E-stability vis-à-vis determinacy in regime-switching models

Determinacy and E-Stability with Interest Rate Rules at the Zero Lower Bound

“…In most of the above-mentioned papers, the monetary policy regime is expected to last forever, but a determinate equilibrium may exist in models with an inflation targeting policymaker and persistent, transitory passive monetary regimes, as shown by Cho (2016) and Barthélemy and Marx (2019), and earlier considered by Davig and Leeper (2007). 5 Similarly, Mertens and Ravn (2014) and Christiano et al (2018) show when expectations are stable under adaptive learning in economies that are subject to a one-time transient ZLB regime ;and McClung (2020) shows that Markov-switching models with an inflation targeting central bank and recurring interest rate peg regimes can admit E-stable REE if interest peg regimes are not expected to last too long. 6 Therefore, the ability of policymakers to manage expectations subject to interest peg regimes such as ZLB events depends crucially on the expected duration and frequency of the interest peg regime.…”

“…Proposition 1 of McClung (2020) derives the E-stability conditions under which agents may learn an MSV solution (17) by estimating ( 21) in real time and making forecasts contingent on these estimates using contemporaneous information. When the E-stability conditions fail (i.e.…”

“…Therefore, the regions of the model parameter space that generate E-stable MSV solutions are larger than those of the parameter space that generate a unique mean-square stable REE. In fact, McClung (2020) shows that the E-stability conditions for models of the form (16) can be significantly weaker than the determinacy conditions.…”

“…Recursive approaches to estimating (21) are briefly discussed in section 5.2 and alsoMcClung (2020).22 Here, and throughout the paper, we use Êt to denote (potentially) non-rational expectations formed under adaptive learning. E t denote rational expectations.…”

“…Note that σ d and σ s are irrelevant for determinacy and E-stability analysis (e.g. seeCho (2021) andMcClung (2020)). …”