Partisanship, Political Knowledge, and the Dunning-Kruger
Effect
Ian G. Anson
University of Maryland, Baltimore County
A widely cited finding in social psychology holds that individuals with low levels of competence will judge
themselves to be higher achieving than they really are. In the present study, I examine how the so-called
“Dunning-Kruger effect” conditions citizens’ perceptions of political knowledgeability. While low performers
on a political knowledge task are expected to engage in overconfident self-placement and self-assessment when
reflecting on their performance, I also expect the increased salience of partisan identities to exacerbate this
phenomenon due to the effects of directional motivated reasoning. Survey experimental results confirm the
Dunning-Kruger effect in the realm of political knowledge. They also show that individuals with moderately
low political expertise rate themselves as increasingly politically knowledgeable when partisan identities are
made salient. This below-average group is also likely to rely on partisan source cues to evaluate the political
knowledge of peers. In a concluding section, I comment on the meaning of these findings for contemporary
debates about rational ignorance, motivated reasoning, and political polarization.
KEY WORDS: political knowledge, Dunning-Kruger, overconfidence, motivated reasoning, partisanship
A widely studied phenomenon in social psychology holds that individuals with low levels of
competence will judge themselves to be more competent than they really are, while those with high
levels of competence will underestimate their excellence (see also Dunning, 2011; Ehrlinger,
Gilovich, & Ross, 2005; Kruger & Dunning, 1999). This so-called “Dunning-Kruger effect” is composed
of several interrelated phenomena thought to occur because individuals vary in their awareness
of “known unknowns” (concepts, skills, or experiences that one is aware of but which have not yet
been mastered) relative to “unknown unknowns” (which fall outside of an individual’s cognizance).
As a result, low achievers are unaware of the extent of their ignorance because they are lacking in
metacognitive skills (e.g., Everson & Tobias, 1998).
This “double burden of incompetence” (Dunning, 2011, p. 260) means that low-performing individuals
often overestimate their own objective performance.1
A second and related aspect of the
Dunning-Kruger effect is that these low achievers will be less capable of rating and comparing peers’
1
Critics of the Dunning-Kruger effect have argued that the phenomenon is most likely to occur not because individuals
lack metacomprehension, but rather because their preexisting knowledge is biased (e.g., Krajc & Ortmann, 2008).
Further empirical demonstrations have shown that individuals’ overconfident self-assessments are not a product of the
distribution of task scores or other statistical artifacts (Schl€osser, Dunning, Johnson, & Kruger, 2013).
1173
0162-895X VC 2018 International Society of Political Psychology
Published by Wiley Periodicals, Inc., 350 Main Street, Malden, MA 02148, USA, 9600 Garsington Road, Oxford, OX4 2DQ,
and PO Box 378 Carlton South, 3053 Victoria, Australia
Political Psychology, Vol. 39, No. 5, 2018
doi: 10.1111/pops.12490
bs_bs_banner
performances (Helzer & Dunning, 2012; Hodges, Regehr, & Martin, 2001).2
Paradoxically, these phenomena
combine to affect the ability of low achievers to overcome their incompetence because they
are unaware that they lag behind others until their objective performance is measured and reported to
them. The only way to increase proficiency is to increase metacognition, and vice versa (Kruger &
Dunning, 1999).
The phenomena which comprise the Dunning-Kruger effect have substantial consequences for
contemporary theories of political knowledge. Political knowledge informs the perceived social utility
of engaging in political discussion (David, 2009; Eveland, 2002; Neuman, 1986), whether to seek
additional knowledge from preferred sources (e.g., Cowen, 2005; Redlawsk, 2004) and whether to
adopt ideologically extreme positions (Ortoleva & Snowberg, 2015a). Overconfident citizens may
become emboldened, making strong political assertions in their social networks and resisting persuasive
counterarguments. Despite these potentially important consequences, modern political scientists
have only occasionally examined the subjects of political over and underconfidence3
and have never
engaged in a direct test of the Dunning-Kruger effect.4
Overconfidence may also influence the way that partisans assess the political knowledge of others
(Leeper & Slothuus, 2014). Partisanship causes individuals to raise “perceptual screens” which
increase the likelihood of believing congenial partisan stereotypes and which also reduce partisans’
willingness to believe disconfirming narratives (e.g., Kunda, 1990). This motivated reasoning process
causes partisans to possess biased characterizations of the political knowledge of outgroup and
ingroup members and may lead to increases in overconfidence when partisanship is made salient.
Existing literature therefore identifies two important outstanding research questions. First, to what
extent do contemporary Americans overestimate their levels of political knowledge? And perhaps
more important given the increasingly polarized context of contemporary political discourse: How do
partisan identities condition this phenomenon?
In the present study, I examine political overconfidence using a survey experimental research
design. Two online survey experiments designed to measure political knowledge corroborate the original
Dunning-Kruger effect in the realms of self-assessment, self-placement, and peer assessment. Further,
the studies experimentally manipulate the salience of partisanship in the minds of respondents by
priming partisanship through simple cues (e.g., Bullock, 2011; Klar, 2013). Results show that the
increased salience of partisanship exacerbates gaps between perceived and actual levels of political
knowledge—but only among partisans with moderately low performance on the knowledge battery.
Party cues also interfere with these low-performing partisans’ ability to accurately assess the political
knowledge of in- and out-party peers. Higher-achieving and extremely low-achieving respondents, in
contrast, are relatively unresponsive to party cues. In a concluding section, I comment on the meaning
of these findings for contemporary theories of political knowledge, partisan motivated reasoning, and
political engagement.
Political Overconfidence
Across the disciplines, a multitude of studies has sought to measure overconfidence on objective
task performance. Scholars have largely focused on three types of self-evaluation. Perhaps the most
popular subject of study is overestimation, in which individuals feel their performance on a task is
2
The Dunning-Kruger thesis also holds that top performers on tasks are likely to discount their exceptionalism, rating
themselves as lower achieving than they should (e.g., Schl€osser et al., 2013).
3
By confidence in political knowledge, I do not refer to political efficacy, which instead relates to the feeling that
political action is possible and impactful (Balch, 1974; Campbell, Gurin, & Miller, 1954).
4
A JSTOR search of major political science journals and a Google Scholar keyword search for “dunning” produced
several articles which mention the theory (e.g., Cowen, 2005; Lassen & Serritzlew, 2011; Weissberg, 2001), but no
study incorporated its expectations in a research design.
2 Anson11581174
better than it actually is (e.g., Sheldon, Dunning, & Ames, 2014). Overplacement, or the “Lake
Wobegon” effect,5
in which individuals rate their performance as higher than average, is also an
important subject in prior literature (e.g., Dunning, Heath, & Suls, 2004; Svenson, 1981). Finally,
studies of overprecision show that many individuals believe they possess more accurate information
about a concept than they actually do (e.g., Ortoleva & Snowberg, 2015a; Soll & Klayman, 2004).
Despite these many advances, no existing study of political knowledge has appraised overestimation
or overplacement from the perspective of the Dunning-Kruger framework.6
Some extant research
designs have examined political overprecision from a cross-sectional perspective: Ortoleva and Snowberg
(2015a) examine this aspect of overconfidence as a predictor of ideological extremeness and
strength of partisanship. The authors develop a theory of imperfect information processing which
demonstrates a “correlational neglect” phenomenon, in which citizens fail to recognize and compensate
for the fact that their own experiences are not highly informative of the consequences of public
policy. As a result, those who suffer from correlational neglect will exhibit strengthened partisanship,
ideological extremism, and increased voter turnout (Ansolabehere, Meredith, & Snowberg, 2014; see
also Ortoleva & Snowberg, 2015b).
A second strand of political science literature has investigated the link between certainty and
knowledgeability in the political realm.7
The majority of such studies has examined the relationship
between confidence in political perceptions and opinions on policies. Druckman (2004) shows that
those citizens who are most susceptible to framing efforts are often highly confident in their political
beliefs—a finding that squares well with existing research on the Dunning-Kruger effect. Pasek,
Sood, and Krosnick (2015) examine perceptions of the Affordable Care Act, showing that citizens
with incorrect knowledge of the law can be further subdivided into those proferring uncertain and certain
responses (Kuklinski & Quirk, 2001; see also Kuklinski, Quirk, Jerit, Schwieder, & Rich, 2000).
The authors argue that confidently incorrect individuals should be construed as possessing misperceptions,
while uncertain respondents acknowledge that they lack the relevant political knowledge to
make claims about an issue. This line of reasoning resonates with recent work on the causes and consequences
of political misperceptions (Flynn, Nyhan, & Reifler, 2017; Jerit, Barabas, & Bolsen, 2006;
Nyhan & Reifler, 2010; Prasad et al., 2009).8
Taking these findings from social psychology and political science together, we would expect
citizens’ self-assessments of political knowledge to correspondingly deviate from their objective performance
on political knowledge batteries. But according to the Dunning-Kruger thesis, this pattern
should be especially evident among those with low levels of political knowledge. These citizens will
lack metacognitive awareness in their political lives, as they are unaware of the “unknown unknowns”
that constitute sophistication in the political realm. This phenomenon is likely exacerbated by the
prevalence of civic norms of engagement and participation in contemporary American society (e.g.,
Galston, 2001; Theiss-Morse & Hibbing, 2005). Many citizens will feel remiss to admit to themselves
that they possess low levels of civic competence because such deficiencies are socially undesirable.
But in addition to these norms, low achievers on political knowledge tests may be chronically lacking
5
On Prairie Home Companion, Garrison Keillor would introduce the fictional town of Lake Wobegon, Minnesota as a
place where “the women are strong, the men are good looking, and all the children are above average.”
6
Political knowledge has been variously defined since the advent of modern political behavior research. It can be differentiated
from political sophistication, a more encompassing concept which relates to knowledge, awareness, and
ideological constraint (e.g., Converse, 1964; Luskin, 1990). According to Luskin (1990), “a person is politically
sophisticated to the extent to which his or her political cognitions are numerous, cut a wide substantive swath, and
are highly organized, or constrained” (p. 332). A more concise description in this vein is political expertise.
7
For a tangential literature on the measurement of political knowledge giving the proclivity of unsure respondents to
select “don’t know” as a response to survey questions, see Mondak and Davis (2001) and Mondak and Anderson
(2004).
8
Gvirsman (2015) also examines how political knowledge conditions “false projection,” a phenomenon in which individuals
use their own beliefs about a political controversy to inform their assessments of majority public opinion on
the issue (in this case, dismantling of Israeli settlements in Palestinian territories.)
Partisanship, Political Knowledge, and the Dunning-Kruger Effect 311591175
in opportunities to engage in discourse with knowledgeable peers about politics (Eveland & Hively,
2009; La Due Lake & Huckfeldt, 1998), which, according to the Dunning-Kruger thesis, may have
otherwise led them to overcome the “double burden of incompetence” that characterizes a lack of
metacognitive ability (Kruger & Dunning, 1999). Thus, we should expect the overplacement component
of the Dunning-Kruger effect to extend to the realm of political knowledge:
H1 (Self-Placement): When asked to report their performance relative to peers on a
political knowledge battery, low achievers will consistently rank themselves as above
average.
However, one salient consideration leads to different theoretical expectations for a political DunningKruger
effect when compared to more general learning and task-performance applications: Many
Americans possess partisan identities which could interfere with accurate self-assessment and
self-placement of political knowledgeability.
Partisan Stereotypes, Overestimation, and Overplacement
Social context has been found to condition the Dunning-Kruger effect across a wide variety of
performance tasks, as an individual’s socially constructed self-image informs their level of over- or
underconfidence due to the prevalence of stereotypes about in- and outgroups (Ariely & Norton,
2007; Camerer, Hogarth, Budescu, & Eckel, 1999; McGlone & Aronson, 2006; McGlone, Aronson,
& Kobrynowicz, 2006). One important class of such “stereotype threats” encompasses racial and gender
identities. For example, women are often susceptible to prevalent stereotypes about the gender
gap in science education, such that priming gender results in a meaningful reduction of confidence in
women’s self-perceptions on science exams (Dunning, Johnson, Ehrlinger, & Kruger, 2003).9
Similarly, priming minority identities results in greater anxiety and negative self-perception among
members of these groups when performing a variety of tasks (e.g., Steele & Aronson, 1995). Because
participants in Dunning-Kruger experiments are sensitive to the preservation of feelings of self-worth
in the face of threatening social contexts, their self-placement and self-assessment can be strongly
conditioned by the activation of social identities.
Political identities are also likely to play an important role in self-assessments of political knowledge.
In the realm of factual perceptions, the motivated reasoning framework shows that partisans are
often unreliable reporters of objective reality (Jerit & Barabas, 2012; Kunda, 1990; Mele, 1997). This
is due in large part to the way that partisans acquire information about current events and conditions.
When partisans engage with new political content, they are driven by both directional and accuracy
goals (e.g., Leeper & Slothuus, 2014). In order to serve directional (partisan-motivated) considerations,
partisans develop and maintain a “preferred-world state” which operates through wishful thinking
in the realm of predictions and assessments (Babad, 1997; Jerit & Barabas, 2012; Parker-Stephen,
2013). This preferred interpretation of reality causes partisans to develop a set of beliefs which contrast
with objective appraisals of reality.
Interpretations of objective facts aside, motivated reasoning can also strongly influence partisans’
perceptions of the characteristics of in- and out-party group members (e.g., Huddy, 2002; Rahn,
1993). Iyengar and Westwood (2015) demonstrate that in recent years, partisan stereotyping has
increased such that partisans often hold remarkably hostile feelings for the opposition (so-called
“negative partisanship”). This party-stereotyping phenomenon means that individuals will engage in
greater directional reasoning if they possess enough awareness to realize that their partisan identity is
9
A separate literature in political science has debated a gender gap in political knowledge. See Dolan (2011) for a
partial review.
4 Anson11601176
meaningful to a particular social interaction (Bolsen, Druckman, & Cook, 2014; Nicholson, 2011,
2012). Party identities can be made more salient simply by mention of partisan in- and outgroups or
by making reference to recognizable party leaders. When partisans expend cognitive resources in service
of a task, they will become even more likely to turn to partisan stereotypes to inform their behavior,
if cues are available to them. They will also become resistant to new information which could
otherwise contradict deleterious party stereotypes (e.g., Nyhan & Reifler, 2010).
The Dunning-Kruger effect is likely to be exacerbated by the salience of party cues, both in terms
of self-placement and in evaluations of others’ performance. When partisans access schema related to
their political identities, they are also likely to be reminded of political knowledgeability as a criterion
for judging the qualities of political in- and outgroups. As the partisan “preferred world state” holds
that ingroup partisans are politically knowledgeable relative to the outgroup (among a host of other
indicators of competence), partisans will become more defensive of their self-placement when asked
to reflect on their political knowledge. Even if partisans can extract a signal about their performance
from the task itself, such directional considerations may come to overwhelm disconfirming performance
signals. This process is analogous to situations in which partisans reject disconfirming information
that is damaging to an ingroup political incumbent in favor of more congenial facts (e.g., Anson,
2016; Bisgaard, 2015).
H2 (Party Cues and Self-Evaluation): When partisan identities are made salient through
partisan cueing, subjects will exhibit increased evidence of overestimation and overplacement
relative to the control group.
While directional motivated reasoning might influence self-appraisals in the aggregate for these
reasons, it is those who are moderately low in political knowledge who are expected to most strongly
condition their self-placement and peer assessments in response to the increasing salience of partisanship.
This expectation derives from the notion that the unsophisticated will struggle to extract performance
signals from their performance. The Dunning-Kruger thesis holds that low achievers fail to
accurately self-place due to an inability to engage in metacognitive reflection on their performance.
Those with extremely low levels of political knowledge are likely to be unaffected by party cues
entirely, because they will be incapable of associating task performance with partisan threat. In this
case, self-placement will stem from neither directional goals nor metacognition—an entirely
“uneducated guess” which entertains neither party cues nor performance signals. Higher achievers are
also likely unaffected by stronger accuracy signal due to their good understanding of what constitutes
successful task performance. The moderately low-achieving group, in contrast to very low and high
acheivers, is most willing to rely upon directional motivations in the face of partisan priming. This
group cannot assess performance using knowledge about the task itself, but its members do understand
the socially undesirable implications of low performance (e.g., Leeper & Slothuus, 2014). A
dearth of interpretable evidence allows directional motivated reasoning to occupy a more central role
in the formation of these low-achievers’ perceptions of performance. We should therefore expect a
pattern in which partisan cueing causes overconfident self-placement to increase among partisans with
moderately low levels of objective performance.
H3a (Task Performance-Cue Interaction): When partisan identities are made salient
through a party cue, moderately low-scoring individuals will be most likely to exhibit
heightened levels of overestimation and overplacement relative to control groups.
H3b (Task Performance-Cue Interaction): When partisan identities are made salient
through a party cue, very low-scoring individuals will be unlikely to exhibit heightened
levels of overestimation and overplacement relative to control groups.
Partisanship, Political Knowledge, and the Dunning-Kruger Effect 511611177
H3c (Task Performance-Cue Interaction): When partisan identities are made salient
through a party cue, average and high-performing individuals will be unlikely to exhibit
heightened levels of overestimation and overplacement relative to control groups.
When evaluating peer performance, partisan source cues should most strongly influence peerperformance
assessments among this moderately low-achieving group as well. Kruger and Dunning
(1999) posit that such subjects will be unable to correctly place the performance of peers when provided
with examples because they cannot recognize when peer performance is of high or low quality
relative to their own. Low-achieving Republicans and Democrats will make stereotype-driven
assumptions about the comparative quality of their own responses relative to a partisan peer. This is
because moderately low-performing partisans, having the ability to employ source cues to assess peer
quality, will assume that if their responses differ from an outgroup peer, they are more likely to be
correct than the peer. In this case, the peer is assumed to be stereotypically deficient in political
knowledge. Likewise, if an ingroup peer’s responses differ from their own, they will be more likely to
consider the possibility that the peer is more knowledgeable than they are (as ingroup peers are
assumed to be knowledgeable).
These biases will be stronger when a respondent has guessed on most of the knowledge questions,
as they will have fewer opportunities to extract a performance signal by comparing across responses
they know to be correct. But when the respondent exhibits extremely low achievement on the knowledge
battery, they will also be less likely to respond to party cues due to their complete lack of political
sophistication. As discussed above, such respondents will possess neither performance signals nor
the ability to interpret cues which fill in for this information. The result will be mere guesswork characterized
by high variance.
In contrast, those with a strong understanding of objective performance will possess a
much more direct performance signal than a reliance upon in- and outgroup stereotypes and
will be capable of making accuracy-driven assessments of peer performance which can
obscure or counteract the effects of directional motivated reasoning. Taken together, we again
expect a situation in which moderately low-performing respondents will be most likely to use
party cues to assess peer performance:
H4a (Moderately Low-Performance Peer Evaluation): When moderately low-scoring
partisans are made aware of the ingroup (outgroup) partisan identity of a fictional highachieving
peer, their assessments of peer performance will increase (decrease).
H4b (Very Low-Performance Peer Evaluation): When very low-scoring partisans are
made aware of the ingroup or outgroup partisan identity of a fictional high-achieving
peer, their assessments of peer performance will be unchanged.
H4c (High-Performance Peer Evaluation): When high- and average-scoring partisans
are made aware of the ingroup or outgroup partisan identity of a fictional highachieving
peer, their assessments of peer performance will be unchanged.
Overall, then, we have assembled a series of expectations which posit biases in the ways
that contemporary Americans estimate the extent of their political knowledge—especially
among those with below-average levels of political knowledge. In order to test the existence
of such patterns, it is necessary to create a research design which measures both the objective
performance and the self-assessments of citizens, while simultaneously manipulating the
salience of political identities. Below, I introduce a methodology designed to accomplish
these tasks.
6 Anson11621178
Research Design
While most tests of the Dunning-Kruger effect have been performed using undergraduate student
samples in laboratory settings, some studies have replicated the effect using online samples (e.g.,
Schl€osser et al., 2013).10
In the present study, I engage in two tests of the Dunning-Kruger effect using
large samples of online survey respondents recruited in April and June of 2017. The first sample was
recruited using Qualtrics’ qBus omnibus program, comprising 1,047 respondents in total. Qualtrics
qBus surveys are online omnibus instruments which recruit panelists from actively managed social
media platforms into large-scale national convenience samples for corporate and academic applications.
Panelists’ demographics resemble Census estimates of age, gender, income, race, and education
(see the online supporting information for summary statistics). The second sample (N 5 1,559) was
recruited using Amazon AWS’ MTurk service.11
Respondents to both surveys completed a political knowledge quiz, which asked five questions
designed to tap knowledge of basic political institutions, awareness of current political conditions, and
ideological differentiation (e.g., Luskin, 1990). These questions were modeled after standard knowledge
batteries found in recent iterations of the American National Election Study (ANES; American
National Election Studies, 2016). See the online supporting information for a closer examination of
this question battery. Following Prior, Sood, and Khanna (2015), a preamble first assuaged respondents
not to worry about their performance (while instructing them not to cheat): “Please try to answer
this Political Quiz to the best of your ability. There is NO penalty for incorrect answers.” Basic institutional
knowledge was assessed using questions which asked respondents to identify the number of
years served by a Senator, as well as the name of the current Secretary of Energy, from four possibilities.
Ideological differentiation was assessed by asking respondents to identify which party is more
conservative on the issue of health care. Respondents were also asked which political party currently
controls the House of Representatives and on which of four different programs the Federal government
spends the least. Together, these questions form an additive scale of political knowledge which
ranges from 0 to 5. See the online supporting information for a more detailed discussion.
In Study 1, immediately following the political knowledge battery, respondents were divided into
two groups. The control group performed a simple ordering task designed to take approximately the
same amount of time as the treatment group to assuage concerns that the treated group would experience
increased satisficing due to the length of the treatment (Krosnick, 1991).12
The treatment group
was instead exposed to a univalent partisan priming task modeled after recent studies by Klar (2013)
and Lavine, Johnston, and Steenbergen (2012). The task asked respondents to “think about Americans
who identify as members of the two major political parties” and to select attributes which they associate
with partisan groups. These words include positive considerations such as “intelligent,”
“compassionate,” and “hardworking,” in addition to those with negative valence, such as
“irresponsible” and “misguided.” This priming cue is designed to heighten the salience of partisanship
in the minds of the treated group relative to other identities. After treatment exposure, groups were
asked to provide a self-assessment of their mastery of the knowledge battery (“excellent,” “very
good,” “good,” “fair,” or “poor”) and to rate their performance relative to others who had taken the
test (self-placement) by identifying their perceived performance quintile.
10
This approach is also consistent with prior research on political knowledge and certainty (e.g., Pasek et al., 2015).
11
An ongoing debate in the social sciences has questioned the validity of results derived from MTurk workers. However,
the majority of this literature is optimistic regarding the attention levels, effort, and generalizability of data drawn
from MTurk samples (e.g., Berinsky, Huber, & Lenz, 2012; Levay, Freese, & Druckman, 2016). Further, additional
literature has recently shown that poststratification weights can be applied to MTurk samples to help assuage concerns
of unrepresentativeness (Huff & Tingley, 2015).
12
Respondents were asked to rank the size of various common objects from largest to smallest.
Partisanship, Political Knowledge, and the Dunning-Kruger Effect 711631179
Study 2’s design similarly exposed respondents to the political knowledge battery and subsequently
measured their self-assessment and self-placement. However, after completing this selfassessment,
subjects were next asked to evaluate the performance of another (fictional) respondent.
This design follows the work of Hodges et al. (2001), who use a “similar grading” task to demonstrate
that those with low performance were also poor evaluators of the performance of peers. The fictional
respondent’s answers were presented to the subject in tandem with basic demographic information,
including age, gender, and political partisanship. These vignettes were randomly varied to identify the
fictional respondent as either a Republican or a Democrat, reflecting not a univalent party-priming
cue, but a simple affiliation-based source cue (Bullock, 2011; Goren, Federico, & Kittilson, 2009).
After reading through the respondent’s answers, subjects were asked to evaluate the fictional respondent’s
performance on the battery. In both studies, partisanship was measured on a 7-point scale, following
the standard ANES question wording (American National Election Studies, 2016). Leaning
partisans were included in Republican and Democratic groups (e.g., Klar & Krupnikov, 2016). Party
identification was measured prior to a distractor task battery and subsequent treatment exposure in
keeping with recent literature cautioning against stratifying on posttreatment covariates (Montgomery,
Nyhan, & Torres, 2016). For more information on question wording and experimental design, see the
online supporting information.
Treatment-control contrasts are measured in the aggregate using a simple t-test design, in addition
to an interactive model which measures treatment effects across task performance quartile in Study 1.
In Study 2, an interactive model estimates the effects of the treatments on the outcomes of interest
across task performance and partisanship.13
For ease of interpretation, subgroup results are also presented
using simple t-test comparisons; full models are presented in the online supporting
information.
STUDY 1
A Political Dunning-Kruger Effect
Study 1 examines Hypotheses 1–3 using data from the 2017 qBus sample (N 5 1,047). The
results of this study corroborate the original findings of Kruger and Dunning (1999), by demonstrating
that on average, individuals with low scores on the political knowledge battery substantially overestimate
the perceived quality of their performance and their perceived performance relative to others.
Before examining the effects of party cues on this pattern, I assess the objective performance of
respondents on the political knowledge battery. Below, Figure 1 demonstrates the distribution of the
number of correct responses achieved by the 2017 qBus sample.
The task was designed to be difficult to master, an expectation that we see fulfilled in Figure 1
above. Only around 10.6% of the sample correctly answered all five questions, while the modal score
was three correct responses (26.9% of the sample). The mean score on the test was 2.63 out of 5. But
to assess respondents’ perceptions of their performance relative to this objective measure, we should
also examine the respondents’ self-placement scores in Figure 2.
Figure 2 replicates the original plot of perceived and observed task performance from Dunning
and Kruger (1999).14
This plot demonstrates that overall, low-performing respondents were quite optimistic
in their self-perceptions. The solid line in the figure shows the performance quintile achieved
by respondents having answered a given number of questions correctly. We see respondents achieve a
13
Consistent with Franco, Malhotra, Simonovits, and Zigerell (2017), survey weights are not employed in the findings
presented below; poststratification-weighted results are robust to these findings.
14
For a tabular representation of the data analysis seen in Figure 2, please see Table 2 in the online supporting information
for results of a linear regression model.
8 Anson11641180
higher performance quintile with every additional correct answer, with the exception of the very best
performers (four out of five correct answers were sufficient to place in the top 20%). The dashed line
shows the mean self-placement score for respondents at each level of quiz performance.
Figure 2 shows us that respondents at the lowest levels of performance suffered from a high
degree of overplacement, a finding which supports Hypothesis 1. Respondents at the very lowest performance
levels ranked themselves, on average, in the fourth quintile. Those exhibiting average or
near-average performance, with scores of 2 or 3, similarly thought of themselves as having performed
much better than average. Only at the highest echelons of performance does self-placement rise
Figure 1. Distribution of correct knowledge battery responses, 2017 qBus sample.
Figure 2. Observed vs. perceived performance quintile, 2017 qBus sample. For this graphic and all subsequent graphics,
vertical lines denote 95% confidence intervals.
Partisanship, Political Knowledge, and the Dunning-Kruger Effect 911651181
further, though the group of respondents scoring 5 out of 5 on the task still underestimated their performance
on average.
This preliminary view of the distinction between perceived and objectively measured performance
on the task offers initial confirmation of the Dunning-Kruger phenomenon. As can be readily
assumed from the 95% confidence interval estimates seen in Figure 2, one-sample t-tests also reject
the null hypothesis that poor performers’ self-placement estimates include the theoretical median of 3
out of 5 (p < 0.01 for all performance groups). Even respondents with observed scores of 0 out of 5
rated themselves as significantly more politically knowledgeable than average—well within the fourth
quintile (t 5 4.64, p < 0.01). These results provide strong evidence in support of Hypotheses 1.15
Skeptical observers of this replication might look to measurement issues to explain the reason for
the overplacement pattern—indeed, the Dunning-Kruger effect has come under substantial scrutiny
for these reasons in recent years (e.g., Krajc & Ortmann, 2008). Critics have suggested the potential
for floor and/or ceiling effects, in which the tendency of respondents to moderate comes from the
uncertainty space of the response, as well as low performers making assumptions about the distribution
of “peer” responses. The most important critiques have argued that citizens exposed to DunningKruger
tasks will often exhibit a floor effect stemming from an assumed “backwards-J” performance
curve. On very difficult tasks where the median respondent performs objectively poorly, participants
might correctly infer that the vast majority of performers are unskilled, thus producing a distorted
understanding of self-placement due to the task difficulty. But as Schl€osser et al. (2013) and the present
study’s Figure 1 demonstrate, respondents’ performance on many Dunning-Kruger tasks (including
the knowledge battery) more closely approximates Normal rather than J-shaped curves, meaning
that respondents’ expectations about the distribution of responses are not likely to be a product of the
difficulty of the task. Essentially, the nature of the task itself should not drive the results seen here,
meaning that respondents’ assumptions about the task difficulty will be a product of their (lack of)
metacognition. Correspondingly, measurement issues should be less influential on the observed overplacement
phenomenon than the psychological self-regard of respondents.
This first glance at the phenomenon obscures important differences between treated and control
groups in Study 1, as the salience of partisanship is expected to influence the distribution of selfplacement.
I assess Hypothesis 2 below through a comparison of experimental groups.
Partisan Cueing and Overconfidence
Figure 3, below, demonstrates the experimental effects of partisan priming on overplacement and
overestimation.16
Respondents who engaged in the partisan-cueing task after completing the quiz had,
on average, a perceived mastery score that was 0.14 points higher than those in the control group (on
a 5-point scale; t 5 1.91, p 5 0.06.) Correspondingly, those receiving the partisan-priming task also
rated their performance as 0.14 quintiles higher than the control on average (t5 2.08, p 5 0.04). It
appears that when asking respondents to reflect on the qualities and defects of partisan groups, willingness
to engage in overplacement significantly and substantially increased in the qBus sample,
while overestimation increased to a similar but nonsignificant degree. These findings work to corroborate
Hypothesis 2 when it comes to self-placement. They also provide suggestive evidence in favor of
Hypothesis 2 from the perspective of self-assessment, though given that self-placement is by
15
Interestingly, we also see evidence of underplacement among the highest performers, an expectation of the DunningKruger
effect which states that these respondents will assume their exceptional knowledgeability is more common
than it really is. This underplacement could stem from such individuals’ reflection upon their social networks, which
are likely similarly nonmedian in task performance.
16
For tabular regression results which correspond to Figure 3, please see the online supporting information Tables 3 and
4.
10 Anson11661182
definition comparative, it seems sensible that self-placement would exhibit more consistent increases
in response to the party cue than self-assessment.
While Figure 3 provides us with a straightforward presentation of average effects across the full
sample of respondents, it only lends partial insight into the phenomenon in question. To investigate
further, I present Figure 4, which, much like Figure 2, shows mean self-placement scores across levels
of objective performance. Figure 4, however, subdivides treated and control groups by observed performance
quartiles.17
The fourth quartile comprises those who answered four to five questions
Figure 3. Comparison of treated and control self-placement across quiz score, 2017 qBus sample.
Figure 4. Comparison of treated and control self-placement across quiz score, 2017 qBus sample.
17
Quartiles are used to assess performance in order to retain sufficient subgroup sample size to permit comparisons. For
regression results pertaining to Figure 4, please see Table 4 in the online supporting information.
Partisanship, Political Knowledge, and the Dunning-Kruger Effect 1111671183
correctly, the third encompasses those answering three questions correctly, the second quartile features
two correct answers, and the lowest performance quartile includes respondents answering zeroto-one
questions correctly. This subgroup analysis allows us to evaluate Hypothesis 3, which asserts
that below-average performers will be most susceptible to the effects of party cues.
A first glance at Figure 4 shows that the pattern of overplacement across treated and
control groups varies considerably across performance groups. As performance declines,
treated units’ mean self-placement, denoted by the rightmost point in each panel, diverges
from the self-placement of the control units, until these estimates converge again for the
lowest quartile. Panels featuring hollow point estimates of perceived performance show
ranges of performance for which the 95% confidence intervals overlap enough to produce
nonsignificant estimates. This was true of the performers in the highest quartile, who correctly
answered four to five questions correctly (t 5 20.62, p 5 0.53). It appears that those
subjects with strong performance on the task were relatively immune to partisan cues, a
finding which comports well with theoretical expectations.
Instead, it is treated units with below-average performance who are more likely than analogous
units in the control condition to engage in overplacement, a finding which provides additional evidence
in support of Hypothesis 3. Respondents in the second quartile, correctly answering two out of
five questions, exhibited a strong treatment effect (t 5 2.48, p < 0.01). To a lesser and nonsignificant
extent, treated subjects in the third quartile also witnessed an increase in self-placement relative to the
control group (t5 1.64, p 5 0.10). Those at the low end of the performance distribution (answering
zero-to-one questions correctly), however, were immune to identity cues (perhaps because they lack
the minimal awareness to connect such cues to performance estimation, as discussed above; t 5 0.22,
p 5 0.82). Overall, Study 1 provides evidence that below-average performers were especially susceptible
to overplacement when primed using simple party cues, while other groups were less influenced
by the partisan-cueing task.
STUDY 2
Performance Assessment in Political Context
Thus far, results from Study 1 have revealed that low performers on a knowledge battery react to
simple party cues by increasing their self-placement. In the following sections, I examine partisan
source cues as a second type of cue which works to condition the Dunning-Kruger effect. Specifically,
Study 2 provides evidence that among those with moderately low performance on political knowledge
tasks, peer assessments are also strongly influenced by partisan identities. However, these effects are
not consistent across Republican and Democratic groups, a surprising finding which is deserving of
further discussion.
In Study 2, MTurk respondents completed the knowledge battery and assessed their own performance.
These self-placement and self-assessment scores deviate from respondents’ objective performance
in a pattern analogous to the findings presented above. Figure 5 shows this pattern.
In Figure 5, those with quiz scores of 0, 1, and 2 substantially overestimate their performance by
more than a full quintile.18
However, while this pattern of overplacement among low-performers replicates
in the MTurk sample, we observe a curious depression in self-placement across almost all
respondents relative to the qBus sample. No group is willing to rate themselves higher than around
the 4th quartile on average, while most respondents rate themselves near or below the median. It is
unclear why MTurk respondents would be so much less confident in their responses relative to the
18
For regression results pertaining to Figure 5, see Table 8 in the online supporting information.
12 Anson11681184
qBus sample, although it may be that MTurk users expected many of their peers to cheat or to possess
high levels of competence. We also know that MTurk respondents are highly concerned about payment
due to the prevalence of attention checks in academic MTurk studies, possibly prompting these
subjects to engage in more honest self-appraisals (Berinsky, Margolis, & Sances, 2014; Hauser &
Schwarz, 2016). Regardless, the findings presented in Figure 1 show that low performers do not
understand the extent of their lack of competence even when accounting for this “negative intercept
shift.” Once again, we see evidence of the Dunning-Kruger effect.
Study 2 goes further to assess peer-performance evaluations. After completion of the knowledge
task, respondents were asked to evaluate the political knowledge of a (high-performing) fictional
Republican or Democrat. Expectations related to Hypothesis 4 are assessed by Figure 6 below.19
This
graphic shows the average rating of a (fictional) peer’s performance on the same five-item knowledge
battery implemented in Study 1. Figure 6, much like Figure 4 above, is faceted into four panels for
ease of interpretation, depicting the relationships under consideration across objective performance
quartiles.20
Each facet contains four estimates with 95% confidence intervals: Democrats’ assessments
of Democratic and Republican peers’ performance (the far left and center-left bars within each panel,
respectively) and Republicans’ assessments of Democratic and Republican peers’ performance (the
far-right and center-right bars). Point estimates of peer evaluations are represented by D’s (the mean
evaluation of a Democratic peer) and R’s (showing the mean evaluation of a Republican peer). In this
way, Figure 6 visualizes a three-way interaction between respondent’s party, peer’s party, and objective
task performance.
The rightmost panel of Figure 6 shows that irrespective of respondent or peer partisanship,
respondents with the highest levels of political knowledge were able to correctly identify that the peer
was high-performing (in all conditions, the fictional peer “answered” all five questions correctly). No
Figure 5. Observed vs. perceived performance quintile, 2017 qBus sample.
19
For regression results pertaining to Figure 6, see Table 9 in the online supporting information.
20
The modal response among the MTurk sample was identical to that of the qBus sample: three answers correct out of
five.
Partisanship, Political Knowledge, and the Dunning-Kruger Effect 1311691185
cross-party or treatment contrasts approach conventional levels of significance when examining the
peer ratings of high-achieving respondents—a null finding which is expected given Hypothesis 4.
This is consistent with the expectation that high performers will possess a good understanding of
objective peer performance thanks to their own performance. However, looking across the point estimates
towards the leftmost facets of the figure, we see initial evidence that respondents at lower levels
of objective performance struggle to identify the fictional peer as excellent. The mean peer assessment
among Quartile 2 and Quartile 3 respondents is around 0.63 points lower than that of above-average
respondents (t 5 15.65; p < 0.01), reflecting the expectation that these respondents are generally less
capable of extracting an accurate performance signal from the peer’s objective performance. Further,
respondents in Quartile 1 exhibited very high variance in their responses, despite mean evaluations
which were lower on average than other groups.
But in addition to these overall decreases in peer ratings as knowledge decreases, Figure 6 allows
us to evaluate the extent of directional motivated reasoning in peer evaluations through an assessment
of treatment effects. Within the Q3 performance group, Democrats (t5 0.42, p 5 0.68) and Republicans
(t 5 20.30, p 5 0.76) show little evidence of a reliance upon party cues to assess the political
knowledge of a peer as the peer’s partisanship varies. This is especially interesting given that these
groups are generally much worse than the high performers at recognizing the peer’s excellence, as
seen above. High performers in Q4 are similarly unaffected by the treatment, consistent with Hypothesis
4 (Republicans: t5 0.61, p 5 0.54; Democrats: t 5 0.04, p 5 0.97). And again consistent with
Hypothesis 4, very low performers in Q1 also exhibit high variance and little evidence of a treatment
effect (Republicans: t 5 20.07, p 5 0.95; Democrats: t 5 20.49, p 5 0.63).
However, this pattern contrasts with that of moderately low-performing partisans, for whom the
expected treatment effects are most evident. In terms of cross-party differences in reaction to the fictional
peer, we see evidence that moderately low-performing Democrats rated the Republican peer as
substantially less knowledgeable than did moderately low-performing Republicans (t 5 22.20,
p 5 0.03). Analogously, this group of Democrats saw their ingroup peer, the fictional Democrat, as
more knowledgeable than did Republicans (t 5 2.98, p < 0.01). But when assessing within-party treatment
effects, variation in the treatment condition yielded a much larger contrast in performance evaluations
among Republican respondents compared to Democratic respondents. Low-performing
Democrats rated the Democratic peer as roughly 0.22 points better than the Republican peer on
Figure 6. Perceptions of peer performance across party and treatment, 2017 MTurk sample.
14 Anson11701186
average (on a 5-point scale; t 5 1.53, p 5 0.13). This finding is in the expected direction, though it
does not approach conventional levels of statistical significance. The weak pattern for Democrats
stands in contrast to the striking treatment contrast exhibited by moderately low-performing Republicans,
who rated the Democratic peer as roughly 20.60 points worse than the Republican peer on average
(t 5 23.52, p < 0.01).
We can attend to this surprising asymmetry with a host of speculative explanations. We could
cautiously separate these explanations into two categories: measurement issues and asymmetries in
the ways in which partisans engage with the world around them. The first set of explanations largely
stem from the existing literature on MTurk samples—especially studies which have suggested that
Republican MTurk users are unlikely to resemble the average Republican (e.g., Huff & Tingley,
2015). While recent findings have shown that MTurk samples tend to skew liberal in general (e.g.,
Levay et al., 2016), it is also possible that Republican self-identifiers on MTurk batteries will be differently
susceptible to directional cues for less well-known reasons. The second rationale for this phenomenon,
the existence of meaningful partisan asymmetries, I take up in the discussion below.
CONCLUSIONS
More than half a century after Downs (1957) asserted a theory of rational ignorance in American
democracy, evidence continues to mount that most citizens know little about the institutions, actors,
and processes of American government (e.g., Delli Carpini & Keeter, 1997). Among those lacking in,
but not entirely devoid of political knowledge, the present findings speak to an ignorance of ignorance—Kruger
and Dunning’s (1999) “double burden of incompetence”—which also allows for partisanship
to exacerbate political overconfidence. This result is normatively worrying from the
perspective of citizens’ ability to self-correct, as it may be that rationally ignorant Americans are especially
confident that they are better informed than many of their (partisan) peers. The rationally ignorant
fail to overcome their ignorance not just because they face steep costs and lack incentives to
improve, but because they are unaware that they are relatively ignorant. They become increasingly
hardened to the possibility that they are uninformed when partisan identities are activated, a commonplace
feature of most contemporary political discussion. Social norms of civic engagement and participation
(which are prevalent even among the moderately unsophisticated) are unlikely to turn these
citizens away from rational ignorance, because they are likely satisfied by the assumption that their
knowledge outpaces their peers.
We do, however, see evidence that among even the modestly knowledgeable, party cues do less
to exacerbate overplacement on political knowledge tasks. The directional goals of motivated reasoning
may still be strong among partisans in this group, especially because pretreatment effects are notoriously
difficult to evaluate (e.g., Druckman & Leeper, 2012). But these citizens can also pursue
accuracy motivations thanks to their ability to better estimate the true distribution of responses in the
population. This finding assuages concerns that partisans’ overconfidence will harm the quality of
political discussion and information acquisition, as politically knowledgeable individuals will be the
most likely to engage in such behavior in the first place (e.g., David, 2009). But much like the phenomenon
of selective exposure to political content (e.g., Prior, 2007), the notion that the welleducated
are accurate judges of political competence does little to assuage our concerns about those
for whom political knowledge is more substantially lacking.
We also see evidence in Study 2 that Republicans use partisan cues to judge peers’ political
knowledge to a greater extent than do Democrats. Such a thesis speaks to the burgeoning literature on
“asymmetric polarization,” which finds that Republicans have become more committed ideologues
than Democrats in recent years (Grossmann & Hopkins, 2016). But more broadly, the findings resonate
with recent studies on the nature of “factual polarization” in the American electorate (e.g.,
Partisanship, Political Knowledge, and the Dunning-Kruger Effect 1511711187
Bisgaard, 2015; Jerit & Barabas, 2012; Parker-Stephen, 2013). Americans are increasingly likely to
report starkly different sets of factual beliefs about political actors and current conditions as a function
of their partisanship. Online discussion is rife with unsubstantiated rumors, which propagate through
polarized political discussion networks. While the extent of these phenomena and their effects are the
subjects of active scholarly debate (e.g., Berinsky, 2017), the present findings speak to one potential
pathway through which relatively uninformed partisans might come to reinforce their nonmedian factual
beliefs. Echo chambers can grow more cloistered when political overconfidence and party cues
interact to negatively influence citizens’ reliance on the political knowledgeability of their peers.
While we might otherwise expect that interaction with knowledgeable peers would cause selfcorrection
among the moderately uninformed, the strong influence of selective media exposure could
be unchecked by this form of discourse. Rumors, misinformation, viral stories, and partisan framing
would be left immune to serious metacognitive reflection. As Kruger and Dunning (1999) and others
have shown, the “double burden” of incompetence is paradoxical because it is self-reinforcing—a
notion that resonates with current work on political polarization.
These studies possess several inherent methodological limitations and shortcomings, which caution
against an uncritical interpretation of the present analyses. While a power analysis (see the online
supporting information) demonstrates that the sample sizes in both studies are appropriate, analyses
which necessitate scrutiny of partisan and performance subgroups are occasionally close to being
underpowered. We might also inevitably question the design of the knowledge battery, despite evidence
(in the online supporting information) that task performance strongly correlates with known
predictors of political knowledge like education. And one might wonder about the experience of
respondents reflecting upon the task in an online survey-experimental setting—especially the possibility
that some of the respondents cheated on the task despite being told not to. These concerns more
broadly relate to critiques of the external validity of online survey experiments, which likely apply to
the present study. While it is hard to imagine a research design which could evaluate the DunningKruger
effect without subjecting respondents to a highly artificial knowledge assessment, we must
acknowledge that it requires some imagination to connect these effects to the respondent’s real-world
political environment.
Nevertheless, the present findings foreshadow several downstream consequences of the DunningKruger
effect for theories of political participation, partisan motivated reasoning, and political discussion.
Future research is poised to further explore how political overconfidence influences political
engagement, ideological commitment, and extremism, especially when citizens communicate with
homogeneous and diverse social networks. It is likely that those with low political knowledge will
overestimate their performance relative to their (rare) discussion partners, especially when those partners
identify with the out-party. And following the example of Ortoleva and Snowberg (2015a), pressing
questions emerge regarding the degree to which overplacement in political knowledge might lead
to the adoption of more extreme ideological positions, as well as more unshakeable partisan identities.
However, much work remains to further theorize and empirically investigate these important patterns.
ACKNOWLEDGMENTS
The author wishes to thank Carolyn Forestiere, William Blake, Simon Stacey, Marc Hetherington,
Laura Bucci, Chris Anson, Chris DeSante, Elizabeth Holz, Tom Schaller, and Tyson
King-Meadows for their comments and assistance on earlier versions of this work. All errors and
omissions remain my own. Correspondence concerning this article should be addressed to Ian G.
Anson, Department of Political Science, University of Maryland, Baltimore County, 1000 Hilltop
Cir., 305 PUP, Baltimore, MD 21250. E-mail: iganson@umbc.edu
16 Anson11721188
REFERENCES
American National Election Studies. (2016). The ANES guide to public opinion and political behavior. Ann Arbor:
University of Michigan, Center for Political Studies. Retrieved from www.electionstudies.org.
Ansolabehere, S., Meredith, M., & Snowberg, E. (2014). Mecro-economic voting: Local information and microperceptions
of the macro-economy. Economics & Politics, 26(3), 380–410.
Anson, I. G. (2016). Just the facts? Partisan media and the political conditioning of economic perceptions. Political
Research Quarterly, 69(3), 444–456.
Ariely, D., & Norton, M. I. (2007). Psychology and experimental economics: A gap in abstraction. Current Directions
in Psychological Science, 16(6), 336–339.
Babad, E. (1997). Wishful thinking among voters: Motivational and cognitive influences. International Journal of Public
Opinion Research, 9(2), 105–125.
Balch, G. I. (1974). Multiple indicators in survey research: The concept “sense of political efficacy.” Political
Methodology, 1, 1–43.
Berinsky, A. J. (2017). Telling the truth about believing the lies? Evidence for the limited prevalence of expressive survey
responding. Journal of Politics, 80(1). dx.doi.org/10.1086/694258.
Berinsky, A. J., Huber, G. A., & Lenz, G. S. (2012). Evaluating online labor markets for experimental research:
Amazon.com’s Mechanical Turk. Political Analysis, 20(3), 351–368.
Berinsky, A. J., Margolis, M. F., & Sances, M. W. (2014). Separating the shirkers from the workers? Making sure
respondents pay attention on self-administered surveys. American Journal of Political Science, 58(3), 739–753.
Bisgaard, M. (2015). Bias will find a way: Economic perceptions, attributions of blame, and partisan-motivated reasoning
during crisis. Journal of Politics, 77(3), 849–860.
Bolsen, T., Druckman, J. N., & Cook, F. L. (2014). The influence of partisan motivated reasoning on public opinion.
Political Behavior, 36(2), 235–262.
Bullock, J. G. (2011). Elite influence on public opinion in an informed electorate. American Political Science Review,
105(03), 496–515.
Camerer, C. F., Hogarth, R. M., Budescu, D. V., & Eckel, C. (1999). The effects of financial incentives in experiments:
A review and capital-labor-production framework. In B. Fischhoff & C. F. Manski (Eds.), Elicitation of preferences
(pp. 7–48). Heidelberg, Germany: SpringerNetherlands.
Campbell, A., Gurin, G., & Miller, W. (1954). The voter decides. Evanston, IL: Row, Peterson.
Converse, P. E. (1964). The nature of belief systems in mass publics. In D. Apter (Ed.), Ideology and discontent
(pp. 206–261). New York, NY: Free Press of Glencoe.
Cowen, T. (2005). Self-deception as the root of political failure. Public Choice, 124(3), 437–451.
David, C. C. (2009). Learning political information from the news: A closer look at the role of motivation. Journal of
Communication, 59(2), 243–261.
Delli Carpini, M. X., & Keeter, S. (1997). What Americans know about politics and why it matters. New Haven, CT:
Yale University Press.
Dolan, K. (2011). Do women and men know different things? Measuring gender differences in political knowledge.
Journal of Politics, 73(1), 97–107.
Downs, A. (1957). An economic theory of democracy. New York, NY: Harper.
Druckman, J. N. (2004). Political preference formation: Competition, deliberation, and the (ir) relevance of framing
effects. American Political Science Review, 98(04), 671–686.
Druckman, J. N., & Leeper, T. J. (2012). Learning more from political communication experiments: Pretreatment and its
effects. American Journal of Political Science, 56(4), 875–896.
Dunning, D. (2011). The Dunning-Kruger effect: On being ignorant of one’s own ignorance. Advances in Experimental
Social Psychology, 44, 247–296.
Dunning, D., Heath, C., & Suls, J. M. (2004). Flawed self-assessment implications for health, education, and the
workplace. Psychological Science in the Public Interest, 5(3), 69–106.
Dunning, D., Johnson, K., Ehrlinger, J., & Kruger, J. (2003). Why people fail to recognize their own incompetence.
Current Directions in Psychological Science, 12(3), 83–87.
Ehrlinger, J., Gilovich, T., & Ross, L. (2005). Peering into the bias blind spot: People’s assessments of bias in themselves
and others. Personality and Social Psychology Bulletin, 31(5), 680–692.
Eveland Jr, W. P. (2002). News information processing as mediator of the relationship between motivations and political
knowledge. Journalism & Mass Communication Quarterly, 79(1), 26–40.
Partisanship, Political Knowledge, and the Dunning-Kruger Effect 1711731189
Eveland Jr., W. P., & Hively, M. H. (2009). Political discussion frequency, network size, and “heterogeneity” of discussion
as predictors of political knowledge and participation. Journal of Communication, 59(2), 205–224.
Everson, H. T., & Tobias, S. (1998). The ability to estimate knowledge and performance in college: A metacognitive
analysis. Instructional Science, 26(1), 65–79.
Flynn, D., Nyhan, B., & Reifler, J. (2017). The nature and origins of misperceptions: Understanding false and unsupported
beliefs about politics. Political Psychology, 38(S1), 127–150.
Franco, A., Malhotra, N., Simonovits, G., & Zigerell, L. J. (2017). Developing standards for post-hoc weighting in
population-based survey experiments. Journal of Experimental Political Science, 4(2), 161–172.
Galston, W. A. (2001). Political knowledge, political engagement, and civic education. Annual Review of Political
Science, 4(1), 217–234.
Goren, P., Federico, C. M., & Kittilson, M. C. (2009). Source cues, partisan identities, and political value expression.
American Journal of Political Science, 53(4), 805–820.
Grossmann, M., & Hopkins, D. A. (2016). Asymmetric politics: Ideological Republicans and group interest Democrats.
Oxford, United Kingdom: Oxford University Press.
Gvirsman, S. D. (2015). Testing our quasi-statistical sense: News use, political knowledge, and false projection. Political
Psychology, 36(6), 729–747.
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(1), 400–407.
Helzer, E. G., & Dunning, D. (2012). Why and when peer prediction is superior to self-prediction: The weight given to
future aspiration versus past achievement. Journal of Personality and Social Psychology, 103(1), 38–53.
Hodges, B., Regehr, G., & Martin, D. (2001). Difficulties in recognizing one’s own incompetence: Novice physicians
who are unskilled and unaware of it. Academic Medicine, 76(10), S87–S89.
Huddy, L. (2002). Context and meaning in social identity theory: A response to Oakes. Political Psychology, 23(4),
825–838.
Huff, C., & Tingley, D. (2015). “Who are these people?” Evaluating the demographic characteristics and political preferences
of MTurk survey respondents. Research & Politics, 2(3). https://doi.org/10.1177/2053168015604648.
Iyengar, S., & Westwood, S. J. (2015). Fear and loathing across party lines: New evidence on group polarization.
American Journal of Political Science, 59(3), 690–707.
Jerit, J., & Barabas, J. (2012). Partisan perceptual bias and the information environment. Journal of Politics, 74(3),
672–684.
Jerit, J., Barabas, J., & Bolsen, T. (2006). Citizens, knowledge, and the information environment. American Journal of
Political Science, 50(2), 266–282.
Klar, S. (2013). The influence of competing identity primes on political preferences. Journal of Politics, 75(4),
1108–1124.
Klar, S., & Krupnikov, Y. (2016). Independent politics: How American disdain for parties leads to political inaction.
Cambridge, United Kingdom: Cambridge University Press.
Krajc, M., & Ortmann, A. (2008). Are the unskilled really that unaware? An alternative explanation. Journal of
Economic Psychology, 29(5), 724–738.
Krosnick, J. A. (1991). Response strategies for coping with the cognitive demands of attitude measures in surveys.
Applied Cognitive Psychology, 5(3), 213–236.
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(6), 1121–1134.
Kuklinski, J. H., & Quirk, P. J. (2001). Conceptual foundations of citizen competence. Political Behavior, 23(3),
285–311.
Kuklinski, J. H., Quirk, P. J., Jerit, J., Schwieder, D., & Rich, R. F. (2000). Misinformation and the currency of democratic
citizenship. Journal of Politics, 62(3), 790–816.
Kunda, Z. (1990). The case for motivated reasoning. Psychological Bulletin, 108(3), 480–498.
La Due Lake, R., & Huckfeldt, R. (1998). Social capital, social networks, and political participation. Political Psychology,
19(3), 567–584.
Lassen, D. D., & Serritzlew, S. (2011). Jurisdiction size and local democracy: Evidence on internal political efficacy
from large-scale municipal reform. American Political Science Review, 105(02), 238–258.
Lavine, H. G., Johnston, C. D., & Steenbergen, M. R. (2012). The ambivalent partisan: How critical loyalty promotes
democracy. Oxford, United Kingdom: Oxford University Press.
Leeper, T. J., & Slothuus, R. (2014). Political parties, motivated reasoning, and public opinion formation. Political
Psychology, 35(S1), 129–156.
18 Anson11741190
Levay, K. E., Freese, J., & Druckman, J. N. (2016). The demographic and political composition of Mechanical Turk
samples. SAGE Open, 6(1). https://doi.org/10.1177/2158244016636433.
Luskin, R. C. (1990). Explaining political sophistication. Political Behavior, 12(4), 331–361.
McGlone, M. S., & Aronson, J. (2006). Stereotype threat, identity salience, and spatial reasoning. Journal of Applied
Developmental Psychology, 27(5), 486–493.
McGlone, M. S., Aronson, J., & Kobrynowicz, D. (2006). Stereotype threat and the gender gap in political knowledge.
Psychology of Women Quarterly, 30(4), 392–398.
Mele, A. R. (1997). Real self-deception. Behavioral and Brain Sciences, 20(1), 91–136.
Mondak, J. J., & Anderson, M. R. (2004). The knowledge gap: A reexamination of gender-based differences in political
knowledge. Journal of Politics, 66(2), 492–512.
Mondak, J. J., & Davis, B. C. (2001). Asked and answered: Knowledge levels when we will not take “don’t know” for
an answer. Political Behavior, 23(3), 199–224.
Montgomery, J. M., Nyhan, B., & Torres, M. (2016). How conditioning on post-treatment variables can ruin your experiment,
and what to do about it. Working paper, Washington University in St. Louis.
Neuman, W. R. (1986). The paradox of mass politics: Knowledge and opinion in the American electorate. Cambridge,
MA: Harvard University Press.
Nicholson, S. P. (2011). Dominating cues and the limits of elite influence. Journal of Politics, 73(04), 1165–1177.
Nicholson, S. P. (2012). Polarizing cues. American Journal of Political Science, 56(1), 52–66.
Nyhan, B., & Reifler, J. (2010). When corrections fail: The persistence of political misperceptions. Political Behavior,
32(2), 303–330.
Ortoleva, P., & Snowberg, E. (2015a). Overconfidence in political behavior. American Economic Review, 105(2),
504–535.
Ortoleva, P., & Snowberg, E. (2015b). Are conservatives overconfident? European Journal of Political Economy, 40,
333–344.
Parker-Stephen, E. (2013). Tides of disagreement: How reality facilitates (and inhibits) partisan public opinion. Journal
of Politics, 75(4), 1077–1088.
Pasek, J., Sood, G., & Krosnick, J. A. (2015). Misinformed about the affordable care act? Leveraging certainty to assess
the prevalence of misperceptions. Journal of Communication, 65(4), 660–673.
Prasad, M., Perrin, A. J., Bezila, K., Hoffman, S. G., Kindleberger, K., Manturuk, K., & Powers, A. S. (2009). “There
must be a reason”: Osama, Saddam, and inferred justification. Sociological Inquiry, 79(2), 142–162.
Prior, M. (2007). Post-broadcast democracy: How media choice increases inequality in political involvement and polarizes
elections. Cambridge, United Kingdom: Cambridge University Press.
Prior, M., Sood, G., & Khanna, K. (2015). You cannot be serious: The impact of accuracy incentives on partisan bias in
reports of economic perceptions. Quarterly Journal of Political Science, 10(4), 489–518.
Rahn, W. M. (1993). The role of partisan stereotypes in information processing about political candidates. American
Journal of Political Science, 37, 472–496.
Redlawsk, D. P. (2004). What voters do: Information search during election campaigns. Political Psychology, 25(4),
595–610.
Schl€osser, T., Dunning, D., Johnson, K. L., & Kruger, J. (2013). How unaware are the unskilled? Empirical tests of the
“signal extraction” counterexplanation for the Dunning–Kruger effect in self-evaluation of performance. Journal of
Economic Psychology, 39, 85–100.
Sheldon, O. J., Dunning, D., & Ames, D. R. (2014). Emotionally unskilled, unaware, and uninterested in learning more:
Reactions to feedback about deficits in emotional intelligence. Journal of Applied Psychology, 99(1), 125–137.
Soll, J. B., & Klayman, J. (2004). Overconfidence in interval estimates. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 30(2), 299–314.
Steele, C. M., & Aronson, J. (1995). Stereotype threat and the intellectual test performance of African Americans.
Journal of Personality and Social Psychology, 69(5), 797–811.
Svenson, O. (1981). Are we all less risky and more skillful than our fellow drivers? Acta Psychologica, 47(2), 143–148.
Theiss-Morse, E., & Hibbing, J. R. (2005). Citizenship and civic engagement. Annual Review of Political Science, 8,
227–249.
Weissberg, R. (2001). Democratic political competence: Clearing the underbrush and a controversial proposal. Political
Behavior, 23(3), 257–284.
Partisanship, Political Knowledge, and the Dunning-Kruger Effect 1911751191
SUPPORTING INFORMATION
Additional supporting information may be found in the online version of this article at the publisher’s
website:
1. Question Wording
2. Demographic Information and Descriptive Statistics
3. Tabular Results
4. Evaluation of Political Knowledge Question Battery
5. Power Analysis
20 Anson11761192