2020 5th International Conference on Green Technology and Sustainable Development (GTSD)

A New Approach for Analyzing and Predicting

Carbon Dioxide Emissions: Case study of Vietnam
Le Thi Giang Khuu Manh Dat Nguyen Xuan Hiep Sam Nguyen-Xuan
Faculty of Commerce Faculty of Commerce Faculty of Commerce Faculty of Information
University of Finance - University of Finance - University of Finance - Technology
Marketing Marketing Marketing Posts and Telecoms Institute
HCM city, Vietnam HCM city, Vietnam HCM city, Vietnam of Technology
[email protected] [email protected] [email protected] HCM city, Vietnam
[email protected]

Abstract— Vietnam has the potential to affect to air pollution However, the trade conflict between China and the US has
such as increasing carbon-dioxide emission from the rapid the potential to affect to extension of industry section of
extension of gross domestic products, foreign direct investment, Vietnam’s economy because foreign manufacturers in China
and manufacturing sections. Thus, this work analyzes the
relationships between the variables and proposed a dynamic have been moving in Vietnam. While this trend can become
model for predicting carbon dioxide emissions in Vietnam based a good chance for economic growth of Vietnam, it has the
on the relationships. Moreover, the work presents trend potential to affect to air pollution such as increasing carbon-
prediction of carbon dioxide emissions in Vietnam based on dioxide emission from the rapid extension of industry section.
multiple regression model. The work shown that there is a Examining impacts of the trend in advance is necessary to
statistically significant positive association between carbon- adapt current economic model relating with environmental
dioxide emission and manufacturing. However, the other
economic sectors, gross domestic product, and foreign direct
sustainability in Vietnam.
investment, are weaker impact on carbon-dioxide emission. In general, the relationship between CO2E and economic
growth can be modelled by the environmental Kuznets curve
Keywords—multiple regression (MR); gross domestic products
(GDP); manufacturing (MAN); foreign direct investment (FDI); [3, 4]. The environment Kuznets curve (EKC) shows an
carbon-dioxide emission (CO2E). inverted U shape relationship between economic growth and
environmental degradation. In [5], a study examined
I. INTRODUCTION relationship among CO2E, income, energy consumption,
Recently, many developing countries all over the world are trade openness, financial development, and so on for 151
looking for a new economic model which addresses the issue countries. The study examination is followed by ECK model
of economic growth and environmental protection. Therefore, [4].
they need an approach, which figure out the relationship
To clear understanding the relationship between CO2E
between economic growth and environmental protection.
Basically, the study is known as the green growth model and gross domestic products (GDP), manufacturing (MAN)
(GGM) and it is adopted by the ministerial conference on and foreign direct investment (FDI) sectors, an analysis of the
environment and development (MCED) [1]. The approach is relationship must show. The aim of the work is to examine
complex because several economic sectors come into play, is there any a significant and positive correlation between
including carbon-dioxide emission, manufacturing, gross CO2E and the other sectors and a significant increase in
domestic product, foreign direct investment, and so on. CO2E due to FDI and MAN. Because data for this work is
The GGM model plays an important role to achieve equal collected from (The World Bank data), however, the sample
pay because it allows Vietnamese government pay attention periods of the resource for CO2E, GDP, MAN, FDI are
on investments in industrial development as well as different. Thus, we proposed new approach to optimal the
environmental protection. To improve economic growth data by varying K values.
rapidly, the present suggestions focus on maximizing the In addition, we analyze and investigate correlation
benefit of industrial sector while light on the other sectors. coefficients for each K value. Each scenario is used to reveal
The policies have led to the rapidly increasing energy relationship between CO2E and other sectors. Moreover,
consumption such as fossil fuel, natural gas, etc. In [2], predicting of CO2E in the future for each K value will be
carbon-dioxide emission (CO2E) is produced significantly shown and compared with business as usual (BAU) in the
due to the manufacturing activity. Caused by environmental work. The contribution make our proposed model is different
degradation problems, the quality of Vietnam’s growth has to previous works [5-9]. Take the results into account,
declined. Therefore, Vietnam must take greater efforts to Vietnamese government may pay more attention on
improve policies and accurate measures the relationship investments in industrial development as well as environment
between CO2E and economic growth. problems.

978-1-7281-9982-5/20/$31.00 ©2020 IEEE 500

Authorized licensed use limited to: East Carolina University. Downloaded on June 16,2021 at 10:21:59 UTC from IEEE Xplore. Restrictions apply.
 2020 5th International Conference on Green Technology and Sustainable Development (GTSD)

The rest of this paper is organized as follows: Section 2

introduces the related works, section 3 is the methodology
framework of our proposal, then section 4 is our results on
different scenarios, and the last section shows conclusions
and some policy implications are provided.
II. RELATED WORKS
A. Economy affects CO2 emissions
In [10], economic activity has divided into three sectors
which are agriculture, industry , and service. The agriculture Fig. 1. Carbon dioxide emissions of Vietnam from 1986 to 2014 [14].
sector of the economy is the sector of an economy making
direct use of natural resources. The sector is the process of
producing food, feed, fiber, and other goods by the systematic
raising of plants and animals. In contrast, industry is the sector
of economy concerned with production of goods including
fuels and fertilizers. The last one, the service sector, produces
services. The agriculture sector is usually most important in
less-developed countries and typically less important in
industrial countries and all of them are a component of the
GDP of a nation.
The economic growth leads to a shift from agriculture
Fig. 2. The economic growth in Vietnam from 1986 to 2018 [14].
sector to industry sector, which may consume natural
resources and contribute CO2E. Since the beginning of the
industrial revolution have produced large percentage increase
in the atmospheric concentration of carbon dioxide [11, 12].
This implies that rapid economic growth is necessary to
tackle poverty, and problems related with basic standard of
life. However, the shifting economic from agriculture sector
to industrial sector caused the depletion of natural resource
and the deterioration of the environment.
A consequence of the rapid economic growths generated
unsustainable development. This has threatened our life and Fig. 3. Manufacturing of Vietnam from 1986 to 2018 [14].
survival of generations in the future. Until now, there had been
struggling with definition of sustainability. In [13],
sustainable development is defined as development that meets
"the needs of the present without compromising the ability of
future generations to meet their own needs". Therefore, to
achieve both economic growth and environmental protection,
international community has sought effective ways to make
sustainable growth.
B. Context of Vietnam
After renovation and opening market, the average CO2E
of Vietnam increased more than five times from 1986 to
Fig. 4. The FDI inflows of Vietnam from 1986 to 2018 [14].
2014. As the most important engines after renovating and
opening market, FDI and manufacturing hold the main role We consider historical values of all sections. As shown in
for economic growth in Vietnam, however, the engines are Fig. 1 and Fig. 2, the CO2E and economic growth have been
not only effect on GDP but also contribute to the rising of rising rapidly within 30 years. Thus, the Vietnam government
CO2E. Given growing concerns over CO2E, and the climate has been focusing on the problems and aims to reduce GHG
change effects, analyzing and predicting of the engines emissions by 8 percent below the business as usual (BAU)
contribute to CO2E will provide recommendations and scenario by 2030 [4]. On the other hand, the effect of FDI and
suggestions for long-term policies to implement frameworks manufacturing also increases gradually with the increase of
for foreign trade actions. Therefore, the answers to questions the CO2E, these are showed in Fig. 3 and Fig. 4. Over long-
such as how strong the engines contribute to the rising of term outlook to achieve BAU for CO2E, there is a
CO2E and what is the relationship among them are critical relationship among CO2E, FDI, and manufacturing in which
metrics for resolving above concerns. the coefficients of FDI and manufacturing will indicate how
strong the engines impact on CO2E.

501

Authorized licensed use limited to: East Carolina University. Downloaded on June 16,2021 at 10:21:59 UTC from IEEE Xplore. Restrictions apply.
 2020 5th International Conference on Green Technology and Sustainable Development (GTSD)

III. METHODOLOGY has a unit measurement of tons per capita, GDP is gross
domestic product per capita in US dollars, MAN is
A. Multiple regression equations
manufacturing added value in US dollars, and FDI represents
In [15], there are several models which presented the per capita of foreign direct investment in US dollars.
relationship between economic growth, energy, and CO2
emissions. To express a relation between CO2E and GDP, TABLE I. THE SUMMARY OF THE VARIABLES FROM 1986 TO
manufacturing, and FDI, a mapping from the sectors to CO2E 2018.
in which CO2E is a dependent variable and GDP,
Year CO2E GDP MAN FDI
manufacturing, and FDI are independent variables. We
1986 0.3706802 26336616250 5891663779 40000
propose a modeled form, shown in equation (1) as following:
1987 0.407674883 36658108850 8200176618 10363703.7
‫ ܧʹܱܥ‬ൌ ݂ሺ‫ܲܦܩ‬ǡ ‫ܰܣܯ‬ǡ ‫ܫܦܨ‬ሻ(1)
1988 0.355998429 25423812649 4590135743 7680000
where, CO2E is denoted as values of CO2 emission (tons), 1989 0.263108391 6293304975 953639655.7 4070000
GDP observation represents for economic growth, MAN is
1990 0.31487431 6471740806 793175839.4 180000000
manufacturing refers to industries, and FDI refers to direct
1991 0.308941403 9613369520 1259650117 375190278
investment equity flows. Given any value of the predictors,
we can estimate the CO2E by taking the logarithmic form on 1992 0.302998931 9866990236 1518898000 473945856
both sides using the multiple regression model [16]. The 1993 0.318211065 13180953598 1999349773 926303715
equation (1) is then rewritten into logarithm as following: 1994 0.356138727 16286433533 2428725001 1944515936
Žሺ‫ݕ‬
ෝ ሻ ൌ ߚ଴ ൅ ߚଵ Žሺ‫ܲܦܩ‬ሻ ൅ ߚଶ ݈݊ሺ‫ܰܣܯ‬ሻ ൅ ߚଷ Žሺ‫ܫܦܨ‬ሻ(2) 1995 0.388334428 20736164459 3108993366 1780400000

where β0 is regression constant, and βi (i =1,2,3) is the 1996 0.455743274 24657470575 3742635769 2395000000
regression coefficient to be determined from the variable 1997 0.584708333 26843700442 4425119856 2220000000
factors of GDP, MAN, and FDI. 1998 0.608242811 27209602050 4665812621 1671000000
There are several ways to find the coefficients for the best 1999 0.603434888 28683659007 5075377296 1412000000
multiple regression, the most common method to measure 2000 0.671308552 31172518403 5750363582 1298000000
closeness is to minimize the residual sum of squares (rss). 2001 0.757220692 32685198735 6425801618 1300000000
Thus, the difference between the ith expected value‫ݕ‬ෞప and the
2002 0.868419917 35064105501 7173075035 1400000000
ith predicted value ‫ݕ‬௜ is presented ith residual, ߳௜ ൌ ‫ݕ‬௜ െ ‫ݕ‬ෝప .
We define the rss as: 2003 0.957054376 39552513316 8040061270 1450000000
2004 1.090129494 45427854693 9238854312 1610000000
‫ ݏݏݎ‬ሺߚ௜ ሻ ൌ ‹ሺσ௡௜ୀଵ ߳௜ଶ ሻ(3)
௠௜௡ 2005 1.170708252 57633255618 10848471132 1954000000
where ߳௜ ሺ݅ ൌ ͳǡʹǡ ͵ሻ a vector of residual terms. 2006 1.214236115 66371664817 12863310252 2400000000
As showed in equation (3), if rss is small, it means that the 2007 1.227733963 77414425532 15003236547 6700000000
model predictions are very close to the actual values and 2008 1.368139953 99130304099 18418071125 9579000000
versa. To solve the optimization problem, we transformed 2009 1.476993533 1.06E+11 19401780477 7600000000
the problem into linear equations by using matrix operations.
2010 1.622618922 1.16E+11 15008931942 8000000000
Generally, the equation (3) is equivalent as:
2011 1.712240646 1.36E+11 18100756957 7430000000
‫ ݏݏݎ‬ሺߚ௜ ሻ ൌ σ௡௜ୀଵሺ‫ ݕ‬െ ܺߚሻ் ሺ‫ ݕ‬െ ܺߚሻ(4)
௠௜௡ 2012 1.583708122 1.56E+11 20700211254 8368000000
2013 1.622307629 1.71E+11 22832775311 8900000000

‫ݕ‬ଵ ߚଵ ͳ ܺଵଵ  ܺଵଶ ܺଵଷ 2014 1.819893977 1.86E+11 24539530925 9200000000

‫ݕ‬ଶ ߚଶ ͳ ܺଶଵ  ܺଶଶ ܺଶଷ Ǥ 2015 N/A 1.93E+11 26463842087 11800000000
‫ۇ‬Ǥ‫ۊ‬ ‫ۊ ۇ‬ ‫ۇ‬ ‫ۊ‬
Ǥ Ǥ Ǥ Ǥ Ǥ 2016 N/A 2.05E+11 29283700778 12600000000
‫ ݕ‬ൌ ‫ ۈ‬Ǥ ‫ۋ‬, ߚ ൌ ‫ۋ ۈ‬,ܺ ൌ ‫ ۈ‬   ‫ۋ‬
‫ۋ ۈ‬ ‫ ۈ‬Ǥ ‫ۋ‬ ‫ۈ‬Ǥ Ǥ Ǥ Ǥ ‫ۋ‬
2017 N/A 2.24E+11 34308986384 14100000000
Ǥ Ǥ Ǥ Ǥ Ǥ Ǥ
‫ݕۉ‬௡ ‫ی‬ ‫ߚۉ‬௡ ‫ی‬ ‫ܺ ͳۉ‬௡ଵ  ܺ௡ଶ  ܺ௡ଷ ‫ی‬ 2018 N/A 2.45E+11 39225645461 15500000000

where X is data matrix with an extra column of ones on the C. Processes of predicting model
left to account for the intercept, y = (y1, . . ., yn)T, and β = (β1, Since the values of CO2E from 2015 to 2018 are not
. . . , βn)T. available [14]. In order to predict the values, we proposed
linear regression [17] based on observations from table I. Our
B. Data Set and Sample Period algorithm examines different periods, K. Because our data is
The data for this work is collected from world bank data collected from World Bank [14] from 1986. Thus, the work
[14] where the sample period is taken into consideration for examines three periods, corresponding K (5, 10, 15). As
GDP, manufacturing, and FDI from 1986 to 2018 and the results, all CO2E values are completed replace with linear
sample period is taken into consideration for CO2E from prediction, then we go to next step for multiple regression
1986 to 2014. Table I shows the summary of the variables analysis. Intuitively, the proposed algorithm is not only
used in the multiple regression equations. In the table I, CO2E

502

Authorized licensed use limited to: East Carolina University. Downloaded on June 16,2021 at 10:21:59 UTC from IEEE Xplore. Restrictions apply.
 2020 5th International Conference on Green Technology and Sustainable Development (GTSD)

optimum replacement values of CO2E data from 2015 to 2018 0.905. The greater R-squared values, the smaller error
but also accumulated growth rate of CO2E data. between the observed values and expected values. In our
model, F-statistic is 108.7 which follows an F distribution
The main idea behind the proposal help us to figure out [18] and Prob (F-statistic) so that we can say the overall
how the different time periods will impact on dynamic multiple regression is significant. All results for K = 15 are
change of CO2E in the future. Based on these results, presented in table III.
advanced policies for each period will be recommended to
Table IV presents the coefficients of the intercept and the
attract either foreign investment in industrial section or not.
constant values for multiple regression. In addition, the other
Thus, our hope is to improve maximum chances to build a coefficients such as standard error (std err), t statistic model
model economic to achieved rapid development and and its P value are presented. Standard errors of the
environment protection. Our main contribution is twofold 1) coefficients we will calculate the covariance-variance factor,
optimum replacement values of CO2E data, and 2) multiple t statistic is given by the ratio of the coefficient (or factor) of
regression analysis to show the relationships between CO2E the predictor variable of interest, and its corresponding
and GDP, MAN, and FDI which is presented in Fig. 5. standard error. The confidence interval is the range of values
we would expect to find the parameter of interest and a
Input (incomplete Output (completed smaller confidence interval suggests that we are confident
CO2E values) CO2E values)
linear regression about the value of the estimated coefficient. In the table IV,
with different K three of the coefficients are significant at the 5% level, and
std err is the standard deviation of its sampling F distribution.
a. Process of CO2E prediction
GDP TABLE III. MODEL SUMMARY OF OLS REGRESSION
MAN Multiple regression CO2E Dep. Variable: CO2E R-squared: 0.918
FDI Model: OLS Adj. R-squared: 0.910
b. Proposed model
Method: Least Squares F-statistic: 108.7
Fig. 5. Proposed process for multiple regression analysis Date: Tue, 02 Jun 2020 Prob (F-statistic): 7.11e-16
Time: 15:07:23 Log-Likelihood: 10.412
IV. RESULTS AND DISCUSSION
No. Observations: 33 AIC: -12.82
A. Results
Df Residuals: 29 BIC: -6.838
In the first step, we perform linear regression [17] to
predict the values for a time series data of CO2E from 2015 Df Model: 3
to 2018. The K values are selected for length of time series Covariance Type: nonrobust
(K= 5, 10, 15). An iterative process for each K is presented
until all CO2E values are replaced. As results, all in TABLE IV. THE COEFFICIENTS OF OLS REGRESSION
completed values of CO2E from 2015 to 2018 are completed
and shown in table II. For all selected K sizes, though values coef
std
t P>|t| [0.025 0.975]
of CO2E from 2015 to 2018 for all scenarios increase, each err
K present different fit line through the size of data. This refers const 1.0000 0.033 30.512 0.000 0.933 1.067
that dynamic change of CO2E in the future depend on the GDP 0.2349 0.200 1.175 0.250 -0.174 0.644
policies on each period. MAN 0.3265 0.191 1.713 0.097 -0.063 0.716
TABLE II. VALUES OF CO2E FROM 2015 TO 2018. FDI 0.0337 0.130 0.260 0.797 -0.232 0.299
Year K=5 K = 10 K = 15
2015 1.763538987 1.935047562 1.830832243 TABLE V. ADDITIONAL TESTS IN OLS REGRESSION
2016 1.801972633 2.011980894 1.918060666 Omnibus: 0.544 Durbin-Watson: 0.294
2017 1.835850887 2.087192423 2.003552097 Prob(Omnibus): 0.762 Jarque-Bera (JB): 0.666
2018 1.891462374 2.164784455 2.086096141 Skew: 0.200 Prob(JB): 0.717

In the second step, we perform the multiple regression Kurtosis: 2.430 Cond. No. 13.6
with dependent output CO2E for each K, independent inputs
GDP, MAN, and FDI. The regression summary consists of Table V provides some additional information about the
three tables in which table III present the model summary, residuals of the model such as Omnibus, Skewness, Kurtosis,
table IV shows coefficients, and table V is additional tests. Durbin-Watson, Jarque-Bera (JB), and Cond. No in which
To keep original results, we show all information of the Skew and kurtosis refer to the shape of a distribution,
regression results and focus on the most important sectors for Omnibus test uses skewness and kurtosis to test the null
identifying relationship among coefficients. hypothesis that a distribution is normal, The Durbin-Watson
test is used to detect the presence of autocorrelation, and
R-squared and Adj. R-squared evaluate the scatter of the Jarque-Bera test is a goodness-of-fit test of whether sample
data points around the fitted multiple regression line. Since data have the skewness and kurtosis matching a normal
R-squared values for each K (5, 15, 25) are 0.872, 0.918, and distribution.

503

Authorized licensed use limited to: East Carolina University. Downloaded on June 16,2021 at 10:21:59 UTC from IEEE Xplore. Restrictions apply.
 2020 5th International Conference on Green Technology and Sustainable Development (GTSD)

B. Discussion products, manufacturing sector and foreign direct investment.

By using OLS [18], we find the parameter values for the The empirical results reveal that there is positive relation
predictive multiple regression equation in which constant among carbon dioxide emission and gross domestic products,
value is 1, the other coefficient values for GDP, MAN, and manufacturing sector, and foreign direct investment.
FDI are corresponding 0.2349, 0.3265, and 0.0337. Thus, the However, foreign direct investment and gross domestic
relationship can be modelled as follows: products are weaker impact on carbon-dioxide emission.

ln(CO2E) = 1 + 0.2349*ln(GDP) + 0.3265*ln(MAN) + The future work will investigate and analyze different
0.0337*ln(FDI) (5) manufacturing sections which are impact on carbon-dioxide
emission. Thus, we can focus on how to pay attention on
From the equation (5), if the CO2E will increase one unit, manufacturing sections which contribute smaller carbon-
then the GDP is expected to raise 0.2349 unit, while MAN dioxide emission.
and FDI expected to raise 0.3265 unit and 0.0337 unit.
REFERENCES
Moreover, we can find CO2E target at the year 2030, then we
[1] The Fifth Ministerial Conference on Environment and Development,
control the dependent variables so that they satisfy a target 2005.
reduction of CO2E the equation (5). To make the explain is [2] B. Trinh, K. Kobayashi, N. Q. Thai, N. V. Phong, and P. Le Hoa,
clear, we also show prediction of CO2E based on linear "Analyzing some economic relations based on expansion input-output
regression which is presented in Fig.6. model," International journal of business and management, vol. 7, no.
19, p. 96, 2012.
[3] G. M. Grossman and A. B. Krueger, "Environmental impacts of a North
American free trade agreement," National Bureau of Economic
Research1991.
[4] G. M. Grossman and A. B. J. T. q. j. o. e. Krueger, "Economic growth
and the environment," vol. 110, no. 2, pp. 353-377, 1995.
[5] C. Kiliç and F. J. P. Balan, "Is there an environmental Kuznets
inverted-u shaped curve?," vol. 65, no. 1, pp. 79-94, 2018.
[6] S. Dinda and D. J. E. E. Coondoo, "Income and emission: a panel data-
based cointegration analysis," vol. 57, no. 2, pp. 167-181, 2006.
[7] A. Jalil and M. J. E. E. Feridun, "The impact of growth, energy and
financial development on the environment in China: a cointegration
analysis," vol. 33, no. 2, pp. 284-291, 2011.
[8] M. Nasir and F. U. J. E. P. Rehman, "Environmental Kuznets curve for
Fig. 6. The predicted CO2 emissions to 2030 in Vietnam carbon emissions in Pakistan: an empirical investigation," vol. 39, no.
3, pp. 1857-1864, 2011.
In table VI, we show the correlation coefficients of the [9] H.-T. Pao, H.-C. Yu, and Y.-H. J. E. Yang, "Modeling the CO2
variables. The correlation coefficients are used to measure the emissions, energy use, and economic growth in Russia," vol. 36, no. 8,
strength and direction of the linear relationship among CO2E, pp. 5094-5100, 2011.
[10] A. G. Fisher, "Production, primary, secondary and tertiary," Economic
GDP, MAN, and FDI. Thus, Relationship of CO2E and MAN
record, vol. 15, no. 1, pp. 24-38, 1939.
are stronger than relationship of CO2E and FDI, and [11] T. Blasing and S. Jon, "Current greenhouse gas concentrations,"
relationship of CO2E and GDP. The correlation between two Updated February, 2005.
variables is helpful to determine how well a mutual fund [12] E. Dlugokencky and P. Tans, "Trends in atmospheric carbon dioxide,"
vol. 25, 2013.
performs relative to its benchmark index. [13] G. H. Brundtland, Report of the World Commission on environment
TABLE VI. THE CORRELATION COEFFICIENTS FOR CO2E, GDP, and development:" our common future.". United Nations, 1987.
MAN, AND FDI [14] The World Bank data [Online]. Available: http://www.worldbank.org/
[15] K. Saidi and S. J. E. R. Hammami, "The impact of CO2 emissions and
economic growth on energy consumption in 58 countries," vol. 1, pp.
CO2E GDP MAN FDI 62-70, 2015.
CO2E 1.000000 0.953208 0.955378 0.930073 [16] P. S. Mann, Introductory statistics. John Wiley & Sons, 2007.
[17] R. J. Little and D. B. Rubin, Statistical analysis with missing data. John
GDP 0.953208 1.000000 0.984062 0.965260 Wiley & Sons, 2019.
MAN 0.955378 0.984062 1.000000 0.961701 [18] J. Deng, A. C. Berg, and L. Fei-Fei, "Hierarchical semantic indexing
for large scale image retrieval," in CVPR 2011, 2011, pp. 785-792:
FDI 0.930073 0.965260 0.961701 1.000000 IEEE.

V. CONCLUSIONS AND FUTURE RESEARCH

In this paper, we investigate and analyze relationship
between carbon-dioxide emission and gross domestic

504

Authorized licensed use limited to: East Carolina University. Downloaded on June 16,2021 at 10:21:59 UTC from IEEE Xplore. Restrictions apply.