What Makes You Smart?

A Regression Analysis on Economic Factors Affecting Intelligence Quotient

An Empirical Paper
Presented to
The School of Economics
De La Salle University

In Partial Fulfillment of the Requirements in ECONMET

Submitted by:
Ermitao, Lorenzo Gabriel S.
11300914

Submitted to:
Ms. Neriza Casas

April 20, 2015

Table of Contents

I.

II.

III.
IV.

V.
VI.

VII.
VIII.
IX.

Introduction
Background of the Study
Statement of the Problem
Objectives of the Study
Scope and Limitations
Review of Related Literature
National IQ and Gross Domestic Product per Capita
Body Mass Index
Population Growth
Mean Years of Schooling
Theoretical Framework
Implicit Theories on Intelligence
The Augmented Solow Model
Operational Framework
Description of the Variables
A priori Expectations
Hypothesized Econometric Model
Methodology
Data
Empirical Estimation and Inference Procedures
Econometric Model
Initial Regression
Test for Overall Significance
Test for Presence of Multicollinearity
Test for Presence of Heteroscedasticity
Test for Misspecification Bias
Corrective Measures and Final Regression
Conclusion
Recommendations
Bibliography

Introduction

Background of The Study

When you think about the smartest people in history such as Albert
Einstein and Steven Hawking, there is one thing that they all have in
common high IQs. The smartest people in the world all have very high IQ
scores, but we must ask, what is this ranking exactly?
IQ or Intelligence Quotient is a score from specialized test that in
essence will rank your overall intelligence. The term was first coined in the
1912 book of psychologist William Stern entitled The Psychological Methods
of Intelligence Testing. A pioneer on the psychology of personaly traits,
Stern looked for a comprehensive way to determine general intelligence. He
was able to take test scores as a representation of the mental age of the
participants of the study and divided this by their true age. This ratio showed
how much a persons level of intelligence advanced throughout the years
and this is what we know as IQ now (Stern, 1912).
Currently, there are many tests that are being used by different
psychologists to determine the intelligence quotient of a person; the most
well-known being the Stanford-Binet Test and Weschlers tests for adult and
child intelligence. Using these tests, many educational institutions look for
ways to improve their curriculum on the basis that higher IQ scores means
that their students are smarter. In the same way, IQ tests have developed

throught the years in order to compensate for the increasing amount of new
knowledge and technology available to everyone in the world.
In this age where information always at a ready, understanding
intelligence becomes more and more critical. More information leads to more
knowledge being spread, the consequences of this event may be debated.
However, more knowledge for the general population will help develop the
scoiety as a whole more job opportunities for citizens, more education for
the lower-class populations, and an economy that runs smoother and
produces more. Microeconomists have always pondered and tried to measure
what makes the average worker better in some countries than in other, and
one factor can that set them apart is the knowledge they posses. By
increasing the general intelligence level in the country, the possible benefits
can encompass many sectors of the countrys economy.

Statement of the Problem

Even with the advancements in modern psychology, there are still no
definite measures of true intelligence, simply because it is such a complex
process to understand. However, by continually searching for ways to
improve the intelligence of the average person, there is an obvious initiative
and desire for progress and development of every human being. This paper
looks to study the variables that can have significant effects on the average
intelligence quotient of people in different places around the world. The main
purpose of thus study is to determine and analyze the relationship between

National Intelligence Quotient, Gross Domestic Product per Capita, Body

Mass Index, Population Growth, and Mean Years of Schooling.

Objectives of the Study

The purpose of this paper is to specify which economic factors can
affect a countrys average intelligence quotient. It also looks to present the
correlations between National Intelligence Quotient, Gross Domestic Product
per Capita, Body Mass Index, Population Growth, and Mean Years of
Schooling across a set of selected countries. The secondary objectives are as
follows:
1. To present new research and ideas in the field of Educational
Economics
2. To give more research to policy makers so that they may be able
to create resolutions that work towards reducing the knowledge
and education gap between communities
3. To use and demonstrate current economic theories to construct a
model that will look to explain the possible connection between
the dependent and independent variables

Significance of the Study

According to Tatu Vanhanen (2011), Average general intelligence of
the population provides the most powerful explanation for the differences in
the wealth and poverty of nations. In a world where there is a increasing
gap between the rich and the poor, there is a need for a way to help the less-

fortunate citizens. The study looks to present new research in order to enable
the governments in many third-world countries reduce the gap in wealth. By
reducing this gap, more people will be able to go out of their empoverished
life and be able to contribute to society. If this were to occur, the economies
will likely see more economics growth because the labor force has not just
grown, but the quality of each laborer will increase. More knowledge can
mean increases in wages for the workers, which leads to more income and a
better standard of living thorughout the country.
Scope and Limitations
The research in this paper will employ a cross-sectional dataset that
uses the statistics of National Average Intelligence Quotient, Gross Domestic
Product per Capita (in current US dollars), Body Mass Index, Population
Growth (in Percentage), and Mean Years of Schooling (in years) for the year
2010. The National Average Intelligence Quotient was taken from Lynn and
Vanhanens study of national IQ across countries in their book IQ and the
Wealth of Nations. This variable will be used to measure average
intelligence level of a person. The independent variables, GDP per capita and
population growth, were retrieved from the World Bank Human Development
Indicators. The other two explanatory variables, BMI and mean years of
schooling, were both taken from the United Nation Development Programs
Human Development Report of 2012. The year 2010 was used to ensure that
all data was relevant in the model and because there has not been another
data set containing comprehensive average IQ score that has been made

since then. Furthermore, countries that did not have complete data in the
categories were excluded from the study in order to ensure significant
results. Lastly, 183 observations were used in the data where the different
countries included in this research represent each of these observations.
One thing that is apparent about the study is the fact that it is intuitive
that there is a limit of IQ in actually determining general intelligence.
However, since IQ is widely accepted statistic and there is a lack of data on
other knowledge statistics, it will still be used. It will be shown later,
however, that despite using IQ, there are still very strong economic
implications that may still have future significance. Furthermore, the use of
Body Mass Index in the study will be used to represent the relative health
level of each country. Other indices were considered for the study, however,
not enough data was available for all the countries. In order to give a more
complete study, Body Mass Index was therefore used since statistics were
found for all countries for the year 2010, and many regard Body Mass Index
as a viable index for the health level of a person.

Review of Related Literature

With the significance, objectives, and scope of the study defined in the
previous pages, this section will now focus on the existing literature that
pertains to the different variables. The next section will look to describe
current literature, economic theories, and current knowledge that will help

understand the possible relationships of the variables to provide the research

with a conceptual framework to develop the a-priori expectations.

National IQ and Gross Domestic Product per capita

National IQ scores are a collection of IQ test scores from adults over
the ages of 25 and averaged for a national level. Richard Lynn, a Phd in
Psychology and Tatu Vanhanen, a Professor Emeritus of Political Science,
pioneered this collecting and study of IQ scores. Their 2002 book entitled IQ
and the Wealth of Nations, and their subsequent follow-up book IQ and
Global Inequality published in 2006, looked at the different factors that
affected a countrys IQ and what implications they might have. Their books
have shown that national IQ correlates with gross domestic product per
capitaat 0.82, and with the rate of economic growth from 19501990 at
0.64 (Lynn and Vanhanen, 2002). The authors were able to draw the
conclusions that high IQ countries will have higher GDP per capitas as
compared to countries with a low general IQ level. The authors attributed
these effects to genetic and environmental factors that play a big role in
determining how smart the average person is, one of those environmental
factors being GDP per capita. This theory works hand-in-hand with the
Nature vs. Nurture premise in psychology where both environmental and
biological factors affect different aspects of a newborns development
through life. While both Lynn and Vanhanen followed up their initial study
with another book in 2006 and further subsequent research, not much has

been delved on when it comes to the topic of IQ. As stated earlier, there are
not enough studies that focus on the average IQ score for a country due to
the complex systems that go on in determining what intelligence is, many
psychologists still debate on an exact definition of it. Furthermore, there is
also the problem of the correct measure of intelligence and whether IQ is a
good enough measure to suffice in determining other economic factors.
Nevertheless, the research of Lynn and Vanhanen proves that there is a
correlation in national IQ and gross domestic product per capita; where there
are positive benefits to IQ when the output per worker and the economic
growth is high.

Body Mass Index

With the increasing obesity levels in the world, there have been many
studies that have shown that there are negative effects of this on different
human abilities. A study entitled The Impact of Obesity and Exercise and
Cognitive Aging (Chan, Yan, Payne, 2013) showed how the United States has
been experiencing this effect of obesity. In their study, the researchers say
obesity exacerbates age-related declines in a range of cognitive abilities
(e.g., executive function, memory, and processing speed). With a decline in
basic functions of the brain, it can be seen the health risks of having high
body masses BMIs above 25 kg/m2. Another study done entitled Elevated
body mass index is associated with executive dysfunction in otherwise
healthy adults, studied the effect that a high BMI had on cognitive function

and if this varied as a function of age i.e. higher body mass indexes as you
get older declines your brains abilities more. What the researchers were able
to see as that, despite the age function was not a significant factor, there is a
negative relationship between high body mass indexes with a decline in
cognitive ability across all ages (Gunstad et al., 2006). Declining cognitive
ability can lead to lower performance in standard IQ tests, suggesting a lower
level of general intelligence. This is confirmed by a study published in
Obesity, A Research Journal by Ehrenstein, Mnster, Milstein, Adler, and
Srensen in 2015. Their study was conducted to compare the BMI levels and
IQ scores of 37414 Danish conscripts. Their study showed that there were
mean differences (95% confidence interval) in BPP score were 0.6 (1.1;
0.1) for underweight men; 0.8 (1.1;0.5) for overweight men; and 2.0
(2.4;1.5) for men with obesity, where BPP score is the score from the
Brge Prien's Prve group intelligence test. The scores show a significant 2
point decrease for Danish men who had aboce average body mass indeces.
The studies show that there exists some correlation between body mass
index and average intelligence quotient in countries.

Population Growth
Population Growth is defined by the World Bank as the average
increase in the population from the year before. This set of data has many
economic implications, especially in the field of Macroeconomics. Paul
Romer, in his article published in the Journal of Political Economy in 1986

entitled Increasing Returns and Long Run Growth, he pioneered a growth

model that looked o see the effects of technological progress and how
research and development advanced economic growth in countries. In his
model, he was able to isolate the effect in his equation for the growth rate in
the stock of ideas for his model. The equation is as follows:
gA=

n
1

This equation shows how the growth rate of the stock of ideas in an
economy is a function of population growth,

n . Using this equation, we can

see that increases in the population would lead to an increase in the average
growth of the stock of ideas in an economy. When this happens, there is a
subsequent increase the number of total researchers as well in the economy.
With both the stock of ideas and the number of researchers increasing in an
economy, the total technological progress of an economy can be said to
increase by the Romer Model.
Therefore, we can see that an increase in population will lead to
increases in the total stock of knowledge in the economy, creating more
technological progress in the economy. With the stock of knowledge
increasing, we can infer that the intelligence quotient also increases because
the more knowledge available in the economy leads to the increase in the
average level of knowledge that people have, which is their general level of
intelligence or IQ.

Mean Years of Schooling

Studying the factors that determine intelligence quotient has many
economic implications. Macroeconomists have long discussed the problem of
increasing the productivity in workers and one way to do that is to increase
human capital. Gary Becker explains that by increasing the factors that
create more skilled workers, such as education, a countrys level of
technology and economy will be able to grow. With a growing economy, more
and more investments can be made towards increasing human capital and
consequently, the general level of intelligence of the citizens. This approach
differs from the Romer Model previously discussed because the previous
model assumes that increases in population will lead to immediate increases
in the number of researchers, who in turn will increase the number of new
ideas in an economy. While that is also the case here, another way to
interpret the model is that by directly increasing the level of human capital,
h , in the Solow Model where Human Capital is added, which can be seen in
the equation below:

sk
1

+ g+n
y = Ah

The equation shows how the steady state level of output per worker,
y , can be directly increased by a positive increase in coefficient of human
capital. By following this model, increasing human capital by, for example

10

increasing the level of intelligence of all the workers, will directly affect the
output per person in that country by providing the country with more skilled
workers who will contribute better to the economy. This event can only occur
if there is a definite correlation between the mean years of schooling in a
country with the average level of intelligence in the country (Mankiw, Romer,
& Weil, 1992).

Theoretical Framework

Implicit Theories of Intelligence

The Implicit Theories of Intelligence were developed by Carol Dweck
and they describe the process that people take when the form their own
beliefs about themselves and how these beliefs create certain patterns
within the person. In their study, Dweck was able to show two kinds of views
on intelligence, the entity view and the incremental view. The entity view is a
mindset where people treat intelligence as fixed and stable. Believers of this
usually have a high desire to prove themselves. The incremental view is the
frame of mind where intelligence is though of as malleable or fluid. Students
here seek satisfaction from the process of learning and focus more on the
journey rather than the end means (Dweck, Chiu, & Hong, 1995). The
incremental view of intelligence is to be focused on for this paper because it
shows how people can easily gain knowledge, and thus increase their IQ,

11

when they put a specific focus on gaining that knowledge itself and not
attaining that intelligence for personal gain.

Augmented Solow Model

The Augmented Solow Model, as discussed earlier provides the
framework for changes in both population growth and the stock of
knowledge in a country. An increase in the population growth reduces the
model, which implies that all things equal, the investments into human
capital will not be enough to compensate for the increased number of
people. The stock of knowledge has a positive effect on the total output per
worker in an economy (Mankiw, Romer, & Weil, 1992). By increasing the level
of skill that each worker has, by increasing the average intelligence of each
worker ceteris paribus, the output of each worker increases due to the more
skilled workers in the economy increasing the technological progress of the
country.

Operational Framework

Description of the Variables

Describing each variable is necessary before the initial regression to
determine their relationship with one another is done. This is to better
understand what each variable is measuring and how they may or may not
affect each other. In this analysis, one dependent variable (regressand) and

12

four independent variables (regressors) will be used. The regressand is the

National IQ Scores of 183 countries, which are the average IQ scores
determined by Richard Lynn and Tatu Vanhanen. In their book IQ and the
Wealth of Nations, the researchers gathered the different IQ scores across
all ages and were able to obtain the national averages. The regressors in this
study are Gross Domestic Product per Capita, Body Mass Index, Population
Growth, and Mean Years of Schooling. The table below will show each
variable, its label in Stata, a basic definition, and each variables unit of
measurement.

Variable
National IQ

Gross Domestic
Product (GDP)
Per Capita

Body Mass

Label

Definition
Regressand
nationaliq
This variable will
be used as a
measure of the
average
intelligence level
of a country
Regressors
gdppercap
The variable is
the measure of
the amount of
goods and
services
produced in a
country divided
by its total
population
bmi
The variable is

Measurement
No specified unit

In current US
dollars

No specified unit

13

Index

Population
Growth

popgrowth

Mean Years of
Schooling

meanyears

the measure of a
persons total
body fat in
relation to their
height. In the
study, it will be
used to show the
health level of a
country
The variable is a
measure of how
much a
population
increases or
decreases from
the previous year
The variable is a
measure of an
average number
of years spent in
school for each
country

In Percent

In Years

Table 1. Description of Variables Used

Hypothesized Relationship of the Endogenous Variables

A-Priori Expectations of the Independent and Dependent Variables
Independent Variable

Relationship

Description

Gross Domestic
Product (GDP) Per
Capita

Increases in GDP per

capita will lead to
increases in the average
IQ level in countries
because average

14

Body Mass Index

Population Growth

Mean Years of
Schooling

Table 2. A-Priori Expectations

income levels will

increase which will allow
people to invest more in
education and other
activities that will boost
their IQ levels.
Increases in body mass
indexes lowers the
cognitive ability of the
citizens of a country,
thus lowering the
average IQ level.
An increase in the
population, ceteris
paribus, lowers the level
of investment that can
be made to human
capital. This creates less
skilled workers and a
lower stock of
knowledge in the
country, and a generally
lower IQ level.
An increase in the
average years of
schooling in a country
will increase the national
IQ level because these
are direct investments
to human capital. This
creates more skilled
workers and smarter
citizens in the process.

Hypothesized Econometric Model

Econometrics is defined as quantitative analysis of actual economic
phenomena based on the concurrent development of theory and

15

observation, related by appropriate methods of inference (Samuelson,

Koopmans and Stone, 1954). Using this field of study, economists are able to
use existing theories and prove them using the exact science of mathematics
with models. In this paper, a classic linear regression model with multiple
regressors will be employed. This model calculates for the relationship
between several predictor variables, which are the independent variables,
with one dependent variable. Moreover, the form of the equation is called a
linear-linear model, which will determine how an absolute change in one of
the independent variables creates an absolute change in the dependent
variable.
With the economic theories and intuition previously discussed, the
hypothesized econometric model is able to formulated, which can be seen
below:

nationaliq= 1 + 2 gdppercap 3 bmi 4 popgrowth+ 5 meanyears +ui

Methodology
Data
Name
Afghanistan
Albania
Algeria
Andorra
Angola
Antigua and
Barbuda
Argentina
Armenia
Australia

nationaliq
84
90
83
98
68

gdppercap
561.198
4094.36
4349.57
43722.1
4218.65

bmi
21.7
25.9
24.5
27.025
23

popgrowth
2.45842
-.496462
1.85044
-.960625
3.23595

meanyears
3.2
9.1
7.6
10.4
4.7

70

13017.3

26.655

1.07531

8.9

93
94
98

11460.4
3124.78
51800.9

28.6
25.6
27.6

.872126
-.157056
1.55572

9.8
10.8
12.6

16

Austria
Azerbaijan
Bahamas, The
Bahrain
Bangladesh
Barbados
Belarus
Belgium
Belize
Benin
Bhutan
Bolivia
Bosnia and
Herzegovina
Botswana
Brazil
Brunei
Bulgaria
Burkina Faso
Burundi
Cambodia
Cameroon
Canada
Cape Verde
Central African
Republic
Chad
Chile
China
Colombia
Comoros
Congo,
Democratic
Republic of
the
Congo,
Republic of
the
Costa Rica
Cote d'Ivoire
Croatia
Cuba
Cyprus
Czech
Republic
Denmark
Djibouti
Dominica
Dominican
Republic
Ecuador
Egypt
El Salvador
Equatorial
Guinea

100
87
84
83
82
80
97
99
84
70
80
87

46590.6
5842.81
21941.9
20546
762.804
15812.3
5818.85
44358.3
4527.34
690.002
2211.34
1934.67

26.4
26.2
28.225
27.4
20.2
27.82
27
25.1
28.415
23.4
22.855
27.6

.240394
1.18979
1.68006
4.91075
1.07933
.49696
-.178976
1.13995
2.48663
2.8696
1.74428
1.61984

10.8
11.2
10.9
9.4
5.1
9.4
11.5
10.9
9.3
3.2
2.3
9.2

90

4380.6

25.8

-.195263

8.3

70
87
91
93
68
69
91
64
99
76

6980.36
10978.3
30880.3
6580.81
578.669
219.53
782.619
1145.37
47465.3
3413.26

25
26
23.53
25.7
21.7
21.4
22.3
24.6
26.8
24.24

.89905
.884609
1.55204
-.658275
2.90742
3.37119
1.54835
2.5556
1.11396
.387748

8.8
7.2
8.7
10.6
1.3
2.7
5.8
5.9
12.3
3.5

71

456.563

21.5

1.94232

3.5

68
90
100
84
77

909.3
12681.8
4433.34
6179.77
756.811

21.9
27.8
24
26.6
22.27

3.02686
.931579
.48296
1.39245
2.51781

1.5
9.8
7.5
7.1
2.8

78

346.704

20.8

2.77964

3.1

76

2920.41

22.1

2.87601

6.1

89
69
90
85
91

7773.19
1311.33
13500.9
5701.96
27889

26.8
23.185
26
27.1
26.1

1.47258
1.99723
-.25539
-.062542
1.19697

8.2
4.3
10.8
10.2
11.3

98

19764

26

.291362

12.3

98
68
67

57647.9
1353.19
6926.85

24.9
23.905
26.69

.444197
1.47004
.240569

12.1
3.8
7.7

82

5295.4

27

1.33193

7.3

88
81
80

4636.69
2803.53
3444.46

25.2
28.4
25.4

1.64432
1.67996
.55978

7.6
6.4
6.3

59

16638.1

23.6

2.81384

5.4

17

Eritrea
Estonia
Ethiopia
Fiji
Finland
France
FYROM
Gabon
Gambia, The
Georgia
Germany
Ghana
Greece
Grenada
Guatemala
Guinea
Guinea-Bissau
Guyana
Haiti
Honduras
Hungary
Iceland
India
Indonesia
Iran
Iraq
Ireland
Israel
Italy
Jamaica
Japan
Jordan
Kazakhstan
Kenya
Kiribati
Kuwait
Kyrgyzstan
Laos
Latvia
Lebanon
Lesotho
Liberia
Libya
Lithuania
Luxembourg
Madagascar
Malawi
Malaysia
Maldives
Mali
Malta
Marshall
Islands
Mauritania
Mauritius

85
99
69
85
99
98
91
64
66
94
99
73
92
71
79
67
67
87
67
81
98
101
82
87
84
87
92
95
102
71
105
84
94
80
85
86
90
89
98
82
67
67
83
91
100
82
69
92
81
74
97

368.748
14629.6
343.69
3649.38
46202.4
40706.1
4442.3
9362.11
566.348
2613.76
41723.4
1326.09
26861.5
7365.67
2882.39
435.449
534.148
2873.95
669.187
2078.33
12958.5
41696
1417.07
2946.66
5674.92
4473.71
47900.8
30550.9
35875.7
4917.02
42909.2
4370.72
9070.65
977.779
1539.05
38584.5
880.038
1122.85
11446.5
8755.85
1083.02
326.604
12375.4
11852.2
102857
414.143
359.58
8754.24
6552.48
673.695
19695.3

20.2
24.3
20.3
27.94
26.4
24.5
25.2
24.3
21.9
25
27.1
23.4
27.5
26.24
27.4
22.8
21.7
25.125
24.1
24.8
25.5
26.605
21.7
21.9
25.9
25
25
26.5
25.1
27.3
22.5
27.2
24.3
21.6
30.26
29.5
24.2
24.5
25
26
25.8
24
25.8
25.5
26.76
21.9
22.2
23.4
24.845
22.8
27.395

3.24427
-.228058
2.62588
.94336
.457495
.493689
.0789
2.41068
3.16185
.945436
-.153198
2.38307
-.301076
.364641
2.48923
2.63447
2.27699
.647152
1.33508
2.00602
-.226014
-.143903
1.29285
1.33412
1.24235
2.64952
.544884
1.82675
.307591
.351364
.017961
2.19054
1.41224
2.68667
1.51641
4.84469
1.19286
2.01757
-2.08131
2.1931
.952694
3.5109
1.27094
-2.09694
1.82541
2.80877
2.9769
1.73197
1.87004
3.09817
.491183

3.4
12
2.2
9.9
10.3
11
8.2
7.4
2.8
12.1
12.9
7
10.2
8.6
4.9
1.6
2.3
8.5
4.9
5.4
11.3
10.4
4.4
7.4
7.8
5.6
11.6
12.4
10.1
9.6
11.5
9.9
10.4
6.3
7.8
6.8
9.3
4.6
11.5
7.9
5.9
3.9
7.5
12.4
11.3
5.2
4.2
9.5
5.8
2
9.4

84

3126.52

30.37

.16608

11.7

76
89

977.151
7772.1

25.1
25.3

2.62012
.237887

3.7
8.4

18

Mexico
Micronesia,
Federated
States
Moldova
Mongolia
Morocco
Mozambique
Myanmar (Bur
ma)
Namibia
Nepal
Netherlands
New Zealand
Nicaragua
Niger
Nigeria
Norway
Oman
Pakistan
Panama
Papua New
Guinea
Paraguay
Peru
Philippines
Poland
Portugal
Qatar
Romania
Russia
Rwanda
Saint Kitts and
Nevis
Saint Lucia
Saint Vincent
and the
Grenadines
Samoa
Sao Tome and
Principe
Saudi Arabia
Senegal
Serbia
Seychelles
Sierra Leone
Singapore
Slovakia
Slovenia
Solomon
Islands
Somalia
South Africa
South Korea
Spain

88

8920.69

28.4

1.24935

8.3

84

2838.45

29.67

-.350671

8.8

96
101
84
64

1631.54
2285.65
2822.73
424.135

24.6
27.3
24.7
22

-.099865
1.50477
1.16277
2.56199

9.8
8.3
4.4
3.2

87

799.5

24.3

.755265

74
78
100
99
81
69
84
100
83
84
84

5177.68
595.772
50338.3
32975.6
1535.19
359.801
2310.86
86096.1
20922.7
1023.2
7833.9

22.6
20.8
25
28.2
27.5
22.1
23.1
25.3
24.9
22.6
25.6

1.64118
1.12782
.512923
1.11173
1.36407
3.78802
2.74655
1.24567
5.10701
1.7803
1.70781

6.2
3.2
11.8
12.5
5.8
1.4
5.2
12.6
6.8
4.6
9.4

83

1416.72

23.6

2.27256

3.9

84
85
86
99
95
78
94
97
70

3100.84
5075.48
2135.92
12484.1
22538.7
71510.2
8139.15
10709.8
525.855

25.3
27.3
22.9
25.1
25.8
26.8
24.1
25.4
21.6

1.75435
1.12903
1.68127
.08405
.04591
11.2153
-.593959
.33506
2.87447

7.7
8.8
8.9
11.7
7.8
8.9
10.6
11.7
3.3

67

13227

29.375

1.19329

8.4

62

7014.2

25.95

1.2462

8.3

71

6231.71

26.515

.061309

8.6

88

3456.77

31.98

.714803

10.3

67

1127.98

24.205

2.83857

4.7

84
76
89
86
91
108
96
96

19326.6
998.6
5399.3
10842.8
448.222
46569.7
16509.9
23417.6

27.3
22.8
25.9
26.745
23.8
22.7
26.2
26.5

1.70946
2.84885
-.402006
2.79233
1.94498
1.77066
.093191
.436079

8.5
4.5
9.5
9.4
2.9
10.1
11.6
11.8

84

1294.69

28

2.20537

4.5

68
77
106
98

111.2
7389.96
20976.5
30736

21.5
26.3
25.3
25.7

2.68532
1.45987
.463176
.460408

6.9
9.6
11.8
9.5

19

Sri Lanka
Sudan
Suriname
Swaziland
Sweden
Switzerland
Syria
Tajikistan
Tanzania
Thailand
Togo
Tonga
Trinidad and
Tobago
Tunisia
Turkey
Turkmenistan
Uganda
Ukraine
United Arab
Emirates
United
Kingdom
United States
Uruguay
Uzbekistan
Vanuatu
Venezuela
Vietnam
Yemen
Zambia
Zimbabwe

79
71
89
68
99
101
83
87
72
91
70
86

2400.02
1439.52
8321.39
3261.6
52076.3
74276.7
2808.1
739.732
707.927
4802.66
503.162
3546.78

20.7
22.5
26.67
24.8
25.4
26
25.8
24.3
22.4
23.6
22.8
32.605

.98777
2.2661
.916062
1.64528
.852525
1.04156
2.35467
2.38729
3.01012
.188395
2.59534
.521058

10.8
3.1
7.7
7.1
11.7
12.2
6.6
9.9
5.1
7.3
5.3
9.4

85

15630.1

28.6

.420809

10.8

83
90
87
84
97

4212.15
10135.7
4392.72
553.263
2974

25.9
26.3
25
21.8
24.9

1.02447
1.2505
1.25804
3.35966
-.397285

6.5
7.2
9.9
5.4
11.3

84

33885.9

28

8.95678

9.1

100

38363.4

27

.783889

12.3

.829343
.343146
2.82285
2.34034
1.59737
1.0494
2.37109
3.01039
1.44854

12.9
8.3
10
9
8.6
5.5
2.5
6.5
7.2

98
96
87
84
84
94
85
79
82

48377.4
28.8
11530.6
27.3
1377.08
24.9
2965.75
27.61
13559.1
27.8
1333.58
21.2
1394.53
22.9
1533.28
21.5
723.165
23.7
Table 3. Dataset

The data set employed in this research is a cross-sectional collection

consisting of data from the year 2010. Cross-sectional data is defined as a
set of data taken from one point in time. In the data set, the observations of
National Intelligence Quotients, Gross Domestic Product per Capita, Body
Mass Index, Population Growth, and Mean Years of Schooling for 183
countries are included. The data for these variables were collected from
various sources, such as the book IQ and the Wealth of Nations by Lynn

20

and Vanhanen, The World Bank, and The United Nations Development
Program.

Empirical Estimation and Inference Procedures

For the study, a regression analysis will be undertaken. Regression was
first coined by Francis Galston in his study of how the heights of children all
regressed towards an average. Currently, a modern definition is widely
accepted and it is defined below:
Regression analysis is concerned with the study of the
dependence of one variable, the dependent variable, on one or
more other variables, the explanatory variables, with a view to
estimating and/or predicting the (population) mean or average
value of the former in terms of the known or fixed (in repeated
sampling) values of the latter. (Gujarati, 2003)

Simply put, this paper is trying to see what effects GDP per Capita,
Body Mass Index, Population Growth, and Mean Years of Schooling have on
National Intelligence Quotient. The regression will help us see if the sample
data can be used to explain what is happening with the true population i.e.
the effects of the independent variables on the dependent variable holds
true in all cases.
Before the initial regression, it is apparent to address the problem of
inference. In order to do this, we must validate the a-priori expectations and

21

the fit of the model. By doing this, we can show whether or not each variable
truly has a statistical significance to the model. To validate the a-priori
expectations, the study will use a 95% confidence interval, which is the
standard for most social sciences as compared to a 99% confidence interval
used in the fields of medicine, and the like. If the regressors have p-values
that are less than 0.05, than the variables are deemed significant. When they
are deemed significant, then it is possible to see whether or not the variables
meet the a-priori expectations. To test the how well the model fits, we
interpret the R2 of the regression. R2 is known as the coefficient of
determination, and it measures the percentage of the total variation in the
dependent variable explained by the regression model (Gujarati, 2003).
Having a good R2 entails that model can be well explained by the
independent variables, or how well the line fits for the regression.
For this study, a software called Stata will be employed to conduct a
multiple regression of the equation. In simpler terms, it tries to fit the best
line to a set of data points. The basis for this regression is the classical linear
regression model, which is based on a set of ten assumptions. Using these
assumptions, the researcher looks to find the estimators that are BLUE, or
the Best Linear Unbiased Estimators. In order to do this, the researcher will
regress the model using the Ordinary Least Squares Method a method
attributed to Carl Friedrich Gauss (Gujarati, 2003). The Ordinary Least
Squares looks to minimize the sum of the square residuals. Graphically, this

22

is shown by the attempt to minimize the sum of the distances between the
data points and the fitted line.
After the initial regression, more tests will be done to see if the model
violates any of the assumptions of a classical linear regression model. If any
assumptions are detected, the Ordinary Least Square application to the
equation will not yield the Best Linear Unbiased Estimators. Specific tests will
be employed in the latter part of the paper in order to see if model violates
any of the CLRM assumptions. The following tests that will be used in this
study are as follows: Test for Heteroscedasticity, Test for Multicollinearity,
and the Test for Misspecification Biases. In the case that model shows any of
these violations, corrective measures will be done to the model in order to
counter the effects of these violations.

Econometric Model

Initial Regression
Utilizing the statistical software Stata, The Ordinary Least Square
Method was able to be used. The results are as follows:

Source |
SS
df
MS
Number of obs =
183
-------------+-----------------------------F( 4, 178) = 59.13
Model | 12248.1144
4 3062.02859
Prob > F
= 0.0000
Residual | 9218.2135 178 51.7877163
R-squared
= 0.5706
-------------+-----------------------------Adj R-squared = 0.5609
Total | 21466.3279 182 117.946856
Root MSE
= 7.1964
23

-----------------------------------------------------------------------------nationaliq |
Coef. Std. Err.
t P>|t|
[95% Conf. Interval]
-------------+---------------------------------------------------------------gdppercap | .0001765 .0000393
4.49 0.000
.0000989 .0002541
bmi | -.4545519 .281449 -1.62 0.108 -1.009958 .1008543
popgrowth | -1.742595 .4279474 -4.07 0.000 -2.587098 -.8980917
meanyears | 1.711605 .2939906
5.82 0.000
1.13145 2.291761
_cons | 83.1226 6.329179 13.13 0.000
70.63272 95.61248
------------------------------------------------------------------------------

By running the initial regression on Stata, the researcher was also able
to come up with the coefficients for the hypothesized econometric model.
Substituting the coefficient values with their corresponding variables in the
model, we get:

nationaliq=83.1226+0.0001765 gdppercap0.4545519 bmi1.742595 popgrowth+1.711605 meanyears

To address the problem of inference, the level of significance used will

be equal to 5%, as discussed earlier. Again, if the p-values of the estimates
are lower than the level of significance, there is statistical significance and
the value of the beta-coefficient is significantly different from zero, which
means we reject the null hypothesis. The opposite is true for p-values that
are greater than 0.05, we accept the null hypothesis and say that the
variable has no statistical significance and it has a beta-coefficient value that
is equal to 0. The initial regression shows that three of the regressors, GDP

24

per Capita, Population Growth, and Mean Years of Schooling, are statistically
significant while the variable for Body Mass Index was deemed insignificant.
Looking at the intercept coefficient first, we see that the value is
83.1226. In the model, this is our value for

1 . The intercept coefficient is

the value at which the dependent variable starts on the Y-axis, if you look at
a graph. This implies that, when all of the independent variables are equal to
0, ceteris paribus, then national IQ will have a value of 83.1226 units. On top
of this, the p-value is shown to be equal to 0.000, which is lower than the
level of significance at 0.05 or 5%. This means that there is strong evidence
to reject the null hypothesis, which means we reject that

1=0

and

conclude that the intercept coefficient of the model is not equal to 0 and is
significant to the model. One way to interpret the coefficient in real terms,
however is to say that without the independent variables included in the
model, average IQ in a country may be equivalent to 83.1226 points. This IQ
level in an adult would classify him as below average category, being one
standard deviation, or 15 IQ points away from the mean value of 100
(Minton, 1988). This has some real world application, saying that people
without these factors in life still regress to a certain IQ value or saying if
the independent variables had no affects on a person, his or her IQ would
still be around that 83 points mark. However, statistically interpreting the
intercept coefficient is thought to bear no meaning.
Based from the regression results, GDP per capita is shown to be
significant because its p-value is at 0.000, which means we reject the H0 or

25

the null hypothesis. This shows that the variable is statistically significant to
the model and it has a value that is significantly different from 0. Because
the coefficient of this variable is 0.0001765, this means that GDP per capita
positively affects the National IQ variable, which is consistent with our apriori expectations. To interpret this further, a one-unit increase in a
countrys GDP per Capita also increases the National IQ by 0.0001765 points.
Unfortunately, Body Mass Index did not meet the p-value that would
have made the variable significant. Because the p-value is 0.108, it is does
not fall under the region of being less than the level of significance i.e. less
than 0.05. This means that there is a not a strong argument against the null
hypothesis and we do not reject it, which means that the variable is
insignificant. The a-priori expectations were met since the coefficient of BMI
is negative; meaning it negatively affects National IQ. To interpret the
coefficient, we can say that a unit increase in Body Mass Index of a country
will decrease the National IQ by 0.4545519 points. The insignificance is likely
a result of the Body Mass Index not being the ideal indicator for this model.
As stated earlier in the Scope and Limitations portion of the paper, Body
Mass Index has been used as a measure for the overall health level for a
country. However, for determining IQ, the current BMI of a country would not
seem to fit because previous studies have shown that childrens BMI has an
effect on IQ later in life.
With that being the case, it has confirmed the insignificance of the
variable. Maybe for future reference, another health indicator should be used

26

in the place of the current average Body Mass Index of the different countries
to see what affects IQ in study with cross-sectional data. Despite being
insignificant, not meeting the a-priori expectation has provided some
information about the variable. A study published in the American Journal of
Epidemiology concluded that there is a correlation between IQ and Obesity in
an early age, where children that are obese at an early age show signs of
decreased cognitive ability (Belsky, 2013). However, this may not be the
case when adulthood is taken into consideration, since there may be more
significant factors that affect IQ scores in adults.
The last two variables in the regression, Population Growth and Mean
Years of Schooling, have the coefficient values of -1.742595 and 1.711605
respectively. These values are in line with the a-priori expectations of each
variable, where population growth has a negative effect on national IQ scores
and the average years of schooling has a positive correlation with an
average persons IQ. Population growth is shown to be significant to the
model because its p-value is less than the 5% level of significance. Because
of this, we can reject the null hypothesis that says that the beta-coefficient of
Population Growth is significantly different from zero and is significant to the
regression. We also can see that by increasing the population growth in a
country will lower the average intelligence quotient of the country by
1.742595 points, ceteris paribus. This is in line with the initial assumption
that population growth will negatively affect the national IQ. When high
population growth occurs, there is a decrease in the amount that can be

27

potentially invested in human capital because there are too many people to
compensate for, all else the same. The result of this phenomenon is a lower
economic growth, as well as lower investment in human capital the
probable cause of a decrease in national intelligence qutioent, as stated by
the United Nations Population Fund (UNFPA, 2014). The regression shows as
well that the mean years of schooling is significant with a p-value that is
0.000. Because of this, we again can reject the null hypothesis and say that
this variable is sttistically significant to the model and has a value that is
significantly different from 0. The positive relationship is shown because of
the positive coefficient, meaning that if the mean years of schooling in a
country increased by one year, average IQ would increase by 1.711605
points. This is in line with the a-priori expectations because more years of
school should, in all likelihood, increase the average intelligence quotient of
a person, especially since IQ tests are based on what an average person
should known and have learnt given their respective age.
In the upper-right hand portion of the regression figure above, we can
see other properties of the regression that will also be interpreted. The
important things to note in this portion of the regression figure are the RSquared and the Adjusted R-Squared values. As stated before, the R2 is the
coefficient of determination and it shows the explanatory power of the
model. For the model, we have an R2 value of 0.5706. This means that the
model can explain 57.06% of the total variations in the dependent variable,
in this case National IQ. The closer that the R2 is to 1, the better the fit of

28

the model and a 57% value for R2 shows a good relationship between
National Intelligence Quotient and the dependent variables. However, for
multiple regression, there is a better measure that can be used to determine
the goodness of fit. The Adjusted R2 is another measure for the overall fit of
the model, however it adjusts for the degrees of freedom used when more
and more independent variables are added. This is to counteract the flaw of
R2 where when more independent variables are added, the higher the R2
becomes, which may not necessarily capture the true effects of each
independent variable to the regressand. In the regression, the adjusted R2
has a value of 0.5609, which means that 56.09% of the variations in National
Intelligence Quotient can be explained by the model. Even after adjusting for
the degrees of freedom, the Adjusted R2 still shows a good fit between the
model and the variables.

Test for Overall Significance

To test for the overall significance of the model, the study will use the
Analysis of Variance, or simply the ANOVA table. The ANOVA table tests the
level of significance of the entire model using the F-statistic, which tests all
the variables simultaneously to see if they fit the model. The ANOVA is also
the study the components of the Total Sum of Squares, which is the Sum of
the Explained Sum of Squares and the Residual Sum of Squares. For this
study, if the p-value of the F-statistic is found to be greater than 0.05, then
the null hypothesis were to be accepted, meaning that the coefficients of all

29

of the explanatory variables are insignificant. If the p-value were to be less

than the critical region of 0.05, than the alternative hypothesis that not all of
the coefficients are insignificant would be accepted. In Stata, the ANOVA
table is immediately given when the regress command is used for the initial
regression. The results of the Analysis of Variance is as follows:

Source |
SS
df
MS
Number of obs =
183
-------------+-----------------------------F( 4, 178) = 59.13
Model | 12248.1144
4 3062.02859
Prob > F
= 0.0000
Residual | 9218.2135 178 51.7877163
R-squared
= 0.5706
-------------+-----------------------------Adj R-squared = 0.5609
Total | 21466.3279 182 117.946856
Root MSE
= 7.1964

The F-statistic is shown as F (4,178) = 59.13 which is the F-Statistic

computed the specific degrees of freedom for its numerator and
denominator, shown by the values of 4 and 178 respectively. For the model,
the F-statistic is 59.13. The Prob > F shows that p-value of the calculated Fstatistic and it was determined to be equal to 0.0000, which is significantly
less than 0.05. Therefore, the model can be said to have met the standards
of overall significance. In the next section of the paper, different tests will be
conducted to see if the current model specified violates any of the Classic
Linear Regression Model assumptions.

Test for Presence of Heteroscedasticity

30

The first test that will be conducted to the model is the test for
heteroscedasticity. Heteroscedasticity is the violation of the Classic Linear
Regression Model assumption and it assumes that the variance of the each
disturbance term
E ( u2i ) = 2

ui is not a constant value that or equal to

2 , or

(Gujarati, 2003). This is an important test, especially for this

model because it is one of the usual violations of the CLRM assumptions that
cross-sectional data suffers from. If heteroscedasticity is found in the model,
the Ordinary Least Square estimators will no longer be BLUE due to the
estimates having the same variance of the error term. This means that there
will be observations that contain large disturbance terms, which normally
dont say anything about the model, and gives it equal weight to the
observations with smaller

ui

terms, which provide more statistical

information about the model.

To formally test for heteroscedasticity, the study uses both the
Breusch-Pagan test and Whites Test for Heteroscedasticity. The results of
both tests can be seen below:

Breusch-Pagan / Cook-Weisberg test for heteroskedasticity

Ho: Constant variance
Variables: gdppercap bmi popgrowth meanyears
chi2(4)
=
5.77
Prob > chi2 = 0.2170

White's test for Ho: homoskedasticity

31

against Ha: unrestricted heteroskedasticity

chi2(14)
=
15.54
Prob > chi2 = 0.3424

Both tests show a Chi-square term for each value and show each pvalue for the respective Chi-square. For the model to be homocedastic, the pvalue needs to be greater than 0.05. Unlike the previous sections, the null
hypothesis is the one that needs to be accepted in order for the model to
show that there is no heteroscedasticity. If the alternative hypothesis were to
be accepted, by having p-values less than 0.05, then the model would suffer
from heteroscedasticity (both the null and alternative hypotheses can be
seen in the test results above). With both p-values being greater than the
critical region of 0.05, than we can accept the null hypotheses for both tests,
proving that this model does not suffer from the heteroscedasticity violation.
We can therefore conclude that the model is homoscedastic.

Test for Presence of Multicollinearity

Another important assumption for the Classic Linear Regression Model
is that there is no multicollinearity. Multicollinearity is defined as presence of
an exact linear relationship between any of the OLS estimators. Although
there is always some presence of multicollinearity in every regression model,
which will have no effect on the estimators being BLUE, the significance of
testing for it still is to see if there is a severe case of multicollinearity. If that
32

occurs, then the confidence intervals of the coefficients will tend to be very
wide; the t-statistics will all be insignificant; and the F-statistic and R2 values
will show an overall statistical significance for the model. Because there is no
heteroscedasticity in the model, no corrective measures will need to be
undertaken
To test for the presence of multicollinearity, The Variance Inflation
Factors will be calculated. Gujarati (2003) describes these as the speed at
which variances and covariances increase. The results of the test can be
seen below:

Variable |
VIF
1/VIF
-------------+---------------------meanyears |
2.84 0.352602
gdppercap |
1.69 0.590187
bmi |
1.53 0.655360
popgrowth |
1.45 0.690698
-------------+---------------------Mean VIF |
1.88

When using the Variable Inflation Factors to see whether or not there is
multicollinearity, the general rule is that any VIF that is greater than or equal
to 10 shows the existence of a severe multicollinearity. From the test we can
see that Mean Years of Schooling, GDP per Capita, Body Mass Index, and
Population Growth all have Variance Inflation Factors that are less than 10,
showing that no severe multicollinearity exists in the model. Also, because
most of the values are close to one, the model has a very tolerable case of
multicollinearity, which will not need corrective measures. Because it is such
33

a common violation, there are many ways to correct for it. The first option is
to not do anything to the model and leave it as is. This is a very viable
solution for most because it still leaves you with OLS estimators that are
Blue. Another way to correct for multicollinearity is to drop the variable that
has a VIF that is greater than 10, although this method must be used with
caution since dropping variables could lead to model suffering from the
omitted variable violation. Other methods to correct this violation are, but
not limited to: using a pooled cross-section and time series data set, using a
larger number of observations, or gathering and using the variables from
different data sources.

Test for Presence of Misspecification Bias

The last test that will be employed for this paper is the test for the
presence of misspecification bias. Misspecification occurs commonly when an
observation/s are correctly input in the data set or one or more important
variables are not included in the model, which need to be included or else
the model will be biased and inconsistent (Gujarati, 2003). To test for the
presence of an omitted variable, the Ramsey RESET (Regression
Specification Error Test) will be employed, which is a general test of
specification errors (Gujarati, 2003). This test will show whether or not there
exists an omitted variable bias in the regression model. Using the Stata
software, the results of the Ramsey RESET can be seen below.

34

Ramsey RESET test using powers of the fitted values of nationaliq

Ho: model has no omitted variables
F(3, 175) =
1.89
Prob > F =
0.1331
The test shows both of the formulated hypotheses, where the null
hypothesis entails that the model has no omitted variables and the
alternative hypothesis entails that there are omitted variables. In the results,
the p-value for the test is shown to be equal to 0.1331, which is significantly
greater than the critical region of 0.05. This means we accept the null
hypothesis and reject the alternative hypothesis, proving that there are no
omitted variables in this model.

Corrective Measures and Final Regression

In this section of the paper, the initial regression model will undertake
final adjustments before being presented as the final regression model. The
overall significance test using the ANOVA table and it was shown that the
model satisfies all of the conditions, proving that all of the coefficients are
statistically significant and are all significantly different from zero. The
regression table has also presented this by showing three out of the four
variables are shown to be significant to the model on their own. Furthermore,
the test for heteroscedasticity proved that the model is homoscedastic,
meaning the variance of each disturbance variable is a constant value or
equal to

2 . The test for multicollinearity showed that there is are no

variables that have a perfect linear relationship with each other and that not

35

severe multicollinearity exists in the regression. Lastly, the misspecification

bias test confirmed that no variables were omitted from the regression. After
running all diagnostic tests, the model has satisfied the conditions for all of
the aforementioned tests, the Ordinary Least Square estimators of the model
are all Best Linear Unbiased Estimators. The researcher can now conclude
that no corrective measures will need to be done to the model. Therefore,
the final regression and model will be identical to the initial regression:

Final Regression:
Source |
SS
df
MS
Number of obs =
183
-------------+-----------------------------F( 4, 178) = 59.13
Model | 12248.1144
4 3062.02859
Prob > F
= 0.0000
Residual | 9218.2135 178 51.7877163
R-squared
= 0.5706
-------------+-----------------------------Adj R-squared = 0.5609
Total | 21466.3279 182 117.946856
Root MSE
= 7.1964
-----------------------------------------------------------------------------nationaliq |
Coef. Std. Err.
t P>|t|
[95% Conf. Interval]
-------------+---------------------------------------------------------------gdppercap | .0001765 .0000393
4.49 0.000
.0000989 .0002541
bmi | -.4545519 .281449 -1.62 0.108 -1.009958 .1008543
popgrowth | -1.742595 .4279474 -4.07 0.000 -2.587098 -.8980917
meanyears | 1.711605 .2939906
5.82 0.000
1.13145 2.291761
_cons | 83.1226 6.329179 13.13 0.000
70.63272 95.61248
-----------------------------------------------------------------------------Final Model:

National IQ=83.1226+ 0.0001765 [ GDPperCapita ] 0.4545519 [ Body Mass Index ] 1.742595[ Population Grow

The final regression shows that GDP per Capita and Mean Years of
Schooling have positive effects on a countrys average IQ, whereas

36

Population growth negatively affects it. Moreover, there was only one
insignificant variable in the model, which was Body Mass Index. Even though
the variable went against its a-priori expectation, being contrary to it still
provided significant information in the study. The paper shows many
interesting findings, such as the strong correlations between National
Intelligence Quotient and the significant variables and also the insignificant
relationship of Body Mass Index and average IQ. However, they all have
given due support and justification to the research aforementioned in the
paper.

Conclusion

The purpose of this paper was to study and discover the factors that
affect the average IQ of a country, and try to explain the relationships
between the regressand and the regressors to the best of the researchers
ability. In order to predict the relationship, a linear-linear model was used
that contained four explanatory variables GDP per capita, body mass index,
population growth, and mean years of schooling which were used to study
their effects on the dependent variable. The regressand was the National IQ,
or the average IQ level for every country included in the study and this acted
as the proxy for the general level of intelligence in a given country. The
model used empirical testing, where a multiple regression analysis was done,
where the Ordinary Least Square estimators were found to be BLUE. The

37

model was tested for different violations of the Classic Linear Regression
Model Assumptions, such as tests for overall significance, presence of
heteroscedasticity, multicollinearity, and misspecification bias. The model
was found to not be in violation of any of these assumptions so no corrective
measures were undertaken. Because of this, the initial regression model was
used as the final regression as well. The regression included significant
variables, GDP per capita, Population Growth, and Mean Years of Schooling,
which showed they had significant relationships with the dependent variable.
The last explanatory variable, Body Mass Index, was deemed insignificant by
the regression, although it still provided insights to the research and possible
information for future research on the topic.
Overall, this topic was able to present new research to the field of
Economics. By understanding better the determinants of national IQ, policy
makers can make use of the research to create new policies that look to
increase the overall skill level of the workers in an economy. By doing so, this
could possibly reduce the wealth gap in countries and also provide more
education for the whole population. By looking at current economic theories
and employing econometric models, the paper was able to construct and
interpret in full the relationship between national IQ and the independent
variables.

Recommendations

38

For further research, three recommendat ions can be made to ensure a more in-dept h analysis on the effect s of certain factors on the national IQ. The first recommendation would be to use a pooled cross-sect ional time-series data set. By using this kind of data, the study would be able to show the trends of IQ across all the countries over a period of time to get a better look at the picture as a whole. Using this data set would also allow for more statist ical tests to be done in order to see if there are more variables that affect national IQ more than the variables already employed in this research paper.
Another recommendation would be to find a better healt h indicator for health for this model. Unfortunately, the researcher was not able to find a dat a set for a health indicator that was complete for all countries, as some do not keep tab of very specific stat istical measures. This is unfortunate because there are so many studies that relate that certain health factors could have an affect on the average IQ levels of a person. If future researchers may be able to obtain a better indicator of the healt h level of a population, the model would be able to find a variable that is significant to the model, which would help explain the true relat ionship between health levels in a country wit h their average national income.
Lastly , if a better measurement for average general int elligence were to be found, that future researchers would test that against different economic factors. Even though average IQ per country suffices as an index now, there are still many debates on the validity of the variable to representing the true populat ions level of intelligence. Finding a better indicator for general intelligence will help economist s see the true relationship between econom ic factors and intelligence, the im pacts of which can contain endless possibilities.

Bibliography

Belsky, D., (2013). Is Obesity Associated with a Decline in Intelligence

Quotient During the First Half of the Life Course? American Journal
of Epidemiology: Oxford Press. Retrieved from:
http://aje.oxfordjournals.org/content/early/2013/09/09/aje.kwt135

Chan, J., Yan, J., & Payne, V. (2013). The Impact of Obesity and Exercise on
Cognitive Aging. Frontiers in Aging Neuroscience. Retrieved from:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3869042/

Blanchard, O. & Johnson, D. (2013) Macroeconomics 6th Ed. Prentice Hall

39

Dweck, C., Chiu, C., & Hong, Y. (1995). Implicit theories and their role in
judgments and reactions: A world from two perspectives.
Psychological Inquiry.

Ehrenstein, V., Mnster, A., Milstein, A., Adler, N. E. & Toft Srensen, H.
(2015), Body mass index and cognitive function: Birth cohort effects
in young men. Obesity. doi: 10.1002/oby.21088

Gunstad, J., Paul, R., Cohen, R., Tate, D., Spitznagel, M., & Gordon, E. (2006).
Elevated body mass index is associated with executive dysfunction
in otherwise healthy adults. Journal of Comprehensive Psychiatry.
Retrieved from:
http://www.sciencedirect.com/science/article/pii/S0010440X0600065
4

Gujarati, D., (2003). Basic Econometrics. McGraw Hill Education

Lynn, R. & Vanhanen, T. (2002). IQ and the Wealth of Nations. Praeger

Publishers

Lynn, R. & Vanhanen, T. (2006). IQ and Global Inequality. Praeger Publishers

40

Mankiw, G., Romer, D., & Weil, D. (1992). A Contribution to the Empirics of
Economic Growth. Quarterly Journal of Economics, n.a.

Minton, H., (1988). Lewis M. Terman: Pioneer in Psychological Testing. New

York: New York University Press.

Romer, P. (1986). Increasing Returns and Long Run Growth. Journal of

Political Economy 94.
n.a.

Stern, W., (1912). The Psychological Methods of Intelligence Testing. n.a.

United Nations Deveopment Program, (2012). Mean Years of Schooling of

Adults, in Years. Human Development Report. Retreived from:
http://hdr.undp.org/en/content/mean-years-schooling-adults-years

United Nations Population Fund, (2014). Population and Poverty. UNFPA,

United Nations.

Vanhanen, T., (2011). National IQs and their demographic correlates.

Personality and Individual Differences (Research Journal). Elsevier.

41

Worldbank, (2014). Data on Population Growth and Gross Domestic Product

per capita. Country Indicators. Retreived from: data.worldbank.org

--------------------------------------------------------------------------------------------------------------------name: <unnamed>
log: /Users/renzoermitano/Desktop/RENZO 3RD YEAR COLLEGE/RENZO 3RD YEAR 3RD
TERM/ECONMET/ECONOMETRICS - FINAL
> PAPER.log
log type: text
opened on: 20 Apr 2015, 03:30:40
. use "/Users/renzoermitano/Desktop/RENZO 3RD YEAR COLLEGE/RENZO 3RD YEAR 3RD
TERM/ECONMET/ECONOMETRICS - FINAL PAPER
> .dta"
. summarize nationaliq gdppercap bmi popgrowth meanyears meanyears
Variable |
Obs
Mean Std. Dev.
Min
Max
-------------+-------------------------------------------------------nationaliq |
183 84.67213 10.86033
59
108
gdppercap |
183
12284.8 17667.53
111.2
102857
bmi |
183 25.11792 2.341196
20.2
32.605
popgrowth |
183 1.496042 1.499834 -2.096943 11.21527
meanyears |
183
7.83224 3.055638
1.3
12.9
-------------+-------------------------------------------------------meanyears |
183
7.83224 3.055638
1.3
12.9

42

. describe nationaliq gdppercap bmi popgrowth meanyears

storage display
value
variable name type format
label
variable label
--------------------------------------------------------------------------------------------------------------------nationaliq
int %8.0g
national iq averages per country
gdppercap
float %8.0g
gross domestic product per capita in current us$ for
the year 2010
bmi
float %8.0g
national body mass index averages for adults in 2010
popgrowth
float %8.0g
population growth of countries in the year 2010
meanyears
float %8.0g
average years of schooling of countries in 2010
. tabulate nationaliq gdppercap bmi popgrowth meanyears, col nofreq chi
too many variables specified
r(103);
. regress nationaliq gdppercap bmi popgrowth meanyears
Source |
SS
df
MS
Number of obs =
183
-------------+-----------------------------F( 4, 178) = 59.13
Model | 12248.1144
4 3062.02859
Prob > F
= 0.0000
Residual | 9218.2135 178 51.7877163
R-squared
= 0.5706
-------------+-----------------------------Adj R-squared = 0.5609
Total | 21466.3279 182 117.946856
Root MSE
= 7.1964
-----------------------------------------------------------------------------nationaliq |
Coef. Std. Err.
t P>|t|
[95% Conf. Interval]
-------------+---------------------------------------------------------------gdppercap | .0001765 .0000393
4.49 0.000
.0000989 .0002541
bmi | -.4545519 .281449 -1.62 0.108 -1.009958 .1008543
popgrowth | -1.742595 .4279474 -4.07 0.000 -2.587098 -.8980917
meanyears | 1.711605 .2939906
5.82 0.000
1.13145 2.291761
_cons | 83.1226 6.329179 13.13 0.000
70.63272 95.61248
-----------------------------------------------------------------------------. vif
Variable |
VIF
1/VIF
-------------+---------------------meanyears |
2.84 0.352602
gdppercap |
1.69 0.590187
bmi |
1.53 0.655360
popgrowth |
1.45 0.690698
-------------+---------------------Mean VIF |
1.88
. estat hettest
Breusch-Pagan / Cook-Weisberg test for heteroskedasticity
Ho: Constant variance
Variables: fitted values of nationaliq
chi2(1)
=
3.81
Prob > chi2 = 0.0509

43

. hettest
Breusch-Pagan / Cook-Weisberg test for heteroskedasticity
Ho: Constant variance
Variables: fitted values of nationaliq
chi2(1)
=
3.81
Prob > chi2 = 0.0509
. hettest gdppercap bmi popgrowth meanyears meanyears
Breusch-Pagan / Cook-Weisberg test for heteroskedasticity
Ho: Constant variance
Variables: gdppercap bmi popgrowth meanyears
chi2(4)
=
5.77
Prob > chi2 = 0.2170
. estat imtest, white
White's test for Ho: homoskedasticity
against Ha: unrestricted heteroskedasticity
chi2(14)
=
15.54
Prob > chi2 = 0.3424
Cameron & Trivedi's decomposition of IM-test
--------------------------------------------------Source |
chi2
df
p
---------------------+----------------------------Heteroskedasticity |
15.54
14 0.3424
Skewness |
12.48
4 0.0141
Kurtosis |
6.23
1 0.0126
---------------------+----------------------------Total |
34.24
19 0.0172
--------------------------------------------------. ovtest
Ramsey RESET test using powers of the fitted values of nationaliq
Ho: model has no omitted variables
F(3, 175) =
1.89
Prob > F =
0.1331
. log close
name: <unnamed>
log: /Users/renzoermitano/Desktop/RENZO 3RD YEAR COLLEGE/RENZO 3RD YEAR 3RD
TERM/ECONMET/ECONOMETRICS - FINAL
> PAPER.log
log type: text
closed on: 20 Apr 2015, 03:33:28
---------------------------------------------------------------------------------------------------------------------

44