LeatherSagiKnowledgebase

Regime Switches, Agents' Beliefs, and Post-World War II U.S. Macroeconomic Dynamics

4 min read

Problem

Understanding the causes of the dramatic shifts in U.S. macroeconomic dynamics over the post-WWII period — the Great Inflation of the 1970s, the Great Moderation, and the Great Recession — has been a central question. Two competing explanations exist: “Good Policy” (the Fed’s anti-inflationary stance changed) vs “Good Luck” (exogenous shock volatilities changed). Prior work (Clarida et al. 2000, testing-indeterminacy-application-monetary-policy) supported policy changes, while Sims-Zha (2006) found evidence for volatility changes only. Neither camp accounted for the role of agents’ forward-looking beliefs about future regime changes in shaping equilibrium outcomes.

Key idea

Agents who are aware of the possibility of regime switches form expectations that depend on the transition probabilities between regimes, not just on the regime currently in place. This means the law of motion of the economy under any given regime depends on what agents believe about alternative regimes. The paper introduces beliefs counterfactuals — a new class of counterfactual simulations that modify agents’ beliefs about future regimes (e.g., introducing a hypothetical “Eagle” regime that is even more hawkish than the Hawk) without necessarily changing the actual conduct of monetary policy. These beliefs counterfactuals exploit the beliefs-counterfactual-simulation mechanism unique to MS-DSGE models.

Method

  • Medium-scale Christiano-Eichenbaum-Evans (2005) DSGE model with two independent Markov-switching processes: one for Taylor rule parameters (Hawk vs Dove), one for shock volatilities (high vs low).
  • Solution via Farmer-Waggoner-Zha (2010) method for monetary-policy-regime-switching rational expectations models.
  • Bayesian estimation (Gibbs sampling, 5 chains x 920,000 draws each) on six U.S. quarterly observables 1954:Q3-2009:Q2: real GDP growth, inflation, federal funds rate, consumption growth, investment, and investment price.
  • Model comparison via marginal data densities across restricted specifications (fixed policy, fixed volatility, both switching).
  • Three types of counterfactuals: (1) standard regime-fixed counterfactuals, (2) beliefs counterfactuals introducing a hypothetical Eagle regime, (3) Great Recession counterfactuals.

Results

  • The model strongly favors regime switches in the Taylor rule (Hawk vs Dove), not just in shock volatilities. The Hawk regime has inflation response ~2.39; the Dove regime ~0.95 (sub-unity, passive).
  • U.S. monetary policy history is NOT a simple pre/post-Volcker break. The Dove regime prevailed in the 1970s, the 2008 crisis, and likely the late 1960s. The Hawk regime prevailed most of the 1980s-2000s but also appeared in earlier periods. The pattern is recurrent and stochastic, not a one-time structural break.
  • Beliefs counterfactuals: If agents in the 1970s had anticipated the appointment of an extremely conservative Chairman (Eagle regime), inflation would have been substantially lower AND the sacrifice ratio would have been more favorable — better than simply imposing a hawkish rule without changing beliefs.
  • The Great Recession: the Dove regime probability started rising in 2005, before the crisis. Counterfactuals show that if agents had expected a swift return to aggressive monetary policy, the 2008 drop in output and inflation would have been mitigated.
  • Transition matrix estimates: Hawk persistence ~95%, Dove persistence ~90%.

Limitations

  • The model assumes agents know the transition matrix (rational expectations about regime switches). In practice, agents may learn about it over time.
  • The zero lower bound is not explicitly modeled, limiting interpretation of the 2008-2009 counterfactuals.
  • Fixed inflation target across regimes — only the strength of the response changes, not the target itself.
  • The Eagle regime counterfactual is hypothetical; it is not identified from the data.
  • Only first-order perturbation is used; higher-order terms (which would allow volatility regimes to affect the solution matrices T and R) are not considered.

Open questions

  • How would endogenous transition probabilities (e.g., depending on the level of inflation) change the beliefs-counterfactual results?
  • Can beliefs counterfactuals be extended to the post-2008 period with explicit ZLB modeling?
  • What is the interaction between beliefs about monetary policy regimes and financial market risk premia (not modeled here)?
  • How robust are the Hawk/Dove regime datings to alternative model specifications (e.g., adding financial frictions)?

My take

This is a landmark paper for the MS-DSGE literature and directly relevant to the monetary-policy-regime-switching framework used in the CRE asset pricing model. The key insight — that beliefs about future regimes matter as much as the current regime — has direct implications for the Leather-Sagi model: the cap-rate implications of a regime switch depend not just on the switch itself but on what agents believe about the persistence and alternatives. The beliefs counterfactual methodology is particularly powerful and has no analog in fixed-parameter DSGE models. The finding that U.S. monetary policy switches are recurrent (not a one-time break) supports the use of ergodic Markov chains rather than absorbing states in the CRE model.

Graph View

Backlinks