We seek for determinants of the sources of growth. Using a growth accounting method that accounts for time variations in factor shares, we run growth regressions for a panel of 101 countries between 1950 and 2015. Our methodology takes into account the specific features of the data (namely outliers, heterogeneity, and cross panel correlations) and overcomes most criticisms previously raised on growth regressions. The most important evidence reveals that government current expenditure decreases the factor shares and has no effect on total factor productivity (TFP). Trade affects the TFP and the Biased Technical Change (BTC) components, decreasing the factor shares. Moreover, human capital decreases TFP and increases the BTC contribution to growth. This unveils the channels through which determinants of growth act in influencing economic growth.