Does My Model Predict a Forward Guidance Puzzle?

Abstract

We provide sufficient conditions for when a rational expectations structural model predicts bounded responses of endogenous variables to forward guidance announcements. The conditions coincide with a special case of the well-known (E)xpectation-stability conditions that govern when agents can learn a Rational Expectations Equilibrium. Importantly, we show that the conditions are distinct from the determinacy conditions. We show how the conditions are useful for diagnosing the features of a model that contribute to the Forward Guidance Puzzle and reveal how to construct well-behaved forward guidance predictions in standard medium-scale DSGE models.

Resolving New Keynesian Puzzles

“…A further implication of Figure 4 is that the stabilization outcomes that are possible under optimal policy are not due to the forward guidance puzzle. Gibbs and McClung (2023) provide a sufficient condition to rule out the forward guidance puzzle in an equilibrium. It is straightforward to numerically verify that optimal policy satisfies it under standard calibrations.…”

“…For example, Del Negro, Giannoni and Patterson (2023) makes use of a perpetual youth model to micro found a reduction in the horizon of agents' expectations delivering the same ameliorative effect on forward guidance as imperfect credibility. Further refinements on using credibility to resolve the puzzle are studied by Haberis, Harrison and Waldron (2019) and Gibbs and McClung (2023). They both show that partial credibility of holding rates at zero resolves the forward guidance puzzle.…”

“…In addition to these behavioral approaches, Cochrane (2017, 2023), McClung (2021, Gibbs and McClung (2023), and Diba and Loisel (2021) show that alternative monetary and fiscal policy frameworks also can eliminate the puzzles. The last paper listed shows that when monetary policy is conducted via money supply rules the economy is puzzle free under an interest rate peg.…”

“…The solution method used previously does not yield useful closed form solutions here. We, therefore, consider a special case that is inspired both by Cochrane (2017) and Gibbs and McClung (2023). We assume that 0 ≤ ϕ < 1 but ϕ * = 0, while all else remains the same.…”