**3.1. Causality and cointegration relationship between transport energy consumption, transport activity and economic growth**

An open question for relationship between transport activity, economic growth and environmental effects is the correlation between their trends. Negative environmental effect will be increasing when correlation between economic growth and transport is largely important. Exam of this correlation is important in so far as it provides a several instruments to elaborate an efficient transport policy. In literature, a large majority of studies have interested to determine separately correlation between first, transport activity and economic growth (coupling problem) and secondly transport activity and energy consumption and gas emission. In this chapter we propose a demarche which treats simultaneously both the tow problems. We attempt to include in our analysis dimension of sustainable transport for which actual policies transport have been taken.

In this context, a large number of studies have interested to the problem of coupling. An important number of institutional reports had elaborated like REDEFINE (1999) and SPRITE (2000) in Europe. Otherwise, several scientific studies have been elaborated in two directions. First, studies which estimate relationships between transport of passengers or goods and economic growth using previous traffic models (Meersman and Van De Voodre (1999) for the case of Belgium and Klushershtha and al. (1999) for India). Second, studies which have interested to aggregated indicators in order to estimate the coupling (Baum (2000) and Baum and Kurte (2002) have used the intensity of road transport to measure coupling for the case of Germany and Stead (2001) for the case of Europe).

In order to study the relationship between transport energy consumption, transport activity and economic growth, we should test the stationary of the series. To this end tests which are common use are Augmented Dickey-Fuller (ADF) tests (Dickey and Fuller, 1979) and Phillips-Perron (PP) tests (Phillips and Perron, 1988). The objective is to know if the series are stationary in levels or in some order of differentiation. If the integration of the two series is of the same order, we should test whether the two series are cointegrated over the same period. Analyze of cointegration between the series is often realized through method of Johansen (1988).

The Trace and maximum eigenvalue test provide us the information concerning the presence or nor of the cointegration. Consequently, we can estimate a vector error correction model (VECM) that incorporates variables and variation levels for information on the speed

of adjustment to equilibrium. Engle and Granger (1987) showed that if two series are cointegrated, the VECM for the two series can be written as follows:

$$
\Delta y\_t = \alpha + \sum\_{i=1}^k \beta\_i \Delta y\_{t-i} + \sum\_{i=1}^k \lambda\_i \Delta x\_{t-i} + \eta ECT\_{t-1} + \mu\_{1t} \tag{1}
$$

Transport Intensity and Energy Efficiency: Analysis of Policy Implications of Coupling and Decoupling 277

**3.2. Relationship between transport activity and transport energy consumption :** 

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

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;

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

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)

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,

(Grossman and Krueger, 1991; Bandyopadhyay, 1992; Panayotou, 1997).

**estimation of EKC** 

Storchmann, 2005; Van Dender, 2009).

private cars and its related fuel consumption.

corresponds to an inverted U-shape of the EKC.

$$
\Delta \mathbf{x}\_t = \alpha + \sum\_{i=1}^k \beta\_i \Delta \mathbf{x}\_{t-1} + \sum\_{i=1}^k \lambda\_i \Delta y\_{t-i} + \eta \mathbf{E} \mathbf{C} T\_{t-1} + \mu\_{2t} \tag{2}
$$

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 that measures the magnitude of past disequilibrium. The coefficient η represents the deviation of the dependent variables from the long-run equilibrium. The significance of the explanatory variables coefficient ( λ*<sup>i</sup>* and β*<sup>i</sup>* ) confirms the presence of short-run causality.

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

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 variables in each equation of VECM using the χ<sup>2</sup> -Wald statistics. The Causality long-run 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 can conclude if the two series are coupled or uncoupled.

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 is sustainable or not.
