How to Model Forward Guidance and Address a Larger Puzzle

Forward guidance during the zero lower bound period is typically modeled as news that alters the expected liftoff date of the policy rate, assuming that agents do not expect a policy rate hike in near future. Using U.S. high-frequency data, I empirically reject this assumption and show that forward guidance affects the entire term structure of expected rates. Introducing this estimated forward guidance shock in a standard New Keynesian model substantially magnifies the "forward guidance puzzle", i.e. the excessive model-implied response to forward guidance. I show that allowing agents to update their macroeconomic expectations in the pessimistic direction following a forward guidance easing explains this larger puzzle per se, unlike the common approach of introducing a discount parameter due to a deviation from a baseline assumption. In addition, I find that the puzzle can also be explained by sticky information general equilibrium models