Back to Community
Using macro economic data in models - should we use levels or differences?

It is known that for many statistical models (e.g. PCA or classification machine-learning models) one should input the returns and not the price levels. So my question is, assuming we wanted to use macroeconomic data, e.g. inflation, to help modelling financial series, would one need to input them as differences (as in the case of prices with 'returns') or should one use the actual levels in this case? Thanks for your thoughts!