economics

Procedia Environmental Sciences 11 (2011) 1183 – 1188

1878-0296 © 2011 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of the Intelligent Information Technology Application Research Association. doi:10.1016/j.proenv.2011.12.178

Available online at www.sciencedirect.com Available online at www.sciencedirect.com Procedia

Environmental Sciences

Procedia Environmental Sciences 00 (2011) 000–000

www.elsevier.com/locate/procedia

The Relationship between CO2 Emissions, Economic Scale, Technology, Income and Population in China

Guo Hang , Jiang Yuan-sheng* The college of economic & management, Sichuan Agricultural University, Ya’an, Sichuan, P.R. China, 625014

*Corresponding author: [email protected]；[email protected]

Abstract

The main objective of this paper was to attempt to identify the underlying driving forces which affect CO2 emissions. Based on the known IPAT equation, in this paper, economic scale was considered as one of main determinant of CO2 emissions, and the model on the factors was established to analyze the relationship between CO2 emissions, economic scale, technology, income and population in China. The regression results indicated that economic scale, income and population have a positive effect on CO2 emission; the impact of technology on CO2 emission was more complex. In first phase of the new technology, its effect on CO2 emission was positive. While at the later stage, the effect was negative through technology improvement.

Keywords: CO2 emissions; Principal component analysis; Regression analysis

1. Introduction

As one of main contribution to global warming, the issue of CO2 emissions caused more and more concerns of scholars both aboard and at home. Identifying the driving forces, which affect CO2 emissions, has become the hot research topic and some developments have been getting in this study. The equation IPAT originally introduced by Ehrlich expresses the relationship between environmental pressure and humanitarian factors. According to IPAT, the driving force of environmental impact (short for I) could be decomposed into three parts: affluence (short for A), population (short for P) and technology (short for T). Some researchers improved the known equation and put forward the expanded model, ImIPAT model and STIRPAT model [1-2].

Different from IPAT identity, the relationship between technological change, population or affluence and emissions tend to be complex. In order to address this situation, the flexible models are introduced to reflex the nonlinear, polynomial influence of technological change, population or affluence on emissions. Additionally, economic scale is one of determinants affecting emissions, and is utilized as a dependent variable in this paper.

Open access under CC BY-NC-ND license.

1184 Guo Hang and Jiang Yuan-sheng / Procedia Environmental Sciences 11 (2011) 1183 – 1188Author name / Procedia Environmental Sciences 00 (2011) 000–000

2. Model

The impact of economic scale on emissions is to be reflected through its effect on energy consumption. The rapid economic growth led to the increase in demand for energy, and emissions are increasing accordingly; conversely, the increasing need for energy further promoted economic growth by means of fix investment or consumption.

The existing literatures about influence of technology on emissions are mixed. In the industrialization progress, technology advance lead to the changes of consumption patterns from the more material- and energy-intensive manufacturing sector to the more environmentally-friendly service sector, and to the appearance of alternative energy, eventually emission decreases [3-4].In contrast, other studies have found that the more technological innovation causes more investment ,and lead to more emissions.

There are debates on the indicator of T in the previous studies, most used linear variable to denote technology, such as a time trend t or foreign direct investment (FDI) [5-7]. Regarding that CO2 emissions basically originated the 23 industries, and the correlation between Gross Domestic Product from 23 industries (short for GDP23) and technology in a region or nation, T is measured by a polynomial function about the reciprocal of GDP23 in this context.

Based on Maslow’s demand levels theory, people wish to spend a larger fraction of their income on the clean and well-preserved environmental quality as income rises, but at same time, more of goods and services are needed that lead to more emissions afterwards [8-9]. In this context, affluence is measured by weighted average of rural per capita net income and urban per capita disposable income combining the actual conditions in China.

As one of determinants, the influence of population on emissions is divided into two sides. On the one hand, as the population increases, more demand for goods and services lead to more consumption on energy [10-12], and the less forest areas causes more emissions; on the other hand, the more population involves the increase in aware of environment quality.

Now assume that the consumption and production function is same to each one, and the gross population, as dependent variable, is introduced to analyze the influence of population on emissions [13-14].

Through the analysis above, the model is employed as follows：

CO2t=α (1/GDP23, t) 2+β (1/GDP23, t) + γ Incomet + δPt + ζ GDPt + εt (1)

Where CO2t denotes the fossil fuels-induced CO2 emission (in million ton) in year t; economic scale is represented by GDPt , which is the total market value of goods and services produced in a country during year t (including consumption, investment and the government spending), and measured in 0.1 billion RMB (Chinese currency name for short); GDP23,t is the gross domestic product from 23 industries in year t, and 1/GDP23,t is the reciprocal of GDP23,t; Incomet represents per capita income in year t; Pt is the aggregate population in year t; εt is the error term in year t.

3. Data and analysis

3.1 Data

Data used in this study are time series covering the period from 1980 to 2006, including the CO2 emission data, GDP data, the population data and the income per capita. The data of CO2 emission from consumption and flaring of fossil fuels are measured in million tons and from the Energy Information Administration (EIA), the GDP data expressed in RMB 100 million, the relevant data of population and the income data per capita are both from China Statistical Yearbook.

1185Guo Hang and Jiang Yuan-sheng / Procedia Environmental Sciences 11 (2011) 1183 – 1188Author name / Procedia Environmental Sciences 00 (2011) 000–000

3.2 Principal components analysis (PCA)

We select the CO2 emission (Y) as dependent variable, and (1/GDP23,t) 2, 1/GDP23,t, Incomet, Pt and

GDPt as independent variables. The original data of variables must be normalized to make sure the same calibration according to equation (2).

x’ = (x-x’’)/σ2 (2)

Where x’ and x’’ are the standardized form and the average form of the original data x, respectively. σ is the standard deviation of original data.

Table 1 The result of ADF test

Variabl e

value of ADF test style(c ,t ,k) critical value conclusion

y -3.39 (c,t,10) -3.7347** stationary x1 -2.73 (0,0,1) -2.656*** stationary x2 -1.66 (0,0,10) -1.6262* stationary x3 3.44 (c,0,3) -2.9969** stationary x4 -2.04 (0,0,1) -1.9552** stationary x5 3.61 (c,0,4) -3.0038** stationary

*，**and***represent the critical value at significant level of 0.1,0.05 and 0.01, respectively. In the test style,c,t and k are constant, trend and lag items, respectively.

In order to void spurious regression, stationarity of variables employed in equation (1) must be checked by using the augmented Dickey-Fuller (ADF) unit root tests. The results of the unit root tests are showed in Table 1. From the results, it is showed that all the variables used in this model are stationary. Where x1, x2, x3, x4 and x5 represent (1/GDP23,t)

2, 1/GDP23,t, Incomet, Pt and GDPt ,respectively. Then, the correlations between all the variables are evaluated; the results are showed in Table 2.

Table 2 The results of variables’ correlations

Variable x1 x2 x3 x4 x5 x1 1.00 0.84 -0.31 -0.57 -0.29 x2 0.84 1.00 -0.70 -0.91 -0.68 x3 -0.31 -0.70 1.00 0.90 1.00 x4 -0.57 -0.91 0.90 1.00 0.88 x5 -0.29 -0.68 1.00 0.88 1.00

From Table 2, it is known that about 30%, 70% correlation coefficients among the variables are over 0.9 and 0.6, respectively.

Table 3 Summary of principal component analysis

Component Initial Eigen-values Total Variance (%) Cumulative (%)

1st 3.89115 77.823 77.823 2nd 1.00516 20.103 97.926 3rd 0.09910 1.982 99.908 4th 0.00399 0.080 99.988 5th 0.00060 0.012 100.000

1186 Guo Hang and Jiang Yuan-sheng / Procedia Environmental Sciences 11 (2011) 1183 – 1188Author name / Procedia Environmental Sciences 00 (2011) 000–000

Table 3 shows a summary of the five principal components form PCA. The Eigen-value measure the variance accounted for by the corresponding principal component. The percentage is given by the ratio of the individual Eigen-value to the trace of the correlation matrix, and calculation of all possible Eigen- values would account for all the variance of the original variables. Principal components can be ranked accounting to their ability to explain variance in the original data set. A common approach is to select only those with Eigen-values equal to or greater than one or with at least 80% cumulative explained variance [15].

These criteria were adopted for this study. From Table 3, the first and second principal components had Eigen-values greater than one with a cumulative explained variance of 77.82% and 97.93%, respectively. The first and second principal components were retained, and two new set of variables, U1 and U2, which calculated as a linear combination of the original five variables with their coefficient shown in Table 4.

U1 = ∑ v1i xi (3)

U1 = ∑ v2i xi (4)

Where U1 and U2 are the first and the second principal components of the five variables. v1i and v2i are used to denote coefficients of each variable for its corresponding principal component.

Furthermore, the regression equation is established as follows, and the parameters involved in the model are estimated by ordinary least square (OLS) method.

Y = aU1+bU2 (5)

Where a and b represent the parameters of the first and second principal components, respectively.

4. Regression result

We tested the performances of the regression models. An error analysis was conducted by comparing the calculating results from equation (5) with the reality CO2 emission. In order to quantify the differences, mean bias error (MBE) and root mean square error (RMBE) were determined as equation (6) and (7).

MBE= (∑ (Rt –Yt))/n (6)

RMBE = ((∑ (Rt –Yt )2)/n)1/2 (7)

Where Rt is the CO2 emission in year t, which is calculated according to equation (5); Yt is the reality CO2 emission in year t; n is the number of year and n equal to 27.

Table4 Result of OLS Regression

Parameters a t-test b t-test Value 0.914 14.466 -0.262 -4.152 Parameters R2 F-test MBE RMBE Value 0.951 113.25 1.14×10-5 0.101

Table 4 shows the results of OLS regression analysis. The parameters a and b pass the t-test2 at a significance level of 0.005. The Equation (5) passes the F-test3 at a significance level of 0.005. The values

2 Under given condition that degree of freedom is 27, the critical value of the t-test is 3.078 when the fiducially probability is

under 99.5%.

1187Guo Hang and Jiang Yuan-sheng / Procedia Environmental Sciences 11 (2011) 1183 – 1188Author name / Procedia Environmental Sciences 00 (2011) 000–000

of MBE and RMBE indicate calculated results of CO2 emission almost comforts to the actual CO2 emission.

5. Discussions and Conclusion

According to the results from regression analysis, the addition of one more unit of the first principal component increases CO2 emission by 0.91 units and an additional unit of the second principal component leads to a decline in CO2 emission by 0.262 units. Next, equation (8) is employed as follows to analyze the relationship between variable and CO2 emission.

Y’ =∑ (av1i+bv2i) xi (8)

Each variable’s coefficient (av1i+bv2i) is calculated and shown in Table 5. A unit increase in x3, x4 and x5 will cause an increase in CO2 emission by 0.944, 0.9 and 0.938 units, respectively. Nevertheless, an additional unit of x1 and x2 will decline CO2 emission by 0.354 and 0.73 units, respectively. Through analyzing the conclusion of drawing, the growth in economic scale, income and population will boost CO2 emission; the impact of technology on CO2 emission is more complex. In first phase of the new technology, its effect on CO2 emission is positive. While at the later stage, the effect is negative through technology improvement.

Table 5 Coefficient of variables

Variable 1st Component 2nd Component av1i+bv2i x1 -0.106 0.981 -0.354 x2 -0.566 0.813 -0.73 x3 0.976 -0.2 0.944 x4 0.837 -0.515 0.9 x5 0.978 -0.171 0.938

According to combine each variable’s coefficient, economic scale, income and population would cause CO2 emission to increase, and the impact of technology on CO2 emission is more complex. In first phase of the new technology, its effect on CO2 emission is positive. While at the later stage, the effect is negative through technology improvement.

References

[1] Paul E. Waggoner, “Agricultural technology and its societal implications”, Technology in Society, vol. 26, Apr.-Aug. 2004, pp. 123-136, doi:10.1016/j.techsoc.2004.01.024.

[2] R.York, EA. Rosa, T. Dietz, “STIRPAT, IPAT, and ImPACT: analysis tools for unpacking the driving forces of environmental impacts”, Ecological Economics, vol. 46, Oct. 2003, pp. 351-365, doi:10.1016/S0921-8009(03)00188-5.

[3] A. Shi, “The impact of population pressure on global carbon dioxide emissions, 1975–1996: evidence from pooled cross-country data”, Ecological Economics, vol. 44, Feb. 2003, pp. 29–42, doi:10.1016/S0921-8009(02)00223-9.

3 Under given condition that degrees of freedom are 1 and 26, the critical value of the F-test is 7.68 when the fiducially

probability is under 95%.

1188 Guo Hang and Jiang Yuan-sheng / Procedia Environmental Sciences 11 (2011) 1183 – 1188Author name / Procedia Environmental Sciences 00 (2011) 000–000

[4] R.K. Kaufmann, D. Davidsdottir, S. Garnham, P. Pauly, “The determinants of atmospheric SO2 concentrations: reconsi-dering the environmental Kuznets curve”, Ecological Economics, vol. 25, May. 1998, pp. 209–220, doi:10.1016/S0921-8009(97)00181-X.

[5] N. Shafik, Economic development and environmental quality: an econometric analysis, Oxford University Press, London, 1994.

[6] H. Kahn, A. McDonald, Deforestation in Post-War Philippines, University of Chicago Press, Chicago, 1992.

[7] E. Neumayer, “Can natural factors explain any cross-coun-try differences in carbon dioxide emissions?”, Energy Policy, vol. 30, Jan. 2002,pp. 7-12, doi:10.1016/S0301-4215(01)00045-3.

[8] N.Shafik, S.Bandyopadhyay, Economic Growth and Environmental Quality: Time Series and Gross Country Evidence, the policy Research Dissemination Center, Washington, 1992.

[9] David I. Stern, “Environmental Kuznets Curve”, Encyclopaedia of Energy, 3,pp. 517- 525,doi:10.1016/BO-12-176480-X/00454-X.

[10] T. Seldon, D. Song, “Environmental quality and development:Is there a Kuznets curve for air pollution emissions?”, Journal of Environmental Economics and Management, vol. 27, Sep. 1994, pp. 147–162, doi:10.1006/jeem.1994.1031.

[11] H. Kahn, A. McDonald, Deforestation in Post-War Philippines. University of Chicago Press, Chicago, IL, 1994.

[12] M. Tiffen, M. Mortimore, F.Gichuki, “More People, LessErosion: Environmental Recovery in Kenya”, Land Degradation and Development, vol. 5, Apr. 1994, pp. 59-60, doi: 10.1002/ldr.3400050109

[13] L.W. Jiang, C.O’Neill. Brian, “The Energy Transition in Rural China”, International Journal of Global Energy, vol. 21, Jan. 2004, pp. 2-26, doi: 10.1504/IJGEI.2004.004691.

[14] D. Michael, C.O’Neill. Brian, P. Alexia, “Population Aging and Future Carbon Emissions in the United States. ”, Energy Economics, vol. 30, Mar. 2008, pp. 642-675, doi:10.1016/j.eneco.2006.07.002.

[15] JC. Lam, KK.W. Wan, S.L. Wong, TN.T. Lam, “Principal component analysis and long-turn building energy simulation correlation”, Energy Conversion and Management, vol. 51, 2010, pp. 135-139, doi:10.1016/j.enconman.2009.09.004.