The studies have focused on changes in CO2 emissions over different periods, including the COVID-19 pandemic. Even if CO2 emissions are temporarily reduced during the pandemic according to annual figures, this may be misleading. Considering annual figures is important to understand the overall trend, but using data with much higher frequency (e.g., daily) is much better suited to investigate dynamic relationships and external effects. Therefore, this study comprehensively analyzes the association between CO2 emissions and disaggregated electricity generation (EG) sources across the globe by employing the novel wavelet local multiple correlation (WLMC) approach on daily data from 1st January 2020 to 31st March 2023. The results demonstrate that (1) based on the main statistics, daily CO2 emissions range between 69 MtCO2 and 116 MtCO2, indicating that there is an oscillation, but no sharp changes over the analyzed period. (2) based on the baseline regression using the dynamic ordinary least squares (DOLS) approach, the constructed estimation models have a high predictive ability of CO2 emissions, reaching ~ 94%; (3) in the further analysis employing the WLMC approach, there are significant externalities between EG resources, which affect CO2 emissions. The results present novel insights about time- and frequency-varying effects as well as a disaggregated analysis of the effect of EG on CO2 emissions, demonstrating the significance of the energy transition towards clean sources around the world.