Compositional Linear Model Clarifies Country-Level Energy Mix Drivers

The energy mixes, which describe energy consumption structure by fuel type, are complex compositional data that cannot directly apply the multivariate statistical methods. Consequently, we apply a special compositional linear regression model to study how the country-level energy mix is influenced. The main findings of 45 countries by income group over 1990-2019 are: 1) The dependence of energy use on energy resource for each fuel type is confirmed in upper-middleincome countries. 2) Non-fossil use is driven by hydroelectricity resources and economic level in highcome and upper-middle-income countries, and by oil resources in lower-middle-income and low-income countries; 3) The energy transition as economy grows presents two typical patterns: high-income and uppermiddle-income countries shift to non-fossils and natural gas, while the lower-middle-income and low-income shift to coal and oil.