Reconsidering the role of cognitive ability in the acquisition of political knowledge

Paul T. Weith, Central European University <weith_paul@ceu-budapest.edu>

Levente Littvay, Central European University
Andre Krouwel, Free University of Amsterdam

Abstract: The most commonly cited model of knowledge acquisition posits that political knowledge
stems from a triad of intercorrelated factors, broadly defined as motivation, opportunity and cognitive
ability. This study argues that the direct effect of cognitive ability on political knowledge makes little
theoretical sense. We discuss the shortcomings of the common practice of using education as a proxy
for cognitive ability and introduce a direct measure instead. Since ability is rather a catalyzer for
knowledge accrued through other means, we amend the triadic model by introducing educational
attainment as a direct predictor and cognitive ability as a moderator of motivation, opportunity and
education. The model is tested using both the 2012 wave of the Dutch Vote Compass panel and the
2008 ANES. Our results show that the effect of ability on political knowledge is indirect and generally
weak, contrary to what previous studies have found. This finding calls for a reconsideration of the
mechanisms underlying information effects.

Acknowledgements: We express our gratitude to Gabor Toka, Zsolt Enyedi and Manuel
Bosancianu (Central European University), Zoltan Fazekas and Robert Klemmensen (University of
Southern Denmark), Paul DeBell (Ohio State University) and Gaurav Sood (Princeton University) for
their valuable feedback; the Political Behavior Research Group (PolBeRG) for the extensive
discussions on data collection and survey design and Simon Ceulemans (Kieskompas) for helping with
the translation of the questionnaire.
Introduction

Normative theories of representative democracy claim that a politically knowledgeable populace is

crucial for the functioning of democratic institutions, that democratic elections require informed

decisions from the part of the citizens (Dahl, 1979). Our most basic intuitions tend to agree with such a

claim; whoever knows the rules of the game, the players and their strategies, is more likely to play a

good game or coach a good team. Even non-academics have raised awareness of this problem; in fact,

the need for a cognitively engaged citizenry resonates strongly in many if not all social strata. In the

words of little Lisa Simpson from The Simpsons, “the price of freedom is eternal vigilance”.1 Whether

political knowledge is indeed important for democracy is an empirical question that has been often

addressed in the social scientific literature with various degrees of success.

Where does political knowledge come from? Who are the people who acquire political

knowledge and what drives them to pay attention to political news and occupy their minds with

political information? The classic model proposed by Luskin (1990) links the acquisition of political

knowledge to a triad of interconnected and predominantly individual-level factors (henceforth the

triadic model, or the triad): citizens have different cognitive abilities to aid their understanding of

politics, they are unequally motivated and do not have equal opportunities to get informed.

Considering plausible mechanisms through which ability can influence knowledge, we contend

that there is no political information contained in or directly facilitated by any accurately measured

cognitive faculties. Cognitive ability rests within the individual, while political information is external

to them. We propose to extend the model by theoretically arguing and empirically showing that

cognitive ability has a substantively trivial direct effect on the acquisition of knowledge. We question

the validity of common operationalizations of the triad, namely the use of “education” as a proxy for

cognitive ability (Hamil and Lodge, 1986; Carpini and Keeter, 1996) and argue that unlike cognitive

ability, education does indeed have a substantively significant direct effect on knowledge.

1 Often wrongly attributed to Thomas Jefferson, the source of the original quote could not be traced to date
 We propose a revised version of the triadic model that highlights the moderating effect of

cognitive ability on virtually all relationships between the elements of the triad and knowledge. We

contend that there are two mutually compatible mechanisms by which cognitive ability can play a

defining role in information effects (Bartels, 1996). It can filter and code information about things

political into political knowledge (in which case it is causally prior to political knowledge) or it can

govern the usage of memorized knowledge in subsequent political decisions and outcomes (in which

case it moderates the effects of political knowledge). We test the former and argue that if cognitive

ability fails to account for a significant portion of the variation in political knowledge, the latter is the

more likely mechanism.

To test our model, we designed a short cognitive ability test that we included in a panel survey

conducted in the Netherlands with Kieskompas in March 2012 on a sample of 3466 respondents. Our

results replicate well on the 2008 time series of the American National Election Studies despite the lack

of a direct measure of cognitive ability in the data. We use linear regression to test the triadic model;

show that the effect of education on political knowledge remains when controlling for cognitive ability,

and that there is very little additional predictive power that the effect of cognitive ability adds to our

models explaining political knowledge. We interpret the results as evidence in support for alternative

theories of knowledge acquisition and information effects .

Political Knowledge in the triadic model; the role of education

Political knowledge refers to the amount of factually correct information stored in the conscious

memory of citizens (Carpini and Keeter, 1996). The utility of political knowledge is manifold and very

well documented in the political science literature. More informed citizens are more likely to be aware

of the political alternatives facing them (Fournier, 2002, Bartels, 2005; Sturgis and Smith, 2010) and to

vote for parties and candidates who are politically closer to their personal political beliefs (Bartels,

1996; Fishkin and Luskin, 1999; Lau and Redlawsk, 2001; Lau and Redlawsk, 2008; Luskin, 1990).
Even more importantly from a normative democratic point of view, levels of political information were

often found to be significant predictors of the strength and direction of political attitudes, beliefs, vote

choice on the individual level and election results on the aggregate level (Althaus, 1998; Bartels, 1996;

Crampton, 2009; Fishkin and Luskin, 1999; Johnston, 1996).

The recent literature in the field contends that political knowledge springs from a triad of

nonorthogonal factors that jointly contribute to most of the variation in knowledge across individuals:

capability or ability – cognitive ability; opportunity and motivation (Carpini and Keeter, 1996; Luskin,

1990). The triad was proposed by Robert Luskin primarily as a theoretical model with the main

purpose of clarifying conceptual aspects related to political knowledge (Luskin, 1990); its empirical

use is often hindered by the scarcity of available data and the operational complexity of the concepts it

employs.

Opportunity refers to people's exposure to political news and newspapers. While it is

theoretically an environment-level variable (Luskin, 1990), it is often operationalized with variables

related to news media exposure and environment, and the position individuals hold in the social

structure: gender, income, education (Hutchings, 2003; Jerit, Barabas and Bolsen, 2006). The more

available political information is in one's environment, the more likely it is that they will become

knowledgeable about political matters.

Motivation refers to the desire and ambition of individuals to acquire political knowledge and

the diligence they exercise in their pursuit of political information (Carpini and Keeter, 1996). In line

with the extant literature, we operationalize motivation as “political interest” which is, arguably, the

main pathway through which the socio-political and economic environment can influence one's

propensity to get politically informed (Campbell, Converse, Miller and Stokes, 1960; Carpini and

Keeter, 1996; Lazarsfeld, Berelson and Gaudet, 1948; Miller and Rahn, 2002; Van Deth, 2000).

Motivation may overlap with opportunity and capability in its operationalization; while political

interest is the measure of motivation generally agreed upon (Carpini and Keeter, 1996; Kwak, 1999), it
is not uncommon to see it as a function of education and media consumption (Kwak, 1999).

There is no established operationalization of the triad, especially ability. While some use

education as a proxy for cognitive ability (Carpini and Keeter, 1996), others would look at SAT scores

(Boudreau, 2009) or rely on the survey interviewers' assessments of the respondents' cognitive abilities

(Bennett, 2002; Luskin, 1990). Before suggesting alternative measures, we propose a brief discussion

of the potential perils of using the widely available educational attainment variable as

operationalization for cognitive ability.

The association between educational attainment and various political behaviors, attitudes and

the democratic desideratum is well documented in the political science literature. Educated people are

more likely than their less educated peers to conform to most if not all democratic principles that the

normative literature put forward throughout the years: they have a higher propensity to engage in all

forms of political participation (Putnam, 2000), they express less prejudice towards minorities (Wagner

and Zick, 1995) and are generally more politically knowledgeable than their less educated counterparts

(Carpini and Keeter, 1996; Converse, 1964). In light of these well corroborated empirical findings,

Converse even concludes that all “positive” values, attitudes and behaviors are fostered by high levels

of educational attainment (Converse, 1972).

There is little doubt that higher education covaries with multiple democratic attitudes and

behaviors. There may be a causal connection between the former and the latter, but it can also be

argued that one's level of education is merely a reflection of their pre-adult socialization and

experiences, social class and cognitive abilities. Hence, educational attainment may be merely a

channel through which distal variables affect our outcomes of interest (Kam and Palmer, 2008). It

comes as no surprise, thus, that education stands out so often in political science studies as a strong

predictor of any given political attitude or behavior.

The use of education as a proxy for cognitive ability in the triad hinders our ability to

distinguish between the direct, indirect and spurious associations between the elements of the triad and
produces inaccurate estimates of the effects of ability, motivation and opportunity on the acquisition of

political knowledge. We therefore need to strip the model of the endogeneity caused by the ambiguous

status of education in order to see whether and how cognitive ability interacts with opportunity and

motivation in influencing the acquisition of knowledge. To overcome the problem, we propose a more

direct measurement of cognitive ability that improves on the existing operationalization in terms of

both construct and discriminant validity at the cost of a moderate loss of reliability.

Cognitive ability in the triad

Cognitive ability measures were previously found to correlate positively with indicators of job

performance, educational attainment and achievement, income (Jensen, 1998), prosocial attitudes,

openness to experience, pro-democratic values (Hodson and Busseri, 2012; Murphy and Hall, 2011)

and negatively with religiosity (Nyborg, 2009) and delinquency (Jensen, 1998). It appears that what

Converse concluded about education is true for cognitive ability: most “desirable” outcomes in the

social and political sphere seem to associate positively with various operationalizations of cognitive

ability. Given its relative invariance and temporal precedence to socio-political, attitudinal, behavioral

and economic indicators, we argue that cognitive ability is an exceptionally useful concept that is likely

to aid our understanding of directional theories of political knowledge and reduce the inherent

endogeneity of extant operationalizations of the triad.

The capability-opportunity-motivation theory finds cognitive ability to be one of the main

determinants of political knowledge or sophistication (Luskin, 1990): political matters are often too

complex and obscure for the population at large to grasp, understanding them often requires the

assistance of cognitive faculties that are unequally distributed across individuals. But based on this,

cognitive ability carries little or no substance directly linked to political knowledge. It aids citizens in

the comprehension, evaluation, organization, and application of information, but is otherwise

fundamentally removed from the political world outside the individual. There is no political
information contained in cognitive faculties, thus, ability can only assist citizens in their efforts to make

sense of political information but it cannot provide knowledge directly. Conceptually the described

mechanism is a text book example of a moderator.

On the other hand, no decision can be made without a cognitive processing of existing

information. Certainly, a collection of scraps of factual information cannot self-organize into political

attitudes or behaviors, such tasks rest within the cognition of the individual. If, for instance, one seeks

to minimize the ideological distance between their own position and that of the candidate they elect

(Downs, 1957), one would first need to know their own position and that of the competing parties. This

requires the dismissal of misinformation on the matter and the retention of information that is both

relevant and most plausibly accurate. This process involves judgment calls the quality of which will

likely affect the position of candidates as perceived by the citizen, but also the perceived position of the

citizen herself. Furthermore, it is not sufficient to know the positions of all candidates in addition to

one's own position. Finding the closest candidate is not a trivial matter, since different approaches to

measuring distance will yield strikingly different results (Rabinowitz and Macdonald, 1989). Both the

processes that translate factual information into applicable knowledge, and the screening of accurate vs.

inaccurate, relevant vs. irrelevant information capitalize on cognitive tasks (Krathwohl, 2002) that are

likely unequally distributed in the population.

The definition of political knowledge that we employ makes no explicit mention of its analytic

component. As we discussed previously, though, a common implicit assumption is that knowledge

involves not only the ability to retrieve facts from memory but also a deeper articulation of their

relevance and relatedness; direct information effects on any relevant political outcome would be

unlikely otherwise. Cognitive ability tends to enhance the understanding and usage of information

stored in people's memory (Plomin, Pedersen, Lichtenstein, and McClearn, 1994); in order for

knowledge to represent what the literature in the field routinely assumes of it, cognitive ability needs to

account for at least some of the variance in political knowledge. If, however, the assumption does not
hold, and political knowledge is reduced to the indiscriminate storage of political facts in people's

memory, the processing of information can act as a mediator of the effects of knowledge on political

behaviors and attitudes. To illustrate the two separate causal pathways considered, we propose the two

simplified diagrams in Figure 1 below.

Figure 1. Two possible causal pathways

As an educational goal, political knowledge can be structurally organized along the lines of

Bloom's taxonomy: knowledge of facts, concepts, procedures and metacognitive elements, in

ascending order of the complexity of the cognitive processes they employ (Krathwohl, 2002). The more

cognitively engaging tasks of “understanding” and “applying” (Krathwohl, 2002) are indispensable for

the alleged democratic functions of political knowledge, and they are implicit in all theories on

information effects (see, for instance, Lazarsfeld et al., 1948). As suggested in the figure above, there

are two potential roles that cognitive ability may play: it either regulates the acquisition of political

knowledge by moderating the effects of all other determinants of political knowledge, (diagram A) or it

aids in the process of translating already existing knowledge into political attitudes and behaviors

(diagram B). The two causal pathways are certainly not mutually exclusive; a combination of the two

may constitute the correct mechanism underlying information effects.

We build on the mechanism highlighted in diagram A, which conforms to the current

understanding of the role of cognitive ability in the acquisition of political knowledge. We argue that

political knowledge stems from a specific constellation of direct and conditional effects that fall within

the broad categories of a revised capability – opportunity – motivation triad. We acknowledge the

ambiguous status of education, separate it conceptually from cognitive ability and motivation and

include it in the model as a direct predictor on its own. We also recognize that it is not independent of

the other components of the triad either, it can be viewed as part of opportunity in that it fosters

socialization in relatively more politically sophisticated information environments, or as part of

motivation due to the increased exposure to democratic values of good citizenship. We thus keep it

separated from motivation an opportunity and treat it as a separate effect.

We operationalize opportunity using variables related to the individual's position in the social

structure: income, gender and age, and motivation using political interest and newspaper readership.

Older men who are economically affluent are more likely to acquire political knowledge due to them

having the time and the material prerequisites for nurturing their knowledge of things political. People

who spend more time reading newspapers and self-report to being interested in politics display an

elevated disposition to voluntarily expose themselves to political facts. In line with the considerations
discussed in this section, we propose the following theoretical model to be tested (Figure 2.):In line

with our theoretical model, cognitive ability only moderates other effects on the acquisition of political

knowledge; we do not include a direct effect in the theoretical specification (Figure 2.). We assert that

the issues of endogeneity in the extant operationalizations of the theoretical model have thus far made it

painstakingly hard to isolate the effects of individual variables on the acquisition of political

knowledge.

Figure 2. Moderation Model of Knowledge Acquisition

We contend that education is particularly problematic as a proxy for cognitive ability;

interaction effects between cognitive ability and motivation or opportunity cannot be estimated

accurately if ability is measured with education. Furthermore, operationalizing cognitive ability as

educational attainment would bias downwards the estimates for the effects of motivation and

opportunity and bias upwards the estimate for the effect of cognitive ability. A reduced reliance on

proxies enables us to distinguish between the variation in political knowledge that is causally linked to

cognitive ability and the variation that was previously wrongly attributed to it.

For illustrative purposes, we propose a brief discussion of the consequences of a mismeasure of

cognitive ability in modeling the acquisition of political knowledge with the triad. If cognitive ability,
operationalized as education, is found to be the strongest predictor of political knowledge, we can be

tempted to conclude that the variation in political literacy is mostly caused by individual differences

that cannot be easily changed throughout one's life. While people's position in the social, political and

economic community can be improved with affirmative action policies 2, cognitive ability is mostly

invariant within the individual due partially to its genetic heritability (see Plomin and Spinath, 2004;

Plomin et al., 1994), and its effects on any given political outcome can hardly be manipulated without a

radical reshuffling of the rules of functioning of the political community. Can we make politics easier

to grasp to enable citizens with lower levels of cognitive ability to participate meaningfully in its

workings? Rather unlikely. Furthermore, if cognitive ability is found to ameliorate one's position on a

socio-economic indicator that is itself a predictor of political knowledge; can we design a political and

economic system that is indiscriminate with regard to cognitive ability? Perhaps not without a radical

redefinition of merit in a social context; thus, if we overestimate the impact of cognitive ability on

political knowledge we may easily draw unnecessarily fatalistic conclusions.

In light of all these considerations, having a direct measure of cognitive ability appears to be

prerequisite. The most commonly used test for cognitive ability (or Spearman's g, which is an

equivalent concept) is currently the Raven's Progressive Matrices (Raven, 2000), with its variants

Standard Progressive Matrices (SPM), Colored Progressive Matrices (CPM) and Advanced Progressive

Matrices (APM). They measure two main components of cognitive ability: eductive and reproductive

ability (Raven, 2000); in other words the ability to make sense of complex non-verbal information and

to absorb and reproduce information. Cognitive ability tests are often contested for alleged problems of

validity and reliability, yet correlations between cognitive ability scores obtained from different

measurement tools are usually strong and always positive (Court and Raven, 1995). Our direct measure

of cognitive ability builds upon the logic of the Raven's Progressive Matrices but employs a limited

number of items, thus conforming to the requirements of large N surveys (see Appendix 1).
2 Of course, affirmative action policies will not change someone's gender or age, but they can reduce the structural
advantage of currently dominant socio-demographically defined groups
Data

We used an online panel of respondents - Dutch citizens of age 18 and over - who left an email address

and indicated consent for being occasionally re-contacted after completing a Dutch online Vote Advice

Application. Users of this website – over 3 million voters - were asked to opt-in during the Dutch local

elections in March 2010, while the other respondents opted into our panel during the parliamentary

elections in May/June 2010. Since these respondents consented to renewed contact and have used an

election website, we expect above average political interest and also higher propensities to respond to

survey participation. This may constitute a notable departure from representativity; however, type I and

type II errors often diverge, leading to several notable advantages of survey experimental online studies

over classic probability samples (Martinsson, Dahlberg, and Lundmark, 2013; Vavreck and Rivers,

2008).

Our data was collected by recontacting the panel between March 5 and March 15, 2012 as a

separate wave, using the online services of SurveyGizmo (http://www.surveygizmo.com). Out of the

initial sample of 8145 respondents, 3466 participated in this wave of our panel, 70 percent of them

filled out the questionnaire before we sent the first reminder on March 11. A second reminder was sent

on March 13, and the survey was closed on March 15 when the daily increase in the response rate fell

below 1 percent.

Education was measured using a 12- category ordinal variable with labels corresponding to the

educational qualifications specific to the Dutch education system. We recoded the variable following

the academic content of the initial categories: value 0 was assigned to educational categories lower than

the current level of compulsory education, 1 to respondents who did not enroll in any educational

institution after graduating from the compulsory cycle, 2 includes incomplete bachelor's studies and

college studies, 3 includes university graduates and respondents with incomplete graduate studies; 4

includes everyone with a master's or doctoral degree. We standardized it to a 0-1 scale and used the
variable as continuous scale in all subsequent analyses.

In line with our expectations, our sample is highly educated and more interested in politics than

a random sample of Dutch citizens. Fifty-four percent of our sample is in the highest two categories of

education, forty percent in the next two and only six percent in the lowest category. In terms of political

interest, six percent of our sample reported low and very low levels as opposed to fifty-eight percent

who reported high and very high levels, while the remaining thirty-six percent are in the middle

category of “moderately interested”. The average age in our sample is 49.7 and only 25.5 percent are

female.

The questionnaire has four main batteries of questions. The first battery is a political knowledge

quiz of seven questions with no variation of question formats across respondents. The second battery

consists of eight political knowledge questions with formats (open ended, multiple choice and

true/false) varying across respondents on a random basis. The assignment of question formats was done

using the random generator tool provided by SurveyGizmo. The third battery of items includes socio-

economic, demographic and attitudinal variables; additionally, we asked the respondents whether they

remember what they were doing on September 11, 2011 (with an open ended follow up asking them to

write a few words describing what they were doing) to tap into their memory skills. Finally, the seven

cognitive ability items were custom made, following the patterns found in the Raven's Advanced

Progressive Matrices (Raven, 1936). We employed an aggressive strategy of discouraging non-

responses using the “soft-required” option provided by SurveyGizmo: respondents were not allowed to

leave answer fields blank unless they ignored a warning message popping up on top of the page upon

first clicking the “next page” button. However, clicking the “next page” button the second time would

let the respondents navigate away from the respective page regardless of the number of fields left

blank. This helped us maximize the number of guesses, thus increasing the variance in the success of

guessing.

Due to the time constraints associated with long surveys, we used a reduced item cognitive
ability test. It is worth noting, however, that our study relies on the statistical power of large samples to

overcome the loss of reliability that comes with measuring a given construct with fewer items than the

established norm. All seven items are listed in Appendix 1.

Method and Results

We use the additive of correct answers to quiz-like questions for constructing our political knowledge

scales and our cognitive ability scale. Cognitive ability was measured by counting the correct responses

to the six most reliable of the seven items; the seventh item had no contribution to scale reliability and

was dropped due to considerations of parsimony. We obtained a normally distributed scale with a mean

of 2.0 and a standard deviation of 1.52. The Cronbach's Alpha of .51 shows that the internal

consistency of the scale is less than optimal (Cronbach, 1951), yet well above the value we would

intuitively expect from a scale that combines less than one fifth of the number of items that constitute

the norm in cognitive ability measurements (Carroll, 1993). Our scale correlates positively with

education (r=.23) , the two TIPI (Gosling, Rentfrow & Swann, 2003) “openness to experience” items
3

(r=.08), negatively with age (r=-.33) and self-reported religiosity (r=-.12), and is completely

independent from gender (p=.9 for independent sample t-test).

We run OLS regression models, with political knowledge as dependent and cognitive ability,

education, interest, news consumption, income, age and gender as predictors while controlling for

attention, memory and the number of open-ended questions used in the construction of the dependent

variable. Consistent with previous literature on political knowledge, we find a significant positive

relationship between education and political knowledge (see Table I). The significant association

remains even when controlling for cognitive ability, and including the joint effect of cognitive ability

and education does not render the coefficient for education insignificant. Furthermore, we find that the

3 The literature on cognitive ability usually reports stronger correlations with educational attainment (for a comparable
estimate, r=.32, see Johnson, Deary and Iacono, 2009). We attribute the weakness of our correlation to the reduced
number of items for measuring cognitive ability, the strict focus on non-verbal abilities and to convenience sampling.
direct marginal effect of cognitive ability on model fit is modest at best (a difference of less than half a

percentage point in explained residual variance). Education is the better predictor of political

knowledge (which may seem rather counterintuitive if it is indeed a mere proxy for the “cognitive

ability” element of the triad). Our results cast doubt on the existence of a substantively relevant direct

effect of cognitive ability on political knowledge. Furthermore, we interpret these results as

corroborating evidence for our claim that the causal relationship between education and the acquisition

of knowledge does not dwell in the association between education and cognitive ability, but rather in

the intrinsic value of education in the formation of democratic citizens or in the multitude of

environmental and individual factors that drive one's level of educational attainment (Kam and Palmer,

2008).

Table I. Standardized OLS estimates for the comparative effect of education and ability in nested

models4

DV=Knowledge Both omitted Cognitive omitted Education omitted Fully specified

Intercept 0.45* 0.42* 0.42* 0.40*
Cognitive ability 0.05* 0.04*

(0.01) (0.01)
Education 0.07* 0.06*

(0.00) (0.01)
Income 0.05* 0.03* 0.05* 0.03*

(0.00) (0.00) (0.00) (0.00)

Gender -0.02* -0.02* -0.02* -0.02*

(0.00) (0.00) (0.00) (0.00)

Age 0.05* 0.07* 0.07* 0.08*

(0.01) (0.01) (0.01) (0.01)

4 We also ran the same models with a Tobit link function (the knowledge variable is bounded on both sides of the
distribution) and with education as nominal, specified with a series of dummy variables. We find essentially the same
results: none of the models with cognitive ability performs better than the models they are nested in, the significance of
the chi-square test on log likelihood is essentially 1. Education is a valuable addition to each model, with chi-square tests
on model fit significant at p<0.001 level.
Interest 0.12* 0.11* 0.13* 0.11*

(0.01) (0.01) (0.01) (0.01)

Newspapers 0.06* 0.05* 0.06* 0.06*

(0.01) (0.01) (0.01) (0.01)

Memory 0.02* 0.02* 0.02* 0.01*

(0.00) (0.00) (0.00) (0.00)

Attention 0.06* 0.05* 0.05* 0.05*

(0.00) (0.00) (0.00) (0.00)

Open-ended count -0.01* -0.01* -0.01* -0.01*

(0.00) (0.00) (0.00) (0.00)

Adjusted R2 0.171 0.197 0.181 0.200
df 3152 3145 2372 2365
* = p<0.05, p<0.01, p≈0.00; SE in parentheses; all variables are on a 0-1 scale

We find modest decreases in the effect of educational attainment when cognitive ability is

controlled for, but the difference is substantively trivial and statistically insignificant. The slope of age

increases when controlling for ability, which is likely due to the decrease in cognitive ability at old age

and the positive relationship between age and political knowledge. The explanatory power of the

models changes by less than one percent when cognitive ability is included in the specification, leading

to the conclusion that much of the alleged direct effect of cognitive ability is already accounted for by

the variance of other variables. The strong relationship between educational attainment and political

knowledge is mostly due to other causal paths linking formal education to democratic citizenship, and

not to cognitive ability. However, it is worth noting that the drop in model fit upon the exclusion of

both education and cognitive ability from the specification is far greater than the cumulative decrease

in adjusted R squares induced by their separate omissions, thus making it likely that a more complex

function of the two is responsible for some of the residual variance in political knowledge.

We thus move on to fitting our revised version of the triadic model to the data for the different
levels of cognitive ability separately. This will allow us to see whether the effects of the variables

consideredopportunity and motivation are homogenous across ability-defined groups, and it will shed

some light on the interactive effects of cognitive ability.

We find mixed results. While the effects of interest (motivation) and age (opportunity) are

stronger for higher ability respondents, there is not enough evidence to claim that either motivation or

opportunity effects are systematically exacerbated by cognitive ability. The impacts of income, gender,

education and news consumption display no systematic variation across cognitive ability levels (see

Table II below); the acquisition of political knowledge appears causally heterogeneous with regard to

cognitive ability. At best, we can claim that opportunity and motivation translate into political

knowledge at a slightly better rate when cognitive ability is high.

Table II. OLS estimates by levels of cognitive ability

DV= Cognitive= Cognitive= Cognitive= Cognitive= Cognitive= Cognitive= Cognitive=

Knowledge 0 1 2 3 4 5 6
Intercept 0.54* 0.46* 0.40* 0.38* 0.42* 0.45* 0.26*
Education 0.04 0.07* 0.07* 0.08* 0.07* 0.03 0.12*

(0.02) (0.02) (0.01) (0.01) (0.01) (0.02) (0.04)

Income 0.01 0.06* 0.07* 0.01 0.00 0.01 0.00

(0.03) (0.02) (0.01) (0.01) (0.01) (0.01) (0.03)

Gender -0.05* -0.01 -0.03* -0.00 -0.02* -0.03* -0.04*

(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.02)

Age 0.03 0.02 0.05* 0.17* 0.08* 0.12* 0.24*

(0.05) (0.03) (0.02) (0.02) (0.02) (0.03) (0.07)

Interest 0.05 0.10* 0.10* 0.11* 0.12* 0.17* 0.09

(0.04) (0.03) (0.02) (0.02) (0.02) (0.03) (0.05)

Newspaper 0.07 0.03 0.09* 0.04* 0.05* 0.03 0.07

(0.03) (0.02) (0.02) (0.02) (0.01) (0.02) (0.04)

Memory 0.01 0.01 0.02 0.04* 0.02 0.02 0.02

(0.02) (0.01) (0.01) (0.01) (0.01) (0.02) (0.03)

Attention 0.02 0.04* 0.03* 0.07* 0.08* 0.06* 0.15*

(0.02) (0.01) (0.01) (0.01) (0.01) (0.02) (0.04)

Open-end -0.01* -0.00 -0.01* -0.01* -0.02* -0.01* -0.00

count (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)

Adj. R2 0.095 0.152 0.244 0.214 0.206 0.201 0.319
df 224 407 638 650 652 410 104
* = p<0.05, p<0.01, p≈0.00; SE in parentheses; all variables are on a 0-1 scale

The relationship between education and political knowledge is not constant across ability

groups, yet its variation is not monotonous. The slopes are roughly equal for the middle categories of

cognitive ability but significantly stronger for the top category and statistically insignificant in the

lowest and next-to-highest categories. If a better understanding of the joint effect of ability and

education is sought, further analysis is needed.

We proceed to fit an interactive OLS model in which we specify all relevant interaction terms

between cognitive ability and the other elements of the triadic model (results in Table V below). For

each of the levels of cognitive ability, we plot the relationship between education and predicted

knowledge based on the interactive model (see Figure 3 below). The mean of political knowledge

increases with cognitive ability for low-ability groups, but the returns from education are essentially

equal (the slope is constant), as shown on the left side of the plot. However, for higher ability groups a

different pattern appears to hold: higher ability brings higher returns from education. When education

is 0, it is the highest ability group that has the lowest level of political knowledge, followed by the

second highest ability group and the third highest. However, when education is 1 (maximum), the order
is perfectly reversed. This may be due to the dual nature of educational attainment in the theoretical

model: lower levels of education in the lower ability groups may not be chiefly driven by the

respondents' deliberate and voluntary self-exclusion from formal education, whereas low educational

attainment in the higher ability groups can be indicative of the respondents' deliberate withdrawal from

formal education. The effect of education, thus, capitalizes mostly on opportunity for low ability

groups, and on motivation for high ability groups. While higher ability groups with no education may

lack political knowledge due to their disengagement with things social and political, the lack of

education of low ability groups may constitute a social structural disadvantage. However, given the

interaction's failure to reach statistical significance, this interpretation is strictly tentative.

Figure 3. Relationship between education and political knowledge by levels of cognitive ability

Yet tThe most intriguing finding is the change in model fit: if we compare the explanatory

power of the triadic model for the lowest category of ability to that of the highest category, we find that

the latter is increased more than two-fold. While model fit varies between .20 and .24 for the

subsamples of ability higher than 1 and lower than 6, it is as low as .09 for the lowest ability group and
as high as .32 for the highest ability group. This result entails that we can predict what drives the

acquisition of political knowledge for people of high cognitive ability, but we can only guess what

makes a low-ability respondent become knowledgeable.

Replication on Representative Data: ANES 2008

The generalizability of our results may be hindered by the self-selection inherent to all online

surveys. While the Netherlands is one of the few countries where nearly all citizens have internet

access (94% according to the the Eurostat Report STAT/12/185, dated 18 December 2012), our sample

is evidently skewed towards higher education and higher levels of political interest. The reasons to

doubt the validity of our findings on such grounds are scarce; there is no published evidence

demonstrating that the moderation effect of cognitive ability on the relationships between opportunity,

motivation and political knowledge should be contingent upon variations at the level of political

interest, education, gender or other considerations that render our sample non-representative. However,

we acknowledge that the burden of proof rests on the promoter of any scientific theory. We thus

propose a partial replication of our findings on representative data, with the caveat that a full

replication could not be performed due to the lack of publicly available direct measurements of

cognitive ability in large N political surveys.

To this end, we use the post-electoral 2008 Time Series study of the ANES. The ANES

interviewers are instructed to rate the intelligence of the respondents at the end of the interview. While

admittedly imperfect, this measure of cognitive ability was previously shown to correlate with

validated measures as strongly as other measures correlate with each other (see Luskin, 1990). We find

that “apparent intelligence” correlates at .47 with education, and .38 with political knowledge, but it

shows no significant relationship with religiosity (church attendance), thus casting some doubt on the

impartiality of the interviewers' judgment. The plausibility of a halo effect (Thorndike, 1920) is also
supported by the differential size of the correlation between educational attainment and political

knowledge as rated by the interviewer (.41) versus directly measured (.38). The use of the interviewer's

rating as opposed to a direct measure of cognitive ability can conceivably inflate the observed direct

relationship with political knowledge to the extent that both ability and knowledge are on the same par

with regard to social desirability. Given our theoretical expectations described previously, this

association would increase our type II errors rather than induce bias, by artificially inflating the

estimates for the direct effect of ability on knowledge.

For consistency purposes, all the variables that we used were selected from the post-electoral

survey, with the exception of basic demographics that are not likely to be affected by the electoral

campaign. Considering that the advent of internet media may have had a role in shaping the

information environment of present day democratic citizens (Carpini and Keeter, 2003), the use of a

relatively recent ANES study for replication seems sensible.5 The ANES database contains all the

variables relevant for our study, with the exception of “attentiveness”, “memory” and “question

format”, which were not central to our theory. For more information on the construction of our

knowledge variable see Appendix II.

Table III. OLS estimates for the triadic model fit to ANES 20086

DV: Both Excluded Cognitive Ability Education Full Model

Knowledge Excluded Excluded

Intercept -0.15* -0.21* -0.27* -0.28*
Cognitive 0.08* 0.04*

Ability (0.01) (0.01)

Education 0.36* 0.30*

(0.03) (0.03)

5 Robert Luskin used the ANES1976 in his 1990 study. We chose to use the most recent ANES data due to the radical
change in the information environment of citizens over the last few decades.
6 Tobit coefficients follow the same pattern, we reported OLS for ease of comparison and consistency with other results
presented in the paper. All differences in fit statistics are significant with the Tobit models.
Opportunity
Age 0.19* 0.23* 0.19* 0.22*

(0.03) (0.03) (0.03) (0.03)

Income 0.40* 0.22* 0.32* 0.19*

(0.03) (0.03) (0.03) (0.03)

Gender 0.08* -0.09* 0.07* -0.09*

(0.01) (0.01) (0.01) (0.01)

Motivation
Newspaper 0.04 0.01 0.02 -0.02

(0.02) (0.02) (0.02) (0.02)

Interest 0.28* 0.25* 0.25* 0.19*

(0.02) (0.02) (0.02) (0.02)

Adjusted R 2
0.25 0.31 0.29 0.33
df 1269 1265 1265 1260
* = p<0.05, p<0.01, p≈0.00; SE in parentheses; all variables are on a 0-1 scale

We made no further changes to the models and we made sure the coding of the variables is in a

manner that would make the results in Table III. below comparable to the ones presented in Table I

above (0=minimum, 1=maximum).

Consistent with our previous results, we find that the effect of cognitive ability on the

acquisition of political knowledge is rather small; excluding the variable from the equation only

decreases the fit by a dismal .02 in adjusted R square, whereas the effect of excluding educational

attainment from the model is roughly double in size. Excluding both variables from the model bears a

drop in model fit that is greater than the cumulated impact of their separate omissions, suggesting that

their effects are likely interactive. We find a change in the slope of education upon the exclusion of

ability from the model that is proportionally equal to what we found on our Dutch sample. In contrast

with our previous findings, cognitive ability appears to capture a sizable portion of the effect of

political interest, the slope of which increases by more than one fourth when cognitive ability is not

included in the model. All in all, we find stronger effects for cognitive ability on the ANES sample than
on the Kieskompas sample, yet the conclusion that education cannot serve as a good proxy for

cognitive ability is inescapable, and the existence of interactive effects appears plausible. We thus

proceed to fitting our full model to the data split along the levels of cognitive ability. Table IV below

summarizes our results.

Table IV. OLS estimates for ANES 2008 by levels of cognitive ability

DV = Cognitive Ability Cognitive Ability =2 Cognitive Ability = Cognitive Ability =4

Knowledge ≤1 3
Intercept 0.13 -0.14* -0.17* -0.19*
Education -0.06 0.30* 0.34* 0.22*

(0.15) (0.05) (0.05) (0.08)

Opportunity
Age 0.00 0.24* 0.23* 0.22*

(0.12) (0.05) (0.06) (0.09)

Income 0.08 0.14* 0.25* 0.30*

(0.11) (0.05) (0.06) (0.10)

Gender 0.02 -0.10* -0.05* -0.12*

(0.05) (0.02) (0.02) (0.03)

Motivation
Newspaper 0.22* -0.01 -0.04 0.11

(0.11) (0.04) (0.04) (0.06)

Interest 0.03 0.19* 0.27* 0.29*

(0.07) (0.04) (0.04) (0.07)

Adjusted R 2
0.00 0.16 0.28 0.30
df 79 556 426 192
* = p<0.05, p<0.01, p≈0.00; SE in parentheses; all variables are on a 0-1 scale

We fail again at explaining the variation in political knowledge among the respondents with

minimal level of apparent cognitive ability; however, the explanatory power of our model increases

steadily for higher levels of cognitive ability. This pattern is even more striking here than on the Dutch
data. We also find that motivation plays a stronger role in the acquisition of knowledge in the higher

ability groups, but this is only true for political interest and not for newspaper readership. In fact, we

consistently find no effect of newspaper readership on political knowledge in the ANES sample, which

is potentially due to the falling rates of printed press readership and the rise of online media.

We find that the effect of education does not change monotonously over the categories of

cognitive ability, but the effect is significantly weaker in the highest ability group than in the previous

two (the lowest ability group is remarkably resistant to empirical scrutiny, therefore it provides little

insight for comparison). We interpret this as a corroboration of our finding that higher education may in

fact compensate for lower abilities. Finally, the greatest divergence between our findings on the two

data sets lies in the effect of income, which appears strongest in the higher ability group of the ANES

sample and provides some additional support for the elevated returns of opportunity in higher ability

groups.

The effect of education deserves closer attention. Once again, we use an interactive model

(Table V below) to generate predicted values for political knowledge across the levels of cognitive

ability and we plot the effect of education on political knowledge for each category of cognitive ability

(Figure 4). We find that the means of political knowledge are always higher for higher ability groups,

but the slopes of education vary in a fashion partially consistent with what we found on our Dutch data.

The highest two categories of cognitive ability produce the strongest effects of education, whereas the

next two categories have somewhat milder slopes. The effect of education is weakest in the lowest

ability group; however, the estimate is too noisy for consideration, as seen in the distribution of

bootstrapped coefficients shown on the right side of Figure 4.

 Figure 4. Effect of education on knowledge by levels of cognitive ability. ANES 2008.

Table V. Full interactive models fit to both data

DV = Knowledge Kieskompas data ANES 2008

Intercept 0.46* -0.007
Education 0.06* 0.30*
Cognitive Ability -0.07 -0.22
Gender -0.02* 0.05
Age 0.04 0.15
Income 0.04* -0.00
Interest 0.06* 0.07
Newspaper 0.06* -0.05
Attention 0.05*
Memory 0.02*
Open-ended count -0.01*
Cognitive Ability * Education 0.00 -0.00
Cognitive Ability * Income -0.02 0.31
Cognitive Ability * Age 0.11* 0.10
Cognitive Ability * Gender -0.00 0.04
Cognitive Ability * Newspaper -0.00 0.10
Cognitive Ability * Interest 0.12* 0.25*
2
Adjusted R 0.20 0.33
df 2359 1255
* = p<0.05, p<0.01, p≈0.00; SE in parentheses; all variables are on a 0-1 scale

Finally, we find that the only joint effect of cognitive ability that is significant on both databases

is the one with political interest. The effect of income is close to reaching statistical significance on the

ANES database but fails due to the sizable residuals in the lowest ability subsample. We find that some

of the effects of motivation and opportunity on political knowledge are stronger for higher ability

groups, but the variation in slopes is neither substantively nor statistically significant. The interactive

model does not fit the data better than the models with cognitive ability excluded (R2=0.200 vs

R2=0.197 on Kieskompas data, R2=0.33 vs R2=0.31 on ANES data)7, thus making it doubtful that

cognitive ability would play a substantively relevant role in the acquisition of political knowledge. The

estimates for the full interactive model are presented in Table V above.

Our results stand in sharp contrast with the assertion that cognitive ability exacerbates all other

effects on the acquisition of political knowledge (Luskin, 1990) and they cast doubt on the claim that

ability is the strongest direct predictor of knowledge. Our findings demonstrate that education is not a

good proxy for cognitive ability (Carpini and Keeter, 1996) and suggest that a more complex structure

of underlying factors drive the acquisition of political knowledge. Cognitive ability, however, does not

make a sizable contribution to explaining residual variance in political knowledge.

Discussion and Conclusions

The acquisition of political knowledge is said to be driven by a triad of interrelated factors,

broadly defined by Robert Luskin (1990) with the capability – opportunity – motivation model. We

revise the initial model by specifying the moderating effects of cognitive ability on all other effects and

changing the status of educational attainment from indicator of ability (Carpini and Keeter, 1996) to a

7 Equivalent Tobit models display the same patterns, yet the interactive model fits significantly better than the simpler one
when the difference in degrees of freedom is not accounted for. The difference becomes insignificant when AICs are
considered, as opposed to log likelihood
separate direct effect. We discussed at considerable lengths the likely implications of the use of

imperfect proxies in interactive models , such as educational attainment for cognitive ability, and

demonstrated that improving the validity of the triadic model and reducing its inherent endogeneity

problem is in fact feasible using large N surveys.

We designed a short test of cognitive ability that we included in a panel survey on a

convenience sample of Dutch citizens. We fitted the revised triadic model to our data and found that

few of the relevant predictors of political knowledge differ significantly across ability-defined groups.

The impact of motivation and opportunity on knowledge is not systematically stronger for higher

ability respondents. Our partial replication on ANES 2008 data corroborates our initial results and

provides additional nuances to be taken into consideration. Furthermore, we find that the political

literacy of the more cognitively capable is a lot more predictable than that of their less cognitively

capable peers.

Our findings have significant methodological, conceptual and normative implications. The use

of education as a proxy for cognitive ability leads to the hasty conclusion that ability is the main

driving force of political knowledge; and it hinders our ability to interpret interaction effects altogether.

Such overestimations and misinterpretations of the impact of ability distort our understanding of the

origins of political literacy and can contribute to a fatalistic view of the prospects of representative

democracy (assuming that the functioning of democracy is contingent upon the political sophistication

of its citizens); after all, cognitive ability is the one element of the triad that cannot be externally

manipulated by any feasible means. In this regard, our results are mostly encouraging; cognitive ability

has a rather weak contribution to the acquisition of political knowledge.

While it reaches statistical significance in our main effects model on both Vote Compass and

ANES 2008 data, cognitive ability accounts for a dismal 0.3% of the variance in political knowledge

even when all the relevant joint effects are specified. These findings call for a reconsideration of the

importance of the cognitive ability component of knowledge acquisition and raises questions about the
nature of political knowledge. If higher levels of cognitive ability do not contribute to higher levels of

knowledge,8 neither directly nor in concert with other motivational or socioeconomic traits, political

knowledge is reduced to the mere reproduction of often insufficiently understood facts, anecdotes and

technical information.

We view cognitive ability as an analytic skill that governs all mental operations involving

interconnected or linkable information, thus rendering it indispensable for any decision process. Its

effects on any outcome, however, can only be conditional upon exogenous variables germane to the

decision under study. As a text book example of a moderator, it can attenuate or exacerbate the effects

of motivation and opportunity on the acquisition of political knowledge; conversely, it can moderate

the effects of political knowledge on its common covariates (turnout, partisanship, opinionation, etc.).

We find little support for the former, thus making the latter the more plausible mechanism. Information

effects on political attitudes and behaviors likely occur by the joint contribution of political knowledge

and cognitive ability, rather than as direct effects of political knowledge. This, however, remains to be

tested.

The substantive implication of this assertion is rather optimistic. The generally weak

information effects documented in the literature (Althaus, 1998; Bartels, 1996) appear to be at odds

with the normative theories of representative democracy and with our basic intuitions. However, they

may underestimate real effects by assuming that the knowledge of respondents is readily usable as

input for complex decision processes, thus overlooking the theorized moderation of cognitive ability on

all information effects. If properly specified as interactions with cognitive ability, information effects

on aggregate vote choice (Bartels, 1996) or political attitudes (Althaus, 1998) may prove to be as strong

as theorized.

8 We do find weak effects of cognitive ability on political knowledge; the discussion here focuses on the counterfactual
situation in which the effect would be essentially nil. This is certainly an imperfect approximation of our actual finding;
nevertheless, we find the discussion meaningful for illustrative purposes.
References:

Althaus, S. L. (1998). Information Effects in Collective Preferences. The American Political Science

Review, 92(3), 545-558.

Bartels, L. M. (1996). Uninformed Votes: Information Effects in Presidential Elections. American

Journal of Political Science, 40(1), 194-230.

Bartels, L. M. (2005). Homer Gets a Tax Cut: Inequality and Public Policy in the American Mind.

Perspectives on Politics, 3(1), 15.

Bennett, S. E. (2002). Americans' Exposure to Political Talk Radio and Their Knowledge of Public

Affairs. Journal of Broadcasting & Electronic Media, 46(1), 74-86.

Campbell, A., Converse, P., Miller, W., & Stokes, D. (1960). The American Voter. New York: Wiley.

Converse, P. E. (1964). The Nature of Belief Systems in Mass Publics. In Apter, D. E. (Ed). Ideology

and Discontent. New York: Free Press.

Converse, P. E. (1972). Change in the American Electorate. In Campbell, A., and Converse, P. E. (Ed).

The Human Meaning of Social Change. New York: Russell Sage Foundation: 263-338.

Carroll, J.B. (1993) Human Cognitive Abilities: A Survey of Factor-analytic Studies, UK: Cambridge

University Press.

Crampton, E. (2009). Political Ignorance and Policy Preference. Working paper posted in the SSRN

eLibrary.

Cronbach L. J. (1951) Coefficient Alpha and the Internal Structure of Tests. Psychometrika, 16, 297-

334.

Delli Carpini, M. X., & Keeter, S. (2003). The Internet and an Informed Citizenry. In Anderson D. M.

& Cornfield M. (Eds). The Civic Web: Online Politics and Democratic Values, Lanham, MD:

Rowman & Littlefield, 129–153.

Delli Carpini, M. X., & Keeter, S. (1996). What Americans Know about Politics and Why It Matters.

New Haven, Conn.: Yale University Press.

Delli Carpini, M. X., & Keeter, S. (1993). Measuring Political Knowledge: Putting First Things First.

American Journal of Political Science, 37(4), 1179-1206.

Fishkin, J. S., and Luskin, R. C. (1999). Bringing Deliberation to the Democratic Dialogue. In.

McCombs M. E, & Reynolds A. (Eds). The Poll With a Human Face: The National Issues

Convention Experiment in Political Communication, Hillsdale: Lawrence Erlbaum Associates,

3-38.

Fournier, P (2002). The Uninformed Canadian Voter. In Everitt J., & O'Neill, B. (Eds). Citizen Politics:

Research and Theory in Canadian Political Behaviour, Oxford: Oxford University Press, 92-

109.

Gosling, S. D., Rentfrow, P. J., & Swann, W. B., Jr. (2003.) A Very Brief Measure of the Big-Five

Personality Domains. Journal of Research in Personality 37, 504-528.

Hodson, G., and Busseri, A. M. (2012). Bright Minds and Dark Attitudes: Lower Cognitive Ability

Predicts Greater Prejudice Through Right-Wing Ideology and Low Intergroup Contact.

Psychological Science 23 (2), 187-195.

Hutchings, V. L. (2003). Public Opinion and Democratic Accountability: How Citizens Learn About

Politics. Princeton, NJ: Princeton University Press.

Jensen, A. R. (1998). The g Factor: the Science of Mental Ability. Westport, CT: Greenwood

Publishing.

Jerit, J., Barabas, J. , and Bolsen, T. (2006). Citizens, Knowledge and the Information Environment.

American Journal of Political Science 50(2): 266-282.

Johnson, W., I. J. Deary, Iacono, W. G. (2009). Genetic and Environmental Transactions Underlying

Educational Attainment. Intelligence 37(5): 466-478.

Johnston, R. (1996). The Challenge of Direct Democracy: The 1992 Canadian Referendum. Montreal ;

Buffalo: Mcgill-Queen's University Press.

Krathwohl, D. R. (2002). A Revision of Bloom's Taxonomy: An Overview. Theory Into Practice, 41

 (4), 212-218.

Lau, R. R., & Redlawsk, D. P. (2001). Advantages and Disadvantages of Cognitive Heuristics in

Political Decision Making. American Journal of Political Science, 45(4), 951-971.

Lau, R. R., & Redlawsk, D. P. (2008). Older but Wiser? Effects of Age on Political Cognition. The

Journal of Politics, 70(1), 168-185.

Lazarsfeld, P. F., Berelson, B., & Gaudet, H. (1948). The People's Choice, 2nd ed. New York:

Columbia University Press.

Lupia, A. (1994). Shortcuts Versus Encyclopedias: Information and Voting Behavior in California

Insurance Reform Elections. The American Political Science Review, 88(1), 63-76.

Luskin, R. C. (1990). Explaining Political Sophistication. Political Behavior, 12 (4), 331-361.

Martinsson, J., Dahlberg, S., & Lundmark, S.O. (2013). Comparing Probability and Non-probability

Samples in a Country with Almost Full Internet Coverage. Paper presented at the Annual

Meeting of the Midwest Political Science Association, Chicago, IL, 10-14 April 2013.

Miller, Joanne M. & Wendy Rahn. (2002). Identity-Based Feelings, Beliefs and Actions: How Being

Influences Doing. Presented at the annual meeting of the International Society of Political

Psychology, Berlin, Germany.

Murphy, N. A., Hall, J. A. (2011). Intelligence and Interpersonal Sensitivity: A Meta-Analysis.

Intelligence, 39, 54–63.

Nyborg, H. (2009). The Intelligence – Religiosity Nexus: A Representative Study of White Adolescent

Americans. Intelligence 37(1), 81-93.

Plomin, R. and Spinath, F. M. (2004). Intelligence: Genetics, Genes, and Genomics. Journal of

Personality and Social Psychology. 86(1), 112-129.

Plomin, R.; Pedersen, N. L.; Lichtenstein, P.; McClearn, G. E. (1994). Variability and Stability in

Cognitive Abilities are Largely Genetic Later in Life. Behavior Genetics, 24 (3), 207–215.

Prior, Markus. (2005). News v Entertainment: How Increasing Media Choice Widens the Gap in
 Political Knowledge and Turnout. American Journal of Political Science, 49(3), 577-592.

Rabinowitz, George, and Stuart Elaine Macdonald. (1989). A Directional Theory of Voting. American

Political

Science Review, 83(1), 93–121.

Raven, J. C. (1936). Mental Tests Used in Genetic Studies: The Performance of Related Individuals on

Tests Mainly Educative and Mainly Reproductive. MSc Thesis, University of London.

Sturgis, P., & Smith, P. (2010). Fictitious Issues Revisited: Political Interest, Knowledge and the

Generation of Nonattitudes. Political Studies, 58(1), 66-84.

Thorndike, E.L. (1920). A constant Error in Psychological Ratings. Journal of Applied Psychology,

4(1), 25-29.

van Deth, J. W. (2000). Interesting but Irrelevant; Social Capital and the Saliency of Politics in Western

Europe. European Journal of Political Research, 37(2), 115-147.

Vavreck, L. & Rivers, D. (2008). The 2006 Cooperative Congressional Election Study. Journal of

Elections, Public Opinion & Parties, 18(4), 355 – 366.

Appendix I

Item 1. Correct answer :5

Item 2. Correct answer: 5
Item 3. Correct answer: 7
Item 4. Correct answer: 8
Item 5. Correct answer: 7
Item 6. Correct answer: 8

Appendix II
Variables in ANES 2008:
Cognitive Ability:
R's apparent intelligence: '1. Very high', '2. Fairly high', '3.Average', '4. Fairly low', '5. Very low' .
Reversed scale used
Political Knowledge:
1. Do you happen to know which party had the most members in the House of
Representatives in Washington BEFORE the election (this/last) month?
CORRECT ANSWER: Democratic Party
2. Do you happen to know which party had the most members in the U.S. Senate BEFORE
the election (this/last) month?
CORRECT ANSWER: Democratic Party
 3. As far as you know, what is the current unemployment rate in the United States, that is,
of the adults in the United States who want to work, what percent of them would you
guess are now unemployed and looking for a job?
CORRECT ANSWER: any response between 6 and 8
4. Which party is more conservative?
CORRECT ANSWER: Republicans
The scale we used is the count of correct answers to these four items (0-4)