Predictions in 'a more or less steady state' are just as worthless and unfalsifiable, because the 'steady state' never happens regardless of the size of the shock. Most 'smart models' are getting beat by dumb time-series methods like ARMA even in low-shock periods.