tiểu luận kinh tế lượng FACTORS AFFECT THE AMOUNT OF CO2 EMISSIONS IN DEVELOPING COUNTRIES IN 2014

FOREIGN TRADE UNIVERSITY
The faculty of International Economics
-----šš&šš-----

ECONOMETRICS ASSIGNMENT
FACTORS AFFECT THE AMOUNT OF CO2 EMISSIONS IN
DEVELOPING COUNTRIES IN 2014

STUDENT NAME – ID:
Nguyễn Thu Ngân – 1815520207
Nguyễn Cao Hoài Anh – 1815520151
Nguyễn Minh Anh – 1810520152
CLASS: E6 + E7 – K57 JIB

SUPERVISOR: Dr. Tu Thuy Anh

Hanoi, October, 2019


INDEX

PREFACE................................................................................................................................................... 1
Chapter 1: LITERATURE REVIEW........................................................................................... 2
1.1

Theories................................................................................................................................. 2

1.2

Empirical researches...................................................................................................... 4

Chapter 2: METHODOLOGY........................................................................................................ 7
2.1 Methodology used.................................................................................................................. 7
2.2 Constructing econometrics model.................................................................................. 7
2.3 Data overview.......................................................................................................................... 9
2.4 Data description...................................................................................................................... 9
Chap 3: TEST ASSUMPTIONS & STATISTICAL INFERENCES...........................11
3.1 Model 1..................................................................................................................................... 12

3.2 Model 2..................................................................................................................................... 15
3.3 Model 3..……………………………………………………………………………16

Chapter 4: RESULT ANALYSIS AND POLICY IMPLICATIONS........................... 17
4.1 Result analysis....................................................................................................................... 17
4.2 Policy implication................................................................................................................ 18
CONCLUSION...................................................................................................................................... 19
APPENDIX.............................................................................................................................................. 20
REFERENCES...................................................................................................................................... 26


PREFACE

In the last three decades, the threat of global warming and climate change has
been the major on-going concern for all societies from developing countries to
developed countries. The principle reason is originated from greenhouse gas effect, of
which Carbon dioxide (CO2) is regarded to be the main source. Global climate change
has altered water supplies and weather patterns, changed the growing season for food
crops and threatened coastal communities with increasing sea levels. According to our
research, in recent years, developing countries is responsible for more than 60% of
CO2 emissions due to industrialization and urbanization.
Facing the challenge to find out solutions to balance between sustainable economic
development and harmfulness to the environment, our group decide to examine “Factors
affect the amount of CO2 emissions in developing countries in 2014”.

In the report, we will apply what we learn in Econometrics course to investigate
into this matter. The report is divided into 4 main parts:
Chapter 1: Literature review about the relationship between CO2 emissions and GDP
per capita, population growth, energy use.
Chapter 2: Methodology
Chapter 3: Statistical inferences and test assumptions
Chapter 4: Result analysis and policy implication
Finally, we would like to express our sincere thanks to the dedicated guidance
from Mrs. Tu Thuy Anh and Mrs. Chu Thi Mai Phuong.
Due to our limited knowledge, there are certainly some deficiencies in our
report. We look forward to receive comments and suggestions from you to make our
research more completely.

1


Chapter 1: LITERATURE REVIEW
1.1 Theories
1.1.1 CO2 emissions (metric tons per capita)
1.1.1.1 Definition and roles of CO2 (Carbon dioxide)
Carbon dioxide (chemical formula CO2) is a colorless gas with a density about
60% higher than that of dry air. Carbon dioxide consists of a carbon atom covalently
double bonded to two oxygen atoms. It occurs naturally in Earth's atmosphere as a
trace gas.

CO2 one of the most important gases on the earth because plants use it to
produce carbohydrates in a process called photosynthesis. Since humans and animals
depend on plants for food, photosynthesis is necessary for the survival of life on earth.
However, CO2 can also have negative effects. As CO2 builds up in our atmosphere it
has a warming effect that could change the earth’s climate. Indoors, CO2 levels easily
rise above the recommended amount which has adverse effects.
1.1.1.2 What is CO2 emission?
CO2 emission are those stemming from the burning of fossil fuels and the
manufacture of cement. They include carbon dioxide produced during consumption of
solid, liquid, and gas fuels and gas flaring.
1.1.2 Factors affect CO2 emissions
There are a lot of factors that affect CO2 emissions. However, in this research,

we focus mainly on three significant ones, which are energy use, population growth
and GDP per capita.
1.1.2.1 Energy use
Energy use is the amount of energy or power used (kg of oil equivalent per capita).
1.1.2.2 Population growth

Population growth is an increase in the number of people that reside in a country,
state, county, or city. To determine whether there has been population growth, the
following formula is used: (birth rate + immigration) – (death rate + emigration).
Businesses and governmental bodies use this information to make determinations
about investing in certain communities or regions.

2


1.1.2.3 GDP per capita
GDP per capita is a measure of a country's economic output that accounts for its
number of people. It divides the country's gross domestic product by its total
population. That makes it a good measurement of a country's standard of living. It tells
you how prosperous a country feels to each of its citizens.
1.1.3 Theories about the relationships between energy use, population growth,
GDP per capita and CO2 emissions
1.1.3.1 Energy use and CO2 emissions
Sustainable development (SD) implies the balancing of economic and social

development with environmental protection: the so-called ‘Three Pillars’ model. In the
long term, Planet Earth will impose its own constraints on the use of its physical
resources and on the absorption of contaminants, whilst the ‘laws’ of the natural
sciences (such as those arising from thermodynamics) and human creativity will limit
the potential for new technological developments. SD is a process or journey toward
the destination of ‘sustainability’. It is a key concept when examining energy use and
associated emissions, and has foundations in engineering, economics, ecology and
social science.
1.1.3.2 Population growth and CO2 emissions
The impact of population change on environmental stress was posited by Ehrlich
(1968) and Holder and Ehrlich (1974) in the form of an equation relating environmental
impact to the production of population size, affluence, and environmental impact per unit

of economic activity known as “IPAT”. IPAT is useful framework for assessing the
anthropogenic environmental change, particularly the impacts of population, affluence,
and technology on the environmental change (CO2 emissions).
1.1.3.3 GDP per capita and CO2 emissions

The Environmental Kuznets Curve hypothesis, it was generally assumed that
rich economies destroyed the environment at a faster pace than poorer countries.
However, with the Environmental Kuznets Curve, the relationship between the
environment’s health and the economy has been reanalyzed.
The idea is that as economic development growth occurs, the environment will
worsen until a certain point where the country reaches a specific average income.
Then money is invested back into the environment, and the ecosystem is restored.

3


Critics argue that economic growth doesn’t always lead to a better environment
and sometimes the opposite may actually be true.

1.2 Empirical researches

1.2.1 Empirical research on effects of energy use on CO2 emissions
Thao and Chon [1] stated that energy use has a positive impact on economy, but not
to the environment. Energy use is widely known as the main reason for global warming

and climate change to happen, particularly the consumption of fossil energy. The
environmental adverse effects of such energy used are not only coming from the energy
consumption but also from the exploitation process. Meanwhile, the renewable energy
consumption has a negative relationship to CO2 emissions, which means that an increase
in the consumption of renewable energy will reduce CO2 emissions. Additionally, Ito [2]
found that fossil energy consumption has a negative impact on economic growth in
developing countries, and renewable energy consumption has a positive effect on
economic growth. In this case, the consumption of fossil energy can cause pollution and
environmental damage because the remaining burning of fossil energy is harmful to the
environment; while, the renewable energy residue is considered more environmentally
friendly. Moreover, Shafei and Ruhul [3] who conducted a study on OECD countries on
the Kuznets Curve Hypothesis (EKC) between urbanization and

4


CO2 emissions found that nonrenewable energy consumption has a positive relationship
to CO2 emissions, which means that an increase in non-renewable energy consumption
will increase CO2 emissions. In contrast, renewable energy consumption has a negative
relationship to CO2 emissions, once again, it consolidates the conclusion that an increase
in the consumption of renewable energy will reduce CO2 emissions.

1.2.2 Empirical research on effects of population growth on CO2 emissions
The role of population pressure on environmental quality can be traced back to
the early debate on the relationship between population and natural resources. Malthus

(1798 [1970]) was concerned with increasing population growth, which put pressure
on limited source of land. Because of a lower marginal product of labor, the potential
growth in food supply could not keep up with that of the population. He predicted that
if mankind did not exercise preventive checks, population growth would be curtailed
by welfare checks (poverty, disease, famine and war). Boserup (1981) held the
opposite view, which argues that high population densities were a prerequisite for
technological innovation in 4 agriculture. The technological innovation made possible
the increased yields and more efficient distribution of food. It could then enable the
natural environment to support a large population at the same level of welfare.
The impact of population growth on environment quality is obvious. Each person
in a population makes some demand on the energy for the essentials of life—food,
water, clothing, shelter, and so on. If all else is equal, the greater the number of people,

the greater the demands on energy. Birdsall (1992) specified two mechanisms through
which population growth could contribute to greenhouse gas emissions. First, a larger
population could result in increased demand for energy for power, industry, and
transportation, hence the increasing fossil fuel emissions. Second, population growth
could contribute to greenhouse gas emissions through its effect on deforestation. The
destruction of the forests, changes in land use, and combustion of fuel wood could
significantly contribute to greenhouse gas emissions.
Thus, two questions remain to be addressed fully and empirically: (1) does
population pressure have a net impact on carbon dioxide emissions holding constant
the affluence and technology? and (2) has population pressure exhibited a greater
impact in developing countries than in developed countries?

5


1.2.3 Empirical research on effects of GDP per capita on CO2 emissions
Three Totally Different Environmental/GDP Curves (2012) - In this paper Bratt
compares three different theories explaining the connection between environmental
degradation and GDP. The theories discussed are the Environmental Kuznets curve
(EKC), the Brundtland curve and the Daly curve. All three hypotheses recognize that
the level of GDP affect the environmental degradation, but in different ways. The EKC
hypothesis argues that an increasing level of GDP would initially increase pollution
until a certain level of GDP, at which the level of pollution starts to decrease. The
relationship between environmental degradation and economic growth is in the case of

the EKC graphically shown as an inverted U-shape. The Brundtland curve theory
provides another picture, where the graphical form is the opposite, U-shaped, which
implies the poorest and wealthiest countries to have the highest levels of pollution.
The Daly curve theory suggests increasing levels of pollution with an increasing GDP
that keeps on going, without any turning point. Bratt points out that the three different
environmental/GDP curves deals with different aspects of environmental degradation.
The EKC hypothesis could be used when measuring emissions or concentration. The
Brundtland curve could be used when measuring production and the Daly curve when
measuring consumption. Bratt’s final conclusion is that even though either curve could
be true, the most possible scenario seems to be a positive, monotonic relationship
between environmental degradation and GDP.
In summary, many research studies have been conducted in areas related to this

study. However, a major part of the researches conducted were on the relationship
between CO2 and just one other factor such as GDP per capita, population growth or
energy use. There are not many existing studies that specifically examine the effect of
GDP per capita, energy use, population growth, all together on CO 2 emissions,
especially in developing countries. In this research we will use the existed data and
linear regression to analyze the relationship between CO2 emissions and three other
factors: GDP per capita, energy use and population growth.

6


Chapter 2: METHODOLOGY

2.1 Methodology used
2.1.1 Methodology in collecting data
The collected data are secondary data, mixed data, which indicate information of
the fundamental factors concerning the amount of CO2 emissions (metric tons per
capita): GDP per capita, energy use, population growth. The secondary data were
gathered from prestigious and reliable source of information - World Bank.
2.1.2 Methodology in processing data
Using Gretl in order to process data cursorily then calculate the correlation
matrix among variables.
2.1.3 Methodology in researching
Using Gretl to bring out regression models by using Ordinary Least Squares
method (OLS) to estimate the parameter of multi-variables regression models. As a

result, we can:
- Depend on variance inflation factor (VIF) to identify multicollinearity
- Test Normality of residual
- Use white test to test heteroscedasticity
- Conduct Breusch-Godfrey to identify the correlations
- Use F-test to evaluate the concordance model
- Use T-test to evaluate the confidence interval
2.2 Constructing econometrics model
To demonstrate the relationship between the amount of CO2 emissions and other
factors, the regression function can be constructed as follows:
(PRF): Y=β1+β2EU+β3popgrowth+β4GDPpc+µi
(SRF): #=

& &

&

&

+ EU+ pop-growth+ GDPpc+еi

Where:
● Dependent variable: Y - The amount of CO2 emissions, measured in metric tons per
capita.

● Independent variables:
- EU: Energy use, measured in kg of oil equivalent per capita.

7


-

Researchers found that energy consumption is the long-run causes for CO2
emissions. For example, the burning of fossil fuels such as gasoline, coal, oil,
natural gas in combustion reactions results in the production of carbon dioxide.
Pop-growth: Population growth, measured in %.

-

Theoretically, population growth is believed to increase greenhouse gas
emissions, particularly CO2 emissions through the increase in human activities.
GDP pc: GDP per capita, measured in US$.
As a country’s GDPpc increases, so does its production of carbon dioxide into
the atmosphere. Human activity, which often leads to increased GDP such as
goods production and services, frequently produces carbon dioxide emissions.
For example, most goods and services involve some use of energy, often in the
form of coal or petroleum. Therefore, as the amount of produced goods
increases, the amount of fossil fuels spent also increases.

Exhibition 2.1 Variables explanation
Name

Dependent
variable

Y

Meaning
The amount of CO2
emissions (metric
tons per capita)

Signal

+

As a country’s GDP per
capita increases, so does its
production of carbon
dioxide per capita into the
atmosphere.

+

Energy consumption is the
long-run causes for CO2
emissions.

+

Population growth increases
greenhouse gas emissions,
particularly CO2 emissions
through the increase in
human activities.

Gross Domestic
GDP pc

Independent

EU

variables

Popgrowth

Product per capita

The amount of
energy or power
used (kg of oil
equivalent per
capita)
Population growth is
the increase in the
number of
individuals in a
population.

8

Explanation


2.3 Data overview
-

This set of data is a secondary one, as they are collected from a given source:
Data source: />The structure of Economic data: cross-sectional data

2.4 Data description
2.4.1 Summary statistics
By using Summary statistics command on Gretl, we have:
Exhibition 2.2 Summary Statistics, using the observations 1 – 81
Variable

Mean

Median

S.D.

Min

Max

CO2

4.88

2.60

6.59

0.100

34.0

66.3

1.44e+04

-1.70

6.70

428.

4.41e+04

EU

1.97e+03 1.02e+03 2.53e+03

popgrowth
GDPpc

1.46

1.30

1.36

8.00e+03 5.61e+03 8.82e+03

From the data, we can infer that:
-

Energy use per capita: The average energy use per capita is 1.97×10ˆ3 kg of oil
equivalent per capita, the minimum one is 66.3 kg of oil equivalent per capita and

the maximum one is 1.44×10ˆ4 kg of oil equivalent per capita.

-

Pop-growth: The average population growth is 1.46%, the minimum one is
-1.70% and the maximum one is 6.70%.

-

GDPpc: The average GDP per capita is 8×10ˆ5 US$, the minimum one is 428
US$ and the maximum one is 4.41×10ˆ4 US$.

2.4.2 Table of correlation matrix
Exhibition 2.3 Correlation coefficients, using the observations 1 - 81
5% critical value (two-tailed) = 0.2185 for n = 81

CO2
1.0000

EU

GDP PC

POPGROWTH

0.9843

0.8284

0.0261

CO2

1.0000

0.8034

0.0256

EU

1.0000

0.0154

GDPpc

1.0000

popgrowth

9


Evaluation:
Correlation between dependent variable and independent variables:
- r(Y, EU) = 0.9843: CO2 emissions and energy use have a very strong, uphill
relationship (energy use affects 98.43% of CO2 emissions)
- r(Y, popgrowth) = 0.0261: CO2 emissions and population growth have a very weak,
uphill relationship (population growth affects only 2.61% of CO2 emissions)

-

r(Y, GDPpc) = 0.8284: CO2 emissions and GDP per capita have a strong, uphill
relationship (GDP pc affects 82.84% of CO2 emissions).
In summary, all the correlations above are appropriate with theories.
Correlation among independent variables:
-

r(EU, GDPpc) = 0.8034: Energy use and GDPpc have strong weak, uphill
relationship.
r(EU, popgrowth) = 0.0256: Energy use and pop-growth have a very weak, uphill
relationship.

r(GDPpc, popgrowth) = 0.0154: GDP per capita and population growth have a
very weak, uphill relationship.

10


Chap 3: TEST ASSUMPTIONS & STATISTICAL INFERENCES
After running the regression, the result is summarized in the following table (the
s.e is showed in parentheses)
Variables
X2 (energy use)

Model 1
0.002 ***
(8.226e-05)

Model 2

Model 3 (Robust s.e)
0.002 ***
(0.0001)

X3 (Population
growth)

0.007
(0.091)

-0.052
(0.033)

0.007
(0.089)

X4 (GDP per cap)

7.919e-05 ***
(2.356e-05)

7.919e-05 ***
(2.96e-05)

I_X2 (l_EU)

0.99 ***
(0.083)

l_X4 (l_GDPpc)

0.276 ***
(0.079)

const
N
2
R
Autocorrelation
Heteroskedasticity
Multicollinearity
VIF:

• EU
• GDPpc
• Popgrowth
• l_EU
• l_GDPpc
Normality
assumption

Conclusion

-0.378 *
(0.212)

-8.426 ***
(0.378)

-0.378 **
(0.17)

81
0.973
Not violated
Indicate
heteroskedasticity

p-value = 0.004630
Not violated
2.822

81
0.915
Not violated
Indicate
heteroskedasticity
p-value = 0.022445
Not violated

81
0.973
Not violated
p-value = 0.004630
but not affect the
estimated result
Not violated
2.822

2.821
1.001

1.039
3.585
3.632

Not indicate a normal
distribution, but not
affect estimated result
p-value = 0.000
The model is significant
but it does not pass all
assumptions so we do
not use model 1

Indicate a normal
distribution
p-value = 0.074
Do not use model 2
because the p-value
in heteroskedasticity
is too small

Exhibition 3 Table of estimated result

11

2.821
1.001

Not indicate a normal
distribution, but not
affect estimated result
p-value = 0.000
Use model 3


3.1 Model 1

3.1.1 Overview of the regression model
Based on the data collected from the table, the sample regression function is established:
(SRF): , = -0.378 + 0.002*EU + 0.007*popgrowth + (7.919e-05)*GDPpc

It can be inferred that:
Energy use and GDPpc both have statistically significant effects on the amount
of CO2 emissions (metric tons per capita) at the 1% significant level (as all p-values
are smaller than 0.01) while the variable population growth does not have. In
particular, those effects can be specified by the regression coefficients following:
&

β .= -0.378: When all the independent variables are zero, the expected amount of CO2 emissions is -0.378.

&

β /= 0.002: When energy use (kg of oil equavilent per capita) increases by one, the expected amount of CO2 emissions (metric
tons per capita) increases by 0.2%, ceteris paribus.
β&0= 0.007: When the population growth (%) increases by one, the expected amount of CO2 emissions (metric tons per capita) increases by 0.7%, ceteris paribus.

β&1= 7.919e-05: when GDP per capita ($US) increases by one, the expected amount of CO2 emissions (metric tons per capita) increases by 7.919e-03%, ceteris
paribus.

The coefficient of determination R-squared = 0.973: all independent variables (EU,
popgrowth, GDPpc) jointly explain 97.3% of the variation in the dependent variable
(CO2).

3.1.2 Statistical Inferences
3.1.2.1 Statistical significance of coefficients
Applying p-value method, we can see that:
-

p-value of popgrowth = 0.938 > 0.1 => Population growth is not an important
determinant of CO2 emissions.

-

p-value of energyuse < 0.0001 => EU is an important determinant of CO2
emissions at a 1% significance level

-

p-value of GDPpc = 0.001 < 0.01 => GDP per capita is an important determinant
of CO2 emissions at a 1% significance level

Therefor, all coefficients are statistically significant, except for popgrowth.

12


3.1.2.2 Confidence intervals for coefficients

Definition: A 95% two-sided confidence interval for the coefficient is the set of values
of b that cannot be rejected by a 5% two-sided hypothesis test.
Using Gretl, we define the confidence interval for 3 coefficients as follows (Ceteris
paribus, with 95% confidence interval):
-

When popgrowth (%) increases by 1 unit, CO2 emission increases by a range
from 0.0021% to 0.0025%.

-

When EU (kg of oil equivalent per cap) increases by 1 unit, CO2 emission

increases by a range from -0.174kg per capita to 0.188 kg per capita.

-

When GDPpc ($US) increases by 1 unit, CO2 emission increases by a range from
3.22703e-05$ to 0.0001$.

3.1.2.3 Concordance of regression model
F-test = 918.8297, p-value = 4.18e-41 < α = 0.05
Therefore, we can conclude that the regression model is concordant.
3.1.3 Test Multicollinearity
Multicollinearity is the high degree of correlation amongst the explanatory

variables, which may make it difficult to separate out the effects of the individual
regressors, standard errors may be overestimated and t-value depressed. The problem
of Multicollinearity can be detected by examining the correlation matrix of regressors
and carry out auxiliary regressions amongst them.
In Gretl, we can examine by using the command View – Correlation Matrix. If
the correlation between each independent variable is bigger than 80%, there will be
high Multicollinearity amongst variables but it does not affect the statistical inference.
If the correlation amongst independent variable is bigger than 1, there will be a perfect
multicollinearity, which affect the estimation result.
The result from Exhibition 2.2 Correlation Matrix shows that:
-

The correlation between population growth and energy use is 0.026
The correlation between population growth and GDPpc is 0.015
The correlation between GDPpc and energy use is 0.803
13


The correlation between population growth – energy use and population growth
– GDPpc are all smaller than 0.8, while the correlation between GDPpc and energy
use is bigger than 0.8, which means it has a high multicollinearity. However, as it is
not perfect multicollinearity that Multicollinearity is not a significant problem.
Another way to test the Multicollinearity is to use the Collinearity command in
Gretl to know the VIF (variance inflation factor). If VIF is bigger than 10, there will

be a perfect multicollinearity. The result from Gretl shows:
VIF EU = 2.822 < 10
VIF popgrowth = 1.001 < 10
VIF GDPpc = 2.821 < 10
As the VIF is lower than 10, indicating that Multicollinearity is not a worrisome
problem for this set of data.
3.1.4 Test Autocorrelation
Model with cross-sectional data do not need to check autocorrelation.
3.1.5 Test normality assumption
Using the command Normality of Residual in Gretl, the result of p-value =
0.000. As the result shows: p-value = 0.000 < 0.05. It can be concluded that Model 1
does not follow the normal distribution.

The cause of this problem maybe come from the nature of data.
About the cure, we see that the number of observations in model 1 is 91
observations. As the Central Limit Theorem established, in some situations, when
independent random variables are added, their properly normalized sum tends toward
a normal distribution (informally a "bell curve") even if the original variables
themselves are not normally distributed. So that even though in model 1, it does not
follow the normal distribution, the estimates will asymptotically be normal.
3.1.6 Test Heteroskedasticity
Heteroskedasticity indicates that the variance of the error term is not constant,
which makes the least squares results no longer efficient and t tests and F tests results
may be misleading. The problem of Heteroskedasticity can be detected by plotting the
residuals against each of the regressors, most popularly the White’s test. It can be

remedied by looking back to the model – look for other missing variables. In Gretl the
White’s test command is used. As the result shows, the p-value = 0.0046.
14


At the 5% significance level, there is enough evidence (p-value = 0.0046 < 0.05)
to conclude that this set of data meets the problem of Heteroskedasticity.
To fix the problem, we decide to use the logarithm of dependent variable (CO2
emissions) and independent variables (EU and GDPpc), which will be specified in the
Model 2.
To sum up, it can be said that Model 1 is not perfect for estimation as it indicates
the Heteroskedasticity, so that we decide to establish another model (log-log model).

3.2 Model 2
Based on the data collected, the sample regression function is established:
(SRF): = -8.426 + 0.99*l_EU - 0.052*popgrowth + 0.276*l_GDPpc
3.2.1 Test Multicollinearity
Using the command Collinearity in Gretl, we have the result as follows:

VIF
= 1.039 < 10
VIF
= 3.585 < 10
VIF
= 3.632 < 10

As the VIF is lower than 10, indicating that Multicollinearity is not a worrisome
energyuse

popgrowth

GDPpc

problem for this set of data.
3.2.2 Test Autocorrelation: model with cross-sectional data do not need to check
autocorrelation
3.2.3 Test normality assumption
Using the command Normality of Residual in Gretl, the result shows that:

p-value = 0.07445 > 0.05 => Abnormal distribution is fixed when we use model 2
3.2.4 Test Heteroskedasticity
Heteroskedasticity indicates that the variance of the error term is not constant,
which makes the least squares results no longer efficient; t tests and F tests results may
be misleading. The problem of Heteroskedasticity can be detected by plotting the
residuals against each of the regressors, most popularly the White’s test.
In order to fix heteroskedasticity, we use log transformation method. In this case,
we will take log for CO2 emissions, GDP per capita and energy use, as their units are
not %.
Using Gretl, we have the result that p-value = 0.022445 < α = 0.05

15

Therefore, the model indicates heteroskedasticity. To fix this, we decide to use
the OLS model with Robust standard errors as specified in Model 3
3.3 Model 3 (using Robust standard errors)
Check Heteroskedasticity:
Using Gretl, we have the result that p-value = 0.00463 < α = 0.05, which shows
that the model indicates heteroskedasticity. However, as we use Robust standard
errors, which are not related to residuals, that it does not affect the estimated result.
From the regression model using Robust standard errors, we draw a conclusion:
In 2014, the amount of CO2 emissions (metric tons per capita) in developing
countries around the world was affected by GDP per capita, population growth rate

and the amount of energy use per capita, as described in the model 3:
, = -0.378 + 0.002*EU + 0.007*popgrowth + (7.919e-05)*GDPpc

The model is consistent with all the assumptions (Multicollinearity,
Autocorrelation, Heteroskedasticity, Normality Residuals) and has statistical
significance.
The research process helps to answer the questions raised in the preface: How
GDP per capita, population growth and energy use affect CO2 emissions in
developing countries in 2014.

16

Chapter 4: RESULT ANALYSIS AND POLICY IMPLICATIONS
4.1 Result analysis
From the estimator result, GDP per capita and energy use are 2 variables that are
significant to the amount of CO2 emissions.
The correlation between GDP per capita and CO2 emissions is positive, which
meets the expectation that GDP per cap is a factor affects CO2 emissions, following
the EKC hypothesis. This can be explained that almost all developing countries had
not reached the peak of Kuznets Curve in 2014, so that as GDP per capita increases, so
does the amount of CO2 emissions per capita.
The coefficient of GDPpc is (7.919e-05), which stands for: If the GDP per capita
increases by $1, the amount of CO2 emissions increases by 7.919.10^(-5) metric tons

per capita. It can be understood that human activities, such as producing goods
production and services (which involve use of energy in form of oil, coal, petroleum),
which often leads to increased GDP, frequently produce CO2 emissions. In 2014, as
the global economy stably developed, especially in developing countries that as the
amount of produced goods increase, the amount of fossil fuels spent also increases.
The correlation between the amount of energy use and CO2 emissions is
positive, which meets the expectation that energy use per capita is a factor affects the
amount of CO2 emissions, following the “sustainable development”. This correlation
is consistent with the empirical researches about the relation between energy use and
environment, such as: Thao and Chon (2016), Ito (2017), Shafei and Ruhul (2013).
The coefficient of EU is 0.002, which stands for: If the amount of energy use
increases by 1 kg per capita, the amount of CO2 emissions increases by 0.002 metric

tons per capita. In 2014, the resources expoilation activities in developing countries
are popular, which leads to a rise on CO2 emissions as well.
Besides, it can be concluded that the coefficient of independent variable
population growth is not significant to the amount of CO2 emissions. To explain for
this, we believe that the Ordinary Least Square is not a suitable model to research
about the relation between population growth rate and the amount of CO2 emissions,
which results in an erroneous estimation result. Additionally, the data we use are
cross-sectional in nature, so that our study cannot address the issue of whether the
impact of population growth on emissions could vary across countries with different
levels of economic development, even though they are all developing countries.

17

4.2 Policy implication
From the estimated result, it can be said that in order to reduce the amount of
CO2 emissions, we have to pay attention to the amount of factors which affect it.
Firstly, with GDP per capita, we know that not only in 2014 but also in recent
days, economic development is one of the top priorities in every countries. In
developing countries, as the hypothesis Environmental Kuznets Curve showed,
because these countries are on the way to develop but did not reach the peak of
development yet, means that the keep-rising GDP per capita necessarily leads to an
increase in CO2 emissions. Therefor, the amount of CO2 will only decrease once they
reach the peak of economic development. To achieve that, it will be a long process for

all developing countries. This raises needs for economic development policies from
government to swiftly increase GDP per cap, such as: take advantage of global open
market; make policies in management to encourage business; solve political problems;
practise family-planning; pay more attention to education…
In addition, about the energy use, recently exploitation of resources and energy is
a way that many developing countries use to make economic development. The result
from our estimator could be an evidence for countries’ government to examine their
national energy consumption, the dependence of economy on energy and the trade of
between energy – environment. It is impossible to prevent these exploitation activities,
however, there are ways to exploit it efficiently, which is known as “energy
efficiency”. Varying from countries, government should adopt the appropriated
policies, such as: Energy star federal tax credits for consumer energy efficiency; use

renewable energy sources; green economy development to ensure sustainaility…
These are some of the implications that we can point out from the estimated
result. There are stills shortcomings in our research process, such as the lack of time
and the limitation of our knowledge, that we could not find out a suitable model to
estimate the impact of population growth on CO2 emissions.

18


CONCLUSION
The above research has given us a clear view on the effects of population
growth, energy use and GDP per capita to the amount of CO2 emissions in 2014.

From the model examination, we have comprehensive assessments about the influence
of each variable, its meaning to the dependent variable. This result is also the
evidence for governments of developing countries to evaluate the amount of CO2
emissions emitted to the environment, which demonstrates the necessary role for
developing countries in enabling transitions to the low-carbon economy needed to
limit global temperature. Developing countries can be motivated to engage in more
ambitious emission reductions by eliminating fossil fuel subsidies, utilizing novel
technology options like soil carbon capture… However, the focus on developing
countries should not take interest away from what developed countries can and should
do to meet their targets. International policies calling for transitions to low-carbon
development are required to be cognizant of the inequalities that underlie the global
economic order, where developing countries are the most disadvantaged due to limited

capacity and technology. To ensure that a balance between developed and developing
county commitments and efforts is achieved, climate policies that encapsulate the
different principles that have proven to be effective will have to be adopted.

19


APPENDIX
Appendix 1. (Model 1) Ordinary Least Square regressor running
Model 1: OLS, using observations 1-81
Dependent variable: CO2
Coefficient

Std. Error

t-ratio

p-value

const

−0.378199

0.212392

−1.781

0.0789

EU

0.00234346

8.22578e-05

28.49

<0.0001

popgrowth

0.00713192

0.0912444

0.07816

0.9379

GDPpc

7.91894e-05

2.35626e-05

3.361

0.0012

*

***
***

Mean dependent var

4.882716

S.D. dependent var

6.589780

Sum squared resid

94.40630

S.E. of regression

1.107274

R-squared

0.972825

Adjusted R-squared

0.971766

F(3, 77)

918.8297

P-value(F)

3.62e-60

−121.1369

Akaike criterion

250.2739

259.8517

Hannan-Quinn

254.1166

Log-likelihood

Schwarz criterion

Appendix 2. (Model 1) Confidence Interval Result
t(77, 0.025) = 1.991
Variable

Coefficient

95 confidence interval

const

-0.378199

(-0.801125, 0.0447271)

EU

0.00234346

(0.00217967, 0.00250726)

popgrowth

0.00713192

(-0.174559, 0.188823)

GDPpc

7.91894e-05

(3.22703e-05, 0.000126109)

Appendix 3. (Model 1) Command Analysis - Collinearity on Gretl
Variance Inflation Factors

Minimum possible value = 1.0
Values > 10.0 may indicate a collinearity problem
EU
popgrowth
GDPpc

2.822
1.001
2.821

VIF(j) = 1/(1 - R(j)^2), where R(j) is the multiple correlation coefficient
between variable j and the other independent variables

Belsley-Kuh-Welsch collinearity diagnostics:
lambda

cond

--- variance proportions --const
EU
popgrowth

20

GDPpc

2.947
0.705
0.235
0.113

1.000
2.045
3.539
5.114

0.030
0.054
0.863
0.053

0.019
0.078
0.050
0.853

0.032
0.296

0.665
0.007

0.018
0.050
0.001
0.931

lambda = eigenvalues of X'X, largest to smallest
cond = condition index
note: variance proportions columns sum to 1.0

Appendix 4. (Model 1) Command Test – Normality of Residual on Gretl
Frequency distribution for uhat1, obs 1-81
number of bins = 9, mean = -2.9277e-15, sd = 1.10727
interval

midpt

frequency

rel.

cum.

1.23%

-3.4701 < -3.4701

-3.9255

1

1.23%

- -2.5592
-2.5592 - -1.6484

-1.6484 - -0.73764
-0.73764 - 0.17317
****************
0.17317 - 1.0840
1.0840 - 1.9948
1.9948 - 2.9056
>= 2.9056

-3.0147
-2.1038
-1.1930
-0.28224

0.62857

1
2
8
37
22

1.23%
2.47%
9.88%
45.68%

27.16%

1.5394
2.4502
3.3610

6
3
1

7.41%
3.70%

1.23%

Test for null hypothesis of normal distribution:
Chi-square(2) = 25.937 with p-value 0.00000

21

2.47%
4.94% ***
14.81%
60.49%
87.65% *********

95.06% **
98.77% *
100.00%


Appendix 5. (Model 1) Command Test – Heteroskedasticity – White’s test on Gretl
White's test for heteroskedasticity
OLS, using observations 1-81
Dependent variable: uhat^2
coefficient
std. error
t-ratio

p-value
--------------------------------------------------------------const
−1.89562
0.865094
−2.191
0.0317
EU
0.00126866
0.000644878
1.967
0.0531
popgrowth

1.12496
0.420134
2.678
0.0092
GDPpc
0.000135130 0.000132093
1.023
0.3098
sq_EU
−8.75517e-08 5.57872e-08
−1.569
0.1210

X2_X3
0.000183496 0.000248911
0.7372
0.4634
X2_X4
5.68806e-11 5.12568e-08
0.001110
0.9991
sq_popgrowth
−0.179590
0.100455
−1.788

0.0781
X3_X4
−5.85932e-05 5.30745e-05
−1.104
0.2733
sq_GDPpc
−2.91815e-09 8.42358e-09
−0.3464
0.7300

**
*

***

*

Unadjusted R-squared = 0.293822
Test statistic: TR^2 = 23.799557,
with p-value = P(Chi-square(9) > 23.799557) = 0.004630

Appendix 6. (Model 2) OLS Regressor Running
Model 2: OLS, using observations 1-81
Dependent variable: l_CO2
Coefficient

Std. Error

t-ratio

p-value

const

−8.42630

0.378054

−22.29

<0.0001

***

l_EU

0.990318

0.0827411

11.97

<0.0001

***

popgrowth

−0.0519048

0.0328186

−1.582

0.1178

l_GDPpc

0.275930

0.0799032

3.453

0.0009

***

Mean dependent var

0.834311

S.D. dependent var

1.314631

Sum squared resid

11.75781

S.E. of regression

0.390767

R-squared

0.914959

Adjusted R-squared

0.911646

F(3, 77)

276.1483

P-value(F)

4.18e-41

Log-likelihood

−36.77180

Akaike criterion

81.54359

Schwarz criterion

91.12139

Hannan-Quinn

85.38633

22


Appendix 7. (Model 2) Command Analysis – Collinearity on Gretl
Variance Inflation Factors
Minimum possible value = 1.0
Values > 10.0 may indicate a collinearity problem

l_EU

popgrowth

3.585
1.039
l_GDPpc
3.632

VIF(j) = 1/(1 - R(j)^2), where R(j) is the multiple correlation coefficient
between variable j and the other independent variables
Belsley-Kuh-Welsch collinearity diagnostics:

lambda

cond

3.588
0.399
0.010
0.002

1.000
2.998
18.951

38.258

--- variance proportions --- l_GDPpc
const
l_EU
popgrowth
0.001
0.000
0.023
0.000
0.002
0.001

0.897
0.001
0.869
0.148
0.070
0.027
0.128
0.851
0.009
0.972

lambda = eigenvalues of X'X, largest to smallest

cond = condition index
note: variance proportions columns sum to 1.0

Appendix 8. (Model 2) Command Test – Normality of Residual on Gretl
Frequency distribution for uhat2, obs 1-81
number of bins = 9, mean = -2.98115e-15, sd = 0.390767
interval
-0.97516

< -0.97516

-0.74479 -0.51442 -0.28406 -0.053689 0.17668 0.40705 >=

-0.74479
-0.51442
-0.28406
-0.053689
0.17668
0.40705
0.63741
0.63741

midpt
-1.0903

-0.85997
-0.62961
-0.39924
-0.16887
0.061495
0.29186
0.52223
0.75260

frequency

rel.

cum.

1

1.23%

1.23%

2
6
9

14
20
20
7
2

2.47%
7.41%
11.11%
17.28%
24.69%
24.69%

8.64%
2.47%

Test for null hypothesis of normal distribution:
Chi-square(2) = 5.195 with p-value 0.07445

3.70% **
11.11%
22.22% ***
39.51% ******
64.20% ********
88.89% ********

97.53% ***
100.00%