Bootstrapping Unit Root Test for Small Sample: Application on Gross Domestic Product and Carbon Dioxide Emission in Bangladesh

Abstract

power of conventional unit root tests is very low and suffers a severe size distortion problem in small sample time series. A simulation-based study is done to extact the perJo nance of bootstrap on Augmented Dickey-Fuller (ADF) qnd Covariance Augmented Dickey-Fuller (CADF) tests. It is found that for small sample (n = j0) bootstrapped CADF test perforns relativeLy better than all other tests. We perform these test on GDP per capita and Carbon Dioxide (COz) enission per capita of Bangladesh where data are available frort 1972-2000 i.e. the data size is only 29. The test result shows that CO2 emission per capita and GDP per capita in Bangladesh are unit root process, GDP per capita performs better as a covariate for applying CADF test on CO2 emicin in Bangladesh.