COMPUTING MARKOV-PERFECT OPTIMAL POLICIES IN BUSINESS-CYCLE MODELS