**4. Data, model, hypotheses, and methodology**

#### **4.1 Data, model and hypotheses**

The empirical investigation aims to examine the role of ICT in reducing the impact of GDP on CO2 emissions using an ARDL model in the case of Tunisia and Morocco over the period 1980–2018. To do this, we will estimate, on first hand, the direct impact of GDP on CO2 emissions. Therefore, the representation of our models is illustrated:

$$\text{COZ}\_{\text{ir}} = \alpha\_{\text{o}} + \alpha\_{\text{i}} \text{GDP}\_{\text{ir}} + \alpha \text{2X}\_{\text{il}} + \varepsilon\_{\text{ir}} \tag{1}$$

$$\text{COZ}\_{\text{it}} = \alpha\_0 + \alpha\_1 \text{ICT}\_{\text{it}} + \alpha \text{2X}\_{\text{it}} + \varepsilon\_{\text{it}} \tag{2}$$

Where i represents each country in the panel and t indicates the time period; **CO2** refers to the CO2 emissions; **GDP** is the GDP per capita growth (annual %); X refers to the explanatory variables that are: **inf** is the Consumer price index (2010 = 100); **find** is the Domestic credit to private sector (% of GDP); **trade** is the sum of Imports of goods and services (% of GDP) and Exports of goods and services (% of GDP); **invst** is the Gross fixed capital formation (% of GDP).

The expected sign of GDP is positive because economic growth accelerates the level of pollution while the level of CO2 emissions increases (Danish et al. [14]). The expected sign of ∝1 should be positive. This is essential because with the

increase in GDP, people consume more goods; more industries develop so the level of CO2 emissions increases.

On second hand, we will estimate the direct impact of ICT on CO2 emissions. So, in order to study the nature of the relationship between ICT and CO2 emissions, we include ICT proxies, namely, **mcs** is the Mobile cellular subscriptions (per 100 people); **fbs** is the Fixed broadband subscriptions (per 100 people); **fts** Fixed telephone subscriptions (per 100 people); **internet** is the Individuals using the Internet (% of population).

Finally, we study the impact of the interaction between ICT and economic growth on ameliorating the environmental quality. So, we will introduce each time the interaction between GDP and one of the measures of ICT. The inclusion of these interactions allows us to examine whether growing economies are increasing the use of ICTs in different sectors to expand economic activities. We then try to examine whether the increased use of ICTs with growing economic growth positively or negatively affects environmental quality.

So, the representation of the models is illustrated:

$$\text{CO2}\_{\text{ir}} = \alpha\_0 + \alpha\_1 \text{GDP}\_{\text{ir}} + \alpha\_2 \text{ICT}\_{\text{ir}} + \alpha\_3 \text{GDP} \ast \text{ICT} + \alpha\_4 X\_{\text{ir}} + \varepsilon\_{\text{ir}} \tag{3}$$

Hypotheses


#### **4.2 Methodology**

In order to empirically examine whether ICT and economic growth have an effect on CO2 emissions in Morocco and Tunisia, and whether the interaction between ICT and economic growth affects the environmental quality, we applied an Autoregressive Distributed Lag (ARDL) model. This approach is proposed by and subsequently it was modified by introducing the bounds testing approaches. This technique is effective for many reasons. Firstly, it examines the short- and longterm relationships between the different variables that do not have the same order of integration. Secondly, it can solve the problems associated with autocorrelation and omitted variables. Finally, this approach can be useful for a small sample size application.

Before the data are further analyzed, it is necessary to demonstrate the stationarity of all variables. In fact, in order to arrive at robust empirical results, all estimated variables should be non-unit root. In this case, we use four unit root tests Augmented Dickey-Fuller (ADF), Phillips-Perron (PP 1988), Dickey-Fuller GLS (DF-GLS), and KPSS (Kwiatkowski-Phillips-Schmidt-Shin) unit root; whose critical threshold is 5%, and with a null hypothesis (H0) of non stationarity of the variable. The ARDL bounds test is based on the assumption that the variables are I(0) or I(1). So we use the unit root tests in order to make sure that the variables are not I(2) because if variables are I(2), we cannot interpret the values of Fi statistics provided by.

The next step is to test the presence of cointegrating relationships among the variables. To do this, we use the bounds test that is mainly based on the joint Fi statistics whose asymptotic distribution is non-standard under the null hypothesis of no cointegration. Once cointegration was established, we estimate the long- and short-run relationship between ICT, economic growth, and CO2 emissions using ARDL model. In order to obtain the dynamic parameters in the short run, we estimate a correction error model associated with long-run estimates (Odhiambo [35]). The short-run causal effect was represented by Fi statistics on explanatory variables, while the t statistic on the lagged error correction coefficient represents the causal relationship in the long run.

Finally, we examine the causal relationship between ICT, economic growth, and CO2 emissions using Toda-Yamamoto Granger causality test. This method is based on the estimation of augmented VAR model (k + dmax) where k is the optimal time lag on the first VAR model and dmax is the maximum integrated order on system's variables (VAR model). To do so, it is necessary, firstly, to determine the integration order for each series using AIC and SC criteria. If the integration order is different

we get the maximum (dmax) and we create a VAR model (VAR (k + dmax)) on series levels regardless of integration order that we found. However, if we have the same integration order, we continue on cointegration test using Johansen methodology.
