**3.2. Relationship between transport activity and transport energy consumption : estimation of EKC**

276 Energy Efficiency – The Innovative Ways for Smart Energy, the Future Towards Modern Utilities

β

β

that measures the magnitude of past disequilibrium. The coefficient

tests successfully, the estimation of the VECM gives the cointegrating vector.

λ*<sup>i</sup>* and β

= =

*k k*

= =

*i i*

1 1

*i i*

*k k*

1 1

cointegrated, the VECM for the two series can be written as follows:

α

α

explanatory variables coefficient (

variables in each equation of VECM using the

can conclude if the two series are coupled or uncoupled.

is sustainable or not.

of adjustment to equilibrium. Engle and Granger (1987) showed that if two series are

λ

λ

*t i t i ti t t*

*t i ti i ti t t*

Δ=+ Δ + Δ + + 1 1

Δ=+ Δ + Δ + + <sup>1</sup> 1 2

In Eqs. (1) and (2), PCTS and PCGDP (or per capita transport energy consumption PCTEC) represent per capita transport services and per capita GDP, respectively, whereas ΔPCTS and ΔPCGDP are the differences in these variables that capture their short-run disturbances and k is the number of lags. µ1t and µ2t are the error terms. *ECT* is the error correction term

deviation of the dependent variables from the long-run equilibrium. The significance of the

The robustness of the VECM is evaluated by using the normality residual test of Jarque-Bera, the Portmanteau auto-correlation test, the autocorrelation LM test, and the White homoscedasticity test. All these tests help us to accept or not the null hypothesis of no serial correlation. The normality residual test statistics of Jarque-Bera indicate if we accept or not the null hypothesis of normality of the residuals. The joint test statistics of the White homoscedasticity test with the no cross terms indicates if we accept or not the null hypothesis of non-heteroscadasticity at a 5% confidence level. If the model passes all the

After tests of cointegration, Granger causality test should be applied in order to exam the causality relationship between series. The sources of causation can be identified from the significance test of independent variables coefficients in the VECM. Regarding the causality of the short- run, we can test the nullity of the parameters associated with independent

can be tested by the significance of the speed of adjustment. We use the t-statistics on the coefficients of the ECT indicate the signicance of the long-run causal effects. The test give us the values of the speed of adjustment coefficients in the two equations of the PCTS and PCGDP which indicate if any deviation of the balance of long run of the value of the growth rate of the income tends to accelerate to adjust themselves with the shock and to return on its level of balance in a way faster than the rate of growth of the transport services. The validation of the first equation makes it possible to affirm that it is better to explain the GDP by the transport services than the transport services by the income. After testing cointegration and causality between transport activity evolution and economic growth, we

The same procedure can be applied between transport activity and transport energy in order to determine the relationship between them and so to conclude if the transport sector

χ

 η

 η

*y y x ECT* (1)

*x x y ECT* (2)

− −−

− −−

μ

μ

*<sup>i</sup>* ) confirms the presence of short-run causality.

η

<sup>2</sup> -Wald statistics. The Causality long-run

represents the

The genesis of the EKC can be traced back to Kuznets (1955), who originally discussed the relationship between economic growth and income inequity and suggested when per capita increase, income inequity increases also at first stage but after a certain level, starts decreases. This relationship follows an inverted-U curve and has is known as the Kuznets curve. Since the early 1990s, this curve measurement has progressed to become more used in analysis of relationship between economic development and environment quality (Grossman and Krueger, 1991; Bandyopadhyay, 1992; Panayotou, 1997).

Sustainable transportation system literature has more focused on negative environmental impacts caused by transport activity. Increase of traffic leads to increase of energy consumption and so degradation of air quality. Several factors can explain the increase of traffic, such as the economic growth, growth of population, urbanization, change in the lifestyles, increase of road infrastructures, etc. All these factors can lead to increase of passenger and freight mobility (Carlsson-Kanyama and Lindén, 1999; Ramanathan, 2000; Storchmann, 2005; Van Dender, 2009).

In the EKC literature there is a few studies which focus on the relationship between transport-related energy consumption and gas emissions and economic growth. Among these studies, we can quote the study elaborated by Cole et al. (1997) which examine the between per capita income and local air pollutants, and between energy consumption from transport sector and traffic for European countries over the period 1970-1992. They conclude that EKC relationship exist for local air pollutants from transport. Hilton and Levinson (1998) test the existence of EKC for plumb emissions from transport sector for 48 countries during 20 years. They find that their relationship with economic growth supports an EKC and explain their evolution by the increase of the private cars use. Two types of factors are mentioned by the authors: first, the pollutant fuel intensity (pollutant content per fuel type) and second, vehicle fuel intensity (energy efficient vehicle). Khan (1998) shows the existing of an urban EKC (UEKC) for hydrocarbon emissions from urban traffic in California State. He explains the increase of these emissions through the growth of personal mobility per private cars and its related fuel consumption.

Recently, Rupasingha et al. (2004) examine the urban polluting emissions of 3029 American counties using the EKC model and urban size and daily mobility as important determinants. Liddle (2004) examines the EKC relationship between per capita road energy consumption and per capita income, using IEA statistics. They conclude that hypothesis of an inverted-U curve are not existed and then EKC can't be proved. Tanishita (2006) examines the existing of the EKC for energy intensity from passenger transportation and concerning a set of data during the period 1980-1995. He finds that the relationship between the energy intensity of private and public transportation and the per capita Gross Regional Product (GRP) corresponds to an inverted U-shape of the EKC.

In the EKC literature, many functional form of EKC model are presented. Some studies consider only a cubic equation of income per capita as those of Grossman and Krueger (1991,

1995) and Harbaugh et al. (2002), or a quadratic equation, such as studies of Selden and Song (1994), Holtz-Heakin and Selden (1995), and Stern et al., (1996). Other econometric studies estimate several empirical models (linear, quadratic, cubic). The significance of the cube per capita income is based on the assumption of an N relation. Moreover, two principal types of curves coexist in the EKC literature. The first is the "diachronic" one which used with times series and aims to study the evolution of transport sector energy consumption comparatively with the evolution of income. However, the second is the "synchronic" one which used with cross-section data. The quadratic functional form of EKC is assuming the traditional EKC functional form. Then, we present the following regression model to describe the interaction between economic growth and transport sector energy consumption:

$$\ln\left(\mathcal{E}\right) = \alpha\_0 + \alpha\_1 \ln\left(\text{PCGDP}\right) + \alpha\_2 \left[\ln\left(\text{PCGDP}\right)^2\right] \tag{3}$$

Transport Intensity and Energy Efficiency: Analysis of Policy Implications of Coupling and Decoupling 279

some solutions some sustainable policy options to reduce energy intensity. Majority of studies have interested on consumption of fossil fuels namely, diesel and gasoline. Ang and Zhang (2000) have proposed a review of literature for this decomposition analysis. Banister and Stead (2002) have studied the energy efficiency and economic efficiency of transport sector for European countries. They have found that passengers-kilometer is not a driving factor for nine European countries; tonne-kilometer is a driving factor of energy transport efficiency growth for six countries and of energy transport efficiency deterioration for eight countries. Zhang et al. (2011) have decomposed the energy consumption in Chinese transportation sector and have found that the transportation activity effect is the important contributor to increase energy consumption in the transportation sector and the energy

Technique of energy decomposition is largely useful in sustainable transport studies. Several methods of decomposition have been proposed in the literature such as the refined Laspeyres techniques (Lin *et al*., 2008), the Arithmetic Mean Divisia Index (AMDI) and the Logarithmic Mean Divisia Index (LMDI) techniques (Ang, 2005; Liu *et al*., 2007,

In this section we derive the methodology to decompose transport sector energy consumption growth to the influencing factors, such as demographic, economic and urban characteristics. Examples include vehicle fuel intensity, vehicles intensity, urbanization, per capita GDP, motorization, road network length, modal mix, fuel mix and more other factors

*it*

where subscripts *i*, *j* and *t* refer to fuel type (e.g., diesel, gasoline and GPL), transport mode (e.g., road, rail, air and water) and year, respectively. In order to decomposing transport energy intensity into the contribution factors, several formulations can be proposed according the factors integrated in the decomposition and the transport services type (freight transport or passenger transport). It should be note that choice of potential factors is based on availability of time series data and the causal relationship test. For example Eq. 1

*TE TE TE TS GDP RV TEI*

were *MM* refers to modal mix which indicates the share of fuel consumption by a mode in total transport energy consumption, *EI* refers to transportation energy intensity per mode

*t it it t it t*

*TEI MM EI TI VI M* (8)

*GDP TE TS GDP RV POP*

*TEI TEI* (6)

(7)

intensity effect drivers significantly the decrease of energy consumption.

Hatzigeorgiou *et al*., 2008; Timilsina and Shrestha, 2009).

Transport energy intensity in year *t*( )*<sup>t</sup> TEI* , can be expressed as:

= = ×× × × *ijt ijt it it t it*

 *ijt* = ×××× *jt it it it it ijt*

*<sup>t</sup>* <sup>=</sup> *ijt*

*ijt*

which will be discussed in the section 4.

can be expressed as

Eq.2 can also be rewritten as

were *E* refers to energy consumption from transport activity and treated as dependent variable, *PCGDP* refers to per capita gross domestic production (PPP) and treated as independent variable and ln indicates natural logarithmic transformation. If the regression coefficient α <sup>2</sup> is negative, functional form of regression model corresponds to the standard EKC model.

The inverted-U relationship implies that energy consumption is reduced and so environmental quality improves beyond a certain threshold of income per capita. Lind and Mehlum (2007) present the basic properties that must satisfy the U relation. Their main idea insists that the U- inverted should have a positive slope at the beginning of the turning point and negative thereafter. This condition ensures that the endpoint is in the range of data. This condition can be written as follows:

$$2\alpha\_1 + 2\alpha\_2 \ln PCGDP\_{Min} \ge 0 \tag{4}$$

$$2\alpha\_1 + 2\alpha\_2 \ln \text{PCGDP}\_{\text{Max}} \le 0 \tag{5}$$

where ln *Min PCGDP* and ln *Max PCGDP* are, respectively, the minimal and maximum values of the variable ln*PCGDP* .

#### **3.3. Methodology of decomposition analysis**

In order to determine coupling relationship between increase of transport activity and economic growth and also transport energy consumption between, the analysis based on time series models and EKC models are considered as an aggregated analysis which not provides an explanation of the sources of coupling problem and growth of energy consumption. To this end, more existing studies have explained this relationship by decomposing aggregated variables of transport demand (transport intensity) and transport energy demand (transport energy intensity) into coupling and decoupling factors. The main objective of this approach is to identify the main factors that influence energy consumption of the transport sector, especially of road mode, to evaluate their impacts and to propose some solutions some sustainable policy options to reduce energy intensity. Majority of studies have interested on consumption of fossil fuels namely, diesel and gasoline. Ang and Zhang (2000) have proposed a review of literature for this decomposition analysis. Banister and Stead (2002) have studied the energy efficiency and economic efficiency of transport sector for European countries. They have found that passengers-kilometer is not a driving factor for nine European countries; tonne-kilometer is a driving factor of energy transport efficiency growth for six countries and of energy transport efficiency deterioration for eight countries. Zhang et al. (2011) have decomposed the energy consumption in Chinese transportation sector and have found that the transportation activity effect is the important contributor to increase energy consumption in the transportation sector and the energy intensity effect drivers significantly the decrease of energy consumption.

Technique of energy decomposition is largely useful in sustainable transport studies. Several methods of decomposition have been proposed in the literature such as the refined Laspeyres techniques (Lin *et al*., 2008), the Arithmetic Mean Divisia Index (AMDI) and the Logarithmic Mean Divisia Index (LMDI) techniques (Ang, 2005; Liu *et al*., 2007, Hatzigeorgiou *et al*., 2008; Timilsina and Shrestha, 2009).

In this section we derive the methodology to decompose transport sector energy consumption growth to the influencing factors, such as demographic, economic and urban characteristics. Examples include vehicle fuel intensity, vehicles intensity, urbanization, per capita GDP, motorization, road network length, modal mix, fuel mix and more other factors which will be discussed in the section 4.

Transport energy intensity in year *t*( )*<sup>t</sup> TEI* , can be expressed as:

$$TEI\_t = \sum\_{it} TEI\_{ijt} \tag{6}$$

where subscripts *i*, *j* and *t* refer to fuel type (e.g., diesel, gasoline and GPL), transport mode (e.g., road, rail, air and water) and year, respectively. In order to decomposing transport energy intensity into the contribution factors, several formulations can be proposed according the factors integrated in the decomposition and the transport services type (freight transport or passenger transport). It should be note that choice of potential factors is based on availability of time series data and the causal relationship test. For example Eq. 1 can be expressed as

$$\text{TEI}\_{\vec{\text{}}\vec{t}} = \frac{\text{TE}\_{\vec{\text{}}\vec{t}}}{\text{GDP}\_{t}} = \frac{\text{TE}\_{\vec{\text{}}\vec{t}t}}{\text{TE}\_{\vec{\text{}}t}} \times \frac{\text{TE}\_{\vec{\text{}}t}}{\text{TS}\_{\vec{\text{}}t}} \times \frac{\text{TS}\_{\vec{\text{}}t}}{\text{GDP}\_{t}} \times \frac{\text{GDP}\_{t}}{\text{RV}\_{\vec{\text{}}t}} \times \frac{\text{RV}\_{\vec{\text{}}t}}{\text{POP}\_{t}} \tag{7}$$

Eq.2 can also be rewritten as

278 Energy Efficiency – The Innovative Ways for Smart Energy, the Future Towards Modern Utilities

between economic growth and transport sector energy consumption:

 () ( ) ( ) α α

α

α

**3.3. Methodology of decomposition analysis** 

 α

 α

coefficient

EKC model.

α

condition can be written as follows:

of the variable ln*PCGDP* .

1995) and Harbaugh et al. (2002), or a quadratic equation, such as studies of Selden and Song (1994), Holtz-Heakin and Selden (1995), and Stern et al., (1996). Other econometric studies estimate several empirical models (linear, quadratic, cubic). The significance of the cube per capita income is based on the assumption of an N relation. Moreover, two principal types of curves coexist in the EKC literature. The first is the "diachronic" one which used with times series and aims to study the evolution of transport sector energy consumption comparatively with the evolution of income. However, the second is the "synchronic" one which used with cross-section data. The quadratic functional form of EKC is assuming the traditional EKC functional form. Then, we present the following regression model to describe the interaction

> α= + <sup>+</sup>

<sup>2</sup> is negative, functional form of regression model corresponds to the standard

were *E* refers to energy consumption from transport activity and treated as dependent variable, *PCGDP* refers to per capita gross domestic production (PPP) and treated as independent variable and ln indicates natural logarithmic transformation. If the regression

The inverted-U relationship implies that energy consumption is reduced and so environmental quality improves beyond a certain threshold of income per capita. Lind and Mehlum (2007) present the basic properties that must satisfy the U relation. Their main idea insists that the U- inverted should have a positive slope at the beginning of the turning point and negative thereafter. This condition ensures that the endpoint is in the range of data. This

where ln *Min PCGDP* and ln *Max PCGDP* are, respectively, the minimal and maximum values

In order to determine coupling relationship between increase of transport activity and economic growth and also transport energy consumption between, the analysis based on time series models and EKC models are considered as an aggregated analysis which not provides an explanation of the sources of coupling problem and growth of energy consumption. To this end, more existing studies have explained this relationship by decomposing aggregated variables of transport demand (transport intensity) and transport energy demand (transport energy intensity) into coupling and decoupling factors. The main objective of this approach is to identify the main factors that influence energy consumption of the transport sector, especially of road mode, to evaluate their impacts and to propose

0 1 <sup>2</sup> ln E ln *PCGDP PCGDP* ln (3)

2

+ ≥ 1 2 2 ln*PCGDPMin* 0 (4)

+ ≤ 1 2 2 ln*PCGDPMax* 0 (5)

$$TEI\_{\vec{\eta}t} = \sum\_{\vec{\eta}t} \mathbf{M} \mathbf{M}\_{\vec{\eta}t} \times EI\_{\vec{\mu}t} \times TI\_{\vec{\mu}t} \times VI\_{\vec{\mu}t} \times M\_{\vec{\mu}t} \tag{8}$$

were *MM* refers to modal mix which indicates the share of fuel consumption by a mode in total transport energy consumption, *EI* refers to transportation energy intensity per mode which indicated the demand of energy to produce unit of transport services per mode, *TI* represents transport intensity per mode which indicates the demand of modal transport services to produce unit of GDP, *VI* refers to vehicle intensity and measures the demand of vehicles by mode to produce unit of GDP and *M* refers to rate of motorization. The change in these factors summarizes their direct and indirect impacts on change in transport energy intensity. The indirect impacts pass through the influence of demographic, economic and urban characteristics on transport energy intensity.

Transport Intensity and Energy Efficiency: Analysis of Policy Implications of Coupling and Decoupling 281

*r2*= 0.42

*r2*= 0.39

countries.

Mongolia.

10%).

10%.

Elasticity = 0.71

Existence of a long-term relationship between transport intensity and energy intensity in road freight transport.

Existence of an inverted-U form is significant for CO and NOx: \$23.739 at 2000 constant price. \$27.433 at 2000 constant price.

Existence of an inverted-U shape function of income for road transport:

\$21,402 at 2000 constant price.

Transport energy efficiency is

of emission growth for 7 Asian

Economic growth and population growth are the critical factors in the growth of transportation sector CO2 emissions in all countries except

Pass-km is not driving factor for nine European countries (with an impact of

Tones-km is a driving factor of energy transport efficiency growth for six European countries (with an impact more than 10%) and of energy transport efficiency deterioration for eight countries (with an impact lower than

considering an important driving factor

**Causality relationship analysis** 

**Environmental Curve of Kuznets** Study Variables The estimated turning point

**Decomposing analysis method** 

Study Variables Results

Change in CO2 emission from transport for Asian countries. Driving factors: CO2 intensity of a

transportation mode, share of fuel consumption by a mode in total transport sector energy

consumption, transportation energy intensity, economic activity as captured by per capita GDP and

Change in transport energy intensity for European country. Driving factors: passengerskilometer and tones-kilometer.

**Table 1.** Examples of empirical results of transport energy efficiency analysis

fuel, share of a fuel in a

population.

Study Variables Result

Transport energy efficiency and CO2 emission efficiency in road

transport energy efficiency and efficiency of vehicle usage and

Energy intensity in road freight transport and transport intensity

Polluting emission due to the urban daily mobility (quadratic form).

freight transport.

Ubaidillah, 2011 Trend of CO emission from road transport.

Léonardi and Baumgartner, 2004

2011

2009

Mraihi and Hourabi,

Meuni and Pouyann,

Timilsina and Shrestha, 2009

Banister and Stead,

2002

Using LMDI method because their advantages comparatively to Laspeyres techniques as shown by (Ang, 2004)2, change in transport energy intensity (Δ*TEI ijt* ) between two periods can be attributed to effects of:


Consequently,

$$\Delta TEI\_{\rm j\bar{t}} \equiv RT\mathbb{E}\left(T\right) - RT\mathbb{E}\left(0\right) \equiv MM\_{\rm eff} + EI\_{\rm eff} + TI\_{\rm eff} + VI\_{\rm eff} + M\_{\rm eff} \tag{9}$$

Then, effects can be calculated for example for *MMeff* as:

$$\Delta TEI\_{ijt} \equiv \left[ TEI\left(T\right) - RTE\left(0\right) \right] \ln \left[ MM\left(T\right) / MM\left(0\right) \right] / \left[ TEI\left(T\right) - RTE\left(0\right) \right] \tag{10}$$

Growth of transport-related energy intensity can be analyzed among its sensibility to changes in the named direct factors.
