The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Discussion of Evans and Honkapohja, “Robust Learning Stability.” James Bullard President and CEO Federal Reserve Bank of St. Louis1 28 May 2008 Frontiers in Monetary Theory and Policy—IMES, BOJ 1 Views expressed are those of the author and do not necessarily reflect official positions of the FOMC or the Federal Reserve System. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Learning and monetary policy Rational expectations is a cornerstone of modern macroeconomic theory. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Learning and monetary policy Rational expectations is a cornerstone of modern macroeconomic theory. Rational expectations equilibrium is at the heart of the learning literature. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Learning and monetary policy Rational expectations is a cornerstone of modern macroeconomic theory. Rational expectations equilibrium is at the heart of the learning literature. But rational expectations is not a practical assumption. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Learning and monetary policy Rational expectations is a cornerstone of modern macroeconomic theory. Rational expectations equilibrium is at the heart of the learning literature. But rational expectations is not a practical assumption. Actual private sector expectations are unlikely to be precisely consistent with REE. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Learning and monetary policy Rational expectations is a cornerstone of modern macroeconomic theory. Rational expectations equilibrium is at the heart of the learning literature. But rational expectations is not a practical assumption. Actual private sector expectations are unlikely to be precisely consistent with REE. That is still all right ... Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Learning and monetary policy Rational expectations is a cornerstone of modern macroeconomic theory. Rational expectations equilibrium is at the heart of the learning literature. But rational expectations is not a practical assumption. Actual private sector expectations are unlikely to be precisely consistent with REE. That is still all right ... ... so long as small expectational deviations from RE dissipate, instead of accumulating. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Learning and monetary policy Rational expectations is a cornerstone of modern macroeconomic theory. Rational expectations equilibrium is at the heart of the learning literature. But rational expectations is not a practical assumption. Actual private sector expectations are unlikely to be precisely consistent with REE. That is still all right ... ... so long as small expectational deviations from RE dissipate, instead of accumulating. That condition is known as expectational stability. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses What the literature has shown REE may or may not be expectationally stable in New Keynesian models. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry What the literature has shown REE may or may not be expectationally stable in New Keynesian models. In the simplest New Keynesian model, the Taylor Principle is associated with E-stability. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry What the literature has shown REE may or may not be expectationally stable in New Keynesian models. In the simplest New Keynesian model, the Taylor Principle is associated with E-stability. Failure of the Taylor Principle is associated with expectational instability. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry What the literature has shown REE may or may not be expectationally stable in New Keynesian models. In the simplest New Keynesian model, the Taylor Principle is associated with E-stability. Failure of the Taylor Principle is associated with expectational instability. That result assumes Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry What the literature has shown REE may or may not be expectationally stable in New Keynesian models. In the simplest New Keynesian model, the Taylor Principle is associated with E-stability. Failure of the Taylor Principle is associated with expectational instability. That result assumes contemporaneous data specification of the policy rule. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry What the literature has shown REE may or may not be expectationally stable in New Keynesian models. In the simplest New Keynesian model, the Taylor Principle is associated with E-stability. Failure of the Taylor Principle is associated with expectational instability. That result assumes contemporaneous data specification of the policy rule. decreasing gain in the learning rule. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry What the literature has shown REE may or may not be expectationally stable in New Keynesian models. In the simplest New Keynesian model, the Taylor Principle is associated with E-stability. Failure of the Taylor Principle is associated with expectational instability. That result assumes contemporaneous data specification of the policy rule. decreasing gain in the learning rule. Comforting. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses What the authors do Instrument rules plus various forms of “optimal” policy. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses What the authors do Instrument rules plus various forms of “optimal” policy. Replace decreasing gain with constant gain. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses What the authors do Instrument rules plus various forms of “optimal” policy. Replace decreasing gain with constant gain. Agents are trying to robustly track the changing environment in which they operate. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry What the authors do Instrument rules plus various forms of “optimal” policy. Replace decreasing gain with constant gain. Agents are trying to robustly track the changing environment in which they operate. Concentrate on operational rules in the sense of McCallum. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry What the authors do Instrument rules plus various forms of “optimal” policy. Replace decreasing gain with constant gain. Agents are trying to robustly track the changing environment in which they operate. Concentrate on operational rules in the sense of McCallum. Keep policymaker information in line with reality. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry What the authors do Instrument rules plus various forms of “optimal” policy. Replace decreasing gain with constant gain. Agents are trying to robustly track the changing environment in which they operate. Concentrate on operational rules in the sense of McCallum. Keep policymaker information in line with reality. In particular, contemporaneous values of output and inflation are not known when policy decisions are made. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Main findings There are many recommended good or “optimal” policies for this model ... Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Main findings There are many recommended good or “optimal” policies for this model ... ... but most produce expectational instability in this setting. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Main findings There are many recommended good or “optimal” policies for this model ... ... but most produce expectational instability in this setting. Policymakers following these recommended approaches in this environment would be surprised to find that the economy does not coordinate on the intended equilibrium. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Main findings There are many recommended good or “optimal” policies for this model ... ... but most produce expectational instability in this setting. Policymakers following these recommended approaches in this environment would be surprised to find that the economy does not coordinate on the intended equilibrium. To obtain expectational stability, use the expectations-based rules of Evans-Honkapohja (2003, 2006). Main ideas What the authors do Interpretations Unstable is not optimal Other analyses How to think about E-stability Stability is not a common mode of analysis in macroeconomics and monetary policy. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry How to think about E-stability Stability is not a common mode of analysis in macroeconomics and monetary policy. The actual learning dynamics would only be observed in an unstable situtation. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry How to think about E-stability Stability is not a common mode of analysis in macroeconomics and monetary policy. The actual learning dynamics would only be observed in an unstable situtation. Otherwise convergence occurs. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry How to think about E-stability Stability is not a common mode of analysis in macroeconomics and monetary policy. The actual learning dynamics would only be observed in an unstable situtation. Otherwise convergence occurs. We are talking about locally to the REE. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry How to think about E-stability Stability is not a common mode of analysis in macroeconomics and monetary policy. The actual learning dynamics would only be observed in an unstable situtation. Otherwise convergence occurs. We are talking about locally to the REE. If policy is executed correctly ... Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry How to think about E-stability Stability is not a common mode of analysis in macroeconomics and monetary policy. The actual learning dynamics would only be observed in an unstable situtation. Otherwise convergence occurs. We are talking about locally to the REE. If policy is executed correctly ... ... observers would see only the REE values of key macro variables. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry How to think about E-stability Stability is not a common mode of analysis in macroeconomics and monetary policy. The actual learning dynamics would only be observed in an unstable situtation. Otherwise convergence occurs. We are talking about locally to the REE. If policy is executed correctly ... ... observers would see only the REE values of key macro variables. Learning would appear to be unimportant, even though it is enforcing the REE. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry How to think about E-stability Stability is not a common mode of analysis in macroeconomics and monetary policy. The actual learning dynamics would only be observed in an unstable situtation. Otherwise convergence occurs. We are talking about locally to the REE. If policy is executed correctly ... ... observers would see only the REE values of key macro variables. Learning would appear to be unimportant, even though it is enforcing the REE. One analogy: off-equilibrium play in games. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry How to think about E-stability Stability is not a common mode of analysis in macroeconomics and monetary policy. The actual learning dynamics would only be observed in an unstable situtation. Otherwise convergence occurs. We are talking about locally to the REE. If policy is executed correctly ... ... observers would see only the REE values of key macro variables. Learning would appear to be unimportant, even though it is enforcing the REE. One analogy: off-equilibrium play in games. Another analogy: default punishment in models with endogenous debt constraints. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Expectational instability Some might worry that the unstable case is “not really observed.” Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Expectational instability Some might worry that the unstable case is “not really observed.” Indeed, we would not expect to see a lot of these cases ... Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Expectational instability Some might worry that the unstable case is “not really observed.” Indeed, we would not expect to see a lot of these cases ... ... policymakers would have to abandon their policies. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Expectational instability Some might worry that the unstable case is “not really observed.” Indeed, we would not expect to see a lot of these cases ... ... policymakers would have to abandon their policies. But consider Figure 2 in the paper. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Expectational instability Some might worry that the unstable case is “not really observed.” Indeed, we would not expect to see a lot of these cases ... ... policymakers would have to abandon their policies. But consider Figure 2 in the paper. Inflation and the output gap behave as if the REE has been attained for many periods. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Expectational instability Some might worry that the unstable case is “not really observed.” Indeed, we would not expect to see a lot of these cases ... ... policymakers would have to abandon their policies. But consider Figure 2 in the paper. Inflation and the output gap behave as if the REE has been attained for many periods. But eventually, expectational errors accumulate and drive the economy away from the targeted equilibrium. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Expectational instability Some might worry that the unstable case is “not really observed.” Indeed, we would not expect to see a lot of these cases ... ... policymakers would have to abandon their policies. But consider Figure 2 in the paper. Inflation and the output gap behave as if the REE has been attained for many periods. But eventually, expectational errors accumulate and drive the economy away from the targeted equilibrium. Does it happen? Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Expectational instability Some might worry that the unstable case is “not really observed.” Indeed, we would not expect to see a lot of these cases ... ... policymakers would have to abandon their policies. But consider Figure 2 in the paper. Inflation and the output gap behave as if the REE has been attained for many periods. But eventually, expectational errors accumulate and drive the economy away from the targeted equilibrium. Does it happen? Consider the breakdown of Bretton Woods. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Expectational instability Some might worry that the unstable case is “not really observed.” Indeed, we would not expect to see a lot of these cases ... ... policymakers would have to abandon their policies. But consider Figure 2 in the paper. Inflation and the output gap behave as if the REE has been attained for many periods. But eventually, expectational errors accumulate and drive the economy away from the targeted equilibrium. Does it happen? Consider the breakdown of Bretton Woods. And maybe we should worry about Sweden, as we will see below. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Taylor-type rules Some potential for instability in the calibrated case. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Taylor-type rules Some potential for instability in the calibrated case. Taylor rule fairs better than some other recommended rules studied later in the paper. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Taylor-type rules Some potential for instability in the calibrated case. Taylor rule fairs better than some other recommended rules studied later in the paper. Instability requires the combination of operational rules with constant gain learning. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Duffy and Xiao The Duffy/Xiao rule is a Taylor-type rule. Other analyses Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Duffy and Xiao The Duffy/Xiao rule is a Taylor-type rule. It represents optimal policy under discretion if the policymaker has a preference for interest rate smoothing. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Duffy and Xiao The Duffy/Xiao rule is a Taylor-type rule. It represents optimal policy under discretion if the policymaker has a preference for interest rate smoothing. Results for commitment case similar. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Duffy and Xiao The Duffy/Xiao rule is a Taylor-type rule. It represents optimal policy under discretion if the policymaker has a preference for interest rate smoothing. Results for commitment case similar. The rule is it = ϕλ ϕαx πt + xt . αi αi (1) Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Duffy and Xiao The Duffy/Xiao rule is a Taylor-type rule. It represents optimal policy under discretion if the policymaker has a preference for interest rate smoothing. Results for commitment case similar. The rule is it = ϕλ ϕαx πt + xt . αi αi Produces determinacy if αi sufficiently large. (1) Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Duffy and Xiao The Duffy/Xiao rule is a Taylor-type rule. It represents optimal policy under discretion if the policymaker has a preference for interest rate smoothing. Results for commitment case similar. The rule is it = ϕλ ϕαx πt + xt . αi αi Produces determinacy if αi sufficiently large. Produces expectational instability if αi sufficiently small. (1) Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Duffy and Xiao The Duffy/Xiao rule is a Taylor-type rule. It represents optimal policy under discretion if the policymaker has a preference for interest rate smoothing. Results for commitment case similar. The rule is it = ϕλ ϕαx πt + xt . αi αi Produces determinacy if αi sufficiently large. Produces expectational instability if αi sufficiently small. Similar results for Duffy/Xiao under commitment. (1) Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Duffy and Xiao The Duffy/Xiao rule is a Taylor-type rule. It represents optimal policy under discretion if the policymaker has a preference for interest rate smoothing. Results for commitment case similar. The rule is it = ϕλ ϕαx πt + xt . αi αi Produces determinacy if αi sufficiently large. Produces expectational instability if αi sufficiently small. Similar results for Duffy/Xiao under commitment. Worrisome. (1) Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Duffy and Xiao The Duffy/Xiao rule is a Taylor-type rule. It represents optimal policy under discretion if the policymaker has a preference for interest rate smoothing. Results for commitment case similar. The rule is it = ϕλ ϕαx πt + xt . αi αi Produces determinacy if αi sufficiently large. Produces expectational instability if αi sufficiently small. Similar results for Duffy/Xiao under commitment. Worrisome. The expectations-based approach of Evans and Honkapohja solves this problem and provides robust expectational stability. (1) Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Commitment No interest rate smoothing, timeless perspective, goal is to implement the FOC. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Commitment No interest rate smoothing, timeless perspective, goal is to implement the FOC. Svensson and Woodford (2005) recommended rule equation (18) in the paper. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Commitment No interest rate smoothing, timeless perspective, goal is to implement the FOC. Svensson and Woodford (2005) recommended rule equation (18) in the paper. Operational versions can be associated with instability under learning for reasonable gain parameters. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Commitment No interest rate smoothing, timeless perspective, goal is to implement the FOC. Svensson and Woodford (2005) recommended rule equation (18) in the paper. Operational versions can be associated with instability under learning for reasonable gain parameters. Worrisome for Sweden? Figure 2? Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Commitment No interest rate smoothing, timeless perspective, goal is to implement the FOC. Svensson and Woodford (2005) recommended rule equation (18) in the paper. Operational versions can be associated with instability under learning for reasonable gain parameters. Worrisome for Sweden? Figure 2? There is nothing optimal about instability. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Woodford NRE Woodford (2008) has considered an alternative approach to checking the robustness of policy to the possibility that expectations may not be fully rational. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Woodford NRE Woodford (2008) has considered an alternative approach to checking the robustness of policy to the possibility that expectations may not be fully rational. Avoids committing to a particular recursive algorithm to describe learning. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Woodford NRE Woodford (2008) has considered an alternative approach to checking the robustness of policy to the possibility that expectations may not be fully rational. Avoids committing to a particular recursive algorithm to describe learning. But, expectational stability still plays a role in that analysis. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Bayesian approaches Many feel that recursive learning should be Bayesian. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Bayesian approaches Many feel that recursive learning should be Bayesian. Bullard and Suda (2008). Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Bayesian approaches Many feel that recursive learning should be Bayesian. Bullard and Suda (2008). Standard recursive learning exercise, but replace classical econometricians with Bayesian econometricians. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Bayesian approaches Many feel that recursive learning should be Bayesian. Bullard and Suda (2008). Standard recursive learning exercise, but replace classical econometricians with Bayesian econometricians. Main results still hold: Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Bayesian approaches Many feel that recursive learning should be Bayesian. Bullard and Suda (2008). Standard recursive learning exercise, but replace classical econometricians with Bayesian econometricians. Main results still hold: “extra” term in the actual law of motion. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Bayesian approaches Many feel that recursive learning should be Bayesian. Bullard and Suda (2008). Standard recursive learning exercise, but replace classical econometricians with Bayesian econometricians. Main results still hold: “extra” term in the actual law of motion. expectational stability conditions unchanged. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Bayesian approaches Many feel that recursive learning should be Bayesian. Bullard and Suda (2008). Standard recursive learning exercise, but replace classical econometricians with Bayesian econometricians. Main results still hold: “extra” term in the actual law of motion. expectational stability conditions unchanged. Stability still an issue. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Responding to expectations Policymakers seem to do this at times ... Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Responding to expectations Policymakers seem to do this at times ... but it is not that clear how to map this idea into reality. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Responding to expectations Policymakers seem to do this at times ... but it is not that clear how to map this idea into reality. Measurement issues. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Responding to expectations Policymakers seem to do this at times ... but it is not that clear how to map this idea into reality. Measurement issues. Potential games. Why worry Main ideas What the authors do Interpretations Unstable is not optimal Final thoughts Very nice paper in a very nice conference. Other analyses Why worry Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Final thoughts Very nice paper in a very nice conference. Thanks to the Bank of Japan for inviting me to participate. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Final thoughts Very nice paper in a very nice conference. Thanks to the Bank of Japan for inviting me to participate. I appreciate the attention to stability issues, which I think are insufficiently analyzed in macroeconomics. Main ideas What the authors do Interpretations Unstable is not optimal Other analyses Why worry Final thoughts Very nice paper in a very nice conference. Thanks to the Bank of Japan for inviting me to participate. I appreciate the attention to stability issues, which I think are insufficiently analyzed in macroeconomics. Instability can produce the “big ticket losses” that policymakers really worry about.