I have attached a very simple example file that illustrates what I found out in the meantime with the help of a colleague, namely that STEADY_STATE switches in the period that a change in the exogenous variable is known. The example file should be self-explanatory.
Note, you can run the code for ANTICIPATION = 0 where the permanent shock is unanticipated or ANTICIPATION = 1 where (if I coded this correctly), the shock is permanent but happens only at t=10. Also in this latter case STEADY_STATE() switches to the final SS in t=1.
PermShock.mod (503 Bytes)