A note on the optimal speed of transition : Aghion and Blanchard revisited