the test reasonably acceptable (above the midpoint) if
women outperform men. However, in the condition in which men outperform women, more
liberal ideology predicted lower test acceptability (b = -.18), t = -2.10, p = .037.
Table 3
Test acceptability ratings regressed on Sex condition (0: Men Outperform,
1: Women Outperform), ideology, and the interaction, controlling for sex
β
t
p
95% CI
semipartial r
Sex
-0.28
-4.33
<.001
-1.28, -.48
-.28
Condition
.23
3.46
.001
.30, 1.11
.23
Ideology
-.19
-2.13
.034
-.34, -.01
-.14
Condition x Ideology
.13
1.52
.129
-.06, .43
.10
Categorical. We created a categorical ideology variable for Conservatives (those who
responded 1-3 on the 7-point ideology scale; n = 62), Moderates (those who responded 4; n =
57), and Liberals (those who responded 5-7; n = 86).
7
We analyzed the 2 (Sex condition) x 3
7
This coding scheme is how we created a categorical ideology variable in this study and all
subsequent studies.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
29
(categorical ideology) interaction on acceptability ratings in a Univariate Analysis of Variance
(ANOVA).
There was a significant main effect of Sex condition, indicating that all participants
objected more to a test favoring men than a test favoring women. The main effect of ideology
and the interaction were not significant. However, consistent with the continuous results, simple
contrasts revealed the largest (and a significant) difference between experimental conditions for
Liberals (see Table 4 and Figure 3). Specifically, Liberals rated the test as significantly more
acceptable if women outperform men than if men outperform women, p = .004, Cohen’s d = .64,
whereas Moderates, p = .507, Cohen’s d = .17, and Conservatives, p = .118, Cohen’s d = .44,
demonstrated no such difference (though note Conservatives were trending in a similar direction
as Liberals).
In the Women Outperform condition, no groups significantly differed, ps > .557. In the
Men Outperform condition, Liberals rated the test as marginally less acceptable than Moderates,
p = .099, and somewhat (though not significantly) less acceptable than Conservatives, p = .152;
whereas there were virtually no differences between Conservatives and Moderates in this
condition, p = .791.
Table 4
Categorical ideology, Sex condition (0: Men Outperform;
1: Women Outperform), and the interaction on test acceptability
F
p
p2
Condition
8.11
.005
.039
Ideology
0.78
.460
.008
Condition x Ideology
0.94
.393
.009
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
30
Figure 3. Test acceptability by Sex condition within each ideological group. Error bars are
standard errors.
Moderated Mediation and Mediations. Equalitarianism mediated the interactive effect
of Sex condition and (continuous) ideology on test acceptability, based on PROCESS model 5
(10,000 resamples; Hayes, 2013),
8
specifying ideology as the independent variable and Sex
condition as the moderator, 95% CI [-.24, -.07]. To model this interaction simply, we then tested
simple mediations within each condition. Confirming the results of the moderated mediation,
equalitarianism did not mediate the (non)effect of ideology on test acceptability in the condition
in which women outperform men, 95% CI [-.20, .01], but it did mediate the influence of
ideology on test acceptability in the condition in which men outperform women, 95% CI [-.36, -
.12]. As can be seen in Figure 4, higher equalitarianism fully accounted for the relationship
8
This is how we tested moderated mediation in this study and all subsequent studies.
3
3.5
4
4.5
5
5.5
6
Liberals
Moderates
Conservatives
Men Outperform
Women Outperform
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
31
between more liberal ideology and lower ratings of test acceptability in the condition in which
men outperform women on the test.
Figure 4. Influence of ideology (higher values = more liberal) on test acceptability, mediated by
equalitarianism in the condition in which men outperform women.
Discussion
Study 2 found a general pattern of biased evaluation. Across the full sample, participants
objected to a test more if men outperformed women than if women outperformed men.
Consistent with our predictions, this was strongest (and significant) among Liberals. Also
consistent with predictions, Liberals objected to the test more than Conservatives only in the
condition in which men outperformed women. However, the full interactions did not reach
statistical significance. The upcoming studies shed more light on this pattern.
We should address an important challenge to our argument. Perhaps Liberals are not
biased at all, but rather are using some Bayesian-type reasoning. More women than men are
going to college, and women tend to earn higher GPAs in college than men, so perhaps it is
rational to conclude that a college test that favored men is sexist and unfair (it contradicts real
b = .28, se = .05, p <.0001
Test acceptability
Equalitarianism
Ideology
b = -.79, se = .14, p <.0001
b = .03, se = .09, p =.717
(b = -.19, se = .09, p =.038)
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
32
base rates). Upcoming studies seek to address this limitation by examining sex differences in IQ
(men and women score similarly on IQ tests) and by using within-subjects designs.
It is also possible that Liberals were not biased against the validity of the test per se, but
rather concerned about the explicitly stated downstream consequences (i.e., that fewer women
would be admitted to college). If so, Liberals were not biased against the test, but were rationally
concerned about the potential deleterious consequences to women. To address this objection, the
upcoming studies avoided manipulating downstream consequences of differences and focused
only the stated differences themselves.
Another plausible objection to the bias argument is that Liberals were using a different
but equally rational prior that altered their response patterns when men outperform women: The
base rate of sexism. If society is more sexist against women than against men, then perhaps it is
rational to conclude that a test or policy that favors men is likely less fair and more sexist than a
test or policy that favors women. We address this objection as fully as we can in Studies 6 and 7.
The best methodological strategies to ensure that one is measuring bias are to use matched
materials, to ask questions about the matched information and not the manipulated information
(Ditto et al., 2019a), to use examples in which base rates go against the no bias explanation, and
to use within-subjects designs. We improved upon all these in upcoming studies.
Study 3
Study 3 was similar to Study 2 but focused on race instead of gender. Study 3 also sought
to minimize potential Bayesian counter-explanations for the bias by having participants evaluate
the credibility of identical scientific arguments that only differed in their conclusions. In both
conditions, participants read an argument about the discovery of a gene that was associated with
higher IQ scores and that may explain intelligence differences between Blacks and Whites. The
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
33
only difference between conditions was whether the gene explained why Blacks score higher on
IQ tests than Whites or why Whites score higher on IQ tests than Blacks.
We once again expected that Liberals would display bias such that they would evaluate
the credibility of the argument more unfavorably if the gene was said to explain why Whites
have higher IQs than Blacks than vice versa. We again expected that ideological differences in
argument credibility ratings would be largest in the condition that casts a victims’ group in a less
favorable light than a privileged group such that Liberals would be particularly motivated to
disparage information that suggests that Whites have higher IQs than Blacks (relative to
Conservatives).
Method
Participants. U.S. participants (Mage = 37.65, SD = 12.65; 118 female; 159 White, 17
Asian, 13 Latino, 12 Black, 1 Middle Eastern) were recruited via MTurk. As in Study 2, we
aimed for 100 participants per condition (200 total); 202 participated. Participants were slightly
above the midpoint on liberalism (M = 4.55) and equalitarianism (M = 4.81), and these were
positively correlated, r = .53, p < .001.
Procedure. Procedures were identical to Study 2 (equalitarian scale
= .92), except
participants read a different vignette and responded to different questions in response to the
vignette. This vignette was an ostensible The New York Times science article, which described
research about the discovery of a gene that might explain racial differences in IQ. We used a
racially neutral name, Tom Berry (and used this name in all studies that used a variation of this
vignette). Participants were randomly assigned to read that this gene might explain either why
Whites score higher on IQ tests than Blacks, or why Blacks score higher on IQ tests than Whites:
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
34
Researchers from a large research institution have discovered a gene that might explain
intelligence differences between Blacks and Whites. For many years, researchers have
found that Blacks/(Whites) score higher on certain intelligence tests than
Whites/(Blacks). Tom Berry and his colleagues have tried to find genetic causes for the
disparity in intelligence scores, arguing that environmental explanations cannot explain
the IQ gap. "There is simply no reasonable environmental explanation for the IQ gap that
we can find or that other researchers have proposed," Dr. Berry explained.
Berry and his team think they have an answer. They isolated a gene on the 21st
chromosome that is reliably associated with higher IQ scores. The gene polymorphism,
called THS-56RR, was first found in 1999, but researchers didn't know that it was related
to higher IQ scores. Berry and his team found that it was strongly related to IQ scores.
They also found that the gene is much more common in American Blacks/(Whites) than
Whites/(Blacks). "About 93% of Blacks/(Whites) carry the gene," Dr. Berry said,
"whereas only 10% of Whites/(Blacks) carry it. We really think this might explain the IQ
gap."
Participants responded to the news article on six questions (“How credible do you find
Dr. Berry’s argument?”, “Do you believe Dr. Berry’s argument?”, “Is Dr. Berry’s argument
racist?” [reversed], “Is Dr. Berry’s argument logical?”, “How important is this research?”, and
“Do you think we should fund more of this type of research?”) rated on 7-point scales from Not
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
35
at all to Very much so (first four questions) or Not at all to Extremely/Definitely, which were
combined into an index of argument credibility,
= .92.
Results
Continuous. We regressed argument credibility ratings on the Race condition, ideology
(centered), and the interaction. As can be seen in Table 5, there was a significant main effect of
the Race condition such that the argument that the gene could account for racial differences in
intelligence was considered more credible if the gene explained why Blacks are more intelligent
than Whites (M = 3.61, SD = 1.38) than if the gene explained why Whites are more intelligent
than Blacks (M = 3.15, SD = 1.59). There was also a main effect of ideology such that more
liberalism was associated with lower credibility ratings.
There was also a statistically significant interaction between the Race condition and
ideology. Consistent with predictions, simple slopes one standard deviation above and below the
mean revealed that more liberal participants found the argument more credible if the gene
explained why Blacks have higher IQ than Whites than if it explained why Whites have higher
IQ than Blacks (b = 1.04), t = 3.40, p = .001. In contrast, more conservative participants found
the argument equally credible regardless of whether it explained Blacks’ or Whites’ higher
intelligence (b = -0.12), t = -0.40, p = .693.
Examining the interaction another way, in the Blacks Higher condition, ideology was
unrelated to argument credibility ratings (b = 0.11), t = 1.16, p = .248. Both Liberals and
Conservatives found the argument reasonably credible if the gene explained why Blacks have
higher IQs than Whites. However, as predicted, in the Whites Higher condition, more liberal
ideology predicted lower argument credibility ratings (b = -0.22), t = -2.84, p = .005.
Table 5
Argument credibility ratings regressed on Race condition (0: Whites Higher;
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
36
1: Blacks Higher), ideology, and the interaction
β
t
p
95% CI
semipartial r
Condition
.15
2.13
.034
.04, .89
.15
Ideology
-.25
-2.91
.004
-.37, -.07
-.20
Condition x Ideology
.23
2.71
.007
.09, .58
.19
Categorical. We again created an ideological category variable (Conservatives n = 56,
Moderates n = 37, Liberals n = 109) and analyzed the 2 (Race condition) x 3 (categorical
ideology) interaction on credibility ratings in an ANOVA. There was no main effect of Race
condition, nor ideology, but similar to the continuous results, there was a marginal interaction
(see Table 6 and Figure 5).
All simple contrasts demonstrated the expected pattern of results. Specifically, Liberals
rated the argument as significantly more credible in the Blacks Higher condition than the Whites
Higher condition, p = .005, Cohen’s d = .58. This difference was smaller and not significant for
Moderates, p = .400, Cohen’s d = .58, and slightly (but not significantly) in the opposite
direction for Conservatives, p = .498, Cohen’s d = .20. In the Blacks Higher condition, no groups
significantly differed, ps > .134. In the Whites Higher condition, Liberals rated the argument as
significantly less credible than Moderates, p = .040, and Conservatives, p = .030; Moderates and
Conservatives did not differ, p = .856.
Table 6
Categorical ideology, Race condition (0: Whites Higher;
1: Blacks Higher), and the interaction on argument credibility
F
p
p2
Condition
1.90
.170
.010
Ideology
1.67
.192
.017
Condition x Ideology
2.44
.090
.024
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
37
Figure 5. Argument Credibility by Race condition within each ideological group. Error bars are
standard errors.
Moderated Mediation and Simple Mediations. We next tested whether equalitarianism
mediated the interactive effect of Race condition and ideology on argument credibility,
specifying ideology as the independent variable and Race condition as the moderator. As
expected, higher equalitarianism mediated the interactive effect, 95% CI [-.24, -.07]. Simple
mediations within each condition confirmed the results of the moderated mediation:
equalitarianism did not mediate the (non)effect of ideology on argument credibility in the Blacks
Higher condition, 95% CI [-.10, .10], but did mediate the influence of ideology on argument
credibility in the Whites Higher condition, 95% CI [-.42, -.16]. As can be seen in Figure 6,
higher equalitarianism fully accounted for the relationship between more liberal ideology and
lower ratings of argument credibility in the condition in which Whites were said to have a higher
average IQ than Blacks.
2
2.5
3
3.5
4
4.5
5
Liberals
Moderates
Conservatives
Whites Higher
Blacks Higher
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
38
Figure 6. Influence of ideology (higher values = more liberal) on argument credibility, mediated
by equalitarianism in the Whites Higher condition.
Discussion
Study 3 replicated the basic pattern of results of Study 2 with materials more resistant to
potential Bayesian-type counterarguments. Specifically, it found that Liberals, but not
Conservatives, were biased against genetic explanations for IQ differences between Whites and
Blacks when Whites were said to outperform Blacks (on average).
These findings do not rule out an objection about pervasive racism. That is, one could
argue that modern society is rife with racism and that therefore any explanation, any test, any
policy, that appears to disfavor Blacks is likely to be unfair and racist. This does seem a plausible
objection to some of our questions (e.g., “should this research be funded?”), but it seems less
plausible to raise this objection to other questions (e.g., “Is Dr. Barry’s argument logical?”). In
the upcoming studies, we dropped the objectionable questions, and in Study 6, we seek to
examine the plausibility of this alternate explanation with a within-subjects design.
Study 4
Study 4 sought to replicate and extend the results of Study 3 by including an Equal
condition, in which it was said that a gene explained individual differences in intelligence, that
b = .39, se = .05, p <.0001
Argument
Credibility
Equalitarianism
Ideology
b = -.70, se = .13, p <.0001
b = .06, se = .09, p =.523
(b = -.22, se = .08, p =.006)
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
39
the gene was found in equal degrees in both Blacks and Whites, and that this explains why
Blacks and Whites score similarly on intelligence tests. We added this condition to explore
whether Liberals are motivated to reject the conclusion that Whites have higher IQs than Blacks,
motivated to accept the conclusion that Blacks have higher IQs than Whites, or perhaps
motivated to reject both but to different degrees relative to an Equal condition. Consistent with
our hypothesis, we predicted that Liberals would find the Equal condition most credible,
followed by Blacks higher and then last by Whites higher. Regarding Conservatives, our main
prediction (as in previous studies) was that they would be more accepting of the argument when
the privileged group (here, Whites) is said to have higher IQs than the victims’ group (here,
Blacks) relative to Liberals, though we did not have predictions for Conservatives regarding
differences between conditions.
Method
Participants. U.S. participants (Mage = 36.96, SD = 12.34; 233 female; 341 White, 48
Asian, 34 Black, 28 Latino, 1 Middle Eastern) were recruited via MTurk. Because of the addition
of the Equal condition, we aimed for 150 participants per condition (450 total); 452 participated.
Participants were slightly above the midpoint on liberalism (M = 4.49) and equalitarianism (M =
4.70), and these were positively correlated, r = .54, p < .001.
Procedure. Methods were identical to Study 3 (equalitarian scale
= .92) with two
exceptions. First, we used only the one question from Study 3 that was the least vulnerable to
Bayesian counter-explanation: “Is Dr. Berry’s argument logical?” We also added two additional
items that should be minimally vulnerable to Bayesian counter-explanations: “How reasonable
do you find Dr. Berry’s argument?” and “How plausible is it that a gene could explain IQ
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
40
differences?”, which were combined into an index of argument credibility, = .91. Second, an
Equal condition was also included (pasted below).
9
Researchers from a large research institution have discovered a gene that might explain
intelligence similarities among Blacks and Whites. For many years, researchers have
found that Whites and Blacks score similarly on certain intelligence tests. Tom Berry and
his colleagues have tried to find genetic causes for intelligence scores, arguing that
environmental factors cannot explain IQ. "There is simply no reasonable environmental
explanation for IQ differences within races that we can find or that other researchers
have proposed," Dr. Berry explained.
Berry and his team think they have an answer. They isolated a gene on the 21st
chromosome that is reliably associated with higher IQ scores. The gene polymorphism,
called THS-56RR, was first found in 1999, but researchers didn't know that it was related
to higher IQ scores. Berry and his team found that it was strongly related to IQ scores.
They also found that the gene is equally common in American Whites and Blacks. "About
60-65% of both Whites and Blacks carry the gene," Dr. Berry said, "We really think this
might explain similarities in intelligence scores between them."
9
There were a few other trivial changes that apply to Studies 4-7: Order of procedures was fixed
rather than randomized (science article and DVs came first, then the equalitarianism scale, then
demographics), some unrelated and unreported demographic questions were removed (e.g.,
relationship status, sexual orientation), and open-ended suspicion probes and comment boxes
were added.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
41
Results
Continuous. We regressed argument credibility ratings on the Race condition dummy
coded with the Equal condition as the reference category, ideology (centered), and the
interactions. As can be seen in Table 7, there were significant main effects for both dummy
variables such that participants rated the arguments as less credible if the gene explained why
Whites have higher IQ than Blacks or if the gene explained why Blacks have higher IQ than
Whites (relative to the Equal condition). There was no main effect of ideology.
There was no significant interaction between the Blacks Higher dummy variable and
ideology. But, as expected, there was a significant interaction between the Whites Higher
dummy variable and ideology. Simple slopes at each level of the Whites Higher dummy variable
revealed that in the condition in which Whites were said to have a higher IQ than Blacks, more
liberal ideology predicted lower ratings of credibility (b = -.24), t = -3.76, p < .001. In the other
conditions, ideology was unrelated to credibility ratings (b = -.03), t = -.41, p = .680.
Table 7
Argument credibility ratings regressed on dummy coded Race conditions, ideology,
and the interactions
β
t
p
95% CI
semipartial r
Blacks Higher
-.21
-3.95
<.001
-1.02, -.34
-.18
Whites Higher
-.23
-4.43
<.001
-1.10, -.42
-.20
Ideology
-.04
-0.42
.673
-.18, .12
-.02
Black x Ideology
.02
0.35
.724
-.17, .24
.02
White x Ideology
-.14
-2.05
.041
-.40, -.01
-.09
Categorical. We again created an ideological category variable for Conservatives (n =
125), Moderates (n = 103), and Liberals (n = 224), and analyzed the 3 (Race condition: Whites
Higher, Blacks Higher, Equal) x 3 (categorical ideology) interaction on credibility ratings in an
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
42
ANOVA. There was a significant main effect of Race condition, a marginal main effect of
ideology, and a significant interaction (see Table 8).
Table 8
Categorical ideology, Race condition (0: Whites Higher;
1: Blacks Higher), and the interaction on argument credibility
F
p
p2
Condition
9.06
<.001
.039
Ideology
2.49
.085
.011
Condition x Ideology
2.57
.037
.023
As can be seen in Figure 7, there were no differences between Liberals, Moderates, and
Conservatives on argument credibility in the Equal or Blacks Higher conditions, ps > .344. In the
Whites Higher condition, Conservatives and Moderates did not differ (p = .648), but Liberals
rated the argument as less credible than both Conservatives (p = .003) and Moderates (p = .002).
Among Conservatives and Moderates, only the Equal and Blacks Higher conditions significantly
differed (ps = .010 and .025, respectively, Cohen’s ds = .57-.58) such that participants rated the
argument that a gene could explain similarities in intelligence among Blacks and Whites as more
credible than when it was said to explain why Blacks have higher IQ than Whites. The Whites
Higher condition fell between the other two conditions and did not significantly differ from
either the Blacks Higher or the Equal condition for Conservatives or Moderates, ps > .117,
Cohen’s ds = .31-.33. Among Liberals, all conditions significantly differed. Liberals rated the
argument as more credible in the Equal condition than the Blacks Higher condition, p = .038,
Cohen’s d = .36, and the Whites Higher condition, p < .001, Cohen’s d = .69, and more credible
in the Blacks Higher condition than the Whites Higher condition, p = .016, Cohen’s d = .39.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
43
Figure 7. Argument credibility by Race condition within each ideological group. Error bars are
standard errors.
Mediations. We next examined whether higher equalitarianism mediated the influence of
ideology on argument credibility ratings within each Race condition. In the Equal and Blacks
Higher conditions, equalitarianism did not mediate the (non-effect) of ideology on argument
credibility ratings, 95% CI [-.12, .09] and 95% CI [-.10, .13], respectively. As can be seen in
Figure 8, and consistent with all results thus far, in the Whites Higher condition, higher
equalitarianism fully mediated the influence of more liberal ideology on lower argument
credibility ratings, 95% CI [-.28, -.08].
2
2.5
3
3.5
4
4.5
5
Liberals
Moderates
Conservatives
Equal
Blacks Higher
Whites Higher
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
44
Figure 8. Influence of ideology (higher values = more liberal) on argument credibility, mediated
by equalitarianism in the Whites Higher condition.
Discussion
Results were mostly consistent with predictions. As in Studies 2 and 3, ideological
differences in argument credibility only emerged in the condition in which the privileged group
was portrayed more favorably, such that Liberals found the Whites Higher argument less
credible than Moderates and Conservatives. And higher equalitarianism mediated the influence
of ideology on lower credibility ratings in the Whites Higher condition. Liberals, Moderates, and
Conservatives did not significantly differ in their credibility ratings of the Blacks Higher or
Equal arguments.
Also consistent with predictions (and Studies 2-3), Liberals found the Whites Higher
argument less credible than the Equal and Blacks Higher arguments. Conservatives (and
Moderates) showed no (significant) difference in credibility ratings between the Whites Higher
and Blacks Higher arguments.
It may seem surprising that all ideological groups (Conservatives, Moderates, and
Liberals) rated the Equal argument as more credible than the other arguments (though note, not
significantly more than Whites Higher for Conservatives and Moderates). This suggests that all
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
45
groups have some preference for group equality, and perhaps simply Conservatives and
Moderates are somewhat more willing to accept that that might not be the case (or Liberals are
somewhat more unwilling), especially if those differences favor the privileged group. While
apparently people across the political spectrum preferred to hear that the races have equal IQs,
Liberals stood out in rejecting the message of higher average intelligence among Whites than
Blacks. They were readier to accept that Blacks have higher average IQs than Whites, whereas
Moderates and Conservatives showed a (non-significant) tendency toward the reverse.
Study 5
Study 5 explored biases when the targets are men (privileged group) and women
(victims’ group), instead of Blacks and Whites. Study 5 replicated the methods of Study 4
exactly, but manipulated sex rather than race. The objective psychometric facts would incline a
purely data-driven person toward regarding the two as roughly equal, but it may be equally
reasonable to conclude that men have slightly higher IQs than women or vice versa. Large-scale
comparisons of intelligence test performance suggest that adult men and women have nearly
equal intelligence, with the male mean being very slightly higher. Women outperform men in
school, whereas men slightly outperform women on the SAT. There is also a substantial
difference in variance, with more men at both extremes, and so someone exposed to more
exemplars of either extreme might generalize mistakenly.
However, we expected that people would answer based more on their prejudices than on
published IQ data. We predicted that Liberals in particular would evince bias such that they
would evaluate the Men Higher argument as less credible than the Equal or Women Higher
arguments, due to their protective concern for women as a victim class. Furthermore, we
expected Liberals to rate the Men Higher argument as less credible than Conservatives, and that
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
46
higher equalitarianism would mediate the influence of more liberal ideology on lower credibility
ratings in the Men Higher condition.
We were less confident and more uncertain about our predictions for Conservatives. But,
we suspected that Conservatives might demonstrate a slight preference for the Equal argument
(as in Study 4) over the other two arguments, and possibly also a slight preference for the
Women Higher argument over the Men Higher argument (as in Study 2).
Method
Participants. U.S. participants (Mage = 36.42, SD = 11.52; 254 female; 353 White, 35
Black, 32 Asian, 31 Latino, 2 Middle Eastern) were recruited via MTurk. We again aimed for
450 participants (150 per condition); 454 participated. Participants were slightly above the
midpoint on liberalism (M = 4.51) and equalitarianism (M = 4.78), and these were positively
correlated, r = .51, p < .001.
Procedure. Methods were identical to Study 3 (equalitarian scale
= .90; argument
credibility, = .91) with one exception: we manipulated which sex was said to have a higher IQ
(or that the sexes have roughly equal IQs) instead of which race.
Results
Continuous. We regressed argument credibility ratings on the Sex condition dummy
coded with the Equal condition as the reference category, ideology (centered), and the
interactions. As can be seen in Table 9, there was only a main effect of the Men Higher dummy
variable, such that participants rated the Men Higher argument as less credible than the other
arguments. No other effects were significant (including the expected Men Higher x ideology
interaction).
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
47
Nonetheless, the simple slopes at each level of the Men Higher dummy variable revealed
that in the condition in which Men were said to have a higher IQ than Women, more liberal
ideology predicted marginally lower credibility ratings (b = -.12), t = -1.90, p = .058 (consistent
with predictions). In the other conditions, ideology was unrelated to credibility ratings (b = -.07),
t = -1.11, p = .269, (consistent with predictions, though note this relationship was trending in the
same direction as the Men Higher condition, hence, the non-significant interaction).
Table 9
Argument credibility ratings regressed on dummy coded Sex conditions, ideology,
and the interactions
β
t
p
95% CI
semipartial r
Women Higher
-.04
-0.38
.474
-.43, .20
-.03
Men Higher
-.23
-4.38
<.001
-1.02, -.39
-.20
Ideology
-.09
-1.08
.280
-.20, .06
-.05
Women x Ideology
-.04
-0.58
.563
-.23, .13
-.03
Men x Ideology
-.04
-0.57
.573
-.23, .13
-.03
Categorical. We again created an ideological category variable for Conservatives (n =
132), Moderates (n = 82), and Liberals (n = 239), and analyzed the 3 (Sex condition: Men
Higher, Women Higher, Equal) x 3 (categorical ideology) interaction on credibility ratings in an
ANOVA. There was a significant main effect of Sex condition, a marginal main effect of
ideology, and again unexpectedly, no significant interaction (see Table 10).
Table 10
Categorical ideology, Race condition (0: Men Higher;
1: Women Higher), and the interaction on argument credibility
F
p
p2
Condition
7.84
<.001
.034
Ideology
2.99
.051
.013
Condition x Ideology
0.30
.876
.003
As can be seen in Figure 9, consistent with predictions and all studies thus far, there were
no differences between Liberals, Moderates, and Conservatives on argument credibility in the
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
48
Equal or Victims’ Group (here, women) Higher conditions, ps > .107. In the Men Higher
condition, there was a marginal main effect such that Liberals evaluated the argument as less
credible than Conservatives, p = .076, weakly consistent with predictions (and consistent with all
studies thus far and upcoming Study 7).
Liberals generally displayed the expected pattern of results: They rated the Equal
condition as the most credible, followed by Women Higher, followed by the Men Higher.
Liberals did not significantly differ between the Equal and Women Higher conditions, p = .310,
Cohen’s d = .17 (we did not have a strong prediction, here, but thought Liberals would rate the
Equal condition as most credible). Consistent with predictions, Liberals rated the Men Higher
argument as significantly less credible than both the Women Higher, p = .005, Cohen’s d = .41,
and Equal arguments, p <.001, Cohen’s d = .59.
Conservatives also did not differ between the Equal and Women Higher conditions, p =
.865, Cohen’s d = .04, but unexpectedly, rated the Men Higher argument as marginally less
credible than the Equal argument, p = .076, Cohen’s d = .39, and significantly less credible than
the Women Higher argument, p = .048, Cohen’s d = .45. Moderates did not significantly differ
between any of the sex conditions, ps > .185, Cohen’s ds .03- .36, but demonstrated the same
basic pattern as Conservatives and Liberals (i.e., Equal and Women Higher roughly equivalent,
and both higher than Men Higher),
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
49
Figure 9. Argument credibility by Sex condition within each ideological group. Error bars are
standard errors.
Mediations. We next examined whether equalitarianism scores mediated the influence of
ideology on argument credibility ratings within each Sex condition. As expected, and consistent
with Studies 2-4), in the condition in which the victims’ group was said to be higher (Women, in
this case), there was no significant mediation, 95% CI [-.06, .09].
As can be seen in Figure 10, and as expected, in the Men Higher condition, higher
equalitarianism fully mediated the marginal influence of more liberal ideology on lower
argument credibility ratings, 95% CI [-.23, -.02].
2.5
3
3.5
4
4.5
5
5.5
Liberals
Moderates
Conservatives
Equal
Women Higher
Men Higher
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
50
Figure 10. Influence of ideology (higher values = more liberal) on argument credibility,
mediated by equalitarianism in the Whites Higher condition.
Unexpectedly (and unlike Study 4), in the Equal condition, equalitarianism mediated the
influence of ideology on argument credibility ratings, 95% CI [.02, .16]; see Figure 11. The
meaning of this significant mediation was not immediately obvious to us. Consistent with
previous results, higher liberalism predicted higher equalitarianism, higher equalitarianism
predicted stronger agreement in the Equal condition (not particularly surprising), but more liberal
ideology predicted lower credibility ratings in the Equal condition (significantly so only after
controlling for equalitarianism). Across all reported studies, Liberals generally found the
vignettes about genetic differences less credible than Conservatives. Perhaps this mediation
pattern reflects this. Higher liberalism is related to lower credibility scores, but Liberals also
score higher in Equalitarianism than Conservatives and so want groups to be equal. Therefore,
when the Equalitarian score is put into the mediation analysis, and thus Liberals’ desire for
equality is accounted for, the negative relationship between Liberalism and argument credibility
becomes significant.
b = .35, se = .04, p <.0001
Argument
Credibility
Equalitarianism
Ideology
b = -.33, se = .15, p =.026
b = .00, se = .09, p =.958
(b = -.12, se = .07, p =.094)
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
51
Figure 11. Influence of ideology (higher values = more liberal) on argument credibility,
mediated by equalitarianism in the Equal condition.
Discussion
Results were partially consistent with predictions. As expected, Liberals appeared biased
against arguments that suggested that a privileged group scores higher on a socially valued trait
than a victims’ group: They rated the Men Higher argument as less credible than the Women
Higher and the Equal arguments. And as in Studies 2-4, ideological differences in credibility
ratings emerged only in the condition in which the privileged group was said to be higher such
that Liberals found the Men Higher argument (marginally) less credible than Conservatives, and
this was mediated by higher equalitarianism scores.
As in Study 4, all ideological groups generally rated the Equal condition as the most
credible. However, unlike Study 4, there were generally no differences in credibility ratings
between the Equal condition and the Victims’ Group (women) Higher condition, for any
ideological group. Also, and surprisingly, all groups rated the Privileged (men) Higher condition
as the least credible (significantly lower than the other two conditions for Liberals and
Conservatives, but not for Moderates). In Study 4, on the other hand, Conservatives and
Moderates showed a slight (but not significant) reverse effect such that they rated the Victims’
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
52
Group Higher (Black) condition as less credible than the Privileged Group Higher (White)
condition.
So far, our results have consistently shown that Liberals are biased against information
that suggests that a privileged group is higher in a socially valued trait than a victims’ group
relative to information that suggests that a victims’ group is higher or that the two groups are
equal. Our results have also consistently shown that higher liberalism scores predict lower
credibility ratings of vignettes that suggest that suggest that a privileged group is higher in a
socially valued trait than a victims’ group, and this has been consistently mediated by scores on
our equalitarianism scale.
However, the story for Conservatives is more mixed and more difficult to summarize. In
this study, Conservatives’ results looked like Liberals’ results. However, in previous studies that
used race instead of sex, Conservatives either evinced no bias or appeared to “favor” the
Privileged Group (White) over the Victims’ group (Black) (though they demonstrated the
strongest preference for Equal). There are multiple possible explanations for Conservatives’
pattern of results, such as that they are racially biased against Blacks and sexually biased against
men; or that they believe that which race or sex is said to be higher is a valid input into argument
credibility evaluations (see PRE principle); or that our results for Conservatives are false
positives (given the inconsistency of the size and direction of these differences for
Conservatives). Though we cannot address all these explanations assiduously in a few studies,
and though the primary focus of this paper is on Liberal bias, which has been very consistent and
predictable in all studieswe do explore them further in Studies 6 and 7.
Of course, there are possible objections to our interpretation of our results thus far, the
two most serious are these: (1) Perhaps the results do not show that Liberals are biased but rather
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
53
that they are using appropriate Bayesian reasoning; and/or (2) Perhaps the results do not show
that Liberals are biased but rather that they are appropriately skeptical of the powerful
(privileged groups) when they (or anyone) claim that their group (Whites or Men) scores higher
on a socially valued trait. One might contend that (1) powerful groups in society often forward
narratives, including even putatively scientific narratives, that cast them in a favorable light
while casting less powerful groups in a negative light and (2) Liberals are more sensitive to this
reality than Conservatives.
These alternative hypotheses are difficult to rule out entirely, but we believe that a
within-subjects design is the best tool to do so. Therefore, in Studies 6 and 7, we used such a
design. We explain the logic in more detail below.
Study 6
Study 6 sought to replicate Study 4 and attempt to rule out possible alternative
explanations by parlaying a within-subjects design. For the sake of simplicity, the Equal
condition was dropped from Study 4. All other materials were identical. As discussed in the
introduction, within-subjects designs are useful for studying bias because it allows us to ascertain
whether participants believe they should answer both vignettes consistently. Every participant
gets both vignettes. Some get the Whites Higher first; others get the Blacks Higher first. If they
believe that they should rate them consistently, then they should anchor their second response to
their first. This would suggest that they believe it is biased (or that it looks biased) to rate them
differently. Bias would manifest as an order effect such that if participants see preference
congruent information first, then they would rate both arguments higher (on average) than when
preference incongruent information came first (because they are anchoring their second response
to their first). On the other hand, if they don’t think they should answer them consistently,
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
54
because they think it is rational to let which group is said to be higher influence their judgments
(e.g., because it is right and rational to be skeptical of information that suggests that privileged
groups are higher), then we should not see an order effect, and conclude that perhaps this is not a
bias after all.
To see this more clearly, imagine that we used two vignettes describing identical research
procedures. In one, a scientist concluded, “A squirrel is larger than a bear.” And in the other,
he/she concluded, “A bear is larger than a squirrel.” And then we asked how credible each
procedure was. We might not expect an order effect because people believe that it is rational not
to answer these two statements consistently because one is clearly wrong and the other is clearly
correct. Now imagine two vignettes in which either a very attractive or a very unattractive
woman applied for an office job with the exact same résumé. And then we asked, “how qualified
is the candidate?” Here, we might expect an order effect because participants know that it would
be biased to rate the candidates differently, but they also might have a propensity to rate the
attractive candidate as more qualified. In this case, we would see that when they evaluate the
attractive candidate first, they would evaluate her as relatively qualified, and then upon viewing
the exact same résumé from a less attractive candidate, evaluate her similarly as high; but if they
evaluated the less attractive candidate first, they would evaluate her as relatively less qualified,
and then upon viewing the exact same résumé from the more attractive candidate, evaluate her
similarly as low. The presence of this kind of order effect can suggest that participants are not
basing their evaluations on different rational criteria for the two different conditions, but rather
that they believe it is unreasonable and biased to do so.
For Liberals, we expected an order effect such that they would evaluate both arguments
more favorably if they first read the Blacks Higher argument and then the Whites Higher
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
55
argument than if the arguments were presented in reverse order. We also expected Liberals to
rate both Race conditions more similarly within order condition than between order conditions,
which would indicate that Liberals at least believe it is irrational to evaluate the two arguments
differently, despite evincing this exact bias in the order effect.
For Conservatives, we expected a possible main effect of race (consistent with the
trending but non-significant patterns in Studies 3-4) such that they would rate the Whites Higher
argument somewhat more credible than the Blacks Higher argument. We did not have strong
predictions about whether there would be an order effect for Conservatives, but we did think the
presence or absence of it would be informative for understanding the underlying reasons for a
possible race effect for Conservatives. If we did observe an order effect, this would provide
evidence that Conservatives demonstrate a reverse bias as Liberals, with a preference for
information that portrays high status groups favorably over information that portrays low status
groups favorably (at least on racetheir patterns were trending in the opposite direction for sex).
If we did not observe an order effect, this might suggest that Conservatives have an intuition
about average differences in IQ scores between races (Hunt, 2011; Mackintosh, 2011) and thus
believe it is rational to treat these conditions differently.
Method
Participants. U.S. participants (Mage = 35.41, SD = 11.88; 421 female; 604 White, 83
Black, 75 Asian, 34 Latino, 5 Middle Eastern) were recruited via MTurk. We originally aimed
for 400 participants (401 participated) and analyzed the results after 401. The Order condition x
ideology interaction was trending in an informative direction, but was not statistically significant.
We then conducted a second wave of recruitment a few days later for 400 more participants (800
total); 803 participated. After recruiting these additional participants, observed power = .79 for
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
56
the Order x ideology interaction. Participants were slightly above the midpoint on liberalism (M
= 4.53) and slightly above the midpoint on equalitarianism (M = 4.70), and these were positively
correlated, r = .56, p < .001.
Procedure. Methods were identical to Study 3 with one exception (equalitarian scale,
= .92; Blacks Higher credibility,
= .93; Whites Higher credibility,
= .94): it was conducted
within subjects rather than between. Order of presentation was randomly assigned. After reading
the first argument, they received the direction below before receiving the second:
In the article you just read, we altered the direction of the IQ gap that Dr. Berry was
trying to explain and the results that Dr. Berry found. That is, we changed the article to
say that Dr. Berry was trying to explain why Whites(/Blacks) score higher on certain IQ
tests than Blacks(/Whites), and that he found that 93% of Whites(/Blacks) carry the
intelligence gene whereas only 10% of Blacks(/Whites) carry it.
In reality, Dr. Berry was trying to explain why Blacks(/Whites) score higher on certain
IQ tests than Whites(/Blacks), and he found that 93% of Blacks(/Whites) carry the
intelligence gene whereas only 10% of Whites(/Blacks) carry it.
On the next page, you will read the actual article as it was originally published, and
respond on the same three questions.
Results
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
57
We first entered credibility ratings into a general linear model, with Order condition
(between: Whites Higher First vs. Blacks Higher First), Race condition (within: Whites Higher
vs. Blacks Higher), ideology (centered), and all interactions as predictors. As can be seen in
Table 11, there was a significant main effect of Race condition such that the argument was
perceived as somewhat more credible when the gene explained why Whites score higher on
intelligence tests than Blacks (M = 3.70, SD = 1.72) than vice versa (M = 3.59, SD = 1.65). There
was no main effect of order. There was a main effect of ideology such that more liberal ideology
predicted lower argument credibility ratings. All two-way interactions and the three-way
interaction were statistically significant.
Table 11
The influence of the Race Condition (Whites Higher credibility; Blacks
Higher credibility), Order Condition (0: Whites Higher First; 1: Blacks
Higher First), ideology, and the interactions on argument credibility
F
p
p2
Race condition
26.52
<.001
.033
Order condition
1.49
.223
.002
Ideology
3.49
.002
.026
Race x Order
8.78
.003
.011
Race x Ideology
11.93
<.001
.083
Order x Ideology
2.22
.039
.017
Race x Order x Ideology
5.22
<.001
.038
Conservatives Only
Race condition
25.71
<.001
.113
Order condition
0.38
.540
.002
Race x Order
2.61
.108
.013
Moderates Only
Race condition
3.83
.052
.021
Order condition
2.20
.140
.012
Race x Order
4.49
.035
.024
Liberals Only
Race condition
12.71
<.001
.030
Order condition
5.47
.020
.013
Race x Order
0.11
.737
.000
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
58
As in previous studies, we broke the model down into categorical ideological groups. We
reran the model among only Conservatives (n =204), among only Moderates (n = 183), and
among only Liberals (n =414) with Order condition (between: Whites Higher First vs. Blacks
Higher First), Race condition (within: Whites Higher vs. Blacks Higher), and the two-way
interaction as predictors. As predicted, and as can be seen in Figure 12, Liberals displayed an
order effect such that they rated both arguments (averaged) as more credible if they read the
Blacks Higher argument first and then the Whites Higher (M = 3.63, SD = 1.63) than when the
arguments were presented in the reverse order (M = 3.26, SD = 1.65). However, Liberals also
display a main effect of race such that they rated the Blacks Higher argument as more credible
(M = 3.51, SD = 1.64) than the Whites Higher argument (M = 3.37, SD = 1.67). As can be seen
in Figure 12, simple contrasts revealed that Liberals rated both the Blacks Higher argument, p =
.028, and the Whites Higher argument, p = .020, as more credible in the Blacks Higher First
condition than the Whites Higher First condition. Liberals also rated the argument more credible
in the Blacks Higher condition than the Whites Higher condition regardless of which argument
came first, ps < .023. Note that magnitude of the difference between the two Order conditions
within each Race condition was more than double the magnitude of the difference between each
Race condition within each Order condition. In other words, within each order condition,
Liberals evaluated the Blacks Higher and Whites Higher arguments more similarly to each other
than how similarly they rated the exact same Blacks Higher argument across order conditions
and how similarly they rated the exact same Whites Higher argument across order conditions.
Among Conservatives, there was only a significant main effect of Race condition, such
that Conservatives evaluated the argument as more credible in the Whites Higher condition (M =
4.25, SD = 1.75) than in the Blacks Higher condition (M = 3.67, SD = 1.69), somewhat similarly
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
59
to Studies 3-4, which found trending but non-significant effects in the same direction. The Order
effect and interaction were not significant (see Table 11). As can be seen in Figure 12, simple
contrasts revealed that Conservatives rated the argument more credible in the Whites Higher
condition than the Blacks Higher regardless of which argument came first, ps < .014, and the
order condition had no significant influence on credibility ratings for either the Blacks Higher or
Whites Higher argument, ps > .178.
Among Moderates, there was a marginal main effect of Race condition, such that they
evaluated the argument as more credible in the Whites Higher condition (M = 3.82, SD = 1.65)
than the Blacks Higher condition (M = 3.68, SD = 1.63), similar to Conservatives. The Order
condition was not significant, but the interaction was. As can be seen in Figure 12, simple
contrasts revealed that Moderates rated the Whites Higher argument as more credible when they
saw it second than when they saw it first, p = .044, but evaluated the Blacks Higher argument as
equally credible regardless of order of presentation, p = .241. When the Whites Higher argument
was presented first, Moderates rated the Blacks Higher and Whites Higher arguments roughly
equally, p = .910, but when the Blacks Higher argument was presented first, Moderates rated the
Whites Higher argument as more credible than the Blacks Higher argument, p = .004.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
60
Figure 12. Argument credibility by Race and Order conditions within each ideological group.
Notes. Within each ideological group, the center two bars are credibility ratings of the Blacks
Higher argument and the outer two bars are ratings of the Whites Higher argument; the left two
bars are ratings within the condition in which the Whites Higher argument came first and the
right two bars are ratings within the condition in which the Blacks Higher argument came first.
Moderated Mediations and Mediations. We next tested whether equalitarianism
mediated the interactive effect of Order condition and ideology on argument credibility (within
each Race condition), specifying ideology as the independent variable and Order condition as the
moderator. For Whites Higher credibility ratings, equalitarianism mediated the interactive effect,
95% CI [-.21, -.11]. Unexpectedly, there was also a smaller but significant moderated mediation
for argument credibility in the Blacks Higher condition, 95% CI [-.12, -.02].
We then ran simple mediations within each Order condition. Consistent with the results
of the moderated mediation, equalitarianism mediated the effect of ideology on Whites Higher
argument credibility when they read the Whites Higher argument first, 95% CI [-.27, -.11]. As
can be seen in Figure 13, higher equalitarianism accounted for the relationship between more
2
2.5
3
3.5
4
4.5
5
Liberals
Moderates
Conservatives
Whites Higher First
Blacks Higher Second
Blacks Higher First
Whites Higher Second
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
61
liberal ideology and lower ratings of argument credibility when Whites were higher and that
condition came first.
Figure 13. Influence of ideology (higher values = more liberal) on argument credibility,
mediated by equalitarianism in the Whites Higher condition when this argument came first.
These relationships were somewhat smaller, but generally similar when they read the
Whites Higher argument second, 95% CI [-.19, -.07], such that higher equalitarianism mediated
the relationship between more liberalism and lower ratings of argument credibility that Whites
are higher (see Figure 14).
Figure 14. Influence of ideology (higher values = more liberal) on argument credibility,
mediated by equalitarianism in the Whites Higher condition when this argument came second.
b = .39, se = .03, p <.0001
Argument
Credibility
Equalitarianism
Ideology
b = -.49, se = .09, p <.0001
b = -.09, se = .06, p =.148
(b = -.27, se = .05, p <.001)
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
62
On the Blacks Higher outcome, equalitarianism did not mediate ideology on argument
credibility when the argument came first 95% CI [-.12, .00], but did when the argument came
second 95% CI [-.17, -.01]. As can be seen in Figure 15, higher equalitarianism fully accounted
for the relationship between more liberal ideology and lower ratings of argument credibility
when Blacks were said to be higher and that argument came second.
Figure 15. Influence of ideology (higher values = more liberal) on argument credibility,
mediated by equalitarianism in the Blacks Higher condition when this argument came second.
Discussion
For Liberals, the results were almost exactly consistent with predictions derived from
equalitarianism. They evinced an order effect such that they rated both arguments as more
credible when they received the preference consistent argument (Blacks Higher) first than when
the arguments were presented in reverse order. Within each order condition, there were only
small differences between the Blacks Higher and Whites Higher arguments (though, they did
consistently rate the Blacks Higher argument as slightly more credible), whereas there were
larger differences in argument credibility ratings for the identical Blacks Higher and Whites
Higher arguments between order conditions. This suggests that Liberals believe that the race of
the higher IQ group should not (much) affect their assessment of the argument’s credibility.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
63
However, despite this, our previous results and the order effect in this study show that the race of
the higher group does in fact affect their rating. This supports our contention that the difference
in credibility ratings between race conditions is the result of motivated cognition and constitutes
a bias. Furthermore, and consistent with previous results, more liberal opposition to the Whites
Higher argument was again mediated by higher equalitarianism.
Unexpectedly, we also found that higher equalitarianism mediated the relationship
between more liberal ideology and lower ratings of argument credibility when Blacks were said
to be higher and that argument came second. It might seem surprising at first that Liberals found
it less credible that Blacks have higher IQ than Conservatives did, but it appears consistent with
our theory (though, we did not predict it). When the Blacks Higher argument came second,
participants had already read the Whites Higher argument (which Liberals had evaluated as
relatively non-credible). This lowered the anchor point for Liberals (compared to Conservatives).
When Liberals rated the Whites Higher argument as non-credible, presumably in an effort to
maintain consistency, they then rated the second argument as less credible than did
Conservatives, even though it favored a victims’ group (and, indeed, in the Blacks Higher first
condition, Liberals rated the Blacks Higher argument as somewhat [though not significantly]
more credible than did Conservatives). Therefore, equalitarianism mediated the relationship
between liberal ideology and rating the Blacks Higher argument as less credible when
participants had already read the Whites Higher vignette.
For the first time, Conservatives displayed a significant effect of Race condition, such
that they evaluated the Whites Higher argument as more credible than the Blacks Higher
argument. Though this effect was not significant for Conservatives in Studies 3 and 4, the
difference was in the same direction in those studies, and the difference was fairly large in the
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
64
present study, so this is likely to be a real and replicable effect for Conservatives. The meaning
of this difference is not obvious. Perhaps one’s first reaction to the result would be to accuse
Conservatives of anti-Black bias. And this is certainly possible; however, other results are not so
consistent with an anti-Black bias interpretation. First, in previous studies, Conservatives rated
the Equal condition as slightly more (though not statistically significantly more) credible than the
Whites Higher condition, which is hard to square with a posited anti-Black bias (why then would
they not rate the argument that said Whites score higher than Blacks the most favorably?). Still,
they did rate Blacks Higher as the least credible, which, one might argue, does suggest a kind of
anti-Black bias. But second, Conservatives did not display an order effect, suggesting that they
thought it was rational (or defensible) to rate the stories differently. And this is congruent with
current psychometric data, which show that Whites score somewhat higher than Blacks on
intelligence tests, on average (Hunt, 2011; Mackintosh, 2011; though note, the underlying
reasons for this difference are still debated). Therefore, especially considering they displayed no
order effect, one could argue that Conservatives’ credibility ratings are rational from a Bayesian
perspective. One better way to test explanations for the Conservatives’ results is to choose a trait
on which Whites and Blacks score roughly equally or where differences are ambiguous. If
Conservatives rate a Whites Higher condition as more credible in such an experiment, it would
support an anti-Black bias hypothesis. In the current paper, we were more interested in Liberals,
so we did not run such tests and can therefore only make suggestions and theory-based
arguments. Future studies should fill this lacuna.
Study 7
In Study 6, we provided evidence of Liberal bias against information that appears to
favor a privileged group over a victims’ group because Liberals demonstrated an order effect,
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
65
which weakens the force of Bayesian objections. In the current study, we extended the
investigation further by using the same design but different examples of privileged and victims’
groups: men and women. All methods were identical to Study 6 except instead of manipulating
which race was said to perform better on certain IQ tests, we manipulated which sex was said to
perform better on certain IQ tests (men vs. women). We again predicted an order effect for
Liberals such that they would rate both arguments are less credible when they read the Men
Higher (privileged group) argument first than when they read the Women Higher (victims’
group) argument first. We also expected that equalitarianism would mediate the influence of
more liberal ideology on lower argument credibility ratings that men are more intelligent than
women when this argument came first. We did not have strong predictions about Conservatives.
We guessed that they would roughly match Liberals’ response pattern for the Sex condition (as
they did in Study 3), but we did not know whether to expect an order effect for Conservatives.
Method
Participants. U.S. participants (Mage = 35.84, SD = 12.05; 417 female; 625 White, 62
Black, 67 Asian, 49 Latino, 2 Middle Eastern) were recruited via Mturk. We again aimed for 800
participants total; 805 participated. Participants were slightly above the midpoint on liberalism
(M = 4.54) and slightly above the midpoint on equalitarianism (M = 4.72), and these were
positively correlated, r = .57, p < .001.
Procedure. This study was preregistered: http://aspredicted.org/blind.php?x=rz2fv9. We
followed this preregistration exactly, with the exception that we said that we would report the
results for Moderates in the supplemental materials only, but instead we report them in the main
text as we did in previous studies (though note, we had no a priori predictions about Moderates,
nor are they the focus of this paper). Methods were identical to Study 6 (equalitarianism scale
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
66
= .92; Women Higher credibility = .92; Men Higher credibility = .92) with one exception:
sex was manipulated instead of race (i.e., the words “Whites” and “Blacks” were swapped with
the words “men” and “women”).
Results
We first entered credibility ratings into a general linear model, with Order condition
(between: Men Higher First vs. Women Higher First), Sex condition (within: Men Higher vs.
Women Higher), ideology (centered), and all interactions as predictors. As can be seen in Table
12, there was a significant main effect of Sex condition such that the argument was perceived as
somewhat more credible when the gene explained why women score higher on intelligence tests
than men (M = 4.15, SD = 1.50) than vice versa (M = 3.90, SD = 1.53). There was no main effect
of order. There was a main effect of ideology such that more liberal ideology predicted lower
argument credibility ratings. All two-way interactions were significant or marginal. There was no
significant three-way interaction.
To dissect the three-way interaction further, we again reran the model among only
Conservatives (n =229), only Moderates (n =163) and only Liberals (n =413) with Order
condition (between: Men Higher First vs. Women Higher First), Sex condition (within: Men
Higher vs. Women Higher), and the two-way interaction as predictors. These results are
presented in Table 12 and Figure 16.
Among Liberals, there was a significant main effect of Sex condition, such that Liberals
evaluated the argument as more credible when the gene explained why women score higher on
some intelligence tests than men (M = 4.04, SD = 1.55) than vice versa (M = 3.65, SD = 1.56).
As predicted, there was also a significant effect of Order condition such that Liberals evaluated
the arguments as more credible when they read the Women Higher argument first (M = 4.09, SD
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
67
= 1.47) than when they read the Men Higher argument first (M = 3.62, SD = 1.59). There was
also a marginal interaction such that the difference between the Sex conditions was larger when
participants read the Men Higher argument first. As can be seen in Figure 16, simple contrasts
revealed that Liberals who read the Women Higher argument first rated both the Women Higher
argument, p = .010, and the Men Higher argument, p < .001, as significantly more credible than
those who read the Men Higher argument first. Liberals also rated the Women Higher argument
as more credible in both order conditions, p < .001. Note that the magnitude of the difference in
credibility ratings between the Men Higher condition and Women Higher condition when each
argument came first was nearly ten times the difference between these conditions than when each
came second, which demonstrates a clear attempt to anchor the second judgment to the first.
Among Conservatives, there was a significant main effect of Sex condition, such that
Conservatives evaluated the argument as more credible when the gene explained why women
score higher on some intelligence tests than men (M = 4.42, SD = 1.33) than vice versa (M =
4.28, SD = 1.39). Conservatives displayed similar patterns in Study 5 and Study 2 (though it was
only significant in Study 5). As in Study 6, the Order condition was not significant for
Conservatives. Unexpectedly, there was a significant two-way interaction between Sex and
Order, such that there was only a significant effect of Sex in the Men Higher First condition.
When Conservatives read the Men Higher argument first and then the Women Higher argument,
they rated the Women Higher argument as significantly more credible, p = .001, than the Men
Higher argument. When Conservatives read the Women Higher argument first and then the Men
Higher argument, they rated the arguments as equally credible, p = 1.00. There were no
significant differences in credibility ratings in the Men Higher or Women Higher conditions
between order, ps > .247. Across Studies 2, 5, and 7, Conservatives demonstrated a somewhat
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
68
consistent pattern such that they rated the Women Higher argument as either more credible than
the Men Higher argument, or rated the arguments as equally credible. Conservatives never rated
the Men Higher argument as more credible (nor did any other group).
Among Moderates, neither main effects nor the interaction approached significance, ps >
.658. Moderates rated all arguments virtually equally credible. Thus, across the three studies that
manipulated sex, Moderates consistently rated either the Women Higher argument as more
credible than the Men Higher argument, or rated the two arguments as equally credible (similar
to Conservatives).
Table 12
The influence of the Sex Condition (Men Higher credibility; Women
Higher credibility), Order Condition (0: Men Higher First; 1: Women
Higher First), ideology, and the interactions on argument credibility
F
p
p2
Sex condition
36.17
<.001
.044
Order condition
1.30
.255
.002
Ideology
4.47
<.001
.033
Sex x Order
6.70
.010
.008
Sex x Ideology
3.02
.006
.022
Order x Ideology
2.07
.055
.015
Sex x Order x Ideology
0.72
.631
.005
Conservatives Only
Sex condition
5.44
.021
.023
Order condition
0.10
.747
.000
Sex x Order
5.44
.021
.023
Moderates Only
Sex condition
0.20
.659
.001
Order condition
0.01
.938
.000
Sex x Order
0.02
.880
.000
Liberals Only
Sex condition
61.70
<.001
.131
Order condition
11.28
.001
.027
Sex x Order
3.09
.079
.007
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
69
Figure 16. Argument credibility by Sex and Order conditions within each ideological group.
Notes. Within each ideological group, the center two bars are credibility ratings of the Women
Higher argument and the outer two bars are ratings of the Men Higher argument; the left two
bars are ratings within the condition in which the Men Higher argument came first and the right
two bars are ratings within the condition in which the Women Higher argument came first.
Moderated Mediations and Mediations. We next tested whether equalitarianism
mediated the interactive effect of Order condition and ideology on argument credibility (within
each Sex condition), specifying ideology as the independent variable and Order condition as the
moderator. For Men Higher argument credibility, equalitarianism mediated the interactive effect,
95% CI [-.14, -.05]. There was no significant moderated mediation for Women Higher argument
credibility, 95% CI [-.04, .04].
To model these interactions simply, we then tested simple mediations within each Order
condition. As predicted, equalitarianism mediated the influence of ideology on Men Higher
argument credibility when they read the Men Higher argument first, 95% CI [-.19, -.07]; see
2.5
3
3.5
4
4.5
5
5.5
Liberals
Moderates
Conservatives
Men Higher First
Women Higher Second
Women Higher First
Men Higher Second
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
70
Figure 17. Confirming the results of the moderated mediation, equalitarianism did not mediate
the effect of ideology on Men Higher argument credibility when they read the Men Higher
argument second, 95% CI [-.11, .03].
Figure 17. Influence of ideology (higher values = more liberal) on argument credibility,
mediated by equalitarianism in the Men Higher condition when this argument came first.
As expected, on the Women Higher outcome, equalitarianism did not mediate ideology
on argument credibility in either Order condition, Women Higher First 95% CI [-.04, .09], Men
Higher First 95% CI [-.08, .04].
Discussion
Liberals displayed the expected pattern of results exactly. Specifically, they again
demonstrated an order effect such that they evaluated both arguments more favorably when they
received the preference congruent argument (Women Higher) first than when they received the
preference incongruent argument (Men Higher) first. Even so, they still consistently rated the
Women Higher argument as more credible than the Men Higher argument in both order
conditions, particularly so when they read the Men Higher argument first (i.e., they were willing
to significantly adjust their credibility rating up when the conclusions changed and the argument
said women were actually higher). This suggests that Liberals might think it is acceptable (and
b = .33, se = .03, p <.0001
Argument
Credibility
Equalitarianism
Ideology
b = -.39, se = .08, p <.0001
b = -.12, se = .05, p =.014
(b = -.24, se = .04, p <.001)
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
71
rational) to permit the direction of the sex difference to influence their judgments somewhat,
though not to the extent that sex actually does influence their judgments as demonstrated by the
roughly 2-3 times greater difference in credibility ratings between the arguments when each was
presented first than the differences between the arguments within each order condition. Also as
predicted, and consistent with all previous results, higher equalitarianism mediated the influence
of more liberal ideology on lower argument credibility ratings in the Privileged Group (men)
Higher condition.
Again, results were somewhat less clear for Conservatives. Differences were generally
small. Participants who read the Women Higher argument first rated both arguments virtually
identically. Participants who read the Men Higher argument first demonstrated a small trend
similar to the pattern for Liberals (and the previous studies) such that they adjusted their rating
up somewhat when the argument conclusions changed and said women are actually higher.
However, there was no main effect for order, which would indicate a bias.
Moderates rated all arguments virtually identically regardless of Sex condition or Order
of presentation.
A possible limitation of these within-subjects studies is that the instruction between the
two arguments may have made participants suspicious of the experimenter, thus influencing their
evaluations. It seems that if this were the case, participants would have generally evaluated the
second argument less favorably than the first (suspicion of the argument should make the
argument seem less credible), yet we did not observe this pattern in either Study 6 or 7, thus we
believe this is unlikely. Nonetheless, we hope readers note this limitation in their evaluations of
these studies, and we hope future researchers will try other “cover stories” in within-subjects
designs to rule this possibility out more thoroughly.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
72
Results Summary and Meta-analyses
Table 13 below summarizes the results of all simple contrasts (with effect sizes) between
experimental conditions among Conservatives, Moderates, and Liberals. The table lists which
group was marginally or significantly unfavored by each ideological group (relative to the
favored group). By unfavored, we mean that participants rated vignettes that said that this group
was higher on a socially valued trait (generally, IQ) as less credible than the comparison
condition (either groups are equal or the other group was higher). Although the term “unfavored”
is slightly clumsy, it allowed us to condense a great deal of information into one table.
In general, neither Conservatives nor Liberals appeared to desire that one group perform
better than another group (on a socially valued trait). Rather, both seemed to prefer that both
groups be equal. However, relative to Conservatives, and relative to information that portrayed
victims’ groups more favorably, Liberals were averse to information that portrayed privileged
groups more favorably than victims’ groups. This is the clear and consistent pattern observed in
Table 13. But, and importantly, Liberals didn’t seem biased for information that favored victims’
groups (Blacks, women). Instead, our results support our contention that Liberals prefer that all
demographic groups be roughly similar on socially valued traits. However, if demographic
groups are not similar, Liberals seem particularly averse to the notion that the privileged group
would have a more desirable quality. Despite some inconsistencies across studies for both
Moderates and Conservatives, the meta-analyses below allowed us to identify some likely real
patterns for these groups.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
73
Table 13
Simple contrasts and Cohen's d between indicated binary conditions within each categorical ideological group
for each experimental study
Study
Bias Effect
Ideological Group
Conservatives
Moderates
Liberals
Unfavored
Unfavored
Unfavored
Group
p
d
Group
p
d
Group
p
d
Study 2
Sex Condition
.44
.17
Men
**
.64
Study 3
Race Condition
-.20
.28
Whites
**
.58
Study 4
Race (White/Black)
-.33
-.31
Whites
*
.39
Race (Equal/Black)
Blacks
*
-.58
Blacks
*
-.57
Blacks
*
-.36
Race (Equal/White)
.26
.24
Whites
*
.69
Study 5
Sex (Men/Women)
Men
*
.45
.33
Men
**
.41
Sex (Equal/Women)
.04
-.03
-.17
Sex (Equal/Men)
Men
+
.39
.36
Men
***
.59
Study 6
Race Condition
Blacks
***
-.34
Blacks
+
-.09
Whites
***
.08
Order Condition
-.08
.21
Whites
*
.22
Race Within BH1
Blacks
***
-.50
Blacks
**
-.17
Whites
*
.08
Race Within WH1
Blacks
*
-.22
-.01
Whites
**
.10
BH Between Order
-.19
.12
Whites
*
.21
WH Between Order
.03
Whites
*
.30
Whites
*
.23
Study 7
Sex Condition
Men
*
.10
.03
Men
***
.25
Order Condition
-.04
.01
Men
*
.31
Sex Within WH1
.00
.03
Men
***
.21
Sex Within MH1
Men
**
.21
.01
Men
***
.29
WH Between Order
-.16
.02
Men
**
.26
MH Between Order
.06
.00
Men
***
.37
Notes. Blank cell = No preference at p >.10; +p <.10, *p <.05, **p <.01, ***p <.001.
Shaded cell = Privileged group unfavored; Unshaded cell (if not blank) = Victims' group unfavored.
In Study 6, BH1 = Blacks Higher First condition; WH1 = Whites Higher First condition;
BH = Blacks Higher evaluation; WH = Whites Higher evaluation.
In Study 7, WH1 = Women Higher First condition; MH1 = Men Higher First condition;
WH = Women Higher evaluation; MH = Men Higher evaluation.
Positive Cohen's ds indicate privileged group unfavored; Negative indicates victims' group unfavored on relevant
comparisons. Bold indicates a Cohen's d > .199 (or -.199), the threshold for a "small effect."
Meta-Analyses
As a last step, we conducted 12 mini meta-analyses of the interaction effect between the
experimental manipulations and continuous political ideology as well as of the condition effect
among categorical Conservatives, Moderates, and Liberals for Studies 2 through 7.
Method. We used procedures outlined by Goh, Hall, and Rosenthal (2016). To make
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
74
effect sizes comparable across studies, we dropped the equal conditions from Studies 4 and 5,
and we used only participants’ first evaluation in Studies 6 and 7 so they could be treated as a
between-subjects designs as in Studies 2 through 5. For the interaction effects between the
experimental manipulations and continuous ideology, we used the semipartial rs of the
interactions as effect size estimates. To test the bias effect size among Conservatives, Moderates,
and Liberals separately, we computed r effect sizes from the Ms, SDs, and ns for each
experimental condition in each study. The r effect sizes were then Fisher’s Z transformed to rz.
We used random effects to meta-analyze the overall interactions and overall effect sizes among
Conservatives, Moderates, and Liberals because there were at least two kinds of studies: those
that manipulated race and those that manipulated gender.
We then conducted fixed effects meta-analyses to test the interaction effect and
manipulation effect sizes among each ideological group only on the subset of studies that
manipulated sex (Studies 2, 5, and 7) and on the subset that manipulated race (Studies 3, 4, and
6) separately. Fixed effects is appropriate for these meta-analyses because the studies were very
similar within manipulation type, and thus weighting by sample size is preferred (see Goh et al.,
2016). The rzs were weighted and averaged using the formula: Weighted z = Σ ([N-3] rz) / Σ
(N-3). To estimate statistical significance, we used the Stouffer’s Z test, in which the p values for
each effect size were converted to Zs, combined using the formula: Zcombined = Σ Z / sqrt(k), and
then converted back to ps.
Results. For the fixed effects analyses, we conducted four single-sample t-tests on the
rzs. Consistent with hypotheses, there was a significant effect of the interaction between the
experimental manipulations and continuous political ideology (k=6, n=2,617), rz = .12,
t(5)=4.50, p=.006, 95% CI [.05, .19]. Also consistent with hypotheses, among Liberals, there was
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
75
a significant overall effect of the experimental manipulations (k=6, n=1,320), rz = .23, t(5)=
9.19, p<.001, 95% CI [.17, .30]. Among Conservatives, there was no overall significant effect of
the experimental manipulations (k=6, n=729), rz = .00, t(5)=-0.02, p=.983, 95% CI [-.20, .19].
And among Moderates, there was no overall significant effect of the experimental manipulations
(k=6, n=568), rz = .01, t(5)=0.13, p=.899, 95% CI [-.12, .14].
Among studies that manipulated sex, there was a significant effect of the interaction (k=3,
n=1,313), rz = .10, p=.004. This significant interaction effect reflected the relatively larger effect
size among Liberals than Conservatives. There was a small but significant effect of the sex
manipulation on Conservatives (k=3, n=380), rz = .10, p=.016, and a larger one among Liberals
(k=3, n=654), rz = .27, p<.00001. There was no significant effect of the sex manipulation among
Moderates (k=3, n=279), rz = .06, p=.235.
Among studies that manipulated race, there was again a significant effect of the
interaction (k=3, n=1,304), rz = .16, p<.00001. There was again no significant effect among
Moderates (k=3, n=289), rz = -.01, p=.492 Conservatives displayed a reverse effect such that
they evaluated the argument more positively when it favored Whites than when it favored Blacks
(k=3, n=349), rz = -.18, p=.003. And Liberals displayed the hypothesized effect (k=3, n=666), rz
= .18, p<.00001.
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
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
76
vignettes that stated that one group outperformed another. Second, they were consistently biased
against results that favored a privileged group over a victims’ group (either Whites over Blacks
or men over women). 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 liberalism 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 empirically unexploredbut not undiscussedsource of
bias. Many scholars have notedsome lamenting and some championingthat 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 resistance to human biological variation
and powerful biases about victims’ groups. Therefore, at minimum, our theory is a priori
plausible. But it is also largely congruent with many previous analyses of Liberals and not only
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
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
77
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]), Liberals are
particularly opposed to research on male-favoring sex-differences [Stewart-Williams et al.,
2020]), but it also 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).
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
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
78
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, Reyna, Chambers, Crawford, & Wetherell, 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.
Before expanding our interpretation (and to include Conservatives), discussing possible
alternative explanations, and forwarding some ideas for future directions, we should address
limitations.
Limitations
All studies were conducted on Amazon Mechanical Turk. We used this population
because we could get a more diverse sample than we could from most University participant
pools (Casler, Bickel, & Hackett, 2013). Furthermore, research suggests that Mturk participants
are not insouciant survey takers; they pay attention and provide reliable data (Hauser &
Schwartz, 2016). But this also means that all data were self-report. This comes with standard
problems such as desirability biases. It is certainly possible that Liberals and Conservatives have
slightly different presentational values (Liberals have more equalitarian presentational values)
and that our results simply reflect those values and not biased assessments of our vignettes.
Future work would also benefit from testing these biases among nationally representative
samples, which would have more balanced numbers between Liberals and Conservatives as well
as a more representative spread of Liberals and Conservatives within each ideological group.
Probably the most severe methodological limitation of our investigation was that we
focused on only one socially valued trait: Intelligence. Equalitarianism predicts that Liberals
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
79
(and others who score high on this trait/philosophy) will be biased against information that
suggests that demographic groups differ on almost any socially valued trait; and that they will be
especially biased against information that suggests that a privileged group is “better” than a
victims’ group on such a trait. We focused on intelligence alone, and thus it remains unknown
whether or results would replicate for other traits. However, in the time since we conducted these
studies, a set of scholars have provided some evidence that this bias likely extends to other traits.
For example, Stewart-Williams and colleagues found that people more negatively evaluated
research on sex differences that favored males (that men are better at drawing and lie less often)
than research that drew the reverse conclusions, and that this tendency was stronger among more
liberal participants. And Winegard and colleagues (2019) found that people had greater desire to
censor science that argued that men evolved to be better leaders than women than science that
argued the reverse, especially among Liberals. We hope future research will continue to explore
these patterns with numerous socially valued traits (e.g., self-control, ambition, agreeableness,
criminality, etc.) and across different kinds of groups that are perceived as more advantaged or
disadvantaged. In general, we would predict that the more intensively valued the trait, the more
intense the bias. The exception is traits whose differences across demographic groups are too
conspicuous to deny (thus increasing clarity of the difference and reducing bias) such as physical
strength differences between men and women. Though note, Study 1b did demonstrate that
Liberals agree less with this contention than Conservatives.
Focusing only on intelligence allowed us to scrutinize the bias carefully across different
demographic target groups (sex-based and race-based) while using varied study designs to refine
the theory, replicate the effects, and rule out alternate explanations. Furthermore, intelligence is a
highly socially valued trait, perhaps one of the highest, and discussion about demographic
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
80
differences in intelligence is often morally supercharged, vitriolic, and even counterproductive
(Hunt, 2011). It may be that equalitarian bias plays a role in the unfortunate tone and results of
many such discussions.
Our measure of equalitarianism also has limitations. In the present paper, we focused on
identifying a particular form of biased responding among Liberals, trying to rule out as many
potential alternate explanations as possible (though of course not all of them). We did not,
however, assiduously validate the equalitarianism measure by establishing convergent and
discriminant validity with other scales, establishing retest reliability, or narrowing our set of
items down to a perhaps better (or less redundant) set. Although this scale had high internal
consistency and very consistently mediated our findings across studies, it is possible future
research will identify a construct that can better explain our findings, and perhaps one that is
more established and has already undergone several perfecting revisions over time. In other
words, we may not need an entirely new construct to explain the present findings. We hope
future researchers will seek to compare this measure of equalitarianism to other similar
constructs, such as Social Dominance Orientation (Pratto et al., 1994) or Preference for Merit
(Davey, Bobocel, Hing, & Zanna, 1999), and their relation to biased evaluations of information
with significance to group differences. If our measure of equalitarianism proves uniquely useful
for studying liberal bias, and particularly liberal bias regarding low status groups, we hope future
research will refine the measure and perhaps generate a more concise measure. If not, we hope
our findings regarding the equalitarianism scale will help scholars identify the construct that
better explains the identified liberal bias.
Alternate Explanations
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
81
We can think of at least two serious alternative explanations to our analysis: (1) It is
rational to be skeptical of scientific results which seem to favor the powerful over the oppressed;
and (2) it is rational to believe it more plausible that genetics explain higher IQ scores among
victims' groups than that they explain higher IQ scores among privileged groups. This second
explanation seems compelling because the environment that victims' groups navigate is likely
harsher and less conducive to thriving than the environment that privileged groups experience.
Therefore, if a privileged group scores higher, it could be because of the environment; but if a
victims’ group scores higher, then it must be because of genes (because it couldn’t be caused by
the [worse] environment).
It is almost impossible to rule out completely the first explanation. Powerful people have
doubtlessly used science to justify their privilege and to pacify the less fortunate by claiming that
the indigent “deserve” their lot because of their inferiority (Gould, 1996). So, according to this
argument, when socially conscious and concerned Liberals are confronted with data that claim
that Whites score higher than Blacks on IQ scores because of genetics, they rationally assess it as
implausible (and probably as politically motivated). The same holds for data that claim that men
score higher than women. We believe, however, that the within-subjects design studies at least
mitigate the force of this argument. If Liberals believe they are rational to rate arguments that
favor privileged groups as less credible than arguments that favor victims’ groups, then it is
unclear why we would see an order effect. The order effects suggest that Liberals realize that
they should answer the two vignettes consistently (or at least somewhat consistently), which
suggests that they believe it would be biased not to do so.
It is possible however that the order effects are simply standard anchoring effects
(Furnham & Boo, 2011). That is, perhaps the order effects don’t reflect a concern for appearing
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
82
or being unbiased, but simply reflect a psychological anchoring heuristic. Although this is
certainly possible, Conservatives (and Moderates) did not display an anchoring effect, suggesting
that anchoring is not a necessary outcome of such an experimental design. Furthermore, it strikes
us as rather implausible that such an anchoring should occur. In fact, if anything, if not for
concerns of bias, we would probably predict that a within-subjects design would augment
differences. Consider, for example, a study in which there were two conditions: In one cheetahs
were said to be faster than dogs and in the other, dogs were said to be faster than cheetahs. The
researchers asked participants to rate the plausibility of the stories. If cheetah faster came first
and got a high plausibility rating, it would seem sensible that dogs faster would get an even
lower rating than if it came first, because the participant would think, “Well, I already noted that
Cheetahs were faster…no way can dogs also be faster.” Still, this is a possible counterargument
that future researchers should address in more detail. A helpful reviewer suggested a possible
route of casting further doubt on this alternate explanation in future work: explicitly making the
scientist in the vignette a black woman. This should minimize, at least to some extent, skepticism
that the scientist is motivated by racism or sexism in the conditions in which the privileged group
is said to perform better on the test.
The order effects cast doubt on the second explanation in the same way they do for the
first explanation. However, one might speculate that whereas Liberals are using their knowledge
of alternate environmental explanations that could explain higher IQ scores among privileged
groups than victims’ groups in the between subjects design, the within subjects design compels
them to disregard this knowledge in order to appear consistent in their judgment even though
consistency is not the only rational response pattern. This argument seems to make more sense
when applied to race differences in IQ than to sex differences. Men and women inhabit largely
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
83
the same environment (same socioeconomic status and schools, for example) and, in fact, girls
and women outperform boys in every stage of education, from elementary school through
college (Duckworth & Seligman, 2006). Still, one might contendand Liberals might rationally
believethat they are held back by pernicious and invidious stereotypes or more subtle forms of
sexism (Spencer, Steele, & Quinn, 1999). Our present investigation does not allow us
definitively to rule out this explanation. We believe, as of now, that our equalitarian theory of
bias is a better, more parsimonious explanation, but researchers should continue to pursue these
questions so that we can continue to update our understanding of these biases (or patterns of
responses, if they are, indeed, not biases).
Conservatives (and Moderates)
In the present investigation, we predicted that Liberals would evince a specific and
consistent pattern of biased responding, one that was supported by the data across all studies and
meta-analyses of those studies. We focused on Liberals because they are a relatively
understudied group (Eitan et al., 2018), very little work deliberately explores biases among
Liberals, and recent work has suggested there are likely domains of liberal bias yet to be fully
uncovered (Ditto et al., 2019a, 2019b). However, by including participants across the ideological
spectrum, we were able to discover patterns among Conservatives and Moderates as well.
Although the patterns were less consistent for Conservatives across studies, the meta-
analyses revealed two patterns for Conservatives. First, in contrast to Liberals, Conservatives
rated the argument that a gene explains why Whites score higher on IQ tests as more credible
than the argument that a gene explains why Blacks score higher. As we noted earlier (see
discussion section of Study 6), this might suggest a bias against Blacks, or this might suggest that
Conservatives believe it is rational to treat these two arguments differently. Although it would be
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
84
rash to conclude one way or another, there are at least a few reasons to doubt the anti-Black bias
account. Conservatives rated the Equal condition as the most credible (in Study 4), suggesting a
slight “preference” for equality (although this was not statistically significant). Conservatives did
not display an order effect and so appeared to believe that it was rational to rate the Whites
Higher condition as more credible than the Blacks Higher conditions (from a Bayesian
perspective, this could possibly be a rational response). And, a growing body of recent work
suggests that whereas Liberals are biased in favor of Blacks over Whites, Conservatives’
judgments and evaluations are relatively less influenced by race manipulations (e.g., Axt et al.,
2016; Kteily et al., 2019; Purser & Harper, 2020; Uhlmann et al., 2009; Winegard et al., 2019).
We hope future research will explore the possibility of an anti-black racial bias among
Conservatives across numerous kinds of judgments and evaluationsit might be that they have
biases in particular domains and not others.
Second, like Liberals, Conservatives seemed to have a slight preference for women
having higher IQ than men than vice versa, although to a weaker extent than Liberals.
Unfortunately, their order effect in Study 7 was relatively uninformative. When they received the
women higher argument first, they anchored their second judgment to the first, consistent with a
bias explanation; when they received the men higher argument first, they evaluated the women
higher argument as more credible, less consistent with a bias explanation. Other work seems to
support the idea that Conservatives, like Liberals, have a pro-female bias, just to a lesser degree
(e.g., Winegard et al., 2019; Stewart-Williams et al., 2020). We hope future work will further
explore this possibility as well.
Moderates demonstrated some inconsistent patterns across studies, but the meta-analyses
suggested that they were relatively unaffected by the manipulationsthey evaluated research
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
85
similarly credible regardless of whether findings favored high or low status groups. Such
findings may be consistent with the rigidity-of-the-extreme hypothesis, which suggests that more
extreme partisan attachments, whether more right or more left, are associated with more rigid,
dogmatic, and biased cognition (Zmigrod, 2020). Future work should explore numerous kinds of
biases to further test the possibility that those more centrist tend to display the smallest and
fewest biases.
Overall, whereas our studies demonstrated very consistent patterns for Liberals, they
were somewhat less consistent among Moderates and Conservatives, but we do hope the patterns
we discovered here for Moderates and Conservatives will be useful for generating new
hypotheses regarding the biases of these groups.
Relevance to the World
We believe that our results are germane to many modern political and moral discussions.
Few topics are as explosive and controversial as demographic differences. Many researchers who
have discussed such differences openly have been calumniated and accused of moral treachery
(Winegard & Winegard, 2015). This is probably one reason very few have studied bias about
demographic differences: merely studying the bias may earn them obloquy. However, in a free,
scientifically literate society, people need to have these conversations, with appropriate care and
caution, so that we can best understand humans and best design social policy in a cosmopolitan
society. Some researchers believe that the modern genomic revolution is likely to reveal that
many demographic differences are at least partially genetically caused, and we, as a society, will
be ill-prepared for these results if we are too afraid to talk about such a possibility (Reich, 2018).
We believe that studying and understanding biases regarding such differences might help us to
have such conversations more productively and with fewer accusations of iniquity.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
86
Before briefly discussing some future directions, we should note that equalitarianism
might be a morally rational world-view and the biases that stem from it might be justifiable. One
might argue that even if one accepts that our results stem from bias, one could still argue that
such a bias is rational because racism and sexism are grave threats in the United States and we
need to be vigilant against them. From this perspective, it is better to err on the side of caution
against results that seem to buttress the current system of power or that seem to disparage
historically oppressed and marginalized peoples. We think that is a perfectly sensible argument.
It may be right. Our goal in this paper was to understand equalitarianism and how it contributed
to liberal bias, not to assess it morally. We do think that such biases could lead to indefensible
accusations toward scholars who forward certain kinds of explanations and even informal
censorship of certain kinds of explanations, but we do not doubt that they could be morally
justifiable on whole.
Future Directions
In the future, researchers should expand our investigation by examining more privileged
and victims’ groups (e.g., Muslims, homosexuals, Native Americans). They should also examine
more socially valued traits than intelligence (e.g., self-control, ambition, criminal propensities).
We have assumed that pervasive racism and sexism are appealing explanations for group
differences among Liberals because the alternate explanationthat demographic groups might
have different underlying traits and abilitiesposes a far greater challenge to egalitarian ideals.
Presumably, biology is harder to change than individual and societal prejudices. However, this
might not be true in the future. Advances in gene technology could make it both possible and
commonplace to alter genes in a way that benefits individuals and thus groups. If biological
differences were no longer viewed as a barrier to equality and instead were viewed as easily
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
87
resolved, we suspect there would be less resistance among Liberals to such explanations,
particularly if these explanations provided compelling reason to allocate such technological
resources to relatively low status groups. Future research should investigate this emerging
reality. Technology may obviate our theory.
Another future direction is to explore how ideology relates to beliefs in and motivations
for numerous kinds of explanations for group differences. Here, we explored a possible aversion
to one kind of explanation (a biological one) for one particular kind of group difference
(intelligence) across only two group categories (race and sex). However, there are numerous
possible explanations for group inequalities: social norms, cultural differences, historical
discrimination, natural environment features (e.g., water quality, exposure to natural disasters),
“free choice” (however one defines it), among many others. For example, previous work has
found that Conservatives more than Liberals believe in the concept of “free will” because of their
stronger desire to hold people morally responsible for their bad behaviorand that both Liberals
and Conservatives selectively appeal to this explanation when it is ideologically convenient
(Everett et al., 2020). Likely, which explanations seem the most plausible and desirable to
particular ideologies vary according to which groups and outcomes they have significance to,
and future research should study these beliefs and motivations sedulously, as these variations
likely explain a great deal of ideological conflict and polarization.
Conclusions
For a long time, many social psychologists contended that Conservatives are more biased
than Liberals. Recent scholarship has strongly challenged this argument. Conservatives and
Liberals appear roughly equally biased. Our research adds to this important debate and suggests
one domain in which Liberals demonstrate a consistent bias. Because most social psychologists
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
88
are Liberals and because demographic differences are such an explosive topic, this bias has
remained unstudied and largely unknown (although researchers speculated about it). Our hope
here is to provoke more internal reflection among Liberals and liberal scholars on whether such
biases may affect their own evaluations, discussions, and scholarship.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
89
References
Adorno, T. W., Frenkel-Brunswik, E., Levinson, D. J., & Sanford, R. N. (1950). The
authoritarian personality. New York, NY: Harper and Row.
Altemeyer, B. (1981). Right-wing authoritarianism. Manitoba, Canada: University Press.
Altemeyer, B. (1996). The authoritarian specter. Cambridge, MA: Harvard University Press.
Atran, S., Axelrod, R., & Davis, R. (2007). Sacred barriers to conflict resolution. Science, 317,
1039-1040.
Axt, J. R., Ebersole, C. R., & Nosek, B. A. (2016). An unintentional, robust, and replicable pro-
Black bias in social judgment. Social Cognition, 34, 1-39.
Baron, J., & Jost, J. T. (2019). False equivalence: Are liberals and conservatives in the United
States equally biased?. Perspectives on Psychological Science, 14, 292-303.
Baumeister, R. F., Maranges, H. M., & Vohs, K. D. (2018). Human Self as Information Agent:
Functioning in a Social Environment Based on Shared Meanings. Review of General
Psychology, 22, 36-47.
Bawer, B. (2012). The victims’ revolution: The rise of identity studies and the closing of the
liberal mind. New York: Broadside Books.
Bem, D. (2011). Feeling the future: experimental evidence for anomalous retroactive influences
on cognition and affect. Journal of Personality and Social Psychology, 100, 407-425.
Brandt, M. J., Reyna, C., Chambers, J. R., Crawford, J. T., & Wetherell, G. (2014). The
ideological-conflict hypothesis: Intolerance among both liberals and
conservatives. Current Directions in Psychological Science, 23, 27-34.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
90
Casler, K., Bickel, L., & Hackett, E. (2013). Separate but equal? A comparison of participants
and data gathered via Amazon’s MTurk, social media, and face-to-face behavioral
testing. Computers in Human Behavior, 29, 2156-2160.
Clark, C. J., Liu, B. S., Winegard, B. M., & Ditto, P. H. (2019). Tribalism is human
nature. Current Directions in Psychological Science, 28, 587-592.
Clark, C. J., & Winegard, B. M. (2020). Tribalism in war and peace: The nature and evolution of
ideological epistemology and its significance for modern social science. Psychological
Inquiry, 31, 1-22.
Conrad, K. J., Wright, B. D., McKnight, P., McFall, M., Fontana, A., & Rosenheck, R. (2004).
Comparing traditional and Rasch analyses of the Mississippi PTSD Scale: revealing
limitations of reverse-scored items. Journal of Applied Measurement, 5, 15-30.
Crawford, J. T. (2012). The ideologically objectionable premise model: Predicting biased
political judgments on the left and right. Journal of Experimental Social
Psychology, 48, 138-151.
Crawford J. T. (2014). Ideological symmetries and asymmetries in political intolerance and
prejudice toward political activist groups. Journal of Experimental Social Psychology,
55, 284298.
Crawford, J. T. & Jussim, L. (2017). Frontiers of Social Psychology Series: The Politics
of Social Psychology. New York, NY: Routledge.
Crawford, J. T., Jussim, L., Cain, T. R., & Cohen, F. (2013). Rightwing authoritarianism and
social dominance orientation differentially predict biased evaluations of media
reports. Journal of Applied Social Psychology, 43, 163-174.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
91
Davey, L. M., Bobocel, D. R., Hing, L. S. S., & Zanna, M. P. (1999). Preference for the Merit
Principle Scale: An individual difference measure of distributive justice
preferences. Social Justice Research, 12, 223-240.
Ditto, P. H., & Lopez, D. F. (1992). Motivated skepticism: Use of differential decision criteria
for preferred and nonpreferred conclusions. Journal of Personality and Social
Psychology, 63, 568-584.
Ditto, P. H., Liu, B. S., Clark, C. J., Wojcik, S. P., Chen, E. E., Grady, R. H., ... & Zinger, J. F.
(2019a). At least bias is bipartisan: A meta-analytic comparison of partisan bias in
liberals and conservatives. Perspectives on Psychological Science, 14, 273-291.
Ditto, P. H., Clark, C. J., Liu, B. S., Wojcik, S. P., Chen, E. E., Grady, R. H., ... & Zinger, J. F.
(2019b). Partisan bias and its discontents. Perspectives on Psychological Science, 14,
304-316.
Duarte, J. L., Crawford, J. T., Stern, C., Haidt, J., Jussim, L., & Tetlock, P. E. (2015). Political
diversity will improve social psychological science. Behavioral and Brain Sciences, 38,
e130.
Duckworth, A. L., & Seligman, M. E. (2006). Self-discipline gives girls the edge: Gender in self-
discipline, grades, and achievement test scores. Journal of Educational Psychology, 98,
198-208.
Dupree, C. H., & Fiske, S. T. (in press). Self-presentation in Interracial Settings: The
Competence Downshift by White Liberals. Journal of Personality and Social
Psychology.
Eitan, O., Viganola, D., Inbar, Y., Dreber, A., Johannesson, M., Pfeiffer, T., ... & Uhlmann, E. L.
(2018). Is research in social psychology politically biased? Systematic empirical tests and
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
92
a forecasting survey to address the controversy. Journal of Experimental Social
Psychology, 79, 188-199.
Everett, J. A. C., Clark, C. J., Meindl, P., Luguri, J. B., Earp, B. D., Graham, J., ... & Shariff, A.
F. (2020). Political differences in free will belief are associated with differences in
moralization. Journal of Personality and Social Psychology.
Faul, F., Erdfelder, E., Lang, A. G., & Buchner, A. (2007). G*Power 3: A flexible statistical
power analysis program for the social, behavioral, and biomedical sciences. Behavior
Research Methods, 39, 175-191.
Faul, F., Erdfelder, E., Buchner, A., & Lang, A. G. (2009). Statistical power analyses using
G*Power 3.1: Tests for correlation and regression analyses. Behavior Research Methods,
41, 1149-1160.
Felson, R. B. (1981). Ambiguity and bias in the self-concept. Social Psychology Quarterly, 64-
69.
Frenda, S. J., Knowles, E. D., Saletan, W., & Loftus, E. F. (2013). False memories of fabricated
political events. Journal of Experimental Social Psychology, 49, 280-286.
Furnham, A., & Boo, H. C. (2011). A literature review of the anchoring effect. The Journal of
Socio-Economics, 40(1), 35-42.
Goh, J. X., Hall, J. A., & Rosenthal, R. (2016). Mini meta‐analysis of your own studies: Some
arguments on why and a primer on how. Social and Personality Psychology
Compass, 10, 535-549.
Gould, S. J. (1996). The mismeasure of man. WW Norton & Company.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
93
Graham, J., Haidt, J., Koleva, S., Motyl, M., Iyer, R., Wojcik, S. P., & Ditto, P. H. (2013). Moral
foundations theory: The pragmatic validity of moral pluralism. Advances in Experimental
Social Psychology, 47, 55-130.
Haidt, J. (2011, February 11). The bright future of post-partisan social psychology. Edge.
Retrieved from https://www.edge.org/conversation/jonathan_haidt-the-bright-future-of-
post-partisan-social-psychology
Haidt, J. (2012). The righteous mind: Why good people are divided by politics and religion. New
York: Vintage.
Hannikainen, I. R. (2018). Ideology between the lines: Lay inferences about scientists’ values
and motives. Manuscript submitted for publication.
Hasson, Y., Tamir, M., Brahms, K. S., Cohrs, J. C., & Halperin, E. (2018). Are Liberals and
Conservatives Equally Motivated to Feel Empathy Toward Others? Personality and
Social Psychology Bulletin, 44, 1449-1459.
Hauser, D. J., & Schwarz, N. (2016). Attentive Turkers: MTurk participants perform better on
online attention checks than do subject pool participants. Behavior research methods, 48,
400-407.
Hayes, A. F. (2013). Introduction to mediation, moderation, and condtional process analysis: A
regression-based approach. New York, NY: Gilford Press.
Huddy, L. (2001). From social to political identity: A critical examination of social identity
theory. Political Psychology, 22, 127-156.
Hunt, E. (2011). Human intelligence. New York, NY: Cambridge University Press.
Inbar, Y., & Lammers, J. (2012). Political diversity in social and personality
psychology. Perspectives on Psychological Science, 7, 496-503.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
94
Jeffries, C. H., Hornsey, M. J., Sutton, R. M., Douglas, K. M., & Bain, P. G. (2012). The David
and Goliath principle: Cultural, ideological, and attitudinal underpinnings of the
normative protection of low-status groups from criticism. Personality and Social
Psychology Bulletin, 38, 1053-1065.
Jost, J. T., Glaser, J., Kruglanski, A. W., & Sulloway, F. (2003). Political conservatism as
motivated social cognition. Psychological Bulletin, 129, 339-375.
Jost, J. T., Nosek, B. A., & Gosling, S. D. (2008). Ideology: Its resurgence in social, personality,
and political psychology. Perspectives on Psychological Science, 3(2), 126-136.
Kahan, D. M. (2016). The politically motivated reasoning paradigm, part 1: What politically
motivated reasoning is and how to measure it. Emerging Trends in the Social and
Behavioral Sciences: An Interdisciplinary, Searchable, and Linkable Resource. 116.
Kahan, D. M., & Braman, D. (2006). Cultural cognition and public policy. Yale Law & Policy
Review, 24, 149-172.
Kahneman, D., & Tversky, A. (1984). Choices, values, and frames. American Psychologist, 39,
341-350.
Kruger, J., & Dunning, D. (1999). Unskilled and unaware of it: how difficulties in recognizing
one's own incompetence lead to inflated self-assessments. Journal of Personality and
Social Psychology, 77, 1121-1134
Kteily, N. S., Rocklage, M. D., McClanahan, K., & Ho, A. K. (2019). Political ideology shapes
the amplification of the accomplishments of disadvantaged vs. advantaged group
members. Proceedings of the National Academy of Sciences, 116, 1559-1568.
Kunda, Z. (1990). The case for motivated reasoning. Psychological Bulletin, 108, 480-498.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
95
Lord, C. G., Ross, L., & Lepper, M. R. (1979). Biased assimilation and attitude polarization: The
effects of prior theories on subsequently considered evidence. Journal of Personality and
Social Psychology, 37, 2098-2109.
Lucas, B. J., & Kteily, N. S. (2018). (Anti-) egalitarianism differentially predicts empathy for
members of advantaged versus disadvantaged groups. Journal of Personality and Social
Psychology, 114, 665-692.
Mac Donald, H. (2018). The diversity delusion: How race and gender pandering corrupt the
university and undermine our culture. New York, NY: St. Martin’s Press.
Mackintosh, N. J. (2011). IQ and human intelligence. New York, NY: Oxford University Press.
Paolacci, G., & Chandler, J. (2014). Inside the Turk: Understanding Mechanical Turk as a
participant pool. Current Directions in Psychological Science, 23, 184-188.
Pinker, S. (2003). The blank slate: The modern denial of human nature. New York, NY: Penguin
Books.
Pratto, F., Sidanius, J., Stallworth, L. M., & Malle, B. F. (1994). Social dominance orientation: A
personality variable predicting social and political attitudes. Journal of Personality and
Social Psychology, 67, 741-763.
Pronin, E., Lin, D. Y., & Ross, L. (2002). The bias blind spot: Perceptions of bias in self versus
others. Personality and Social Psychology Bulletin, 28, 369-381.
Purser, H., & Harper, C. (2020). Low system justification drives ideological differences in joke
perception: A critical commentary and re-analysis of Baltiansky et al. (2020).
Unpublished manuscript.
Reich, D. (2018). Who we are and how we got here: Ancient DNA and the new science of the
human past. New York, NY: Pantheon Books.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
96
Rokeach, M. (1956). Political and religious dogmatism: An alternative to the authoritarian
personality. Psychological Monographs: General and Applied, 70, 1-43.
Rushton, J. P. (1995). Race, evolution, and behavior. New Brunswick: Transaction.
SAT: Total group profile report 2016. Retrieved from
https://securemedia.collegeboard.org/digitalServices/pdf/sat/total-group-2016.pdf
Skitka, L. J. (2010). The psychology of moral conviction. Social and Personality Psychology
Compass, 4, 267-281.
Stewart-Williams, S., Thomas, A., Blackburn, J. D., & Chan, C. Y. M. (2020). Reactions to
male-favoring vs. female-favoring sex differences: A preregistered experiment. British
Journal of Psychology.
Taber, C. S., & Lodge, M. (2006). Motivated skepticism in the evaluation of political
beliefs. American Journal of Political Science, 50, 755-769.
Taylor, I. A. (1960). Similarities in the structure of extreme social attitudes. Psychological
Monographs: General and Applied, 74, 1-36.
Tetlock, P. E. (1994). Political psychology or politicized psychology: Is the road to scientific hell
paved with good moral intentions?. Political Psychology, 3, 509-529.
Tetlock, P. E. (2003). Thinking the unthinkable: Sacred values and taboo cognitions.
Trends in Cognitive Sciences, 7, 320-324.
Uhlmann, E. L., Pizarro, D. A., Tannenbaum, D., & Ditto, P. H. (2009). The motivated use of
moral principles. Judgment and Decision making, 4, 476-491.
von Hippel, W., & Buss, D. M. (2017). Do ideologically driven scientific agendas impede the
understanding and acceptance of evolutionary principles in social psychology. In J. T.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
97
Crawford & L. Jussim. Frontiers of Social Psychology Series: The Politics of Social
Psychology (pp. 7-25). New York, NY: Routledge.
Van Bavel, J. J., Reinero, D. A., Harris, E., Robertson, C. E., & Pärnamets, P. (2020). Breaking
groupthink: Why scientific identity and norms mitigate ideological
epistemology. Psychological Inquiry, 31, 66-72.
Van Bavel, J. J., & Pereira, A. (2018). The partisan brain: An Identity-based model of political
belief. Trends in Cognitive Sciences, 22, 213-224.
Winegard, B. M., & Clark, C. J. (2020). Without Contraries is no Progression. Psychological
Inquiry, 31, 94-101.
Winegard, B. M., Clark, C. J., & Bunnel, E. (2019). The ideology of censorship. Manuscript
submitted for publication.
Winegard, B. & Winegard, B. (2015). A social science without sacred values. Working Paper.
Winegard, B. & Winegard, B. (2017, March 27). A tale of two bell curves. Retrieved from
http://quillette.com/2017/03/27/a-tale-of-two-bell-curves/
Winegard, B., & Winegard, B. (2017). Paranoid egalitarian meliorism: An account of political
bias in the social sciences. In J. T. Crawford & L. Jussim. Frontiers of Social Psychology
Series: The Politics of Social Psychology (pp. 193-209). New York: Routledge.
Winegard, B., Winegard, B., & Geary, D. C. (2015). Too paranoid to see progress: Social
psychology is probably liberal, but it doesn’t believe in progress. Commentary on target
article, Political diversity will improve social psychological science (J. Duarte, J.
Crawford, C. Stern, J. Haidt, L. Jussim, & P. E. Tetlock). Behavioral and Brain Sciences,
38, e162.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
98
Zmigrod, L. (2020). The role of cognitive rigidity in political ideologies: theory, evidence, and
future directions. Current Opinion in Behavioral Sciences, 34, 34-39.
EQUALITARIANISM: A SOURCE OF LIBERAL BIAS
99
Appendix
Equalitarianism Measure
Instructions: Please answer the following questions as honestly as you can. Remember, all
answers will be confidential. Use the following scale 1- do not agree all, 4-somewhat agree, 7-
agree completely (so 1 is the lowest level of agreement, and 7 is the highest.)
1. The only reason there are differences between men and women is because society is sexist
2. Differences between men and women in society are caused by discrimination
3. Differences between ethnic groups in society are caused by discrimination
4. Most people are not biased and racism is not a problem anymore*
5. When people assert that men and women are different because of biology, they are usually
trying to justify the status quo
6. People often try to conceal their racism and sexism, but they act that way anyways
7. People often use biology to justify unjust policies that create inequalities
8. Racism is everywhere, even though people say they are not racist
9. Sexism is everywhere, even though people say that they are not sexist
10. People use scientific theories to justify inequalities between groups
11. Men and women have equal abilities on all tasks (for example, mathematics, cooking,
nursing).
12. All ethnic groups have equal abilities on all tasks (for example, mathematics, sports,
creativity)
13. Some differences between men and women are hardwired*
14. Although things are unequal now, if we work really hard, we can make society better and
more fair
15. We should strive to make all groups equal in society
16. We should strive to make men and women equally represented in science fields
17. If we work hard enough, we can ensure that all ethnic groups have equal outcomes
18. Differences among ethnic groups in social outcome are at least partially biologically caused*
*reverse coded items

Supplementary resource (1)

... If protectiveness of females underlies the aversion to male-favouring findings, an implication is that any individual-level variable that predicts a tendency to see women as disadvantaged or as victims of male malfeasance -and thus as requiring greater protection -would predict a stronger aversion. This would include, for instance, a left-leaning political orientation (Winegard et al., 2018) and a belief that men are privileged over women in society (Stewart-Williams et al., 2021). The results of our initial study and direct replication were broadly in line with this proposal (Stewart-Williams et al., 2021, Study 1;Stewart-Williams et al., 2022). ...
... In every case, the effect was found not only among female participants but also among the males. These findings mesh nicely with related research showing that people view claims that males perform better in a given domain as less credible than equivalent claims that females do (Colombo et al., 2016;von Hippel & Buss, 2017;Winegard et al., 2018), and that people are more inclined to censor the former claims (Ashokkumar et al., 2020;Clark et al., 2020Clark et al., , 2022. The aversion to male-favouring sex differences appears to be a robust phenomenon. ...
... The more that one sees women as victims of oppression by men, the more protective one is likely to be of women and thus the more averse to male-favouring findings. People on the left are more likely to see women as victims of male oppression (Winegard et al., 2018), so people on the left are likely to be more averse. In contrast, regardless of one's views on the oppression of women, femalefavouring differences are not a threat to women; thus, left and right would not be expected to differ in their reactions. ...
Article
Full-text available
Two studies (total N = 778) looked at (1) how people react to research finding a sex difference depending on whether the research puts men or women in a better light and (2) how well people can predict the average man and average woman's reactions. Participants read a fictional popular‐science article about fictional research finding either a male‐ or a female‐favouring sex difference. The research was credited to either a male or a female lead researcher. In both studies, both sexes reacted less positively to differences favouring males; in contrast to our earlier research, however, the effect was larger among female participants. Contrary to a widespread expectation, participants did not react less positively to research led by a female. Participants did react less positively, though, to research led by a male when the research reported a male‐favouring difference in a highly valued trait. Participants judged male‐favouring research to be lower in quality than female‐favouring research, apparently in large part because they saw the former as more harmful. In both studies, participants predicted that the average man and woman would exhibit substantial own‐sex favouritism, with both sexes predicting more own‐sex favouritism from the other sex than the other sex predicted from itself. In making these predictions, participants overestimated women's own‐sex favouritism, and got the direction of the effect wrong for men. A greater understanding of the tendency to overestimate gender‐ingroup bias could help quell antagonisms between the sexes.
... In another set of studies, Winegard et al. (2018) document that liberals' judgments favor groups they perceive to be disadvantaged, like women and Black people. Their approach is similar to ours in that they compare judgments of identical situations in which one key demographic characteristics differs (see also Stewart-Williams et al., 2021). ...
... Using such differences between within and between subject judgments is a tool commonly used in psychological studies to draw inferences about conflicting motivations (e.g. Uhlmann et al., 2009;Winegard et al., 2018). ...
... Despite seeing two identical scenarios, respondents judge the behavior of the manager in both scenarios on average 6.4 points more morally wrong if they first saw the scenario with the discriminated woman (11.1 points vs. 4.7 points). Following Uhlmann et al. (2009) and Winegard et al. (2018), we interpret this order effect as evidence that on average respondents themselves think that the gender of the victim should matter less than the between-subject prowomen attitude reveals. The within-subject pro-woman attitude therefore only shows the part of the pro-woman attitude which respondents are comfortable revealing (either to themselves or the researcher). ...
Article
Full-text available
Previous research has shown that people care less about men than about women who are left behind. We show that this finding extends to the domain of labor market discrimination: In identical scenarios, people judge discrimination against women more morally bad than discrimination against men. This result holds in a representative sample of the US population and in a larger but not representative sample of Amazon Mechanical Turk (Mturk) respondents. We test if this gender gap is driven by statistical fairness discrimination, a process in which people use the gender of the victim to draw inferences about other characteristics which matter for their fairness judgments. We test this explanation with a survey experiment in which we explicitly hold information about the victim of discrimination constant. Our results provide only mixed support for the statistical fairness discrimination explanation. In our representative sample, we see no meaningful or significant effect of the information treatments. By contrast, in our Mturk sample, we see that providing additional information partly reduces the effect of the victim’s gender on judgment of the discriminator. While people may engage in statistical fairness discrimination, this process is unlikely to be an exhaustive explanation for why discrimination against women is judged as worse.
... Some research has suggested that right-wing individuals exhibit an anti-female bias (Austin & Jackson, 2019;Christopher & Mull, 2006;Hodson et al., 2017). Other research finds that left-wing individuals may be prone to bias towards groups with low status, including women (Winegard et al., 2018). Overall this body of research indicates that as one moves politically right one becomes less pro-female and more pro-male. ...
... Nonetheless, this hypothesis seems very unlikely. The fact that top publishers and journals are supporting affirmative action in favour of women (Bayazit, 2020;Elsevier, 2021a,b;Lundine et al., 2018;Nature, 2017) would suggest that high performing academics share the same attitudes to sex bias as our surveyed academics who are likely poor performing Moreover, academics at elite institutions are overwhelmingly left-wing, which is associated with having pro-female preferences (Winegard et al., 2018), suggesting editors of top journals are likely to share the same preferences. For example, 39 % of elite American liberal arts colleges have no registered Republican professors (Langbert, 2018). ...
Article
Full-text available
The editorial boards of academic journals overrepresent men, even above their proportion in university faculties. We test whether this sex disparity is caused by anti-female bias, supposing that anti-female discrimination means women must have a higher research output than men to overcome bias against them. We collect a dataset of the research output and sex of 4,319 academics on the editorials boards of 120 journals within four social science disciplines: Anthropology, Psychology, Political Science and Economics. Using a transformation of the h-index as our indicator of research output, we find male research output to be 0.35 standard deviations (p < 0.001) above female research output. However, the gap falls to 0.13 standard deviations (p < 0.001) when years publishing is controlled for. Our results are replicated with alternative dependent variables and using robust regression. We followed up our research with a survey of 231 academics, asking for their attitudes towards discrimination in hiring to editorial boards. Although two-thirds of academics supported no bias, for every 1 academic who supported discrimination in favour of men, 11 supported discrimination in favour of women. Our results were consistent with the hypothesis that academics and journal editors are biased in favour of women, rather than against women.
... Recent work has found that liberals prefer biological group equality for socially valued characteristics (Winegard et al., 2018) and are averse to data that portray higher status groups more favorably than lower status groups on valued traits (e.g., Stewart-Williams et al., 2020;von Hippel & Buss, 2017). These preferences may reflect liberals' otherwise admirable aversion to inequality (Jost et al., 2008) and empathy for lower status groups (e.g., Hasson et al., 2018;Jeffries et al., 2012;Lucas & Kteily, 2018), and may explain why liberals tend to demonstrate biases when evaluating information with significance to groups of varying status. ...
... These preferences may reflect liberals' otherwise admirable aversion to inequality (Jost et al., 2008) and empathy for lower status groups (e.g., Hasson et al., 2018;Jeffries et al., 2012;Lucas & Kteily, 2018), and may explain why liberals tend to demonstrate biases when evaluating information with significance to groups of varying status. For example, liberals more favorably evaluate research on female-favoring sex differences than male-favoring sex differences and more positively judge research on Blackfavoring race differences than White-favoring race differences Stewart-Williams et al., 2020;von Hippel & Buss, 2017;Winegard et al., 2018) and demonstrate prowomen and pro-minority biases in numerous other contexts (e.g., Axt et al., 2016;Dupree & Fiske, 2019;Kteily et al., 2019;Pursur & Harper, 2020;Unzueta et al., 2014). ...
Chapter
Full-text available
The social and behavioral sciences have taken a substantial reputational hit over the past decade. Some highly publicized findings have failed to replicate—and those that do replicate often do so with much smaller effect sizes (Camerer et al., 2018; Nosek et al., 2021). Plus some highly touted “science-based” interventions have failed to produce promised positive social change—even when massive efforts are dedicated to making them work (Singal, 2021). This chapter will lay out our two-tiered hypothesis: (a) the ideological homogeneity of the social sciences has entrenched certain scientific orthodoxies and taboos; (b) these orthodoxies and taboos have protected weak ideas from rigorous scrutiny and contributed to the replication crisis. We also explain how open science practices, although a big step in the right direction, leave many researcher degrees of freedom on the table that can bias methodological decisions and research conclusions. We argue that adversarial collaborations are the next necessary science reform for addressing lingering weaknesses in social scientific norms and can further minimize false positives, expedite scientific corrections, stimulate progress for stalemated scientific debates, and ultimately improve the quality of social scientific outputs.
... Although we did not anticipate this pattern, it does fit with the idea that people are more protective of women than men. People who think that men are greatly privileged over women, and who lean to the left politically, are more likely to see women as downtrodden and as victims of male misbehavior [2,21]. As such, they are more likely to be protective of females in any circumstance where protectiveness might be called for. ...
Article
Full-text available
We report a direct replication of our earlier study looking at how people react to research on sex differences depending on whether the research puts men or women in a better light. Three-hundred-and-three participants read a fictional popular-science article about fabricated research finding that women score higher on a desirable trait/lower on an undesirable one (female-favoring difference) or that men do (male-favoring difference). Consistent with our original study, both sexes reacted less positively to the male-favoring differences, with no difference between men and women in the strength of this effect. Also consistent with our original study, belief in male privilege and a left-leaning political orientation predicted less positive reactions to the male-favoring sex differences; neither variable, however, predicted reactions to the female-favoring sex differences (in the original study, male-privilege belief predicted positive reactions). As well as looking at how participants reacted to the research, we looked at their predictions about how the average man and woman would react. Consistent with our earlier results, participants of both sexes predicted that the average man and woman would exhibit considerable own-sex favoritism. In doing so, they exaggerated the magnitude of the average woman’s own-sex favoritism and predicted strong own-sex favoritism from the average man when in fact the average man exhibited modest other-sex favoritism. A greater awareness of people’s tendency to exaggerate own-sex bias could help to ameliorate conflict between the sexes.
... Moreover, morality is frequently tangled up in the social sciences, and moral concerns are powerful motivators of reasoning Tetlock, Kristel, Elson, Green, & Lerner, 2000). Sometimes accurate conclusions in the social sciences might cause concerns about morally undesirable implications, and people and scholars may then wish to avoid, ignore, disparage, or censor this kind of information, even when it could plausibly be correct (Campbell & Kay, 2014;Clark, Winegard, & Farkas, 2020;Stewart-Williams, Thomas, Blackburn, & Chan, 2019;von Hippel & Buss, 2017;Winegard, Clark, Hasty, & Baumeister, 2018). For just one recent example, a paper by AlShebli, Makovi, and Rahwan (2020) collected a very large sample of mentor and protégé pairs in scientific collaborations and found evidence that female protégés with higher proportions of female mentors were less impactful later in their careers. ...
Chapter
Full-text available
“They all pose as though their real opinions had been discovered and attained through the self-evolving of a cold, pure, divinely indifferent dialectic… whereas, in fact, a prejudiced proposition, idea, or ‘suggestion,’ which is generally their heart's desire abstracted and refined, is defended by them with arguments sought out after the event. They are all advocates who do not wish to be regarded as such, generally astute defenders, also, of their prejudices, which they dub ‘truths,’—and VERY far from having the conscience which bravely admits this to itself...” --Friedrich Nietzsche, Beyond Good and Evil. Scientists are humans. They are smart, ambitious humans, with a peculiar desire to explain and understand the world and a set of principles and procedures that help steer them toward truth. They are humans nonetheless. Their psychology is therefore human psychology. Psychological discoveries in the social sciences—human errors, heuristics, biases, motivations, psychological needs—all apply to scientists in similar if not equal (or possibly even greater) measure.
Article
Full-text available
Behavioral scientists enjoy vast methodological freedom in how they operationalize theoretical constructs. This freedom may promote creativity in designing laboratory paradigms that shed light on real-world phenomena, but it also enables questionable research practices that undercut our collective credibility. Open Science norms impose some discipline but cannot constrain cherry-picking operational definitions that insulate preferred theories from rejection. All too often scholars conduct performative research to score points instead of engaging each other’s strongest arguments—a pattern that allows contradictory claims to fester unresolved for decades. Adversarial collaborations, which call on disputants to co-develop tests of competing hypotheses, are an efficient method of improving our science’s capacity for self-correction and of promoting intellectual competition that exposes false claims. Although individual researchers are often initially reluctant to participate, the research community would be better served by institutionalizing adversarial collaboration into its peer review process.
Article
Full-text available
Available here: https://castfromclay.co.uk/commentary/how-liberal-and-conservative-bias-impacts-policymaking/
Article
Full-text available
I analyze bibliometric data about fundamental physics worldwide from 1970 to now, extracting quantitative data about gender issues. I do not find significant gender differences in hiring rates, hiring timing, career gaps and slowdowns, abandonment rates, citation, and self-citation patterns. Furthermore, various bibliometric indicators (number of fractionally counted papers, citations, etc.) exhibit a productivity gap at hiring moments, at career level, and without integrating over careers. The gap persists after accounting for confounding factors and manifests as an increasing fraction of male authors going from average to top authors in terms of bibliometric indices, with a quantitative shape that can be fitted by higher male variability.
Article
Full-text available
This essay was originally published on Psychology Today: https://www.psychologytoday.com/us/blog/the-antisocial-psychologist/202102/are-liberals-really-more-egalitarian
Article
Full-text available
A contentious debate in political psychology has centred on the role of cognitive rigidity in shaping individuals’ political ideologies and worldviews. Early theories in the 1950s posited that strict ideological doctrines may tend to attract individuals with dispositions towards mental rigidity. This question has persisted: Does psychological rigidity foster a tendency towards ideological extremism? This review evaluates the empirical landscape with respect to the rigidity-of-the-extreme and the rigidity-of-the-right hypotheses and offers conceptual and methodological recommendations for future research avenues. The evidence suggests that cognitive rigidity is linked to ideological extremism, partisanship, and dogmatism across political and non-political ideologies. Advances in the measurement of ideological extremity and cognitive rigidity will facilitate further elucidation regarding how exactly the two hypotheses may be reconciled and why they have been historically placed in a potentially false competition. This synthesis suggests that a scientifically rigorous understanding of the cognitive roots of ideological thinking may be essential for developing effective antidotes to intolerance and intergroup hostility.
Article
Full-text available
We argue that because of a long history of intergroup conflict and competition, humans evolved to be tribal creatures. Tribalism is not inherently bad, but it can lead to ideological thinking and sacred values that distort cognitive processing of putatively objective information in ways that affirm and strengthen the views and well-being of one’s ingroup (and that increase one’s own standing within one’s ingroup). Because of this shared evolutionary history of intergroup conflict, liberals and conservatives likely share the same underlying tribal psychology, which creates the potential for ideologically distorted information processing. Over the past several decades, social scientists have sedulously documented various tribal and ideological psychological tendencies on the political right, and more recent work has documented similar tendencies on the political left. We contend that these tribal tendencies and propensities can lead to ideologically distorted information processing in any group. And this ideological epistemology can become especially problematic for the pursuit of the truth when groups are ideologically homogenous and hold sacred values that might be contradicted by empirical inquiry. Evidence suggests that these conditions might hold for modern social science; therefore, we conclude by exploring potential ideologically driven distortions in the social sciences.
Article
Full-text available
Humans evolved in the context of intense intergroup competition, and groups comprised of loyal members more often succeeded than those that were not. Therefore, selective pressures have consistently sculpted human minds to be "tribal," and group loyalty and concomitant cognitive biases likely exist in all groups. Modern politics is one of the most salient forms of modern coalitional conflict and elicits substantial cognitive biases. Given the common evolutionary history of liberals and conservatives, there is little reason to expect pro-tribe biases to be higher on one side of the political spectrum than the other. We call this the evolutionarily plausible null hypothesis and recent research has supported it. In a recent meta-analysis, liberals and conservatives showed similar levels of partisan bias, and a number of pro-tribe cognitive tendencies often ascribed to conservatives (e.g., intolerance toward dissimilar others) have been found in similar degrees in liberals. We conclude that tribal bias is a natural and nearly ineradicable feature of human cognition, and that no group—not even one’s own—is immune.
Article
Full-text available
Most Whites, particularly sociopolitical liberals, now endorse racial equality. Archival and experimental research reveals a subtle but persistent ironic consequence: White liberals self-present less competence to minorities than to other Whites-that is, they patronize minorities stereotyped as lower status and less competent. In an initial archival demonstration of the competence downshift, Study 1 examined the content of White Republican and Democratic presidential candidates' campaign speeches. Although Republican candidates did not significantly shift language based on audience racial composition, Democratic candidates used less competence-related language to minority audiences than to White audiences. Across 5 experiments (total N = 2,157), White participants responded to a Black or White hypothetical (Studies 2, 3, 4, S1) or ostensibly real (Study 5) interaction partner. Three indicators of self-presentation converged: competence-signaling of vocabulary selected for an assignment, competence-related traits selected for an introduction, and competence-related content of brief, open-ended introductions. Conservatism indicators included self-reported political affiliation (liberal-conservative), Right-Wing Authoritarianism (values-based conservatism), and Social Dominance Orientation (hierarchy-based conservatism). Internal meta-analyses revealed that liberals-but not conservatives-presented less competence to Black interaction partners than to White ones. The simple effect was small but significant across studies, and most reliable for the self-reported measure of conservatism. This possibly unintentional but ultimately patronizing competence-downshift suggests that well-intentioned liberal Whites may draw on low-status/competence stereotypes to affiliate with minorities. (PsycINFO Database Record (c) 2019 APA, all rights reserved).
Article
Full-text available
Recent years have witnessed an increased public outcry in certain quarters about a perceived lack of attention given to successful members of disadvantaged groups relative to equally meritorious members of advantaged groups, exemplified by social media campaigns centered around hashtags, such as #OscarsSoWhite and #WomenAlsoKnowStuff. Focusing on political ideology, we investigate here whether individuals differentially amplify successful targets depending on whether these targets belong to disadvantaged or advantaged groups, behavior that could help alleviate or entrench group-based disparities. Study 1 examines over 500,000 tweets from over 160,000 Twitter users about 46 unambiguously successful targets varying in race (white, black) and gender (male, female): American gold medalists from the 2016 Olympics. Leveraging advances in computational social science, we identify tweeters’ political ideology, race, and gender. Tweets from political liberals were much more likely than those from conservatives to be about successful black (vs. white) and female (vs. male) gold medalists (and especially black females), controlling for tweeters’ own race and gender, and even when tweeters themselves were white or male (i.e., advantaged group members). Studies 2 and 3 provided experimental evidence that liberals are more likely than conservatives to differentially amplify successful members of disadvantaged (vs. advantaged) groups and suggested that this is driven by liberals’ heightened concern with social equality. Addressing theorizing about ideological asymmetries, we observed that political liberals are more responsible than conservatives for differential amplification. Our results highlight ideology’s polarizing power to shape even whose accomplishments we promote, and extend theorizing about behavioral manifestations of egalitarian motives.
Preprint
Full-text available
Both liberals and conservatives accuse their political opponents of partisan bias, but is there empirical evidence that one side of the political aisle is indeed more biased than the other? To address this question, we meta-analyzed the results of 51 experimental studies, involving over 18,000 participants, that examined one form of partisan bias— the tendency to evaluate otherwise identical information more favorably when it supports one’s political beliefs or allegiances than when it challenges those beliefs or allegiances. Two hypotheses based on previous literature were tested: an asymmetry hypothesis (predicting greater partisan bias in conservatives than in liberals) and a symmetry hypothesis (predicting equal levels of partisan bias in liberals and conservatives). Mean overall partisan bias was robust (r = .245), and there was strong support for the symmetry hypothesis: Liberals (r = .235) and conservatives (r = .255) showed no difference in mean levels of bias across studies. Moderator analyses reveal this pattern to be consistent across a number of different methodological variations and political topics. Implications of the current findings for the ongoing ideological symmetry debate and the role of partisan bias in scientific discourse and political conflict are discussed.
Article
On the basis of a meta-analysis of 51 studies, Ditto et al. (this issue, p. 273) conclude that ideological bias is equivalent on the left and right of U.S. politics. In this commentary, we contend that this conclusion does not follow from the review and that Ditto and his colleagues are too quick to embrace a false equivalence between the liberal left and the conservative right. For one thing, the issues, procedures, and materials used in the studies reviewed by Ditto and his colleagues were selected for purposes other than the inspection of ideological asymmetries. Consequently, methodological choices made by researchers were systematically biased to avoid producing differences between liberals and conservatives. We also consider the broader implications of a normative analysis of judgment and decision making and demonstrate that the bias examined by Ditto and his colleagues is not, in fact, an irrational bias, and that it is incoherent to discuss bias in the absence of standards for assessing accuracy and consistency. Other conclusions about domain-general asymmetries in motivated social cognition have suggested that epistemic virtues are more prevalent among liberals than conservatives, and these conclusions are closer to the truth of the matter when it comes to current American politics. Finally, we question the notion that the research literature in psychology is necessarily characterized by liberal bias, as several authors have claimed.