# Relationships between Causal Factors Affecting Future Carbon Dioxide Output from Thailand’s Transportation Sector under the Government’s Sustainability Policy: Expanding the SEM-VECM Model

^{*}

## Abstract

**:**

_{2}emissions from the Thai transportation sector by developing the Structural Equation Modeling-Vector Autoregressive Error Correction Mechanism Model (SEM-VECM Model). This model was created to fill information gaps of older models. In addition, the model provides the unique feature of viable model application for different sectors in various contexts. The model revealed all exogenous variables that have direct and indirect influences over changes in CO

_{2}emissions. The variables show a direct effect at a confidence interval of 99%, including per capita GDP ($\Delta {\mathrm{ln}(\mathrm{GDP})}_{t-1}$), labor growth ($\Delta {\mathrm{ln}(L)}_{t-1}$), urbanization rate factor ($\Delta {\mathrm{ln}(URT)}_{t-1}$), industrial structure ($\Delta {\mathrm{ln}(IS)}_{t-1}$), energy consumption ($\Delta {\mathrm{ln}(EC)}_{t-1}$), foreign direct investment ($\Delta {\mathrm{ln}(FDI)}_{t-1}$), oil price ($\Delta {\mathrm{ln}(OP)}_{t-1}$), and net exports ($\Delta {\mathrm{ln}(X-E)}_{t-1}$). In addition, it was found that every variable in the SEM-VECM model has an indirect effect on changes in CO

_{2}emissions at a confidence interval of 99%. The SEM-VECM model has the ability to adjust to the equilibrium equivalent to 39%. However, it also helps to identify the degree of direct effect that each causal factor has on the others. Specifically, labor growth ($\Delta {\mathrm{ln}(L)}_{t-1}$) had a direct effect on per capita GDP ($\Delta {\mathrm{ln}(\mathrm{GDP})}_{t-1}$) and energy consumption ($\Delta {\mathrm{ln}(EC)}_{t-1}$) at a confidence interval of 99%, while urbanization rate ($\Delta {\mathrm{ln}(URT)}_{t-1}$) had a direct effect on per capita GDP ($\Delta {\mathrm{ln}(\mathrm{GDP})}_{t-1}$), labor growth ($\Delta {\mathrm{ln}(L)}_{t-1}$), and net exports ($\Delta {\mathrm{ln}(X-E)}_{t-1}$) at a confidence interval of 99%. Furthermore, industrial structure ($\Delta {\mathrm{ln}(IS)}_{t-1}$) had a direct effect on per capita GDP ($\Delta {\mathrm{ln}(\mathrm{GDP})}_{t-1}$) at a confidence interval of 99%, whereas energy consumption ($\Delta {\mathrm{ln}(EC)}_{t-1}$) had a direct effect on per capita GDP ($\Delta {\mathrm{ln}(\mathrm{GDP})}_{t-1}$) at a confidence interval of 99%. Foreign direct investment ($\Delta {\mathrm{ln}(FDI)}_{t-1}$) had a direct effect on per capita GDP ($\Delta {\mathrm{ln}(\mathrm{GDP})}_{t-1}$) at a confidence interval of 99%, while oil price ($\Delta {\mathrm{ln}(OP)}_{t-1}$) had a direct effect on industrial structure ($\Delta {\mathrm{ln}(IS)}_{t-1}$), energy consumption ($\Delta {\mathrm{ln}(EC)}_{t-1}$), and net exports ($\Delta {\mathrm{ln}(X-E)}_{t-1}$) at a confidence interval of 99%. Lastly, net exports ($\Delta {\mathrm{ln}(X-E)}_{t-1}$) had a direct effect on per capita GDP ($\Delta {\mathrm{ln}(\mathrm{GDP})}_{t-1}$) at a confidence interval of 99%. The model eliminates the problem of heteroskedasticity, multicollinearity, and autocorrelation. In addition, it was found that the model is white noise. When the SEM-VECM Model was used for 30-year forecasting (2018–2047), it projected that CO

_{2}emissions would increase steadily by 67.04% (2047/2018) or 123.90 Mt CO

_{2}Eq. by 2047. The performance of the SEM-VECM Model was assessed and produced a mean absolute percentage error (MAPE) of 1.21% and root mean square error (RMSE) of 1.02%. When comparing the performance value with the values of other, older models, the SEM-VECM Model was found to be more effective and useful for future research and policy planning for Thailand’s sustainability goals.

## 1. Introduction

_{2}emissions.

_{2}emissions. As a result, it produced 75 Mt CO

_{2}Eq. of greenhouse gasses in 2017 with a growth rate of 17.1% compared to 2016 [4]. The sector was found to produce up to 91.2 percent of all greenhouse gases in 2017, a percentage that has increased steadily since 1997 [3,4,5].

_{2}forecasting model that is accurate, efficient, and effective is required to guide the formulation of policy and planning. To this end, the SEM-VECM Model was developed from existing models to be made available for national planning purposes, as well as for future application in other fields. The researchers reviewed a number of relevant studies focusing on two important components, namely relationship factors and forecasting models, to serve as guidance in the research process. A streamlined review of studies was conducted for the relational investigation. Hu et al. [6] investigated the relationship between energy consumption and economic growth in industrial sectors in China by using first-and-second-generation panel unit root tests, panel co-integration tests, and a system generalized moment method. The study found that, in the short term, there was a unidirectional causal relationship between economic growth and energy consumption. Furthermore, in the long term, a unidirectional causality was found between energy consumption and economic growth. Zhao et al. [7] produced a comparative study investigating the equilibrium relationships and causal relationships between economic growth, electricity consumption, labor force, and capital input in northern China by applying a panel data analysis method based on the Cobb-Douglas production function. Their findings showed all variables to be long-term co-integrated. In addition, bidirectional causal relationships were found between electricity consumption and economic growth in six of these provinces, excluding the Hebei province. The same study also showed a bidirectional relationships between capital input and economic growth, as well as between labor force and economic growth, except for in the Beijing and Hebei province. Armeanu et al. [8] attempted to explore the influence and causal relation between renewable energy and sustainable economic growth in the 28 countries of the European Union (EU) during 2003–2014 by using a multivariate panel data. In their study, they noted that biomass energy had the highest influence on economic growth, while there was an indication of a unidirectional causal relationship in the short and long term between sustainable economic growth and renewable energies. Bandalos [9] examined the preciseness and utility of overall error and error estimators in the structural equation models by using a method of Monte Carlo. In the study, it has shown that the rescaled non-centrality parameter and EFO produced a highly precise estimate of the approximation error and overall error amount. Gómez et al. [10] investigated the linear and nonlinear causality relationship between energy consumption and economic growth in Mexico from 1965 to 2014 by employing unit root with structural breaks, co-integration analysis, and linear and nonlinear causality tests. They concluded that there were long-term linkages among production, capital, labor, and energy. In terms of linear causal links, they extended from total and disaggregated energy consumption to economic growth. Nonlinear causality went from energy consumption, transportation, capital, and labor to output. This result affirmed the importance of input factors in economic activity, and that energy conservation policies would have an impact on economic growth in Mexico.

_{2}emissions, worsening global climate change. In six sub-Saharan African nations, Kivyiro and Arminen [13] analyzed the casual relationships between carbon dioxide emissions, energy consumption, economic growth, and foreign direct investment by implementing an autoregressive distributed lag model and co-integration. The results indicated that all variables are co-integrated in the long term among all countries, and in certain countries, foreign direct investment (FDI) was found to have a greater potential to raise CO

_{2}emissions. In general, unidirectional Granger causality links were observed regarding the relation between the other variables and CO

_{2}emissions. In other studies, the same areas of relational investigation have been explored. Wesseh and Zoumara [14] examined the causal independence between energy consumption and economic growth in Liberia by investigating evidence of a non-parametric bootstrapped causality test. The study contributed to the existence of distinct bidirectional Granger causality between the variables. Additionally, it examined how employment in Liberia influences economic growth, while suggesting the appropriateness of the bootstrap technique. Yoo and Ku [15] investigated the causal relationship between nuclear energy consumption and economic growth in six countries, namely Pakistan, Switzerland, Argentina, Korea, France, and Germany, by using time-series techniques of unit roots, co-integration, and Granger-causality. In this study, it was found that the relationships between the variables among the countries were not uniform. Two types of relationships were observed, presenting both bidirectional causality between nuclear energy consumption and economic growth in Switzerland, and unidirectional causality between economic growth and nuclear energy in the case of France and Pakistan. Moreover, the same causality was found in Korea leading from nuclear energy to economic growth. While two other countries, Argentina and Germany, did not display any of those relationships. Chang et al. [16] examined G6 countries, analyzing the causal relationship between nuclear energy consumption and economic growth in those countries by optimizing the panel Granger causality tests. The findings of the study showed a unidirectional causality from economic growth to nuclear energy consumption across the countries, however, the UK was found to have a bidirectional causality from nuclear energy consumption to economic growth.

_{2}emissions in Beijing from 2005–2011 and 2012–2030 by developing the Long-range Energy Alternatives Planning System (LEAP)-BJ model. The results of the study showed how incremental changes in energy consumption led to fluctuations in total CO

_{2}emissions during 2005–2011. It was estimated that Beijing would reduce total energy consumption by 21.36% and CO

_{2}emissions by 35.37% from 2012 and 2030 if the proposed policies were implemented in full under the POL scenario. Mudarissov and Lee [20] examined the casual relationship between energy consumption and economic growth in Kazakhstan by adapting various methods, including Granger causality, the Vector Error Correction Model, an augmented Dickey–Fuller and Phillips–Perron unit root tests, and a co-integration test. Their findings indicate long-term unidirectional causalities leading from energy consumption (EC) to economic growth, yet also reflect short-term unidirectional causalities leading from economic growth to energy consumption. This indicates the significance of national energy production in boosting the economic growth.

_{2}emissions based on the SEM-VECM Model. Thus, we developed a model to use in sustainable development planning for Thailand in order to maximize useful data outcomes. We used the time series data from 1990 to 2017 in Thailand to do a forecast of CO

_{2}emissions for the next 30 years (2018–2047). The projected results are to be used in the policy planning of Thailand based on the next 30 years’ (2018–2047) sustainable development strategy. The selection of independent variables was based on the framework of the above management strategy. Hence, this research will be deemed useful and beneficial to national management and future applications. The research process was as follows:

- Identify a variable framework according to Structure Equation Modeling [34], where exogenous variables and endogenous variables are extracted to be latent variables and observed variables.
- Choose variables that have a co-integration at the same level to construct a SEM-VECM Model where the relationship of causal factors is both in the short and long term, indicating the direct effect, indirect effect, and total effect of the relationship.
- Examine the developed model regarding its heteroscedasticity, multicollinearity, and autocorrelation.
- Compare the effectiveness of the SEM-VECM Model with other existing models, including Multiple Linear Regression, Gray Model (GM (1,1)), GM-ARIMA Model, Artificial Neural Natural Model (ANN), back propagation neural network (BP Model), and ARIMA Model, through the performance measures of MAPE and RMSE.
- Analyze the relationship and direction parameter estimates of the SEM-VECM Model.
- Forecast CO
_{2}emissions for the next 30 years (2018–2047) using the SEM-VECM Model. The flowchart of the SEM-VECM Model is shown in Figure 1 below.

## 2. The Forecasting Model

#### 2.1. Structure Estimation Modeling-Vector Error Correction Mechanism Model (SEM-VECM Model)

_{X}= correlation coefficient between independent variable and observed variable

_{Y}= correlation coefficient between dependent variable and observed variable

**Pattern 1**

**:**${H}_{0}$ is the maximal number of vectors indicating the long-term co-integration equivalent to $r$. ${H}_{1}$ is the number of vectors indicating the long-term co-integration greater than $r$. In the above, $r=0,1,2,\dots ,n-1$; the statistical value to testify the above assumption is trace statistic $({\lambda}_{trace})$, which can be computed using the equation below:

**Pattern 2**

**:**${H}_{0}$ is the maximal number of vectors indicating the long-term co-integration equivalent to r. ${H}_{1}$ is the number of vectors indicating the long-term co-integration equivalent to $r+1$. In the above, $r=\begin{array}{cccc}0,& 1,& 2,& \dots ,\end{array}n-1$, and the statistical value to testify the above assumption is maximum eigenvalue ${\lambda}_{\mathrm{max}}$, which can be computed using the equation below:

#### 2.2. Measurement of the Forecasting Performance

## 3. Empirical Analysis

#### 3.1. Screening of Influencing Factors for Model Input

_{2}), per capita GDP (GDP), labor growth(L), urbanization rate (UR), industrial structure (IS), energy consumption (EC), foreign direct investment (FDI), oil price (OP), and net exports $(X-E)$. However, all causal factors we used in the structure equation modeling were confirmed to be stationary at the same level according to the Augmented Dickey-Fuller theory. Therefore, this study found that all nine causal factors are stationary at the First Difference I(1), as illustrated in Table 1.

#### 3.2. Analysis of Co-Integration

#### 3.3. Formation of Analysis Modeling with the SEM-VECM Model

_{2}emissions in both direct effect and indirect effect, as can be observed in the following. The factor of per capita GDP ($\Delta {\mathrm{ln}(\mathrm{GDP})}_{t-1}$) had a direct effect on CO

_{2}emissions ($\Delta {\mathrm{ln}({\mathrm{CO}}_{2})}_{t-1}$), equivalent to 82.00%, and its indirect effect was equal to 11.00% with a significance level of 1%. Labor growth ($\Delta {\mathrm{ln}(L)}_{t-1}$) had a direct effect on CO

_{2}emissions ($\Delta {\mathrm{ln}({\mathrm{CO}}_{2})}_{t-1}$) at 31.00%, while its indirect effect was equal to 15.00% with a significance level of 5%. In addition, labor growth ($\Delta {\mathrm{ln}(L)}_{t-1}$) had a direct effect on per capita GDP ($\Delta {\mathrm{ln}(\mathrm{GDP})}_{t-1}$) equivalent to 9.00% with a significance level of 1%, and it also ($\Delta {\mathrm{ln}(L)}_{t-1}$) had a direct effect on energy consumption ($\Delta \mathrm{ln}{(EC)}_{t-1}$) equivalent to 45.00% with a significance level of 1%. The urbanization rate factor ($\Delta {\mathrm{ln}(URT)}_{t-1}$) had a direct effect on CO

_{2}emissions ($\Delta {\mathrm{ln}({\mathrm{CO}}_{2})}_{t-1}$) equivalent to 69.00%, and its indirect effect was equal to 4.00% with a significance level of 1%. Furthermore, urbanization rate ($\Delta {\mathrm{ln}(URT)}_{t-1}$) had a direct effect on per capita GDP ($\Delta {\mathrm{ln}(\mathrm{GDP})}_{t-1}$) equivalent to 64.00% with a significance level of 1%. Urbanization rate ($\Delta {\mathrm{ln}(URT)}_{t-1}$) had a direct effect on labor growth ($\Delta {\mathrm{ln}(L)}_{t-1}$) equivalent to 62.00% with a significance level of 1%, and the urbanization rate ($\Delta {\mathrm{ln}(URT)}_{t-1}$) also had a direct effect on net exports ($\Delta {\mathrm{ln}(X-E)}_{t-1}$) equivalent to 49.00% with a significance level of 1%. As for industrial structure ($\Delta {\mathrm{ln}(IS)}_{t-1}$), it had a direct effect on CO

_{2}emissions ($\Delta {\mathrm{ln}({\mathrm{CO}}_{2})}_{t-1}$) equivalent to 74.00% and its indirect effect was equal to 9.00% with a significance level of 1%. In addition, industrial structure ($\Delta {\mathrm{ln}(IS)}_{t-1}$) had a direct effect on per capita GDP ($\Delta {\mathrm{ln}(\mathrm{GDP})}_{t-1}$) equivalent to 36.00% with a significance level of 1%. Energy consumption ($\Delta {\mathrm{ln}(EC)}_{t-1}$) had a direct effect on CO

_{2}emissions ($\Delta {\mathrm{ln}({\mathrm{CO}}_{2})}_{t-1}$) at 80%, and its indirect effect was equal to 2% with a significance level of 1%. With regards to energy consumption ($\Delta {\mathrm{ln}(EC)}_{t-1}$), it had a direct effect on per capita GDP ($\Delta {\mathrm{ln}(\mathrm{GDP})}_{t-1}$) at 55% with a significance level of 1%. Foreign direct investment ($\Delta {\mathrm{ln}(FDI)}_{t-1}$) had a direct effect on CO

_{2}emissions ($\Delta {\mathrm{ln}({\mathrm{CO}}_{2})}_{t-1}$) at 66%, and its indirect effect was equal to 1% with a significance level of 1%. Furthermore, foreign direct investment ($\Delta {\mathrm{ln}(FDI)}_{t-1}$) had a direct effect on per capita GDP ($\Delta {\mathrm{ln}(\mathrm{GDP})}_{t-1}$) at 32% with a significance level of 1%. Oil price ($\Delta {\mathrm{ln}(OP)}_{t-1}$) had a direct effect on CO

_{2}emissions ($\Delta {\mathrm{ln}({\mathrm{CO}}_{2})}_{t-1}$) equivalent to 47%, and its indirect effect is equal to 5% with a significance level of 5%. Oil price ($\Delta {\mathrm{ln}(OP)}_{t-1}$) had a direct effect on industrial structure ($\Delta {\mathrm{ln}(IS)}_{t-1}$) at 22% with a significance level of 1%, while oil price ($\Delta {\mathrm{ln}(OP)}_{t-1}$) had a direct effect on energy consumption ($\Delta {\mathrm{ln}(EC)}_{t-1}$) at about 43% with a significance level of 1%. Moreover, oil price ($\Delta {\mathrm{ln}(OP)}_{t-1}$) had a direct effect on net exports ($\Delta {\mathrm{ln}(X-E)}_{t-1}$) at about 41% with a significance level of 5%. Net exports ($\Delta {\mathrm{ln}(X-E)}_{t-1}$) had a direct effect on CO

_{2}emissions ($\Delta {\mathrm{ln}({\mathrm{CO}}_{2})}_{t-1}$) at about 35%, and its indirect effect was about 11% with a significance level of 1%. In addition, net exports ($\Delta {\mathrm{ln}(X-E)}_{t-1}$) had a direct effect on per capita GDP ($\Delta {\mathrm{ln}(\mathrm{GDP})}_{t-1}$) at 73% with a significance level of 1%.

#### 3.4. CO_{2} Emission Forecasting Based on the SEM-VECM Model

_{2}emissions in Thailand’s transportation sector for the next 30 years (2018–2047), as shown in Figure 4.

_{2}emissions in Thailand’s transportation sector over the next 30 years (2018 to 2047) are projected to increase with a growth rate of 67.04%, producing a significant amount of CO

_{2}emissions, continuously adding to the greenhouse gas problem.

## 4. Discussion

_{2}emissions in Thailand’s transportation sector over the next 30 years (2018–2047). The SEM-VECM Model was constructed utilizing various relevant forecasting theories. The study found that the SEM-VECM Model is the most efficient model with the lowest MAPE and RMSE as compared to the other models, namely GM-ARIMA Model, ARIMA Model, Gray Model (GM (1,1)), Back propagation neural network (BP Model), Artificial Neural Natural Model (ANN), and Multiple Linear Regression Model.

_{2}emissions from the Thai transportation sector to increase steadily over the projection period. This indicates that CO

_{2}levels will likely increase beyond projections used for the government’s current management strategy. At the same time, the results reflect that the current management plans and approaches will likely not be effective enough to achieve its sustainable development goals. Therefore, the responsible parties must focus on the formulation of new policies and management strategies taking into account each causal factor affecting changes in CO

_{2}emissions in the transportation sector, be it a direct and indirect effect. Otherwise, poor planning could result in serious negative consequences for both the economy and environment.

_{2}emissions from the transportation sector. It was structured based on an analysis of the causal factors, optimizing the Structural Equation Model, and utilizing the VECM Model, which had not been done in previous studies. The researchers chose to use LISREL software integrating Microsoft office to produce the most efficient and effective tool, which is also suitable for application in other sectors. The SEM-VECM Model passed an analysis of co-integration test and error correction mechanism test, which testifies to its ability as an ideal model for checking heteroskedasticity, multicollinearity, and autocorrelation. Throughout the study, it can be observed that the SEM-VECM Model is capable of predicting long-term changes more effectively compared to older models.

_{2}emissions in both the short and long term by sector. However, few studies forecast more than 20 years into the future. The current study integrated advanced statistics in its modeling and made special effort to minimize errors. However, since the model deals with long-term prediction, there are possible factors that could affect its accuracy. Therefore, the paper emphasized the research processes, and used only the specified causal factors with strict criteria in the selection process. If certain variables did not meet the criteria, they were immediately removed from the model. This is another factor which distinguishes this model from other existing forecasting models.

_{2}emissions. It is recommended that the government reviews and revises the relevant current policies based effective implementation of this forecasting model so as to increase managerial efficiency and effectiveness.

## 5. Conclusions

_{2}emissions during the time period (t − 1) at confidence intervals of 99% and 95%. In addition, all factors were found to have an indirect effect on the change of CO

_{2}emissions during the time period (t − 1) at confidence intervals of 99% and 95%, except for the error correction mechanism (ECM) variable. Additionally, the study reveals that all variables are influential over one variable to another, with both direct and indirect effects at confidence intervals of 99% and 95%. Therefore, the use of the SEM-VECM Model for a CO

_{2}emissions prediction is more effective when compared to previously studied models.

## Author Contributions

## Funding

## Acknowledgments

## Conflicts of Interest

## References

- Achawangkul, Y. Thailand’s Alternative Energy Development Plan. Available online: http://www.unescap.org/sites/default/files/MoE%20_%20AE%20policies.pdf (accessed on 1 October 2018).
- Office of the National Economic and Social Development Board (NESDB). Available online: http://www.nesdb.go.th/nesdb_en/more_news.php?cid=154&filename=index (accessed on 1 October 2018).
- National Statistic Office Ministry of Information and Communication Technology. Available online: http://web.nso.go.th/index.htm (accessed on 2 October 2018).
- Department of Alternative Energy Development and Efficiency. Available online: http://www.dede.go.th/ewtadmin/ewt/dede_web/ewt_news.php?nid=47140 (accessed on 2 October 2018).
- Thailand Greenhouse Gas Management Organization (Public Organization). Available online: http://www.tgo.or.th/2015/thai/content.php?s1=7&s2=16&sub3=sub3 (accessed on 2 October 2018).
- Hu, Y.; Guo, D.; Wang, M.; Zhang, X.; Wang, S. The relationship between energy consumption and economic growth: Evidence from China’s industrial sectors. Energies
**2015**, 8, 9392–9406. [Google Scholar] [CrossRef] - Zhao, H.; Zhao, H.; Han, X.; He, Z.; Guo, S. Economic growth, electricity consumption, labor force and capital input: A more comprehensive analysis on North China using panel data. Energies
**2016**, 9, 891. [Google Scholar] [CrossRef] - Armeanu, D.S.; Vintilă, G.; Gherghina, S.C. Does renewable energy drive sustainable economic growth? Multivariate panel data evidence for EU-28 countries. Energies
**2017**, 10, 381. [Google Scholar] [CrossRef] - Bandalos, D.L. Assessing sources of error in structural equation models: The effects of sample size, reliability, and model misspecification. Struct. Equ. Model. Multidiscip. J.
**1997**, 4, 177–192. [Google Scholar] [CrossRef] - Gómez, M.; Ciarreta, A.; Zarraga, A. Linear and nonlinear causality between energy consumption and economic growth: The case of Mexico 1965–2014. Energies
**2018**, 11, 784. [Google Scholar] [CrossRef] - Arango-Miranda, R.; Hausler, R.; Romero-Lopez, R.; Glaus, M.; Ibarra-Zavaleta, S.P. Carbon dioxide emissions, energy consumption and economic growth: A comparative empirical study of selected developed and developing countries. “The Role of Exergy”. Energies
**2018**, 11, 2668. [Google Scholar] [CrossRef] - Chang, C.C. A multivariate causality test of carbon dioxide emissions, energy consumption and economic growth in China. Appl. Energy
**2010**, 87, 3533–3537. [Google Scholar] [CrossRef] - Kivyiro, P.; Arminen, H. Carbon dioxide emissions, energy consumption, economic growth, and foreign direct investment: Causality analysis for Sub-Saharan Africa. Energy
**2014**, 74, 595–606. [Google Scholar] [CrossRef] - Wesseh, P.K., Jr.; Zoumara, B. Causal independence between energy consumption and economic growth in Liberia: Evidence from a non-parametric bootstrapped causality test. Energy Policy
**2012**, 50, 518–527. [Google Scholar] [CrossRef] - Yoo, S.H.; Ku, S.J. Causal relationship between nuclear energy consumption and economic growth: A multi-country analysis. Energy Policy
**2009**, 37, 1905–1913. [Google Scholar] [CrossRef] - Chang, T.; Gatwabuyege, F.; Gupta, R.; Inglesi-Lotz, R.; Manjezi, N.C.; Simo-Kengne, B.D. Causal relationship between nuclear energy consumption and economic growth in G6 countries: Evidence from panel Granger causality tests. Nucl. Energy
**2014**, 77, 187–193. [Google Scholar] [CrossRef][Green Version] - Nasreen, S.; Anwar, S. Causal relationship between trade openness, economic growth and energy consumption: A panel data analysis of Asian countries. Energy Policy
**2014**, 69, 82–91. [Google Scholar] [CrossRef] - Zhixina, Z.; Xin, R. Causal relationships between energy consumption and economic growth. Energy Procedia
**2011**, 5, 2065–2071. [Google Scholar] [CrossRef] - Yu, H.; Pan, S.Y.; Tang, B.J.; Mi, Z.F.; Zhang, Y.; Wei, Y.M. Urban energy consumption and CO
_{2}emissions in Beijing: Current and future. Energy Effic.**2015**, 8, 527–543. [Google Scholar] [CrossRef][Green Version] - Mudarissov, B.A.; Lee, Y. The relationship between energy consumption and economic growth in Kazakhstan. Geosyst. Eng.
**2014**, 17, 63–68. [Google Scholar] [CrossRef] - Li, S.; Li, R. Comparison of forecasting energy consumption in Shandong, China Using the ARIMA model, GM model, and ARIMA-GM model. Sustainability
**2017**, 9, 1181. [Google Scholar] - Ozturk, S.; Ozturk, F. Forecasting energy consumption of Turkey by Arima Model. J. Asian Sci. Res.
**2018**, 8, 52–60. [Google Scholar] [CrossRef] - Sun, W.; He, Y.; Chang, H. Forecasting fossil fuel energy consumption for power generation using QHSA-Based LSSVM model. Energies
**2015**, 8, 939–959. [Google Scholar] [CrossRef] - Yuan, C.; Liu, S.; Fang, Z. Comparison of China’s primary energy consumption forecasting by using ARIMA (the autoregressive integrated moving average) model and GM (1,1) model. Energy
**2016**, 100, 384–390. [Google Scholar] [CrossRef] - Sen, P.; Roy, M.; Pal, P. Application of ARIMA for forecasting energy consumption and GHG emission: A case study of an Indian pig iron manufacturing organization. Energy
**2016**, 116, 1031–1038. [Google Scholar] [CrossRef] - Fan, G.F.; Wang, A.; Hong, W.C. Combining grey model and self-adapting intelligent grey model with genetic algorithm and annual share changes in natural gas demand forecasting. Energies
**2018**, 11, 1625. [Google Scholar] [CrossRef] - Barak, S.; Sadegh, S. Forecasting energy consumption using ensemble ARIMA–ANFIS hybrid algorithm. Int. J. Electr. Power Energy Syst.
**2016**, 82, 92–104. [Google Scholar] [CrossRef] - Okumus, I.; Dinler, A. Current status of wind energy forecasting and a hybrid method for hourly predictions. Energy Convers. Manag.
**2016**, 123, 362–371. [Google Scholar] [CrossRef] - Zeng, B.; Zhou, M.; Zhang, J. Forecasting the energy consumption of China’s manufacturing using a homologous grey prediction model. Sustainability
**2017**, 9, 1975. [Google Scholar] [CrossRef] - Dai, S.; Niu, D.; Li, Y. Forecasting of energy consumption in China based on ensemble empirical mode decomposition and least squares support vector machine optimized by improved shuffled frog leaping algorithm. Appl. Sci.
**2018**, 8, 678. [Google Scholar] [CrossRef] - Liu, B.; Fu, C.; Bielefield, A.; Liu, Y.Q. Forecasting of Chinese primary energy consumption in 2021 with GRU artificial neural network. Energies
**2017**, 10, 1453. [Google Scholar] [CrossRef] - Ma, M.; Su, M.; Li, S.; Jiang, F.; Li, R. Predicting coal consumption in South Africa based on linear (metabolic grey model), nonlinear (non-linear grey model), and combined (metabolic grey model-Autoregressive Integrated Moving Average Model) Models. Sustainability
**2018**, 10, 2552. [Google Scholar] [CrossRef] - Ma, J.; Oppong, A.; Acheampong, K.N.; Abruquah, L.A. Forecasting Renewable Energy Consumption under Zero Assumptions. Sustainability
**2018**, 10, 576. [Google Scholar] [CrossRef] - Barbara, M.B. Structural Equation Modeling with Mplus: Basic Concepts, Application, and Programming; Taylor & Francis Group: New York, NY, USA, 2012. [Google Scholar]
- Dickey, D.A.; Fuller, W.A. Likelihood ratio statistics for autoregressive time series with a unit root. Econometrica
**1981**, 49, 1057–1072. [Google Scholar] [CrossRef] - MacKinnon, J. Critical Values for Cointegration Test in Long-Run Economic Relationships; Engle, R., Granger, C., Eds.; Oxford University Press: Oxford, UK, 1991. [Google Scholar]
- Johansen, S.; Juselius, K. Maximum likelihood estimation and inference on cointegration with applications to the demand for money. Oxf. Bull. Econ. Stat.
**1990**, 52, 169–210. [Google Scholar] [CrossRef] - Johansen, S. Likelihood-Based Inference in Cointegrated Vector Autoregressive Models; Oxford University Press: New York, NY, USA, 1995. [Google Scholar]
- Enders, W. Applied Econometrics Time Series; Wiley Series in Probability and Statistics; University of Alabama: Tuscaloosa, AL, USA, 2010. [Google Scholar]
- Harvey, A.C. Forecasting, Structural Time Series Models and the Kalman Filter; Cambridge University Press: Cambridge, UK, 1989. [Google Scholar]

**Figure 1.**The flowchart of the Structural Equation Modeling-Vector Autoregressive Error Correction Mechanism (SEM-VECM) model.

**Figure 4.**Forecasting results of CO

_{2}emissions from 2018 to 2047 in Thailand’s transportation sector.

ADF Test at First Difference I(1) | MacKinnon Critical Value | |||
---|---|---|---|---|

Variables | Value | 1% | 5% | 10% |

$\Delta \mathrm{ln}({\mathrm{CO}}_{2})$ | −6.79 *** | −4.75 | −3.41 | −2.77 |

$\Delta \mathrm{ln}(\mathrm{GDP})$ | −5.92 *** | −4.75 | −3.41 | −2.77 |

$\Delta \mathrm{ln}(L)$ | −4.77 *** | −4.75 | −3.41 | −2.77 |

$\Delta \mathrm{ln}(UR)$ | −6.51 *** | −4.75 | −3.41 | −2.77 |

$\Delta \mathrm{ln}(IS)$ | −5.99 *** | −4.75 | −3.41 | −2.77 |

$\Delta \mathrm{ln}(EC)$ | −6.47 *** | −4.75 | −3.41 | −2.77 |

$\Delta \mathrm{ln}(FDI)$ | −4.99 *** | −4.75 | −3.41 | −2.77 |

$\Delta \mathrm{ln}(OP)$ | −6.54 *** | −4.75 | −3.41 | −2.77 |

$\Delta \mathrm{ln}(X-E)$ | −6.13 *** | −4.75 | −3.41 | −2.77 |

**Note.**${\mathrm{CO}}_{2}$ is the carbon dioxide emissions; $\mathrm{GDP}$ is the per capita GDP; $L$ is the labor growth; $UR$ is the urbanization rate, $IS$ is the industrial structure, $EC$ is the energy consumption, $FDI$ is the foreign direct investment, $OP$ is the oil price, and $X-E$ is net exports, *** denotes a significance, $\alpha $ = 0.01, compared to the Tau test with the MacKinnon Critical Value, $\Delta $ is the first difference, and $\mathrm{ln}$ is the natural logarithm.

Variables | Hypothesized No of CE(S) | Trace Statistic Test | Max-Eigen Statistic Test | MacKinnon Critical Value | |
---|---|---|---|---|---|

1% | 5% | ||||

$\Delta \mathrm{ln}({\mathrm{CO}}_{2})$, $\Delta \mathrm{ln}(\mathrm{GDP})$, $\Delta \mathrm{ln}(L)$, $\Delta \mathrm{ln}(UR)$, $\Delta \mathrm{ln}(IS)$, $\Delta \mathrm{ln}(EC)$, $\Delta \mathrm{ln}(FDI)$, $\Delta \mathrm{ln}(OP)$, $\Delta \mathrm{ln}(X-E)$ | None *** | 241.65 | 135.09 | 25.25 | 12.50 |

At Most 1 *** | 89.15 | 91.50 | 5.60 | 3.50 |

Dependent Variables | Type of Effect | Independent Variables | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|

$\Delta {\mathbf{ln}(\mathbf{GDP})}_{\mathit{t}-1}$ | $\Delta {\mathbf{ln}(\mathit{L})}_{\mathit{t}-1}$ | $\Delta \mathbf{ln}(\mathit{U}\mathit{R}{)}_{\mathit{t}-1}$ | $\Delta {\mathbf{ln}(\mathit{IS})}_{\mathit{t}-1}$ | $\Delta {\mathbf{ln}(\mathit{EC})}_{\mathit{t}-1}$ | $\Delta {\mathbf{ln}(\mathit{FDI})}_{\mathit{t}-1}$ | $\Delta {\mathbf{ln}(\mathit{OP})}_{\mathit{t}-1}$ | $\mathsf{\Delta}\mathbf{ln}(\mathit{X}-{\mathit{E})}_{\mathit{t}-1}$ | ${\mathit{ECM}}_{\mathit{t}-1}$ | |||

$\Delta {\mathrm{ln}({\mathrm{CO}}_{2})}_{t-1}$ | DE | 0.82 *** | 0.31 ** | 0.69 *** | 0.74 *** | 0.80 *** | 0.66 ** | 0.47 ** | 0.35 *** | 0.39 *** | |

IE | 0.11 *** | 0.15 ** | 0.04 *** | 0.09 *** | 0.02 *** | 0.01 ** | 0.05 ** | 0.11 *** | - | ||

$\Delta {\mathrm{ln}(\mathrm{GDP})}_{t-1}$ | DE | - | 0.09 *** | 0.64 *** | 0.36 *** | 0.55 *** | 0.32 *** | - | 0.73 *** | - | |

IE | - | - | - | - | - | - | - | - | |||

$\Delta {\mathrm{ln}(L)}_{t-1}$ | DE | - | - | 0.62*** | - | - | - | - | - | - | |

IE | - | - | - | - | - | - | - | - | |||

$\Delta {\mathrm{ln}(UR)}_{t-1}$ | DE | - | - | - | - | - | - | - | - | - | |

IE | - | - | - | - | - | - | - | - | - | ||

$\Delta {\mathrm{ln}(IS)}_{t-1}$ | DE | - | - | - | - | - | - | 0.22 *** | - | - | |

IE | - | - | - | - | - | - | - | - | |||

$\Delta {\mathrm{ln}(EC)}_{t-1}$ | DE | - | 0.45 *** | - | - | - | - | 0.43 *** | - | - | |

IE | - | - | - | - | - | - | - | - | - | ||

$\Delta {\mathrm{ln}(FDI)}_{t-1}$ | DE | - | - | - | - | - | - | - | - | - | |

IE | - | - | - | - | - | - | - | - | - | ||

$\Delta {\mathrm{ln}(OP)}_{t-1}$ | DE | - | - | - | - | - | - | - | - | - | |

IE | - | - | - | - | - | - | - | - | - | ||

$\Delta {\mathrm{ln}(X-E)}_{t-1}$ | DE | - | - | 0.49 *** | - | - | - | 0.41 ** | - | - | |

IE | - | - | - | - | - | - | - | - | - |

Forecasting Model | Mean Absolute Percentage Error (MAPE) (%) | Root Mean Square Error (RMSE) (%) |
---|---|---|

Multiple Linear Regression model | 23.09 | 21.39 |

Artificial Neural Natural Model (ANN) | 15.54 | 14.22 |

Back propagation neural network (BP model) | 10.15 | 10.03 |

Gray model (GM (1,1)) | 8.61 | 7.98 |

ARIMA model | 4.97 | 6.07 |

GM-ARIMA Model | 4.63 | 4.09 |

SEM-VECM Model | 1.21 | 1.02 |

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Sutthichaimethee, P.; Ariyasajjakorn, D.
Relationships between Causal Factors Affecting Future Carbon Dioxide Output from Thailand’s Transportation Sector under the Government’s Sustainability Policy: Expanding the SEM-VECM Model. *Resources* **2018**, *7*, 81.
https://doi.org/10.3390/resources7040081

**AMA Style**

Sutthichaimethee P, Ariyasajjakorn D.
Relationships between Causal Factors Affecting Future Carbon Dioxide Output from Thailand’s Transportation Sector under the Government’s Sustainability Policy: Expanding the SEM-VECM Model. *Resources*. 2018; 7(4):81.
https://doi.org/10.3390/resources7040081

**Chicago/Turabian Style**

Sutthichaimethee, Pruethsan, and Danupon Ariyasajjakorn.
2018. "Relationships between Causal Factors Affecting Future Carbon Dioxide Output from Thailand’s Transportation Sector under the Government’s Sustainability Policy: Expanding the SEM-VECM Model" *Resources* 7, no. 4: 81.
https://doi.org/10.3390/resources7040081