You are on page 1of 9

Original Article

Nonlinear Effects of Cognitive Ability


on Economic Productivity
${protocol}://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001/a000247 - Tuesday, July 10, 2018 3:32:01 AM - IP Address:209.107.195.219

A Country-Level Analysis
Thomas R. Coyle,1 Heiner Rindermann,2 Dale Hancock,1 and Jacob Freeman3
1
Department of Psychology, University of Texas at San Antonio, TX, USA
2
Department of Psychology, Chemnitz University of Technology, Chemnitz, Germany
3
Department of Sociology, Social Work, and Anthropology, Utah State University, Logan, UT, USA

Abstract: Cognitive capitalism theory argues that the positive effects of cognitive ability on economic productivity should increase nonlinearly,
with increases in ability amplifying increases in productivity. The theory was tested using country-level indicators of cognitive ability and
productivity. Cognitive ability was based on international student assessments (e.g., Program for International Student Assessment, PISA), and
productivity was based on economic inputs (e.g., scientific achievements and competitiveness) and outputs (e.g., gross domestic product). As
predicted, the effects of cognitive ability on all productivity measures increased nonlinearly at higher levels of ability, suggesting that higher
ability levels disproportionately boost a nation’s productivity. The findings are discussed in light of standard theories of cognitive ability (e.g.,
Spearman’s law of diminishing returns and differentiation theories), and suggest that interventions that boost cognitive ability can have large,
amplifying effects on economic productivity.

Keywords: cognitive capitalism, cognitive ability, economic productivity, nonlinear effects, quadratic effects

Cognitive capitalism theory predicts that the effects of cog- quadratic effect of ability on productivity, and would be
nitive ability on economic productivity increase nonlinearly, depicted on a scatterplot (of productivity against ability)
with increasing returns of cognitive ability to productivity at as a curve whose slope increased from lower to higher
higher ability levels (e.g., Rindermann & Thompson, 2011; levels of ability. Such a pattern would support the predic-
see also, Coyle, Rindermann, & Hancock, 2016; Lubinski, tions of cognitive capitalism, and suggest that policies
2009; Wai, 2013). The theory is based on the assumption boosting cognitive ability through country-level interven-
that higher ability people contribute disproportionately to tions (e.g., better nutrition) can have large, amplifying
economic gains by creating innovations that amplify effects on economic productivity (e.g., Coyle et al., 2016).
productivity. Such a theory contrasts with standard theories To examine the predictions of cognitive capitalism, cogni-
of cognitive ability (e.g., Spearman’s law of diminishing tive ability was measured using three international standard-
returns), which imply that the positive effects of ability on ized tests: Program for International Student Assessment
productivity decrease nonlinearly at higher ability levels (PISA), Trends in International Mathematics and Science
(e.g., Coyle & Rindermann, 2013). These latter theories Study (TIMSS), and Progress in International Reading
are based on the proposition that cognitive ability has Literacy Study (PIRLS). These tests are given to students
diminishing returns to productivity at higher ability levels, throughout the world and can estimate cognitive ability at
because noncognitive factors (e.g., personality traits) drive the country level. The tests yield estimates of cognitive
productivity at higher ability levels. ability that correlate strongly with adult IQs (r  .90) and
Although positive (linear) effects of cognitive ability on also with a general intelligence (g) factor (r  .99; Coyle &
economic productivity have been repeatedly demonstrated Rindermann, 2013; Rindermann, 2007), which drives the
(e.g., Jones, 2016; Meisenberg, 2012), the current study is predictive validity of tests (Jensen, 1998, pp. 270–305).
the first to use country-level data to systematically examine Following prior research (Coyle et al., 2016), cognitive abil-
whether the positive effects of ability on productivity ity levels were estimated for two classes of people in each
increase nonlinearly at higher ability levels. The predicted country: average ability classes, which represent ability
nonlinear pattern would manifest in regression as a positive levels at the 50th percentile, and intellectual classes, which

Ó 2018 Hogrefe Publishing Journal of Individual Differences (2018), 39(1), 39–47


https://doi.org/10.1027/1614-0001/a000247
40 T. R. Coyle et al., Nonlinear Effects of Cognitive Ability

represent ability levels at the 95th percentile. Intellectual years 1995–2012. The assessments measure reading, math,
classes are assumed to contribute disproportionately to and science abilities in 15-year-olds (PISA); math and science
economic productivity, which is strongly influenced by abilities in fourth or eighth grade (TIMSS); and reading
people with the highest ability levels (Park, Lubinski, & abilities in fourth grade (PIRLS). Results from the assess-
Benbow, 2007; Robertson, Smeets, Lubinski, & Benbow, ments were corrected for cross-country differences in school
2010; Wai, 2013). The ability levels of both classes predicted attendance (low attendance decreases test scores) and age at
economic inputs of productivity (science achievements, grade level (cognitive ability increases with age) (Coyle &
innovation, competitiveness, and economic freedom) and Rindermann, 2013, pp. 407–408). In addition, results were
${protocol}://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001/a000247 - Tuesday, July 10, 2018 3:32:01 AM - IP Address:209.107.195.219

economic outputs of productivity (gross domestic product averaged across subject (math, science, and reading), devel-
[GDP] and wealth). opmental level (age or grade), year (1995–2012), and assess-
The main hypothesis was that the effects of cognitive ment (PISA, TIMSS, and PIRLS; Rindermann, Sailer, &
ability on economic productivity would increase nonlinearly Thompson, 2009, pp. 6–7). The averages were standardized
at higher ability levels. In particular, cognitive ability was and converted to Greenwich IQs (UK IQ = 100, SD = 15). The
expected to have positive quadratic effects (beyond linear IQs were used to estimate the ability levels of intellectual
effects) on productivity inputs and outputs. Based on cogni- (95th percentile) and average (50th percentile) classes in
tive capitalism, the effects of cognitive ability were predicted each country. Compared to psychometric IQ tests, student
to be relatively weak at lower ability levels and to strengthen assessments are administered to larger and more represen-
at higher ability levels. The pattern was predicted for both tative samples, increasing their validity as ability estimates
ability classes (average and intellectual classes), whose abil- in countries where psychometric IQ data are sparse (e.g.,
ity levels correlate strongly at the country level (r > .90), and African countries; see Rindermann, 2013).
for all productivity measures, which also correlate strongly
at the country level (r  .74, Coyle et al., 2016). Science, Technology, Engineering, and Math (STEM)
The hypothesis was tested by regressing each measure of Achievement
productivity on the ability levels of different countries. If the STEM achievement was based on patent rates, Nobel Prizes,
effects of cognitive ability on productivity increase nonlin- scientist rates, and high-tech export rates (Cronbach’s
early, then quadratic effects of ability should significantly α = .70). Data were available for 188 countries; 97 countries
predict productivity beyond linear effects. Moreover, the also had cognitive ability data and were used here.
quadratic effects should be represented by a curve whose
slope increases from lower to higher levels of ability. Such Innovation
a pattern, if confirmed, would be the first to show that the Innovation was based on the Global Innovation Index
effects of cognitive ability (on productivity) are nonlinear (GII) of the World Intellectual Property Organization
and would support the predictions of cognitive capitalism. (Dutta & Lanvin, 2013), which measures innovations in
science, technology, and the economy. The GII is based
on seven country-level variables (institutions, human capi-
tal, infrastructure, market sophistication, business sophisti-
Method cation, technology outputs, and creative outputs). GII data
were available in years 2006–2013 for 142 countries;
All data are measured at the country level (N = 99 coun-
95 countries also had cognitive ability data and were used
tries) and were obtained from public sources.1 All variables
here.
have been described elsewhere (Coyle et al., 2016), and are
summarized below. The data used in the analyses, and the Competitiveness
correlations among variables, are provided as electronic Competitiveness was based on the Global Competitiveness
supplementary material (Tables S1 and S2 in Electronic Index (GCI) of the World Economic Forum (Schwab, 2013).
Supplementary Material, ESM 1). The GCI measures factors that determine the level of
productivity of a country. It is based on 12 country-level
factors (quality of institutions, quality of infrastructure,
Variables
macroeconomic environment, health and primary educa-
Cognitive Ability tion, higher education and training, efficient goods markets,
Cognitive ability was based on international student assess- labor market efficiency, financial market efficiency, techno-
ments for PISA, TIMSS, and PIRLS, from 99 countries in logical readiness, market size, business sophistication, and

1
The same data were analyzed by Coyle et al. (2016), who examined whether economic freedom moderated the (linear) effects of cognitive ability
on economic productivity. The study by Coyle et al. (2016) did not examine whether the effects of cognitive ability on productivity increase
nonlinearly at higher levels of ability, the question central to the current study.

Journal of Individual Differences (2018), 39(1), 39–47 Ó 2018 Hogrefe Publishing


T. R. Coyle et al., Nonlinear Effects of Cognitive Ability 41

knowledge and innovation). GCI data were available for Parallel regressions predicted a “criterion G,” which
148 countries; 94 countries also had cognitive ability data measured the variance common to all productivity mea-
and were used here. sures. (By convention, G refers to aggregate-level common
factor variance, whereas g refers to individual-differences
Economic Freedom level common factor variance.) The criterion G was based
Economic freedom was based on the Heritage Foundation on the first unrotated factor in principal axis factoring of
Index of Economic Freedom (Miller, Holmes, & Feulner, all productivity measures and accounted for most of the
2013), which measures rule of law, limited government, variance in the productivity measures (variance = 80%;
${protocol}://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001/a000247 - Tuesday, July 10, 2018 3:32:01 AM - IP Address:209.107.195.219

regulatory efficiency, and open markets; and the Fraser eigenvalue = 4.84; criterion G loadings = .71, .86, .84, .76,
Institute Index (Gwartney, Lawson, & Hall, 2013), which .83, .59, for STEM, GII, GCI, GDP, wealth, freedom, respec-
measures size of government, legal system and property tively). Criterion G factor scores were estimated for each
rights, sound money, trade freedoms, and labor regulations. country using regression, which weighted productivity
The Heritage and Fraser indices were standardized and measures by factor score coefficients (e.g., Tabachnick &
averaged to produce a composite measure of economic Fidell, 2007, pp. 622–625).
freedom (Cronbach’s α = .93). Data were available in year Key statistics were β, which measured standardized
2013 for 182 countries; 97 of these countries also had effects at each step; R2, which measured the variance
cognitive ability data and were used here. explained at each step; and ΔR2, which measured the
change in variance explained from the first to second step.
Gross Domestic Product (GDP) ΔR2 measured the increment in variance explained by
GDP measures the market value of goods and services quadratic effects, beyond linear effects. If the effects of
produced by a country. GDP per capita was obtained for cognitive ability strengthen with increases in ability level
year 2010 and converted to a log scale. The measure was (as predicted by cognitive capitalism), and if these effects
based on data from the Maddison project (Maddison, manifest as quadratic effects, then quadratic effects should
2008; 2010 GDP data from Bolt & van Zanden, 2013). explain significant variance beyond linear effects. In addi-
The final logged GDP variable was available for 161 coun- tion, the sign of the quadratic effects should be positive,
tries; 92 countries also had cognitive ability data and were which (along with positive linear effects) would indicate
used here. that the effects of ability on productivity strengthen at
higher ability levels. All effects are reported as standardized
Wealth coefficients (R2 or ΔR2), with mean effects (MR2 or MΔR2) in
Wealth measures the market value of financial plus nonfi- parentheses.
nancial assets (housing and land), less debts, for a country.
Wealth per adult was obtained for year 2013 and converted
to a log scale (Credit Suisse, 2013). Wealth data were
available for 174 countries; 96 countries also had cognitive Results
ability data and were used here.
Table 1 reports the regressions of productivity measures on
linear effects of cognitive ability (Step 1), followed by quad-
ratic effects of cognitive ability (Step 2). The results of each
Statistical Analyses
analysis are reported in the rows of Table 1, separately for
The main hypothesis was that the effects of cognitive ability average classes (50th percentile ability level) and intellec-
on economic productivity would increase nonlinearly at tual classes (95th percentile ability level). A key statistic is
higher ability levels. To test this hypothesis, regressions the increment in variance explained by quadratic effects
estimated the linear effects of ability in an initial step, beyond linear effects (i.e., ΔR2). If the effects of ability on
followed by quadratic (nonlinear) effects of ability in a productivity increase nonlinearly at higher ability levels
second step. Each analysis regressed a productivity mea- (as predicted by cognitive capitalism), quadratic effects
sure on cognitive ability, separately for intellectual or aver- should explain significant variance beyond linear effects
age classes. To facilitate interpretation, all measures were (and have a positive sign).
standardized and centered (M = 0, SD = 1) prior to analyses Table 1 (Average class) reports regressions for average
(Cohen, Cohen, West, & Aiken, 2003, pp. 201–204). Linear classes (50th percentile ability level). As predicted, cogni-
effects involved a centered ability measure; quadratic tive ability had strong linear effects at Step 1 (MR2 = .43,
effects involved the square of the centered ability measure. range = .21–.62), and modest but significant quadratic
To minimize collinearity, regressions were performed effects at Step 2 (MΔR2 = .06, range = .03–.09). In addition,
separately for each productivity measure. the quadratic effects were slightly stronger for economic

Ó 2018 Hogrefe Publishing Journal of Individual Differences (2018), 39(1), 39–47


42 T. R. Coyle et al., Nonlinear Effects of Cognitive Ability

Table 1. Change in R2 (ΔR2) after adding a quadratic term (Step 2) to a linear term (Step 1) for ability level
Step 1 Step 2 Δ Statistics
Criterion N R2
βlinear Fmodel R2
βlinear βquad Fmodel ΔR 2
ΔF
Average class
STEM 97 .26 .51** 33.48** .34 .69** .34** 24.46** .082 11.68**
Innovation 95 .62 .79** 152.47** .71 .98** .36** 112.63** .089 28.20**
Competitiveness 94 .45 .67** 74.32** .51 .82** .29** 47.54** .064 11.93**
Freedom 97 .21 .45** 24.54** .28 .63** .33** 18.48** .077 10.09**
${protocol}://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001/a000247 - Tuesday, July 10, 2018 3:32:01 AM - IP Address:209.107.195.219

GDP 92 .58 .76** 123.75** .61 .85** .19* 68.20** .026 5.90*
Wealth 96 .44 .66** 73.69** .48 .79** .24** 43.48** .044 7.87**
Criterion G 87 .55 .74** 103.96** .62 .90** .31** 69.10** .072 15.95**
Intellectual class
STEM 97 .29 .54** 39.45** .38 .70** .33** 28.56** .085 12.77**
Innovation 95 .66 .81** 180.41** .73 .96** .30** 123.84** .069 23.54**
Competitiveness 94 .47 .69** 82.79** .52 .80** .25** 50.20** .051 9.73**
Freedom 97 .24 .49** 30.59** .33 .65** .33** 22.86** .084 11.76**
GDP 92 .62 .79** 149.06** .65 .87** .17* 81.57** .024 5.93*
Wealth 96 .48 .69** 86.97** .53 .80** .25** 52.22** .048 9.56**
Criterion G 87 .60 .77** 125.60** .66 .89** .27** 79.73** .059 14.26**
Notes. Average class = average class ability levels (50th percentile); Intellectual class = intellectual class ability levels (95th percentile); STEM = achieve-
ments in science, technology, engineering, and math; Innovation = Global Innovation Index; Competitiveness = Global Competitiveness Index; Free-
dom = economic freedom index; GDP = gross domestic product per capita (log scale); Wealth = wealth per capita; Criterion G = factor scores based on all six
productivity measures. *p < .05, **p < .01 (two-tailed).

inputs (STEM, GII, GCI, and freedom; MΔR2 = .08) than strong correlation between the ability levels of the two
outputs (GDP and wealth; MΔR2 = .04). Results were similar classes (r = .98, Table S1).
for the criterion G (based on all productivity measures), The significant quadratic effects were probed by com-
which also showed significant linear (R2 = .55) and quadra- puting simple effects of cognitive ability on economic
tic effects (ΔR2 = .07). Consistent with predictions, scatter- productivity at five levels of ability: 2 SDs below average,
plots indicated that the effects of cognitive ability increased 1 SD below average, average (country mean), 1 SD above
nonlinearly, with increasing returns of cognitive ability to average, and 2 SDs above average. Simple effects were
productivity at higher ability levels (Figures 1A–1G). computed separately for each productivity measure and
Table 1 (Intellectual class) also reports regressions for the ability class, after zero centering the linear and quadratic
intellectual classes (95th percentile ability level). The results effects of cognitive ability (cf. Cohen et al., 2003, pp.
confirmed the analyses for the average classes. Cognitive 206–207). The simple effects measured the slope of the
ability still had strong linear effects at Step 1 (MR2 = .46, tangent line to the quadratic curve at a particular level of
range = .21–.62), and significant quadratic effects at Step 2 cognitive ability.
(MΔR2 = .06, range = .02–.09). As before, the quadratic Table 2 reports the simple effects, which bolstered the
effects were slightly stronger for economic inputs (MΔR2 = pattern depicted in scatterplots (Figures 1 and 2). Simple
.07) than outputs (MΔR2 = .04). Results were similar for effects of ability on all productivity measures increased
the criterion G, which showed significant linear (R2 = .60) from lowest ( 2 SD, 1 SD) to average ability levels, and
and quadratic effects (ΔR2 = .06). Consistent with theory, stabilized (at stronger values) from average to the highest
scatterplots indicated that the effects of cognitive ability levels (1 SD, 2 SD). The pattern was found for all productiv-
strengthened at higher levels of ability (Figure 2A–2G). ity measures and both ability classes (grand M = .59, .68,
Differences in analogous effects for the two ability classes .77, .80, .81, from lowest to highest ability levels, respec-
(intellectual-average) were very small for linear (MR2 tively). Differences in analogous effects for the two ability
difference = .033) and quadratic effects (MΔR2 difference = classes (intellectual – average) were trivial (Mβ difference =
.004).2 The small differences could be attributed to the .01, range = .02 to .04).

2
Parallel regressions of STEM on cognitive ability were performed after omitting an extreme STEM value in the upper-right quadrant (Figures 1A
and 2A). The value corresponded to Luxembourg, a country with very high-patent rates, which inflated its STEM rating (Rindermann et al., 2009,
Table 1). Consistent with the prior results, the effects of cognitive ability on STEM still increased nonlinearly (and significantly) for average
classes (ΔR2 = .12) and intellectual classes (ΔR2 = .12).

Journal of Individual Differences (2018), 39(1), 39–47 Ó 2018 Hogrefe Publishing


T. R. Coyle et al., Nonlinear Effects of Cognitive Ability 43

(A) (B)
${protocol}://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001/a000247 - Tuesday, July 10, 2018 3:32:01 AM - IP Address:209.107.195.219

(C) (D)

(E) (F)

(G)

Figure 1. Scatterplots of average class ability levels and productivity criteria. Plots are fitted with linear and quadratic curves for the six
productivity measures (Panels A–F) and the criterion G (Panel G). R2 Linear = linear effects. R2 Quadratic = quadratic effects.

Ó 2018 Hogrefe Publishing Journal of Individual Differences (2018), 39(1), 39–47


44 T. R. Coyle et al., Nonlinear Effects of Cognitive Ability
${protocol}://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001/a000247 - Tuesday, July 10, 2018 3:32:01 AM - IP Address:209.107.195.219

Figure 2. Scatterplots of intellectual class ability levels and productivity criteria. Plots are fitted with linear and quadratic curves for the six
productivity measures (Panels A–F) and the criterion G (Panel G). R2 Linear = linear effects. R2 Quadratic = quadratic effects.

Journal of Individual Differences (2018), 39(1), 39–47 Ó 2018 Hogrefe Publishing


T. R. Coyle et al., Nonlinear Effects of Cognitive Ability 45

Table 2. Simple effects (β) between ability levels and productivity unique variance beyond linear effects (Table 1). Moreover,
measures at different levels of ability, for average and intellectual
the effects of ability increased sharply from the lowest to
classes
average ability levels and stabilized (at strong levels)
2 SD 1 SD Average +1 SD +2 SD
thereafter, demonstrating increasing returns of ability to
Average class productivity (Table 2).
STEM .47 .60 .67 .69 .70 The nonlinear effects of cognitive ability replicated for
Innovation .75 .88 .96 .98 .99 both intellectual and average classes (Table 1), suggesting
Competitiveness .54 .62 .73 .77 .79 that the effects apply throughout the ability spectrum.
${protocol}://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001/a000247 - Tuesday, July 10, 2018 3:32:01 AM - IP Address:209.107.195.219

Freedom .35 .45 .55 .59 .61 In addition, the nonlinear effects of ability replicated for
GDP .72 .78 .83 .85 .85 all productivity measures, with slightly stronger effects for
Wealth .59 .67 .74 .77 .78 economic inputs than outputs, suggesting that cognitive
Criterion G .65 .75 .84 .87 .88 ability has slightly greater impact on precursors of produc-
M .58 .68 .76 .79 .80 tivity (i.e., STEM, innovation, competitiveness, freedom).
Intellectual class The similar effects were partly attributable to the strong
STEM .47 .58 .68 .71 .71 relations among productivity measures, as indicated by
Innovation .76 .86 .94 .97 .97 their high loadings on a common productivity factor (i.e.,
Competitiveness .56 .60 .71 .76 .78 criterion G; M loading = .77). This common factor reflected
Freedom .39 .47 .59 .63 .65 variance common to all productivity measures, and yielded
GDP .76 .82 .87 .88 .88 factor scores that showed a pattern similar to the individual
Wealth .62 .69 .77 .79 .80 productivity measures (Tables 1 and 2).
Criterion G .68 .75 .84 .87 .88 The results are the first to demonstrate that the effects of
M .61 .68 .77 .80 .81 cognitive ability on economic productivity increase nonlin-
Grand M .59 .68 .77 .80 .81 early at higher ability levels, a pattern consistent with the
Notes. Average class = average class ability level (50th percentile); Intel- predictions of cognitive capitalism theory (Rindermann,
lectual class = intellectual class ability level (95th percentile);
STEM = achievements in science, technology, engineering, and math;
2012). The results suggest that prior research on country-
Innovation = Global Innovation Index; Competitiveness = Global Competi- level differences in cognitive ability, which has examined
tiveness Index; Freedom = economic freedom index; GDP = gross domestic linear effects, may have inadvertently missed nonlinear
product per capita (log scale); Wealth = wealth per capita; Criterion
G = factor scores based on all six productivity measures; M = mean effect; increases by excluding nonlinear (quadratic) terms from
Grand M = grand mean, averaged across all effects and ability classes. statistical models. The robustness of the results, which
Effects are computed at the following ability levels: 2 SD (2 SD below
replicated with different measures of ability and productiv-
average), 1 SD (1 SD below average), Average (country average), +1 SD
(1 SD above average), and +2 SD (2 SD above average). ity, suggests that future studies should check for nonlinear
effects by including quadratic terms in statistical models.
The pattern of effects is inconsistent with theories of
Discussion intelligence (e.g., Spearman’s law of diminishing returns),
which imply that relations of cognitive ability with various
The current study is the first to use country-level data to criteria should decrease nonlinearly at higher levels of ability
systematically examine whether the positive effects of (cf. Coyle & Rindermann, 2013; see also, Coyle, 2015). Such
cognitive ability on economic productivity increase nonlin- theories are based on the assumption that, at higher ability
early at higher levels of ability. Such a pattern is predicted levels, measures of cognitive ability become less loaded
by cognitive capitalism theory (Rindermann & Thompson, with g (i.e., variance common to tests), which drives predic-
2011), which predicts that higher levels of ability dispropor- tive validity, and more loaded with non-g factors, which
tionately account for higher levels of economic productivity. contribute less to predictive validity (Jensen, 1998,
To test cognitive capitalism, economic productivity was pp. 270–305). The results of the current study support the
regressed on ability levels for intellectual (95th percentile) opposite pattern. The effects of cognitive ability (on all
and average classes (50th percentile), using six measures productivity measures) increased nonlinearly at higher
of productivity (STEM, innovation, competitiveness, ability levels, suggesting that productivity is amplified at
freedom, GDP, and wealth). Linear effects of ability were higher levels of ability.
estimated at an initial step, followed by quadratic effects An alternative interpretation of the results is based on
at a second step, which examined whether the effects of differentiation theories (e.g., Deary et al., 1996; see also,
ability increased nonlinearly at higher levels. Consistent Woodley, 2012).3 Differentiation theories focus on specific
with theory, quadratic effects of ability explained significant abilities (e.g., math, verbal, and spatial), which are partly

3
We thank Michael Woodley of Menie for suggesting the alternative interpretation.

Ó 2018 Hogrefe Publishing Journal of Individual Differences (2018), 39(1), 39–47


46 T. R. Coyle et al., Nonlinear Effects of Cognitive Ability

distinct from g. According to these theories, the effects of In fact, student assessments load strongly on the g factor,
specific abilities increase (and the effects of g decrease) at which is based on diverse measures of ability (λ > .90; Coyle
higher ability levels, a pattern that reflects cognitive special- & Rindermann, 2013; Rindermann, 2007).
ization in narrow domains. This specialization may lead to Future research should explore nonlinear effects of
increased divisions of labor and enhanced comparative ability on noneconomic criteria such as government effec-
advantage and economic efficiency, which may boost tiveness, disease rates, happiness, and well-being, which
economic growth. Cognitive specialization may also con- have been linked to linear effects of cognitive ability (Jones,
tribute to the Flynn effect, which is associated with 2016; see also, Lynn & Vanhanen, 2012; Rindermann,
${protocol}://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001/a000247 - Tuesday, July 10, 2018 3:32:01 AM - IP Address:209.107.195.219

increased effects of specific abilities (rather than gains on Kodila-Tedika, & Christainsen, 2015). Given the assump-
g; te Nijenhuis & van der Flier, 2013), and is predictive of tion that higher ability levels contribute disproportionately
GDP growth, both across countries (Pietschnig & Voracek, to outcomes, the effects of cognitive ability may increase
2015) and over time (Woodley, 2012). Future research nonlinearly for noneconomic criteria.
could test for cognitive specialization by partitioning the Future research should also explore the influence of
effects of g and specific abilities (obtained after removing g) personality traits on ability and economic productivity
and examining their effects at different ability levels. (Heckman & Kautz, 2012). Personality traits include consci-
According to differentiation theories, the effects of specific entiousness and agreeableness, which have been linked to
abilities may increase at higher ability levels, and the productivity at the country level (e.g., Freeman, Coyle, &
pattern of effects may be nonlinear, reflecting increased Baggio, 2016; Stolarski, Zajenkowski, & Meisenberg, 2013)
cognitive specialization at higher ability levels. and may mediate or moderate nonlinear effects of cognitive
The results indicated that the effects of cognitive ability ability. For example, only countries with high levels of consci-
increased disproportionately from the lowest to average abil- entiousness may invest the resources (e.g., research and edu-
ity levels, and stabilized (at strong levels) thereafter. In par- cation) needed to produce the nonlinear effects of ability.
ticular, the effects of cognitive ability changed most rapidly Alternatively, only countries with high levels of agreeable-
at the lower end of the ability spectrum and plateaued at ness may have the levels of cooperation and trust needed
the higher end of the spectrum. Such a pattern suggests that to produce the nonlinear effects of ability (cf. Freeman
boosting cognitive ability through environmental interven- et al., 2016).
tions (e.g., adequate nutrition or better health care) may yield In sum, the current study is the first to show that the
disproportionate gains in productivity at the lower end of the effects of country-level cognitive ability on economic
ability spectrum, where secular increases in ability are most productivity increase nonlinearly at higher ability levels.
rapid (cf. Rindermann & Thompson, 2013). The results replicated for both intellectual (95th percentile)
The results are based on country-level ability estimates, and average (50th percentile) classes, and for all pro-
which have drawn criticism (e.g., Wicherts & Wilhelm, ductivity measures. Future research should examine
2007). One concern is that the ability estimates are based whether the nonlinear effects of ability are moderated by
on student assessments (e.g., PISA, TIMSS, and PIRLS). other factors (e.g., personality traits), and whether the
Student assessments measure cognitive ability in children, effects replicate for noneconomic factors (e.g., happiness
who are not economically productive, and may not and well-being).
accurately measure ability in adults. Such a concern is min-
imized by strong relations between ability estimates from Acknowledgments
student assessments and adult IQs (r  .90, Rindermann, This research was supported by a grant from the National
2012), suggesting that countries with bright children tend Science Foundation’s Interdisciplinary Behavioral and
to have bright adults. A second concern is that country-level Social Science Research competition (Award: 1620457).
ability estimates may be distorted by country-level differ- Portions of the research were presented at the 2014 and
ences in the timing of the Flynn effect (e.g., Pietschnig & 2015 conventions of the Association for Psychological
Voracek, 2015; Rindermann & Thompson, 2013), which Science.
describes secular gains in ability levels. This concern is
minimized by the temporal stability of ability estimates Electronic Supplementary Material
(r  .86, Rindermann, 2012, Figure 2), indicating that The electronic supplementary material is available with the
countries maintain their rank order in ability over time, online version of the article at https://doi.org/10.1027/
which would minimize distortions from the Flynn effect. 1614-0001/a000247
A final issue is the assumption that student assessments
are weak indicators of the cognitive g factor, which ESM 1. Tables S1 and S2 (docx).
measures variance common to cognitive tests and drives The tables report the data used in the analyses and the cor-
the predictive validity of tests (Jensen, 1998, pp. 270–305). relations among the variables.

Journal of Individual Differences (2018), 39(1), 39–47 Ó 2018 Hogrefe Publishing


T. R. Coyle et al., Nonlinear Effects of Cognitive Ability 47

References Pietschnig, J., & Voracek, M. (2015). One century of global IQ gains.
Perspectives on Psychological Science, 10, 282–306. https://
doi.org/10.1177/1745691615577701
Bolt, J., & van Zanden, J. L. (2013). The first update of the
Rindermann, H. (2007). The g-factor of international cognitive
Maddison Project. Re-estimating growth before 1820
ability comparisons: The homogeneity of results in PISA, TIMSS,
(Maddison-Project Working Paper WP-4) Retrieved from www.
PIRLS and IQ-tests across nations. European Journal of
ggdc.net/maddison/maddison-project/data/mpd_2013-01.xlsx
Personality, 21, 667–706. https://doi.org/10.1002/per.634
Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied
Rindermann, H. (2012). Intellectual classes, technological pro-
multiple regression/correlation analysis for the behavioral
gress and economic development: The rise of cognitive capi-
sciences (3rd ed.). Mahwah, NJ: Erlbaum.
talism. Personality and Individual Differences, 53, 108–113.
Coyle, T. R. (2015). Relations among general intelligence (g),
${protocol}://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001/a000247 - Tuesday, July 10, 2018 3:32:01 AM - IP Address:209.107.195.219

https://doi.org/10.1016/j.paid.2011.07.001
aptitude tests, and GPA: Linear effects dominate. Intelligence,
Rindermann, H. (2013). African cognitive ability: Research, results,
53, 16–22. https://doi.org/10.1016/j.intell.2015.08.005
divergences and recommendations. Personality and Individual
Coyle, T. R., & Rindermann, H. (2013). Spearman’s law of dimin-
Differences, 55, 229–233. https://doi.org/10.1016/j.paid.2012.
ishing returns and national ability. Personality and Individual
06.022
Differences, 55, 406–410. https://doi.org/10.1016/j.paid.2013.
Rindermann, H., Kodila-Tedika, O., & Christainsen, C. (2015).
03.023
Cognitive capital, good governance, and the wealth of nations.
Coyle, T. R., Rindermann, H., & Hancock, D. (2016). Cognitive
Intelligence, 51, 98–108. https://doi.org/10.1016/j.intell.2015.
capitalism: Economic freedom moderates the effects of intel-
06.002
lectual and average classes on economic productivity. Psycho-
Rindermann, H., Sailer, M., & Thompson, J. (2009). The impact of
logical Reports, 119, 411–427. https://doi.org/10.1177/
smart fractions, cognitive ability of politicians and average
0033294116659854
competence of peoples on social development. Talent Devel-
Credit Suisse Research Institute. (2013). Global wealth databook
opment and Excellence, 1, 3–25.
2013. Zurich, Switzerland: Credit Suisse.
Rindermann, H., & Thompson, J. (2011). Cognitive capitalism: The
Deary, I. J., Egan, V., Gibson, G. J., Brand, C. R., Austin, E., &
effect of cognitive ability on wealth, as mediated through scientific
Kellaghan, T. (1996). Intelligence and the differentiation
achievement and economic freedom. Psychological Science, 22,
hypothesis. Intelligence, 23, 105–132. https://doi.org/10.1016/
754–763. https://doi.org/10.1177/0956797611407207
S0160-2896(96)90008-2
Rindermann, H., & Thompson, J. (2013). Ability rise in NAEP and
Dutta, S. & Lanvin, B. (Eds.). (2013). The Global Innovation Index
narrowing ethnic gaps? Intelligence, 41, 821–831. https://doi.
2013: The local dynamics of innovation. Retrieved from http://
org/10.1016/j.intell.2013.06.016
www.wipo.int/edocs/pubdocs/en/economics/gii/gii_2013.pdf
Robertson, K. F., Smeets, S., Lubinski, D., & Benbow, C. P. (2010).
Freeman, J., Coyle, T. R., & Baggio, J. A. (2016). The functional
Beyond the threshold hypothesis. Current Directions in Psy-
intelligences proposition. Personality and Individual Differ-
chological Science, 19, 346–351. https://doi.org/10.1177/
ences, 99, 46–55. https://doi.org/10.1016/j.paid.2016.04.057
0963721410391442
Gwartney, J., Lawson, R., & Hall, J. (2013). Economic freedom of the
Schwab, K. (2013). The Global Competitiveness Report 2013–2014.
world. 2013 Annual report. Vancouver, BC: Fraser Institute.
Geneva, Switzerland: World Economic Forum.
Retrieved from https://www.fraserinstitute.org/sites/default/
Stolarski, M., Zajenkowski, M., & Meisenberg, G. (2013). National
files/economic-freedom-of-the-world-2013.pdf
intelligence and personality: Their relationships and impact on
Heckman, J. J., & Kautz, T. (2012). Hard evidence on soft skills.
national economic success. Intelligence, 41, 94–101. https://
Labour Economics, 19, 451–464. https://doi.org/10.1016/
doi.org/10.1016/j.intell.2012.11.003
j.labeco.2012.05.014
Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statis-
Jensen, A. R. (1998). The g factor: The science of mental ability.
tics (5th ed.). Boston, MA: Allyn and Bacon.
Westport, CT: Praeger.
te Nijenhuis, J., & van der Flier, H. (2013). Is the Flynn effect on g?
Jones, G. (2016). Hive mind: How your Nation’s IQ matters so much
A meta-analysis. Intelligence, 41, 802–807. https://doi.org/
more than your own. Redwood City, CA: Stanford University Press.
10.1016/j.intell.2013.03.001
Lubinski, D. (2009). Exceptional cognitive ability: The phenotype.
Wai, J. (2013). Investigating America’s elite: Cognitive ability,
Behavior Genetics, 39, 350–358. https://doi.org/10.1007/
education, and sex differences. Intelligence, 41, 203–211.
s10519-009-9273-0
https://doi.org/10.1016/j.intell.2013.03.005
Lynn, R., & Vanhanen, T. (2012). National IQs: A review of their
Wicherts, J. M., & Wilhelm, O. (2007). What is the national g-
educational, cognitive, economic, political, demographic, sociolog-
factor? European Journal of Personality, 21, 763–765. https://
ical, epidemiological, geographic and climatic correlates. Intelli-
doi.org/10.1002/per.648
gence, 40, 226–234. https://doi.org/10.1016/j.intell.2011.11.004
Woodley, M. A. (2012). The social and scientific temporal corre-
Maddison, A. (2008). Statistics on world population, GDP and per
lates of genotypic intelligence and the Flynn effect. Intelligence,
capita GDP, 1–2008 AD. Retrieved from http://www.ggdc.
40, 189–204. https://doi.org/10.1016/j.intell.2011.12.002
net/maddison/Historical_Statistics/horizontal-file_02-2010.xls
Meisenberg, G. (2012). National IQ and economic outcomes. Received January 14, 2017
Personality and Individual Differences, 53, 103–107. https:// Revision received April 24, 2017
doi.org/10.1016/j.paid.2011.06.022 Accepted May 10, 2017
Miller, T., Holmes, K. R., & Feulner, E. J. (2013). 2013 index of Published online January 12, 2018
economic freedom. Washington, DC: The Heritage Foundation.
Retrieved from www.heritage.org/index/pdf/2013/book/index_ Thomas R. Coyle
2013.pdf Department of Psychology
Park, G., Lubinski, D., & Benbow, C. P. (2007). Contrasting University of Texas at San Antonio
intellectual patterns predict creativity in the arts and sciences: One UTSA Circle
Tracking intellectually precocious youth over 25 years. Psycho- San Antonio, TX 78249
logical Science, 18, 948–952. https://doi.org/10.1111/j.1467- USA
9280.2007.02007.x thomas.coyle@utsa.edu

Ó 2018 Hogrefe Publishing Journal of Individual Differences (2018), 39(1), 39–47

You might also like