# Should stationarity be checked for *both* time series models (ARIMA, ECM…) and causal models with explanatory variables?

I read this paper that discusses if Time Series models or Causal models are the best for forecasting GDP.

I am familiar with unit root tests (ADF, PP) and stationarity test (KPSS) applied to a time series to establish if it is stationary, non-stationary or trend-stationary.

My question is, should stationarity be also verified for causal models (in order to discard spurious regression with very high $$R^2$$? In the linked paper, the author doesn't check it.