According to *Hamilton (1994)*, page 5 f. the effect on y(t+1) from a permanent increase in *w*. Therefore the long-run effect is 1/(1-theta). However, as far is i know the term is far more common for VEC models. Hope this helps a bit.