Active and passive monetary policy regimes coexist with rational expectations

Statement

In a forward-looking New-Keynesian economy with a regime-switching Taylor rule, an “active” monetary policy regime in which the Fed satisfies the Taylor principle (response to expected inflation alpha > 1) and a “passive” regime in which it does not (alpha < 1) can coexist as part of a unique rational-expectations equilibrium, provided the regime transition probabilities are sufficiently persistent and the determinacy conditions of the compound system are satisfied. The passive regime does not produce non-fundamental sunspot dynamics in this setup because rational agents anticipate the eventual return to the active regime and form inflation expectations accordingly. Empirically, the U.S. economy 1970-2008 spends substantial time in both regimes, and the resulting compound model fits both the macro panel and the cross-section of bond yields.

Evidence summary

The supporting evidence is internal to Bikbov & Chernov (2013): they estimate the model on U.S. quarterly 1970-2008 data and report the following.

• The active regime has alpha(1) = 3.53, beta(1) = 2.18, rho(1) = 0.97. The passive regime has alpha(2) = 0.36, beta(2) = 1.27, rho(2) = 0.81.
• Both regimes are persistent: P(active | active) = 98%, P(passive | passive) = 90%. Neither is absorbing.
• The compound rational-expectations system has a unique solution at the estimated parameter point, verified by the regime-switching generalization of the Cho-Moreno forward method (Online Appendix A).
• The smoothed regime probabilities are sharply bimodal in time: the active regime appears in the 1970s portions, the Volcker disinflation, 1991-1995, and 2002-end-of-sample; the passive regime appears in 1973-1975, the monetary experiment 1979-1982, 1988-1991, the internet bubble 1995-2001, and 2005-2007.
• The model matches the macro panel and the cross-section of yields simultaneously, with reasonable measurement-error variance on the long yields.

The economic interpretation is that the Clarida-Gali-Gertler (2000) sunspot-equilibrium puzzle (their pre-Volcker subsample has alpha < 1) is resolved by the regime-switching framework: rational agents in 1970 already anticipated that policy would eventually return to active, so they did not form sunspot expectations during the passive episodes.

Conditions and scope

• The result is sensitive to the specific Taylor-rule specification. Different choices of the response horizon (one-quarter expected inflation vs four-quarter expected inflation), the policy-inertia structure, or the inclusion of additional regressors could change the active/passive estimates.
• The claim is about the U.S. monetary policy 1970-2008. Generalization to the post-2008 zero-lower-bound period or to other central banks is not directly tested.
• The compound rational-expectations equilibrium uniqueness depends on the determinacy conditions of the regime-switching forward method, which are computed numerically and depend on the full set of structural parameters. Small perturbations of the parameters can move the system across the determinacy boundary.
• The “passive does not produce sunspots” intuition depends on rational expectations and on the regime not being absorbing. If agents irrationally believed the passive regime were permanent, sunspot dynamics would re-emerge.

Counter-evidence

• Sims-Zha (2006) find no evidence of changing monetary policy in a macro- only Bayesian VAR. Bikbov-Chernov interpret this as a finite-sample identification problem rather than a substantive contradiction (the policy regime is hard to detect from macro data alone).
• Fernandez-Villaverde et al. (2010) find evidence of changing monetary policy but conclude that the changes had little real effect — partially consistent with the Bikbov-Chernov counterfactual finding that the great moderation was driven by both regime change and lower exogenous shock volatilities.
• Cogley-Sargent (2005) find varying inflation persistence and varying policy, but do not impose the structural Taylor-rule restrictions and do not use yield-curve information.

Linked ideas

• (none yet)

Open questions

• Can the active/passive distinction be re-estimated on the post-2008 ZLB sample using a shadow-rate or threshold-pricing extension?
• Does the persistence of the active regime increase or decrease as the central bank’s inflation target becomes more credible? The BC sample ends before any explicit FOMC inflation target announcement.
• How robust are the active/passive estimates to alternative specifications of risk premia (e.g., pricing the risk of regime change, which BC rule out)?