Nexus of co2 emissions and economic growth in pakistan: analysis by using extended stirpat model

The main purpose of this study is to analyze the relationship between economic growth (GDP per capita) and CO2 emissions in Pakistan. This study applies the theoretical framework of Dietz and Rosa’s STIRPAT Model, widely used for assessing the environment quality. The additional major determinants of CO2 emissions introduced by the extended STIRPAT model include total energy use, Industry, value-added, nancial development, trade openness, and urban population. The empirical results reveal that; total energy use has a positive and signicant relationship with CO2 emissions. The relationship between GDP and CO2 emissions is positive and insignicant in the country. Industry, value-added has an insignicant relationship with CO2 emissions in the country. The urban population has a direct and positive relationship with CO2 emissions. Trade openness has a long-run positive and signicant relationship with CO2 emissions in the country. In general, this case study offers a relevant policy for controlling the enhancement in the CO2 emissions in the selected sample unit; Pakistan, and other similar states that possess the same socioeconomic condition. Auence, and Technology (STIRPAT) Model. Very few studies are conducted to investigate the possible impact of economic growth (GDP per capita) on CO 2 emissions in Pakistan. Moreover, these studies ignore the combined effects of other major contributors to CO 2 emissions, in the country. This study overcomes these limitations by employing an extended STIRPAT model that studies the impact of GDP on CO 2 emissions, with the inclusion of major explanatory variables; industry value-added, total energy use, nancial development, trade openness, and urban population into the analysis. unit Lage Hatemi-J cointegration test approach to the cointegration among variables. to increase robustness of the ndings in the presence of structural breaks, the Hatemi-J cointegration test is ARDL with is to estimate the long-run relationship between CO 2 emissions and explanatory variables; GDP, energy use, nancial development, openness, and urban population. The short-run estimations are by employing Error Correction Model within

Based on the ndings of the Germanwatch Climate Risk Index annual report 2020, the focus area of this inquiry is Pakistan. For this purpose, this study analyzed the nexus of economic growth and CO 2 emissions within the theoretical framework of the STIRPAT model in the country. Moreover, the additional anthropogenic activities that cause CO 2 emissions are incorporated by the extended STIRPAT model, includes energy use, urban population, nancial development, and trade openness.

PROBLEM STATEMENT
Enhancement of CO 2 emissions is one of the alarming issues of the contemporary world because it is one of the major root causes of climate change and global warming. To combat climate change and its impacts controlling CO 2 emissions should be rst taken into consideration. Numbers of efforts had made worldwide to mitigate CO 2 emissions, the focus area of this study is Pakistan. The rationale for choosing this sample unit is based on the Climate Risk Index annual report published by Germanwatch in 2020 [2]. According to Climate Risk Index 2020, Pakistan is the fth most affected country in the 21st century. The mitigation of CO 2 emissions is only possible through the identi cation and control of its determinates. For this purpose, this study is an attempt to analyze the nexus of economic growth and CO 2 emissions within the theoretical framework of the Ditz and Rosa STIRPAT model in the selected sample unit Pakistan.

RESEARCH OBJECTIVES
1. The objective of the research is to inquire about the association of CO 2 emissions with the gross domestic product (GDP) and related measures in Pakistan 2. To nd out whether the association is short-term or long-term betweenCO 2 emissions and GDP.

RESEARCH QUESTIONS
Q1. Why there is an association between CO 2 emissions and GDP per capita in Pakistan?
Q2. What type (long or short run) of a relationship exists between CO 2 emissions and GDP per capita in the country?

SIGNIFICANCE OF THE STUDY
Climate change and global warming are the two immense problems of the present world and one of the major reasons for its happening is CO 2 emissions. This research study analyzed the connection between CO 2 emissions (dependent variable) and GDP. Based on the available literature, it is found that this research topic is very little explored in the context of Pakistan, so this research study is very useful in this regard. The ndings of this study are very useful to the students and researchers who are going to conduct research studies in this particular area of climate change, global warming, CO 2 emissions, and its possible determinants. Moreover, this study enriches the available literature on this speci c topic of analyzing the connection between CO 2 emission and GDP, and other major contributors of CO 2 emissions. It is also very useful for the governments and policymakers while formulating the climate change and environmental policies in Pakistan.

CONTRIBUTION OF THE PAPER
The current research work devotes to the existent body of literature on the determinants of CO 2 by bringing up the STRIPAT model in the context of Pakistan in its rst attempt as far the knowledge of researches, to analyze the interaction between CO 2 emissions and GDP, with the inclusion of other major explanatory variables: Urban population, nancial development, total energy consumption and trade openness. This study updates the previous work by inducting different variables, collectively into a single extended STRIPAT model. Moreover, introducing the dynamic time series econometric estimation models; the Autoregressive Distributed Lag Model in the presence of the structural breaks and the Error Correction Model within ARDL methodology, enhance the quality of the empirical ndings. This study also ensures the robustness of results by incorporating the advanced cointegration technique: Hatime-J cointegration tests, into the analysis.

ORGANIZATION OF THE PAPER
The rest of the paper is organized in this order. The rst portion of the paper is consisting of an introduction, problem statement, research objectives, research questions, and research signi cance. The second portion of the paper is consisting of a literature review and literature gape. The third portion is consisting of theoretical framework, data analysis techniques, type of study, unit of analysis, time horizon, and econometric estimation models. The fourth portion of this study is consisting of empirical results interpretation and analysis in the context of Pakistan. The fth chapter is consisting of a conclusion and relevant policy recommendations for Pakistan, The bibliography and the supporting materials to the study are attached in the appendix.

Literature Review
The association between CO 2 emissions and its driving forces is studied both in developed and developing countries. During these studies, several econometric models are being used to analyze the interaction between CO 2 emissions and their impact factors. As Ghosh and coworkers used the VAR model [5], Begum and coworkers used the ARDL bounds testing approach [6], Hammami and Saidi [7] used the Generalized Method of Moments model and Ghazali and Ali [8] used extended STIRPAT model. This particular study is based on an extended STRPAT model so the literature reviewed in this section will be mostly using the STRIPAT model to analyze the relationships between CO 2 emissions and its driving forces.
Ghazali and Ali. [8] analyzed the impact of different carbon emissions drivers in the new technological advanced countries for the years between 1991 and 2013 by employing an extended STIRPAT model. To study the relationship between CO 2 emissions and economic growth, technology, population, and extensive factors energy mix, energy intensity, the productivity of labor, employment level of urban population and trade openness by using the regression (group means dynamic commonly correlated estimator). To check the robustness of the results of group mean a dynamic common correlated estimator is compared with ordinary least square techniques. According to the ndings of the study carbon emissions intensity, energy intensity, gross domestic products, and population are the main causing agents of CO 2 emissions.
Xue et al. [9] studied the connection between energy intensity, GDP, population and urbanization with the CO 2 emissions in the Yangtze River Delta, China. The time for this study was 1990 to 2011. This study was based on the STIRPAT model. To predict different scenarios of CO 2 emissions between the years, 2015 to 2020 vector machine model was made. According to the ndings, CO 2 emissions had increased for both the time from 1990 to 2011 and 2015 to 2020. Moreover, energy intensity nds to be the main in uencing force of CO 2 emissions while the population is the lesser driving force of CO 2 emissions. But the in uencing position of energy intensity had decreased over time in the long term. The paper suggests that the association between urban population and CO 2 emissions considered being U-shaped. The gross domestic product had more in uence on CO 2 emissions than urbanization and population.
Ahmad et al. [10] studied the interaction between the sector of construction, consumption of energy, gross regional product, urban population, and carbon dioxide emissions by employing the STIRPAT model.
Augmented mean group and dynamic mean group are used to estimate the panel of china and the other three disaggregated regional panels. According to the ndings of the study, the association between carbon dioxide emissions, the construction sector, energy consumption, gross regional product, and urbanization was found to be long-run equilibrium.
Yang et al. [11] stated that the growing trend of CO 2 emissions varied across the years from 1995 to 2014 in Zhejiang, China. This study employed an extended STIRPAT model for its conduction. The results of the study showed that the urbanization level and gross domestic product were the major driving forces of the CO 2 emissions from 1996 to 1999 in Zhejiang, China. While economic development and foreign trade were the major contributors to CO 2 emissions from 2000 to 2014.
It is noticed in the review of the literature that different studies used different determinants of CO 2 emissions. The maximum number of studies that applied the STIRPAT model is carried in developed countries, and very little focus is given to developing regions. Studies conducted to analyze the impact of Economic growth on CO 2 emission in the context of Pakistan ignored the combined effects of other main determinants/ Explanatory variables of CO 2 emissions. Moreover, the structural breaks in the series due to geopolitical and economic changes are overlooked during estimations. In addition, the results of the review studies are not consistent with each other, and vary from country to country and study to study. To ful ll these gaps in the literature, this study is conducted on a selected sample unit; Pakistan.

THEORETICAL FRAMEWORK
Dietz and Rosa's Stochastic Impacts by Regression on Population, A uence and Technology (STIRPAT) model [12] is used to conduct this study. STIRPAT model is not the very rst attempt to study the environmental impact through socioeconomic variables. The STIRPAT model is derived from Ehrlich and Holdren's (1971) IPAT model [13]. Howe ever the IPAT model was reformulated from Duncan's (1959) POET model.

STIRPAT
According to Dietz and Rosa [12], all the in uencing factors included in the IPAT model effects the deviations in the CO 2 emissions. IPAT model was reformulated into a stochastic model by Dietz and Rosa due to the criticism on the very base of the IPAT equation. The criticism of the IPAT model is described in the next paragraph.
Comprehensive criticism had been made on the IPAT model. According to Scholz [14], the IPAT model only inducts demographic and economic forces into the IPAT equation while ignoring the other factors/forces. Moreover, studying the relationship between the variables the concept of proportionality is taken into consideration, which is imposed by the basic principle for accounting equations. York et al. [15]  Where "I" represent environmental impact, "P" represents the population, "A" represents a uence and "T" represents technology.
Dietz and Rosa [12] presented the STIRPAT model to explain the driving forces that are impacting the environment. STIRPAT model simultaneously combines economic, societal, and technological factors to resolve the environmental problems. Here in this study in equation (1), I is the environmental issues that are represented by carbon dioxide emissions, P is the societal parameter that is represented by the urban population. A is the economy parameter that is represented by gross domestic product. T is the technological parameter that is represented by the utilization of renewable energy.
Econometric equation is sciences with the ecological sciences. Because of this character hypothesis in the eld of social sciences that are linked with ecological sciences can be tested through this model. Thus variables from the other elds such as economics, political and cultural sciences can be inducted into the STIRPAT model equation .

Time horizon
This study covers 42 years of data from 1975 to 2016. The period of the study is based on the updated data provided by World Bank till November 1 st, 2020.

Elementary data analysis
Data collection is followed by elementary data analysis. Elementary data analysis in this study includes; a brief description of descriptive statics and the formulation of the PairWise simple correlation coe cient matrix.

Econometric analysis
The econometric analysis of this study includes: 1) Unit root tests 2) Model selection 3) Cointegration tests 4) Dynamic econometric models for the estimations of the long-run and the short-run coe cients. 5) Post estimation diagnostic tests.

Unit root tests
Checking the stationary of data is the rst step to do econometric analysis through dynamic econometric models. For this purpose, conventional unit root tests such as Augmented Dickey-Fuller (ADF) unit root test by Dickey and Fuller [16], Phillips and Perron [17] unit root test is applied. The ndings of the Phillips Perron unit root test are considered superior to those of the ADF unit root test as it eliminates the limitations of serial correlation and heteroscedasticity, proven in Augmented Dickey-Fuller. However, it is observed that the conventional unit root tests may present biased and misleading results in the presence of structural breaks in the data Shahbaz et al. [18]. To overcome the shortcoming of the Augmented Dickey-Fuller unit and Phillips-Perron (PP) unit root test, Zivot and Andrews [19] unit root test is also incorporated accompanied.

Criteria for Model selection
The robust estimation of econometric data can be obtained through the application of dynamic econometric models. The selection of the econometric model is based on the order of integration of data. Figure

Autoregressive Distributed Lag Model
For the sake of estimating the long-run relationship between the dependent variable and independent variable, this study employs the ARDL-bounds co-integration approach, introduced by Pesaran et al. [20]. As the name of the ARDL indicates that ARDL is a distributed lag model that incorporates the lagged values of the dependent variable and the current and lagged values of the explanatory variables. ARDL model takes into account both the exogenous and endogenous variables. Unlike the other models, the ARDL model can be applied in heterogeneous situations: the data is purely stationary at level, at rst order, or the mix of both.
However, the Autoregressive lag distributed model cannot be speci ed at the second difference of the series.
The general Autoregressive Distributed lag model for this study is speci ed as;

ARDL-Bounds test for Cointegration:
The ARDL-Bounds test for cointegration is used to analyze the long-run relationship between variables within ARDL Methodology.
The null hypothesis for ARDL-cointegration states there is no cointegration between variables.
(H 0: constant term= long run coe cients=0) The alternate hypothesis for ARDL-cointegration states there is a cointegration relationship between variables.
The decision to either validate the null hypothesis or alternate hypothesis is based on the F-statics value of the ARDL-bounds test. The F-statics is compared with the upper bound and lower bound critical values introduced by Pesaran et al. [20] and Narayan [21]. The F-statics values higher than the upper bound critical value validates the alternate hypothesis. Whereas, the F-statics value lesser than the lower bound critical value validates the null hypothesis. However, the F-statics value lies between upper and lower bound values neither validate the null hypothesis nor the alternate hypothesis, and the decision remains inconclusive.

ARDL-ECM
Once the cointegration relationship is con rmed between the variables through the ARDL-Bounds test, and long-run coe cients are estimated, the short-run coe cients are estimated through ARDL-ECM estimations.
Then general ARDL-ECM for this study is speci ed as:

Hatemi-J Cointegration tests
To increase the robustness of the ARDL bound estimations, this study incorporates the Hatemij [22] cointegration tests for analyzing the long-run relationships between the variables. The Gauss formulated coding program is employed for Hatemi-J co integration's estimations.
The time-series data consisting of the macroeconomic indicators come across various shocks, put forward by the economic, nancial, and technological changes, regime shifts, and other related factors. Numerous conventional cointegration techniques are used to study the relationship between the macroeconomic variables, but its estimations are not authentic due to structural breaks or shocks within the series. Hatemi-j apprehended the limitations of traditional cointegration tests through the modi ed techniques that analyze the cointegration relationship between the variables in the presence of regime changes or structural breaks endogenously.
Monte Carlo simulation techniques regard the validity of the estimation of Hatemi-j cointegration tests. Gregory takes the lead to analyze the cointegration relationship within variables in the existence of one structural break, and Hatemi-J advanced it to cover two unknown structural breaks. AH-J 3 cointegration test with two unknown structural breaks:

Structural Break and Structural Dummies
The dummies incorporated in the ARDL model are based on the breaking points/structural breaks computed in the Hatemi-J cointegration techniques.
This study used the following structural breaks analysis techniques: 1. Z-Andrew unit root test with structural breaks 2. G-Hansen cointegration test with a regime shift 3. AH-J cointegration test with two unknown structural breaks.
The ZA and GH techniques compute the one unknown structural break, while the Hj de nes two unknown structural breaks. HJ test is the advanced form of Zivot Andrews and Gregory Hansen cointegration techniques, and its estimations are more preferable to the prior one. Hence this study incorporated the Hatmi-J cointegration de ned structural breaks in long-run and short-run econometric analysis.

Elementary Data Analysis
The elementary properties of the data are discussed in this chapter. The values of Jarque-Bera and its probability are placed in the 10th and 11th rows of the table. These values measure the normality of the skewness and kurtosis. Jarque-Bera tests the null hypothesis of normal data distribution. As reported, the Jarque-Bera test statistics probability is higher than 0.05% across the table, so the null hypothesis of the normal distribution is accepted for the entire set of variables.

Unit root tests results
This study used three different unit root tests; Augmented Dickey-Fuller, Phillips Perron, and Zivot Andrews for checking the stationary of the data. The results are displayed and discussed in this section.

Phillips Perron Unit Root Test
The result of the PP unit root test at intercept and intercept & trend are listed in tables 6 and

Zivot Andrew Unit Root Test
The results are listed in table 4 for the Zivot-Andrews unit root test with structural breaks.

Results of cointegration tests:
The result of the H-J cointegration tests with multiple structural breaks is listed in these sections H-J cointegration test with two unknown structural breaks The result of the H-J test for co-integration with two unknown structural breaks is displayed in table 5.
The variables among whom the cointegration relationship is tested include dependent variables; CO 2 emissions and impendent variables: total energy use, nancial development, gross domestic product, trade openness, and the urban population.  Table 5 shows that the Zt and ADF statics values are higher than 1%, 5% and 10% asymptotic critical values, which indicates the cointegration relationship among variables with two unknown structural breaks in the constant and regime shift at 1%, 5% and 10% level, Hence the null hypothesis of no cointegration is

Results for Cointegration Dummies:
The dummies incorporated in the ARDL model are listed in Table 6. Source: Author Calculation base on Hatemi-J cointegration estimation in GAUSS 21 Table 6 contains the list of dummy incorporated into the ARDL model for the estimations of long-run and short-run coe cients. This dummy is used as xed regresses in ARDL to ensure robust econometric estimations in the presence of structural breaks in the series.

Results for Autoregressive Distributed Lag Model
The results for ARDL model are presented in tables 7, 8, and 9 respectively.

Results of ARDL-Bounds test for cointegration:
To analyze the cointegration between the dependent variable and explanatory variables, the ARDL bound test within the ARDL methodology is applied.
The variables between whom the cointegration relationship is tested include the dependent variable: CO 2 emissions, and explanatory variables: total energy use, nancial development, gross domestic product, industry value-added, trade openness, and the urban population.
The results of ARDL-Bounds Cointegration test are listed in table 7. The empirical results of the ARDL long-run relationship coe cients between dependent variable: CO 2 emissions and independent variables: energy use, nancial development, gross domestic product, trade openness, and urban population are presented in table 8. Table-8 Long Run Coe cients Estimates based on selected ARDL (1, 0, 0, 1, 1, 1, and 0  The long-term association between the gross domestic product and CO 2 emission is positive but insigni cant in Pakistan. The results of this analysis are inconsistent with those of [10,[23][24][25].
Similarly, the long-term association between industry-value added and CO 2 emission is negative but insigni cant in Pakistan. This result is consistent with those of Lin et al. [28].
Moreover, the long-run relationship between urban populations is also positive and signi cant in Pakistan.
The results of long-run coe cients indicate that the 1% rise in the urban population increases the CO 2 emissions by 2.36% in the country. Due to enhancement in the urbanization, energy demand will boost up in regards to housing, business units, transportation, and services industry. This condition will lead to high energy consumption, which will ultimately increase CO 2 emissions. The results of this study for the causal relationship between urbanization and CO 2 emission is consistent with the ndings of [10,11,24,27].
It is indicated from the results that the long-term association between trade openness and the emission of CO2 is positive and signi cant in Pakistan. The nding shows that a 1% increase in trade openness increases the CO 2 emissions by 0.17% in the country. Trade openness in this study is the total sum of imports and exports. An increase in exports & imports enhances economic activities and mobilizes the business units to produce more by consuming high energy in the production processes. As discussed earlier in this section more energy consumption means more greenhouse gas emissions. Thus enhancement in trade openness increases CO 2 emissions. Our results support the studies of Wheeler and Martin-Vide [29], and Krueger [30]. However, the long relationship between nancial development and CO2 emissions is insigni cant. Dogan and Turkekul, [31] analysis supports our insigni cant result for the long-run relationship between the two variables.
The long-run relationship between nancial development and CO2 emissions is insigni cant. Dogan and Turkekul, [31] analysis supports our insigni cant result for the long-run relationship between the two variables.
According to ARDL long-run estimations, the major contributor to CO2 emissions in Pakistan is the enhancement in urban population, followed by high energy consumption and an increase in trade openness, respectively.

Autoregressive Distributed Lag Model-ECM estimations
The ARDL-ECM Short-run estimations are presented in table 9.  Table 9 reveals that the value of the error correction term is negative and signi cant, which is the indication of the long-run relationship between the emissions of CO 2 and its explanatory variables in Pakistan.
The error correction term measures the correction of uctuation of the short-run from the long-run equilibrium; this correction of deviation of a short-run from a long-run, towards the long-run equilibrium is called the speed of adjustment. The speed of adjustment or correction of short-run deviation from the longrun equilibrium is 95% each year in Pakistan.
The ndings of short-run coe cients are persistent with those of long-run for the relationship between gross domestic product and CO 2 emissions; industry-value-added and CO 2 emissions. The short-run coe cient for the relationship between trade openness and CO 2 emissions is positive but signi cant, which is inconsistent with the long-run coe cient for the two variables.
The short-run coe cients for the relationship between nancial development and CO 2 emissions; energy use and CO 2 emissions; and urban population and CO 2 emissions are eliminated due to its optimal lag length of zero in selected ARDL (1, 0, 0, 1, 1, 1, 0).model.

Diagnostics tests:
The results of diagnostic tests for ARDL and ARDL-ECM are listed in table 10.  The key results of the study are as follows: 1. Gross domestic product has a positive and insigni cant relationship with CO 2 emissions in the context of Pakistan.
2. Industry, value added has an insigni cant relationship with CO 2 emissions in the country.
3. The total energy use has a positive and signi cant relationship with CO 2 emissions in the country.
4. The urban population has a direct relationship with CO2 emissions in Pakistan.
5. The long-run relationship between nancial development and CO2 emissions is insigni cant in the country.
. Trade openness has a long-run positive relationship with CO 2 emissions in the country.

Policy Recommendations
Based on the results this study has the following relevant policy recommendations: 1. The gross domestic product shows a positive but insigni cant relationship with CO2 emissions in Pakistan. It implies that with the increase in GDP per capita, the CO2 emission also increases in the country. Therefore, it is required to adopt economic policies that promote clean production which will reduce the CO2 emissions caused by the increase in GDP per capita.
2. However, the industry value added has an inverse and non-signi cant relationship with CO 2 emissions in the country, which is a good sign, as the results imply that enhancement in industryvalue addition improves environmental quality in the country.

The relationship between trade openness and CO2 emissions is positive and signi cant in
Pakistan. The result implies that expansion in trade (exports plus imports) increase the CO2 emissions in the country. An increase in exports & imports enhances economic activities and mobilizes the business units to produce more by consuming high energy in the production processes. More energy consumption increases greenhouse gas emissions, which in turn leads to an enhancement in CO2 emissions. Therefore, these countries require adopting export and importing policies that are aligned with environmental protection and sustainability.
4. The relationship between nancial development and CO 2 emissions is positive and signi cant in the country. It infers that when nancial development increases the CO 2 emissions increase. This result implies that the banks' domestic credit giving criteria are not aligned with the protocol or guidelines of environmental stability. Therefore, the country requires adopting credit-giving policies, which are aligned with sustainable environmental practices.
5. The relationship between urban population and CO 2 emissions is positive and signi cant in the country. This is because, at the early stages of urbanization, people rush into the urban area and utilized the available energy resources without any environmental considerations, which enhances the CO2 emissions. Moreover, the process of urbanization accelerates the economic activities which compel the rms to produce more in a competitive environment. As discussed earlier the more production is linked with more consumption of energy. And more energy is consumed more greenhouse gas (CO 2) emissions expel out to the environment. Therefore the country requires urban policies that are aligned with the compact city theory and ecological modernization theory, according to which the urban cities go through a social and institutional transformation in favor of environmental improvement. Moreover, the shift towards a piece of knowledge and service-based economy from agricultural orientation might also reduce CO 2 emission in a long run.
. Energy used prevails a signi cant positive relationship with CO 2 emissions in Pakistan. It implies that the primary energy consumption in the shape of; biomass, animal products, gas, and liquid extracted from biomass, and industrial waste are the major contributors to the CO 2 emissions in the country. Therefore policy shift from nonrenewable energy resources (burning of biomass) towards renewable resources (solar, wind, and hydro energy) is required to alter the adverse effect of total energy use on the environment; CO 2 emissions.  This is a list of supplementary les associated with this preprint. Click to download.