Modeling of Development and Carbon Dioxide Emission in Bangladesh: A Bootstrap Approach

Abstract

In the thesis we study the relationship between CO2 emission per capita and GDP per capita in context of Bangladesh. First, exploratory data analysis (EDA) is used to uncover the hidden information carried by the observed data. Au attempt is made to fit Environmental Kuznets Curve model by means of classical techniques as well as bootstrap techniques. EDA shows that both CO2 Emission per capita and GDP per capita are trended. There is no steady state of the variables within their sample period. This variable does not follow EKC. To test the presence of stochastic trend of the variables we use unit root tests. To mitigate small sample limitations of our data we first design a simulation based study to compare the performance of classical tests with bootstrap tests. Our simulation based result provides that CADF (Covariate Augmented Dickey Fuller Test) test has higher power and BCADF (Bootstrapped CADF) test has less size distortion for testing unit root for small sample of size 30. Unit root tests suggest that both the series are non stationary, i.e., CO2 is accumulating in the atmosphere and GDP is also experiencing accumulation. In the succession of time series modeling ARIMA model is fit­ted to both of the series. Both CO2 emission per capita and GDP per capita of Bangladesh follow ARIMA (0, 1, 1) model. Forecasting by bootstrap produces better results sometimes. In quest of dynamic regression model co integration is checked. Classical, bootstrap and double bootstrap techniques are used to test whether there exists any co integrating relationship among CO2 emission per capita, GDP per capita and it's square. Result shows that these variables are non stationary but not co integrated. There exists no long run equilibrium relationship between CO2 emission per capita and GDP per capita.