Sometimes authors regress the natural logarithms of both the dependent variable (GDP) and all explanatories. Why did they follow this approach? Further, what are the practical implications in interpreting the corresponding results?
Logarithmic transformation reduces the variance of the series and can get rid of collinearity amongst the variables. It can make a non-stationary series stationary. It can improve the fit of the model to the data. There may be other benefits of logarithmic transformation just like a square-root transformation. However interpretation of results can be more difficult to make with these transformations.