ArticlePDF Available

Smart and Poor, or Rich and Dull? A U.S. County-Level Analysis of the Relationship between IQ and Presidential- Election Voting Behavior


Abstract and Figures

A small research stream exists which focuses on relationships between political ideology (inferred from voting behavior) and the intelligence of geopolitical subdivisions such as the 50 U.S. states. With U.S. state-level data, IQ scores positively predict votes cast for Democrats, but only when controlling for state racial composition. Here, however, we explore the relationship between IQ and voting behavior at the level of U.S. counties (approx. n = 3,100). We find that county-level IQ weakly predicts more votes cast for Republicans (r's .07 to .13) and less votes cast for Democrats (r's-.10 to-.14). These small relationships are also found in multiple regression analyses, even when demographic data, social status, and population density appear as covariates (3-4% points more votes cast for Republicans per standard deviation of IQ). The effect of general social status was opposite that of intelligence (6-9% points less votes cast for Republicans per standard deviation of social status), which is surprising considering the very strong positive correlation (r = .77) between IQ and social status. Additionally, racial homogeneity by itself predicts voting Republican; however, when other variables are present in the regression model, homogeneity predicts voting Democrat. Results indicate that aggregate-level relationships between intelligence and voting outcomes are more complex than previously thought.
Content may be subject to copyright.
MANKIND QUARTERLY 2019 60.2 243-255
Smart and Poor, or Rich and Dull? A U.S. County-Level
Analysis of the Relationship between IQ and Presidential-
Election Voting Behavior
Emil O. W. Kirkegaard*
Ulster Institute for Social Research, London, UK
Bryan J. Pesta
Cleveland State University, USA
*Corresponding author email:
A small research stream exists which focuses on relationships
between political ideology (inferred from voting behavior) and the
intelligence of geopolitical sub-divisions such as the 50 U.S. states.
With U.S. state-level data, IQ scores positively predict votes cast for
Democrats, but only when controlling for state racial composition.
Here, however, we explore the relationship between IQ and voting
behavior at the level of U.S. counties (approx. n = 3,100). We find
that county-level IQ weakly predicts more votes cast for Republicans
(r’s .07 to .13) and less votes cast for Democrats (r’s -.10 to -.14).
These small relationships are also found in multiple regression
analyses, even when demographic data, social status, and
population density appear as covariates (3 - 4% points more votes
cast for Republicans per standard deviation of IQ). The effect of
general social status was opposite that of intelligence (6 - 9% points
less votes cast for Republicans per standard deviation of social
status), which is surprising considering the very strong positive
correlation (r = .77) between IQ and social status. Additionally, racial
homogeneity by itself predicts voting Republican; however, when
other variables are present in the regression model, homogeneity
predicts voting Democrat. Results indicate that aggregate-level
relationships between intelligence and voting outcomes are more
complex than previously thought.
Key Words: USA, Voting, Counties, Democrat, Republican,
The relationship between cognitive ability and political opinion has received
increased research attention in recent years. Perhaps unsurprisingly, most of this
work has focused on voting behavior within the United States, examining mean
cognitive ability levels of Republican versus Democrat voters, and self-identified
liberals versus conservatives (Caplan & Miller, 2010; Carl, 2014a,b, 2015a,b;
Ganzach, 2016, 2017, 2018; Ganzach, Hanoch & Choma, 2019; Kemmelmeier,
2008; Kirkegaard, Bjerrekær & Carl, 2017; Lewis & Bates, 2018; Ludeke &
Rasmussen, 2018; Meisenberg, 2015; Onraet et al., 2015; Oskarsson et al.,
2015). In this literature, usually no large gaps exist between the mean IQs of
supporters of different parties. When smaller gaps are found, however, the center
parties are somewhat favored by those with higher IQs. In the United States, only
very small differences exist between the mean IQs of Republican versus
Democrat voters. The direction of this effect depends both on which covariates
(race, education, income, etc.) appear in the models, and on which time period
the data are from.
When studies use one-dimensional scales with self-placement, sometimes
small IQ advantages (about 2-3 IQ points) are seen for “liberals” (in the left-wing,
U.S. sense) over conservatives. Moderates, independents, and centrists usually
score lower than both liberals and conservatives by about four to five IQ points.
This finding could be related to the fact that moderates are generally not strongly
interested in politics, and IQ correlates with political interest and participation (r =
.20 to .30; Deary, Batty & Gale, 2008; Kirkegaard & Bjerrekær, 2016). In the case
of voters and nonvoters, the gap is about 5-10 IQ points (Carl, 2019).
If one instead opts for more fine-grained measures of political ideology, then
more complex patterns emerge. One common approach splits political opinions
into two scales that are theoretically independent, though not necessarily
empirically uncorrelated: economic liberalism / freedom, and social liberalism /
freedom (Nolan chart). The former is concerned with the role of the state in the
economy (e.g., how much tax, which tax forms, government interventions,
regulations, etc.). The latter is concerned with various social and cultural
freedoms (e.g., gay marriage, drug legalization/regulation, prostitution, etc.). The
two scales, though, are usually found to be somewhat correlated. For example,
Carl (2015a) reported a positive correlation of r = .36 between the two scales in a
large U.S. sample, but Kirkegaard, Bjerrekær and Carl (2017) found only a
correlation of .07 in a smaller sample of approximately 250 Danes. This is
somewhat unexpected because at the party level, the two dimensions are often
negatively correlated in Western countries, such that parties favoring less
government influence in the economy also favor many limits on social behavior,
and vice versa. However, this finding seems to depend on political context
because the pattern is apparently reversed in China and many ex-communist,
Eastern European countries (de Regt, Mortelmans & Smits, 2011; Malka et al.,
2014, 2019; Pan & Xu, 2017).
Aside from individual-level analyses, some studies look at voting behaviors
across geo-political divisions of nations. Carl (2018) analyzed data from the
United Kingdom at two levels: 11 regions and 372 local authorities. He found that
economic liberalism was positively related to the mean IQs of the units (r = .70
and .33 for regions and local authorities, respectively). However, Carl (2018) also
found that IQ’s relationship to social liberalism was only weakly positive (r = .21
and .12, for regions and local authorities, respectively). Thus, it appears that the
strength but not the direction of the association strongly depends on the level of
For the United States, both Pesta (2017), and Pesta and McDaniel (2014)
explored state-level relationships between “well-being, IQ, and voting behaviors
for all presidential elections held this century. They found in all elections
nonsignificant, bivariate correlations between state IQ and votes cast for either
Republicans or Democrats. However, when race variables (i.e., percent White,
Black, or Hispanic) also appeared in the regression equations, state IQ
moderately-to-strongly predicted votes cast for Democrats.
For the USA, data also exist for voting behaviors at the county level (N =
approximately 3,100 counties). However, no published studies have looked at
U.S. county-level relationships between IQ and voter behavior. Hence, the
purpose of the present study is to close the gap in this literature by conducting
such a study.
Data and methods
We coded data from the Stanford Education Data Archive (SEDA; Reardon
et al., 2018), available at SEDA
comprises a massive amount of cognitive testing data from many sources,
including NAEP and state tests that have been normed to the same scale. Data
are available at the county level for the years 2009-2015. The scores correlated
.77 on average between these years. We therefore averaged the scores across
years and subjects tested (language and math) to produce a single best estimate
for each county. Note that our measure correlated .86 with IQ estimates from a
previous study (Kirkegaard, 2016), which was based on partially overlapping
data. The IQ variable was standardized to a 0/1 scale (mean/SD). Figure 1 shows
a map of the distribution of average intelligence across counties in the United
Figure 1. Map of average county intelligence in the United States (Hawaii and
Alaska not shown but included in all analyses), n = 3,085. The holes represent
missing data.
Political outcomes
The New York Times (2016) published voting results by U.S. county for the
2008, 2012, and 2016 presidential elections. They reported the percent of votes
cast in each county for the Democrat, Republican, Libertarian and Green Party
candidates. The Green Party, however, had missing data for about 500 counties
where the Green Party candidate did not run. Figure 2 shows the distribution of
percent Democrat votes by U.S. county.
We used the extensive set of covariates compiled by Kirkegaard (2016).
These were compiled by merging various public U.S. surveys, with the majority
of the data coming from the American Community Survey (ACS, We also used Kirkegaard’s
(2016) scores for the general socioeconomic status (SES) factor (S factor). This
variable is a composite formed by factor analyzing 27 diverse indicators of well-
being (e.g. teen birth rate, proportion with at least a bachelor’s degree, smoking
rate, income Gini coefficient). The variable was standardized to a 0/1 scale
(mean/SD). Finally, we computed a racial homogeneity score for each county
based on the probability that two randomly chosen persons will be from the same
race or ethnic group (known as the Simpson or Herfindahl index).
Figure 2. Map of the percent Democrat vote in the 2016 election by U.S. county
(Hawaii and Alaska not shown but included in all analyses), n = 3,111. The holes
represent missing data.
Spatial data
As an additional control variable, we used publicly-available spatial data
(shapefile) from
u-s-current-county-and-equivalent-national-shapefile. This file contains the
borders of all US counties. We computed population density using the spatial data
and the population counts in the covariate datafile. The density variable was
extremely skewed, so the log10 value was taken and normality was achieved.
The study was analyzed in R (3.6.1). All code and data are available for reuse
in the supplementary materials file. The R notebook is available at
248 We used the square root of population size
as weights in line with previous research (Fuerst & Kirkegaard, 2016). We also
outputted the unweighted versions, which are available in the supplementary
materials file. Results here were very similar to those found when using weights.
Table 1 shows bivariate correlations between all study variables. It can be
seen that there are strong correlations between many of them. Of particular
interest are moderate negative correlations between intelligence and percent
Democrat voting, and the reverse for percent Republican voting. Third party data
were not available for 2008 and 2012 (and these were normalized too for
Democrat/Republican vote share), but in 2016 the correlations between vote
share for these and IQ were positive as well.
Table 1. Correlation matrix of main variables for ~3100 US counties. Weighted
by square root of population size.
S 0.77 1.00
White 0.53 0.37 1.00
Hispanic -0.25 -0.08 -0.67 -0.09 1.00
Amerindian -0.23 -0.17 -0.18 -0.10 -0.03 -0.06 1.00
Other -0.02 0.12 -0.20 -0.08 0.02 0.52 0.18 1.00
Homogeneity 0.37 0.20 0.86 -0.56 -0.49 -0.49 -0.08 -0.28 1.00
Dem16 frac -0.11 0.07 -0.65 0.45 0.33 0.53 -0.01 0.19 -0.58
Dem12 frac
Dem08 frac
Rep16 frac 0.07 -0.14 0.61 -0.38 -0.33 -0.53 -0.01 -0.22 0.55
Rep12 frac 0.13 -0.04 0.52 -0.37 -0.23 -0.45 -0.03 -0.19 0.42
Rep08 frac 0.09 -0.08 0.45 -0.31 -0.20 -0.43 -0.03 -0.18 0.34
Green16 frac 0.09 0.35 -0.07 -0.25 0.13 0.38 0.16 0.44 -0.11
Libert16 frac 0.29 0.46 0.23 -0.48 0.05 0.03 0.18 0.19 0.18
Pop. density 0.18 0.23 -0.34 0.28 0.16 0.49 -0.25 0.10 -0.45
Dem16 frac
Dem12 frac
Dem08 frac 0.93 0.98 1.00
Rep16 frac
Rep12 frac
Rep08 frac
Green16 frac 0.45 0.47 0.49 -0.51 -0.50 -0.51 1.00
Libert16 frac -0.01 0.00 0.06 -0.08 -0.03 -0.07 0.47 1.00
Pop. density
Results from bivariate analysis are of course possibly confounded with other
factors, and we see strong correlations between intelligence and other variables
in the table, especially general social status (S), and many of the demographic
variables. Thus, multivariate analysis is warranted to clarify possibly causal
relationships. Table 2 shows the regression results.
The results show multiple things of interest. First, despite the very strong
correlation between intelligence and general social status (r = .77, cf. Table 1), we
see that they have opposite signs in the regression models. Furthermore, IQ is a
positive predictor for Republicans, but negative for others, including third parties,
though the effect size for them is quite small. The effect size for Republicans
against Democrats is sizable at 2.5 to 4.3% points gain / loss per standard
deviation increase in county IQ. Moreover, social status is a quite potent predictor
of votes cast for Democrats, with an effect size of about 6 to 7% points, and -6 to
-9% points for Republicans. The effects of demographic variables are also
substantial. The percent of Blacks within a county predicts more Democrat votes
which is unsurprising because Blacks generally vote about 90% Democrat
( The effect size is
approaching ±1% point, meaning that a 1% point increase in the Black population
results in about a 1% point increase in the percent of votes cast for Democrats.
Table 2. Regression model results for U.S. county-level presidential elections. N
= 3,058-3,059. Intelligence and S are standardized (0/1 mean/SD), demographic
memberships are proportions (i.e. 0-1 range). S = general social status based on
up to 27 indicators. Weighted by the square root of population size. Values in
parentheses are standard errors. * = p < .01.
S 0.072
Black 0.978
Hisp. 0.579
Other 0.603
Homog. 0.365
Dens. 0.085
R2 adj. 0.70 0.53 0.48 0.67 0.51 0.46 0.34 0.39
Hisp., Hispanic; Amer., Amerindian; Homog., homogeneity; Dens., density.
For Hispanics, Amerindians (Native Americans), and Others, it works about
the same way despite the fact that these groups have sizable proportions who
vote Republican. Curiously, the percent of Asians in a county has an effect size
above the ±1% point, which may seem impossible. This suggests that Asians
convert nearby voters to Democrats away from Republicans. We currently have
no non-speculative explanation for this effect.
Next, homogeneity, defined as the chance that two randomly picked persons
are from the same racial group, strongly predicts votes cast for Democrats. Thus,
it appears that (assuming causality), holding the other covariates constant,
increasing the diversity share of a county leads to fewer Democrat votes and more
Republican votes. This is surprising because the correlation between
homogeneity and Democrat votes is strongly negative at r = -.37 to -.58 (cf. Table
1). There is also a very strong suppression effect at play here, which has been
found previously for these kind of data (Pesta, 2017; Pesta & McDaniel, 2014).
Suppression probably results from opposite-signed direct and indirect paths,
where some of the indirect effect is mediated by another predictor in the model.
Finally, the results for third parties were quite weak in comparison and have only
limited interest, so are not discussed further.
Intelligence is indeed related to voting outcomes at the U.S. county level.
Specifically, we found that higher intelligence predicts support for non-Democrat
parties in bivariate analysis. When covariates are introduced, this relationship is
still found but only for the Republicans versus third parties. These results are
interesting considering that previous studies, using state-level data, found that
higher IQ predicted more votes for Democrats (Pesta, 2017; Pesta & McDaniel,
2014), whereas we find the opposite here.
Furthermore, individual level results generally find little relationship between
intelligence and voting for either Democrats or Republicans, especially when
covariates are present (Ganzach, 2016; Meisenberg, 2015). Conversely, here,
the effect size for votes cast for Republicans was not trivial. Specifically, a one
standard deviation increase in intelligence was associated with 3-4% points more
support for the Republican candidate. On the other hand, general social status
(S) predicted support for the Democrat candidate at an even greater magnitude,
6-7% points. The opposite signs of the two predictors is surprising considering
that they are very strongly positively correlated (r = .77). Also of interest was that
racial homogeneity predicted Republican support in bivariate analysis (r = .34 to
.55), but predicted Democrat support in multiple regression (betas -0.38 to -0.48).
The reversal of direction of predictors and the contrast to prior studies are curious
and perhaps alarming since they seem to suggest that suppression effects are
particularly strong in this area of research, and that aggregation paradoxes
(Simpson’s paradox) where results at one level of analysis are inconsistent in
direction with those at a different level are perhaps common.
An interesting trend seen in the model comparisons is that the model fits
seemingly increased over time, from about .46-.48 in 2008 to .67-.70 in 2016.
This indicates that voting behavior has become more predictable from county-
level data, which suggests that political polarization of demographic groups is
increasing. There is in fact good evidence for such increased polarization, which
mainly results from the left moving further left, while the right remaining
approximately the same (Goldberg, 2019a, 2019b; Kaufman, 2019). However,
caution is advised because we only have three elections in the dataset, and
because some of the covariate data are closer to 2016 than to 2008 in age. While
the social variables are available for many decades, the currently available
intelligence data only span 2009 to 2015, so it is not now possible to go back
further in time to cover additional elections at the U.S. county level.
Limitations and Conclusion
First, we did not have data disaggregated by race for voting behavior, so it
was not possible to see how effects might differ specifically by the race of the
voter. Second, the study was cross-sectional, so causality can be backwards for
the variables we considered. Longitudinal data are available for some of the
variables, so they could be used in a future study to better answer causality
questions. Of course, studies of this kind suffer from potential omitted variable
bias, and this may have affected our findings. Finally, we must be careful not to
commit the ecological fallacy here. The effects we observed using aggregate-
level data may not also be found at the level of the individual voter. Aggregate-
level results often differ from individual-level results (see, e.g., Robinson, 1950)
Nonetheless, we found that at the U.S. county level, intelligence predicts
votes cast for Republicans, and this effect persists even when various controls
are added in regression models. The pattern we report here is exactly opposite
that reported by both Pesta (2017), and Pesta and McDaniel (2014), who found
that intelligence predicted votes cast for Democrats (after controlling for race),
albeit using U.S. state-level data. These authors also reported very strong
suppression effects, which we found here as well. Why results differ for U.S.
counties versus states is a puzzle that future research should seek to solve.
Supplementary materials
See, and for code output, see
Caplan, B. & Miller, S.C. (2010). Intelligence makes people think like economists:
Evidence from the General Social Survey. Intelligence 38: 636-647.
Carl, N. (2014a). Cognitive ability and party identity in the United States. Intelligence 47:
Carl, N. (2014b). Verbal intelligence is correlated with socially and economically liberal
beliefs. Intelligence 44: 142-148.
Carl, N. (2015a). Cognitive ability and political beliefs in the United States. Personality
and Individual Differences 83: 245-248.
Carl, N. (2015b). Does intelligence have a U-shaped relationship with leftism? Intelligence
49: 159-170.
Carl, N. (2018). IQ and political attitudes across British regions and local authorities.
Intelligence 69: 169-175.
Carl, N. (2019). Comparing EU knowledge among leave and remain voters: A replication
study. SocArXiv: 15 Sept. 2019.
de Regt, S., Mortelmans, D. & Smits, T. (2011). Left-wing authoritarianism is not a myth,
but a worrisome reality. Evidence from 13 Eastern European countries. Communist and
Post-Communist Studies 44(4): 299-308.
Deary, I.J., Batty, G.D. & Gale, C.R. (2008). Childhood intelligence predicts voter turnout,
voting preferences, and political involvement in adulthood: The 1970 British Cohort Study.
Intelligence 36: 548-555.
Fuerst, J. & Kirkegaard, E.O.W. (2016). Admixture in the Americas: Regional and national
differences. Mankind Quarterly 56: 255-373.
Ganzach, Y. (2016). Cognitive ability and party identity: No important differences between
Democrats and Republicans. Intelligence 58: 18-21.
Ganzach, Y. (2017). Cognitive ability and party affiliation: The role of the formative years
of political socialization. Intelligence 61: 56-62.
Ganzach, Y. (2018). Intelligence and the rationality of political preferences. Intelligence
69: 59-70.
Ganzach, Y., Hanoch, Y. & Choma, B.L. (2019). Attitudes toward presidential candidates
in the 2012 and 2016 American elections: Cognitive ability and support for Trump. Social
Psychological and Personality Science 10: 924-934.
Goldberg, Z. (2019a, June 6). America’s white saviors. Tablet Magazine. Retrieved from
Goldberg, Z. (2019b, July 25). Democrats are turning immigration into a moral ultimatum.
Tablet Magazine. Retrieved from
Kaufman, E. (2019). WHITESHIFT: Populism, Immigration and the Future of White
Majorities. Penguin Books.
Kemmelmeier, M. (2008). Is there a relationship between political orientation and
cognitive ability? A test of three hypotheses in two studies. Personality and Individual
Differences 45: 767-772.
Kirkegaard, E.O.W. (2016). Inequality across US counties: An S factor analysis. Open
Quantitative Sociology & Political Science. Retrieved from
Kirkegaard, E.O.W. & Bjerrekær, J.D. (2016). The OK Cupid dataset: A very large public
dataset of dating site users. Open Differential Psychology. Retrieved from
Kirkegaard, E.O.W., Bjerrekær, J.D. & Carl, N. (2017). Cognitive ability and political
preferences in Denmark. Open Quantitative Sociology & Political Science 1(1). Retrieved
Lewis, G.J. & Bates, T.C. (2018). Higher levels of childhood intelligence predict increased
support for economic conservatism in adulthood. Intelligence 70: 36-41.
Ludeke, S.G. & Rasmussen, S.H.R. (2018). Different political systems suppress or
facilitate the impact of intelligence on how you vote: A comparison of the U.S. and
Denmark. Intelligence 70: 1-6.
Malka, A., Soto, C.J., Inzlicht, M. & Lelkes, Y. (2014). Do needs for security and certainty
predict cultural and economic conservatism? A cross-national analysis. Journal of
Personality and Social Psychology 106: 1031-1051.
Malka, A., Lelkes, Y. & Soto, C.J. (2019). Are cultural and economic conservatism
positively correlated? A large-scale cross-national test. British Journal of Political Science
49: 1045-1069.
Meisenberg, G. (2015). Verbal ability as a predictor of political preferences in the United
States, 19742012. Intelligence 50: 135-143.
Onraet, E., van Hiel, A., Dhont, K., Hodson, G., Schittekatte, M. & De Pauw, S. (2015).
The association of cognitive ability with right-wing ideological attitudes and prejudice: A
meta-analytic review. European Journal of Personality 29: 599-621.
Oskarsson, S., Cesarini, D., Dawes, C.T., Fowler, J.H., Johannesson, M., Magnusson,
P.K.E. & Teorell, J. (2015). Linking genes and political orientations: Testing the cognitive
ability as mediator hypothesis. Political Psychology 36: 649-665.
Pan, J. & Xu, Y. (2017). China’s ideological spectrum. Journal of Politics 80(1): 254-273.
Pesta, B.J. (2017). A U.S. state-level analysis of the presidential election in 2016: IQ,
race, and well-being emerge as mutually-suppressed predictors. Open Differential
Psychology 1(1). Retrieved from
Pesta, B.J. & McDaniel, M.A. (2014). State IQ, well-being and racial composition as
predictors of U.S. presidential election outcomes. Intelligence 42: 107-114.
Reardon, S.F., Ho, A.D., Shear, B.R., Fahle, E.M., Kalogrides, D. & DiSalvo, R. (2018).
Stanford Education Data Archive (Version 2.1). Retrieved from
Robinson, W.S. (1950). Ecological correlations and the behavior of individuals. American
Sociological Review 15: 351-374.
... Finally, some research has found that results may be discordant across levels of analysis [40] . For example, U.S. state-level results may not match U.S. county-level results [41] . For this reason, in their review of regional dif-ferences in intelligence, Lynn et al. [11] urged authors to examine data at multiple levels in order to ensure robustness. ...
Conference Paper
Full-text available
Using a sample of ~3,100 U.S. counties, we tested geoclimatic explanations for why cognitive ability varies across geography. These models posit that geoclimatic factors will strongly predict cognitive ability across geography, even when a variety of common controls appear in the regression equations. Our results generally do not support UV radiation (UVR) based or other geoclimatic models. Specifically, although UVR alone predicted cognitive ability at the U.S. county-level (β = -.33), its validity was markedly reduced in the presence of climatic and demographic covariates (β = -.16), and was reduced even further with a spatial lag (β = -.10). For climate models, average temperature remained a significant predictor in the regression equation containing a spatial lag (β = .35). However, the effect was in the wrong direction relative to typical cold weather hypotheses. Moreover, when we ran the analyses separately by race/ethnicity, no consistent pattern appeared in the models containing the spatial lag. Analyses of gap sizes across counties were also generally inconsistent with predictions from the UVR model. Instead, results seemed to provide support for compositional models.
Full-text available
A dataset of socioeconomic, demographic and geographic data for US counties (N≈3,100) was created by merging data from several sources. A suitable subset of 28 socioeconomic indicators was chosen for analysis. Factor analysis revealed a clear general socioeconomic factor (S factor) which was stable across extraction methods and different samples of indicators (absolute split-half sampling reliability = .85). Self-identified race/ethnicity (SIRE) population percentages were strongly, but non-linearly, related to cognitive ability and S. In general, the effect of White% and Asian% were positive, while those for Black%, Hispanic% and Amerindian% were negative. The effect was unclear for Other/mixed%. The best model consisted of White%, Black%, Asian% and Amerindian% and explained 41/43% of the variance in cognitive ability/S among counties. SIRE homogeneity had a non-linear relationship to S, both with and without taking into account the effects of SIRE variables. Overall, the effect was slightly negative due to low S, high White% areas. Geospatial (latitude, longitude, and elevation) and climatological (temperature, precipitation) predictors were tested in models. In linear regression, they had little incremental validity. However, there was evidence of non-linear relationships. When models were fitted that allowed for non-linear effects of the environmental predictors, they were able to add a moderate amount of incremental validity. LASSO regression, however, suggested that much of this predictive validity was due to overfitting. Furthermore, it was difficult to make causal sense of the results. Spatial patterns in the data were examined using multiple methods, all of which indicated strong spatial autocorrelation for cognitive ability, S and SIRE (k nearest spatial neighbor regression [KNSNR] correlations of .62 to .89). Model residuals were also spatially autocorrelated, and for this reason the models were re-fit controlling for spatial autocorrelation using KNSNR-based residuals and spatial local regression. The results indicated that the effects of SIREs were not due to spatially autocorrelated confounds except possibly for Black% which was about 50% weaker in the controlled analyses. Pseudo-multilevel analyses of both the factor structure of S and the SIRE predictive model showed results consistent with the main analyses. Specifically, the factor structure was similar across levels of analysis (states and counties) and within states. Furthermore, the SIRE predictors had similar betas when examined within each state compared to when analyzed across all states. It was tested whether the relationship between SIREs and S was mediated by cognitive ability. Several methods were used to examine this question and the results were mixed, but generally in line with a partial mediation model. Jensen's method (method of correlated vectors) was used to examine whether the observed relationship between cognitive ability and S scores was plausibly due to the latent S factor. This was strongly supported (r = .91, Nindicators=28). Similarly, it was examined whether the relationship between SIREs and S scores was plausibly due to the latent S factor. This did not appear to be the case.
Full-text available
Multiple studies have reported positive relationships between cognitive ability and preferences for freedom, both at the personal level (e.g. drug use) and the economic (e.g. smaller government). To add to this, we investigated the relationships between cognitive ability and multi-dimensional political preferences in a Danish general population sample (n = 333, n = 259 after quality control). Respondents answered 10 questions pertaining to specific personal issues, 10 questions pertaining to specific economic issues, as well as taking 4 cognitive items. They had previously taken a 5-item cognitive test (ICAR5), and been asked to rate themselves on personal liberalism and economic liberalism. We documented a general factor of personal liberalism across the questions on personal issues, and a general factor of economic liberalism across the questions on economic issues. Self-assessed personal liberalism had a small positive correlation with measured personal liberalism (r = .13), while self-assessed economic liberalism had a moderate positive correlation with measured economic liberalism (r = .49). These findings were in line with our predictions. Contrary to our predictions, however, personal liberalism and economic liberalism had a weak positive correlation (r = .07; 95 % CI = [–.06, .19]) and cognitive ability was only weakly related to both personal liberalism (r = .14; 95 % CI = .02, .26]) and economic liberalism (r = .07; 95 % CI = [–.05, .19]). Corrected for measurement error, the correlations were .23 and .10, respectively. The study was preregistered.
Full-text available
The right-left dimension is ubiquitous in politics, but prior perspectives provide conflicting accounts of whether cultural and economic attitudes are typically aligned on this dimension within mass publics around the world. Using survey data from 99 nations, we find not only that right-left attitude organization is uncommon, but that it is more common for culturally and economically right wing-attitudes to correlate negatively with each other, an attitude structure reflecting a contrast between desires for cultural and economic protection vs. freedom. We examine where, among whom, and why protection-freedom attitude organization outweighs right-left attitude organization, and discuss implications for the psychological bases of ideology, quality of democratic representation, and the rise of extreme right politics in the West.
Using data from the American National Election Studies, we investigated the relationship between cognitive ability and attitudes toward and actual voting for presidential candidates in the 2012 and 2016 U.S. presidential elections (i.e., Romney, Obama, Trump, and Clinton). Isolating this relationship from competing relationships, results showed that verbal ability was a significant negative predictor of support and voting for Trump (but not Romney) and a positive predictor of support and voting for Obama and Clinton. By comparing within and across the election years, our analyses revealed the nature of support for Trump, including that support for Trump was better predicted by lower verbal ability than education or income. In general, these results suggest that the 2016 U.S. presidential election had less to do with party affiliation, income, or education and more to do with basic cognitive ability.
A number of studies have reported that higher intelligence is associated with socially liberal attitudes. Less clear, however, is whether this link extends to economic attitudes, although there are indications that higher intelligence is associated with economically conservative attitudes. Here in two large, longitudinal UK cohorts (each N > 7100) we assessed whether childhood intelligence predicted adulthood economic attitudes. In both cohorts we saw that higher levels of childhood (age 10–11) intelligence were related to higher levels of economic conservatism in adulthood (age 30–33). These effects were robust to the inclusion of potential confounders (sex, parental social class, childhood conduct problems). Moreover, this pathway was at least partially mediated by educational attainment and achieved social class/income. These findings confirm the importance of intelligence as an important phenotype for understanding the origins of economic attitudes. Implications for self-interest and rational-interest theories are discussed.
953), intelligence does not predict the standard oper-ationalization of vote choice in which parties are placed on a single left-vs-right dimension. (Standardized coefficients predicting right-wing vote choice were 0.05 and −0.03, respectively.) However, this apparent non-effect in fact reflects approximately equal and opposite effects of intelligence on vote choice as transmitted through social and economic ideology. In both countries, higher ability predicts left-wing social and right-wing economic views. The impact of intelligence on vote choice is thus most visible in true multi-party systems like Denmark, in which parties do not simply pair similar levels of social and economic conservatism, but instead provide diverse combinations of social and economic ideology. Comparing the parties closest to representing authoritarian egalitarianism (social-right plus economic-left) and libertarianism (social-left plus economic-right), we observed a 0.9 SD intelligence gap.
This paper analyses data on average IQ and four measures of political attitudes at both the regional level (n = 11) and the local authority level in Britain (n = 372). At the regional level, average IQ is positively associated with right-wing economic attitudes and trust in experts, but is not significantly associated with liberal social attitudes or intention to vote Remain in the EU referendum. At the local authority level, average IQ is positively associated with all four measures of political attitudes, although the associations with right-wing economic attitudes and trust in experts are stronger than the associations with liberal social attitudes and intention to vote Remain in the EU referendum. In multivariate models, average SES is a better and more robust predictor of political attitudes than average IQ.
This paper offers a quantitative analysis of ideology in China, to examine how beliefs and preferences are configured, and to measure the extent to which they are bound together by some form of constraint. Based on a large scale online survey, we identify an ideological spectrum characterized by one main dimension bound by a relatively weak constraint using both principal component analysis and models of the item response theory. On one end of this spectrum, preferences for authoritarian rule coincide with support for traditional norms and political distribution of resources. On the other end of this spectrum, preferences for political liberalization coincide with opposition to traditional values and support for markets allocation of resources. Evidence suggests that this latter set of views is more likely found in provinces with higher levels of development and among individuals with higher income and education. We corroborate these findings with a nationally representative sample from the Asian Barometer Survey.