2Department of Data Science and Analytics, Giresun University, Giresun, 28200, Türkiye
Abstract
This research examines Türkiye's carbon dioxide emissions from 1973 to 2021. As carbon dioxide emissions have become one of the key factors in climate change related policy discussion. This study aimed to determine which economic factors or industry sectors affect the level of carbon dioxide emissions. Therefore, robust regression, least squares and generalized least squares techniques are used on carbon dioxide emissions data set. Robust regression method outperforms traditional regression models in comparison to least squares and generalized least squares. Robust regression yields more reliable and efficient results by eliminating the effects of outliers and enhancing the stability of the model. Model performance of these regression estimation techniques are evaluated using mean squared error. Model diagnostics are performed to check the fundamental regression assumptions including heteroscedasticity,
normality, autocorrelation, and multicollinearity. When multicollinearity is present, variables are eliminated using backward elimination method. After elimination of variables, all regression estimation techniques are used again to determine which model fits best. Autocorrelation and heteroscedasticity are corrected if there are present. Our findings show that carbon dioxide emissions are affected by per capita energy use, carbon footprint per capita, and secondary sector’ share of gross domestic product. According to the mean squared error, the robust regression model outperforms all other models. We conclude that Türkiye’s growth has been followed by increasing dependence on fossil fuels. Policies encouraging renewables and energy efficiency should become a priority.