The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Federal Reserve Bank of Chicago Constrained Discretion and Central Bank Transparency Francesco Bianchi and Leonardo Melosi October 2016 WP 2016-15 Constrained Discretion and Central Bank Transparencyy Francesco Bianchi Leonardo Melosi Duke University Federal Reserve Bank of Chicago CEPR and NBER October 2016 Abstract We develop and estimate a general equilibrium model to quantitatively assess the e¤ects and welfare implications of central bank transparency. Monetary policy can deviate from active in‡ation stabilization and agents conduct Bayesian learning about the nature of these deviations. Under constrained discretion, only short deviations occur, agents’ uncertainty about the macroeconomy remains contained, and welfare is high. However, if a deviation persists, uncertainty accelerates and welfare declines. Announcing the future policy course raises uncertainty in the short run by revealing that active in‡ation stabilization will be temporarily abandoned. However, this announcement reduces policy uncertainty and anchors in‡ationary beliefs at the end of the policy. For the U.S., enhancing transparency is found to increase welfare. The same result is found when we relax the assumption of perfectly credible announcements. Keywords: Policy announcement, Bayesian learning, reputation, forward guidance, macroeconomic risk, uncertainty, in‡ation expectations, Markov-switching models, likelihood estimation. JEL classi…cation: E52, D83, C11. y We wish to thank Martin Eichenbaum, Cristina Fuentes-Albero, Jordi Galí, Pablo Guerron-Quintana, Yuriy Gorodnichenko, Narayana Kocherlakota, Frederick Mishkin, Matthias Paustian, Jon Steinsson, Mirko Wiederholt, Michael Woodford, and Tony Yates for helpful comments. We also thank seminar participants at the NBER Summer Institute, Columbia University, UC Berkeley, Duke University, the SED conference in Cyprus, the Bank of England, the ECB Conference “Information, Beliefs and Economic Policy,”the Midwest Macro Meetings 2014, and the Philadelphia Fed. Part of this paper was written while Leonardo Melosi was visiting the Bank of England, whose hospitality is gratefully acknowledged. The views in this paper are solely the responsibility of the authors and should not be interpreted as re‡ecting the views of the Bank of England or that of the Federal Reserve Bank of Chicago or any other person associated with the Federal Reserve System. Francesco Bianchi gratefully acknowledges …nancial support from the National Science Foundation through grant SES-1227397. Correspondence to: Francesco Bianchi, francesco.bianchi@duke.edu, and Leonardo Melosi, lmelosi@frbchi.org. 1 Introduction The last two decades have witnessed two major breakthroughs in the practice of central banking worldwide. First, most central banks have adopted a monetary policy framework that Bernanke and Mishkin (1997) have termed constrained discretion. Bernanke (2003) explains that under constrained discretion, the central bank retains some ‡exibility in deemphasizing in‡ation stabilization so as to pursue alternative short-run objectives such as unemployment stabilization. However, such ‡exibility is constrained to the extent that the central bank should maintain a strong reputation for keeping in‡ation and in‡ation expectations …rmly under control. Second, many countries have taken remarkable steps to make their central bank more transparent (Bernanke et al. 1999 Mishkin 2002 and Campbell et al. 2012). As a result of these changes, the following questions are crucial for modern monetary policymaking. First, for how long can a central bank de-emphasize in‡ation stabilization before the private sector starts fearing a return to a period of high and volatile in‡ation as in 1970s? Second, does transparency play an essential role for e¤ective monetary policymaking? Should a central bank be explicit about the future course of monetary policy? The recent …nancial crisis has triggered a prolonged period of accommodative monetary policy that some members of the Federal Open Market Committee fear could lead to a disanchoring of in‡ation expectations (as an example, see Plosser 2012.) Thus, these questions are at the center of the current policy debate. To address these questions, we develop and estimate a model in which the anti-in‡ationary stance of the central bank can change over time and agents face uncertainty about the nature of deviations from active in‡ation stabilization. When monetary policy alternates between prolonged periods of active in‡ation stabilization (active regime) and short periods during which the emphasis on in‡ation stabilization is reduced (short-lasting passive regime), the model captures the monetary approach described as constrained discretion. However, the central bank can also engage in prolonged deviations from the active regime of the type 1 observed in the 1970s (long-lasting passive regime). Agents in the model are fully rational and able to infer if monetary policy is active or not. However, when the passive rule prevails, they are uncertain about whether the central bank is engaging in a short-lasting or in a long-lasting deviation from the active regime. The central bank can then follow two possible communication strategies: Transparency or no transparency. Under no transparency, the nature of the deviation is not revealed. Under transparency, the duration of short-lasting deviations is announced. Under no transparency, when passive monetary policy prevails, agents conduct Bayesian learning in order to infer the likely duration of the deviation from active monetary policy. Given that the behavior of the monetary authority is unchanged across the two passive regimes, the only way for rational agents to learn about the nature of the deviation consists of keeping track of the number of consecutive deviations. As agents observe more and more realizations of the passive rule, they become increasingly convinced that the long-lasting passive regime is occurring. As a result, the more the central bank deviates from active in‡ation stabilization, the more agents become discouraged about a quick return to the active regime. We solve the model by keeping track of the joint evolution of policymakers’ behavior and agents’beliefs, using the methods developed in Bianchi and Melosi (2016a). In the model, social welfare is shown to be a function of agents uncertainty about future in‡ation and future output gaps. In standard models, monetary policy a¤ects agents’welfare by in‡uencing the unconditional variances of the endogenous variables. In our nonlinear setting, policy actions exert dynamic e¤ects on uncertainty. Therefore, welfare evolves over time in response to the short-run ‡uctuations of uncertainty. To our knowledge, this feature is new in the literature and allows us to study changes in the macroeconomic risk due to policy actions and communication strategies and the associated welfare implications. We measure uncertainty taking into account agents’beliefs about the evolution of monetary policy. As long as the number of deviations from the active regime is low, the increase in uncertainty is very modest and stays in line with the levels implied by the active regime. 2 This is because agents regard the early deviations as temporary. However, as the number of deviations increases and fairly optimistic agents become fairly pessimistic about a quick return to active policies, uncertainty starts increasing and eventually converges to the values implied by the long-lasting passive regime. As a result, for each horizon, our measure of uncertainty is now higher than its long-run value. This is because agents take into account that while in the short run a prolonged period of passive monetary policy will prevail, in the long run the economy will surely visit the active regime again. Therefore, an important result arises: Deviations from the active regime that last only a few periods have no disruptive consequences on welfare because they do not have a large impact on agents’uncertainty regarding future monetary policy. Instead, if a central bank deviates from the active regime for a prolonged period of time, the disanchoring of agents’uncertainty occurs, causing sizable welfare losses. The model under the assumption of no transparency is …tted to U.S. data. We identify prolonged deviations from active monetary policy in the 1960s and the 1970s in line with previous contributions to the literature. However, we also …nd that the Federal Reserve has recurrently engaged in short-lasting passive policies since the early 1980s, supporting the view that constrained discretion has been the predominant approach to U.S. monetary policy in the past three decades. In the analysis, we abstract from the reasons why the Federal Reserve has engaged in such deviations. In fact, we consider these recurrent deviations as a given of our analysis. This approach provides us with a parsimonious, reduced-form framework to estimate the Federal Reserve’s behaviors in the data. Given these estimated behaviors, we evaluate how quickly agents’beliefs respond to policymakers’behaviors and announcements, what this implies for the evolution of uncertainty and welfare, and what the potential gains are from reducing the uncertainty about the future conduct of monetary policy. The paper introduces a practical de…nition of reputation: A central bank has a strong reputation if it is less likely to engage in long-lasting deviations from active policies. We …nd useful to distinguish two related concepts: long-run reputation and short-run reputation. 3 Long-run reputation depends on how frequently the central bank has historically deviated from active policies and for how long. This measure of reputation maps into the estimated transition matrix, which controls the unconditional probability of observing long spans of passive monetary policy and deeply a¤ects the unconditional level of uncertainty in the macroeconomy. Short-run reputation captures agents’beliefs about the conduct of monetary policy in the near future. This second measure of reputation corresponds to a precise statistic: the expected number of consecutive deviations from active monetary policy. To avoid having to constantly distinguish between the two measures of reputation, this last statistic is dubbed pessimism, as it captures how pessimistic agents are about observing a switch to active policy. We often use the term reputation to refer to long-run reputation. While our de…nition of reputation is not exactly as the one used in theory studies (Kydland and Prescott 1977 Barro and Gordon 1983 Faust and Svensson 2001 and Galí and Gertler 2007), it suits well Bernanke’s de…nition of constrained discretion and has the important advantage of being measurable in the data. Bernanke (2003) explains that for constrained discretion to work e¤ectively, the central bank has to “establish a strong commitment to keeping in‡ation low and stable.”In our paper, the strength of this commitment is called long-run reputation. If the Federal Reserve engages in prolonged periods of passive policies, agents become more pessimistic about a return to the active regime. As pessimism increases, so do in‡ation volatility and uncertainty. In Bernanke’s parlance, the Federal Reserve’s “discretion” can become “constrained” in that short-run reputation deteriorates after a prolonged deviation from active policy. The fact that the Federal Reserve conducted a prolonged spell of passive policy in the 1970s has contributed to lowering its reputation in our estimated model. Nevertheless, even though the Federal Reserve’s reputation is not immaculate, the Federal Reserve is found to bene…t from its strong reputation. Based on the estimates, pessimism and, hence, agents’ uncertainty about future in‡ation change vary sluggishly in response to deviations from active monetary policy. This …nding has the important implication that the Federal Reserve can 4 conduct passive policies for a fairly large number of years before the disanchoring of in‡ation expectations and an overall increase in macroeconomic uncertainty occur. However, very prolonged deviations from active policy lead agents to become wary that the central bank has switched to 1970s-type of policies, causing detrimental e¤ects on welfare. While this result implies that the Federal Reserve can successfully implement constrained discretion even without transparency, our …ndings suggest that increasing transparency would improve welfare. The estimated model suggests that the welfare gains from transparency range between 0.54% to 3.74% of steady-state consumption. A transparent central bank systematically announces the duration of any short-lasting deviation from the active regime beforehand, whereas in the case of a long-lasting deviation the exact duration is not known. The implications of such a communication strategy vary based on the nature of the deviation. When the central bank engages in a short-lasting deviation, announcing its duration immediately removes the fear of the 1970s. Under no transparency, instead, agents are not informed about the exact nature of the observed deviation. As a result, whenever a short deviation occurs, ex-ante agents cannot rule out the possibility of a long-lasting deviation of the kind that characterized the 1970s. As a result, ex-post, agents turn out to have overstated the persistence of the observed deviation. How large this e¤ect is depends on the central bank’s reputation. The model allows us to highlight an important trade-o¤ associated with transparency. First, in the short run, being transparent reduces welfare because agents are told that passive monetary policy will prevail for a while and thereby future shocks are expected to have larger e¤ects. Second, as time goes by, agents know that the prolonged period of passive monetary policy is coming to an end. This leads to a reduction in the level of uncertainty at every horizon with an associated improvement in welfare. Notice, that this is exactly the opposite of what occurs when no announcement is made: Agents, in the case of no transparency, become more and more discouraged about the possibility of moving to the active regime and uncertainty increases. To our knowledge, this is the …rst paper that studies this critical 5 trade-o¤ associated with central bank’s announcements through the lens of an estimated dynamic stochastic general equilibrium (DSGE) model. Furthermore, our results are robust to relaxing the assumption that the central bank never lies about the duration of passive monetary policy. This paper makes three main contributions to the existing literature. First, we show how to model recurrent policymakers’ announcements about the central bank’s future reaction function in an estimated DSGE model.1 Second, we show how to characterize and compute social welfare in a Markov switching DSGE model with Bayesian learning and announcements. Interestingly, in our nonlinear framework, welfare captures the macroeconomic risk perceived by the agents as a function of the expected or announced policy decisions. Finally, we estimate a microfounded general equilibrium model with changes in policymakers’behavior and Bayesian learning. To the best of our knowledge, this is the …rst paper that estimates a DSGE model with Markov-switching structural parameters and Bayesian learning. Our learning mechanism implies that agents’beliefs are not invariant to the duration of a certain policy. Therefore, the model captures a very intuitive idea: Agents in the late 1970s were arguably more pessimistic about a quick return to the active regime than they had been in the early 1970s. This feature was not present in previous contributions such as Bianchi (2013) and Davig and Doh (2014a). This paper is part of a broader research agenda that aims to model the evolution of agents’beliefs in general equilibrium models (Bianchi and Melosi 2014, 2016b). Our modeling framework goes beyond the assumption of anticipated utility that is often used in the learning literature.2 Such an assumption implies that agents forecast future events assuming that their beliefs will never change in the future. Instead, agents in our models know that they do not know. Therefore, when forming expectations, they take into account that their beliefs will evolve according to what they will observe in the future. 1 The importance of this type of forward guidance has been recognized by some members of the FOMC. See, for instance, Mester (2014) 2 For some prominent examples see Marcet and Sargent (1989a, 1989b), Cho, Williams, and Sargent (2002) , and Evans and Honkapohja (2001, 2003). 6 Schorfheide (2005) considers an economy in which agents use Bayesian learning to infer changes in a Markov-switching in‡ation target. In that paper agents solve a …ltering problem to disentangle a persistent component from a transitory component. The learning mechanism is treated as external to the model, implying that the model needs to be solved in every period in order to re‡ect the change in agents’beliefs regarding the two components. Consequently, when agents form their beliefs, they do not take into account how their beliefs will change. Furthermore, the method developed in Schorfheide (2005) cannot be immediately extended to models in which agents learn about changes in the stochastic properties of the model’s structural parameters. Eusepi and Preston (2010) study monetary policy communication in a model where agents face uncertainty about the value of model parameters. Cogley, Matthes and Sbordone (2011) address the problem of a newly appointed central bank governor who wants to disin‡ate. Unlike in the last two papers, in our paper regime changes are recurrent, agents learn about the regime in place as opposed to Taylor rule parameters, we do not assume anticipated utility, and we conduct likelihood-based estimation. Our paper also shows that when discrete regime changes are combined with a learning mechanism, a smooth evolution of expectations and uncertainty arises. Therefore, we implicitly connect the literature on discrete regime changes to the literature that models parameter instability as slow-moving processes.3 Our work is also linked to papers that study the transmission of nominal disturbances in general equilibrium models with information frictions, such as Gorodnichenko (2008), Mackowiak and Wiederholt (2009), Mankiw and Reis (2006), Melosi (2014a, forthcoming), and Nimark (2008). Finally, this paper is connected with the literature that studies the macroeconomic e¤ects of forward guidance (Del Negro, Giannoni, and Patterson 2012; Campbell et al. forthcoming). The key innovation of our paper is that forward guidance is about the central bank’s reaction function, whereas in that literature, communication is about future deviations from the monetary policy rule. 3 See, among others, Sims and Zha (2006), Bianchi (2013), Bianchi and Ilut (2013), Davig and Doh (2014a), Liu, Waggoner, and Zha (2011) for the …rst body of literature and then Primiceri (2005), Cogley and Sargent (2005), Fernandez-Villaverde and Rubio-Ramirez (2008), and Justiniano and Primiceri (2008) for the second body of the literature. 7 This paper is organized as follows. In Section 2, we introduce the baseline model. In Section 3, we show how to solve the model under the assumption of no transparency and transparency. In Section 4, the model under the assumption of no transparency is …tted to U.S. data. In Section 5, we assess the welfare implications of introducing transparency. In Section 6, we extend the analysis to imperfectly credible announcements. In Section 7, we assess the robustness of our results. Section 8 concludes. 2 The Model The model is built on Coibion, Gorodnichenko and Wieland (2012), who develop a prototypical New-Keynesian DSGE model with trend in‡ation and partial price indexation. We make two main departures from this standard framework. First, we assume that households and …rms have incomplete information, in a sense to be made clear shortly. Second, we assume parameter instability in the monetary policy rule.4 Households: The representative household maximizes expected utility: E hP 1 t=0 t n ln Ct+j ( + 1) 1 R 1+1= Nit+j i o di jF0 ; where Ct is composite consumption and Nit is labor worked in industry i. The parameter 2 (0; 1) is the discount factor, the parameter 0 is the Frisch elasticity of labor supply. E [ jF0 ] is the expectation operator conditioned on information of private agents available at time 0. The information set Ft contains the history of all model variables but not the history of policy regimes p t that, as we shall show, determine the parameter value of the central bank’s reaction function. 4 Extending the analysis to state-of-the-art monetary DSGE models such as Christiano, Eichenbaum, and Evans (2005) and Smets and Wouters (2007) would be interesting, but it would also imply a signi…cant increase in computational time. Furthermore, we do not have reasons to believe that the main results would change. Bianchi (2013) estimates a version of the Christiano, Eichenbaum, and Evans (2005) model and …nds a sequence of regime changes similar to the one that we recover in this paper. 8 The ‡ow budget constraint of the representative household in period t reads Ct + Bt =Pt R1 0 (Nit Wit =Pt ) di + Bt 1 Rt 1 =Pt + Divt =Pt + Tt =Pt ; where Bt is the stock of one-period government bonds in period t, Rt is the gross nominal interest rate, Pt is the price of the …nal good, Wit is the nominal wage earned from labor in industry i, Tt is real transfers, and Divt are pro…ts from ownership of …rms. Composite consumption in period t is given by the Dixit-Stiglitz aggregator Ct = R1 1 1=" C 0 it di " " 1 ; where Cit is consumption of a di¤erentiated good i in period t and " > 1 determines the elasticity of substitution between consumption goods. The price level is given by Pt = R1 P 1 " di 0 it 1=(1 ") : (1) In every period t, the representative household chooses a consumption vector, labor supply, and bond holdings subject to the sequence of the ‡ow budget constraints and a noPonzi-scheme condition. The representative household takes as given the nominal interest rate, the nominal wages, nominal aggregate pro…ts, nominal lump-sum taxes, and the prices of all consumption goods. Firms: There is a continuum of monopolistically competitive …rms of mass one. Firms are indexed by i. Firm i supplies a di¤erentiated good i. Firms face Calvo-type nominal rigidities and the probability of re-optimizing prices in any given period is given by 1 independent across …rms. We allow for partial price indexation to steady-state in‡ation by …rms that do not re-optimize their prices, with the parameter ! 2 (0; 1) capturing the degree of indexation. Those …rms that are allowed to re-optimize their price choose their price Pit 9 so as to maximize: P1 k k=0 k! Et Qt;t+k Wit+k Nit+k jFt ; Pit Yit+k where Qt;t+k is the stochastic discount factor measuring the time t utility of one unit of consumption good available at time t + k, is the gross steady-state in‡ation rate, Nit is amount of labor hired, and Yit is the amount of di¤erentiated good produced by …rm i. Firms are endowed with an identical technology of production: Yit = Zt Nit : The variable Zt captures exogenous shifts of the marginal costs of production and is assumed to follow a stationary …rst-order autoregressive process in log-di¤erence: ln zt = (1 where zt z ) ln z + z ln zt 1 + Zt =Zt 1 : We refer to the innovations z zt ; zt zt N (0; 1) ; as technology shocks. Firms face a downward-sloping demand function in every period, Yit = (Pit =Pt ) " Yt ; where Pit denotes the price …rm i sells its good at time t. Aggregate labor input is de…ned as Z Nt = 1 1 1=" di Nit " " 1 : 0 Policymakers: There are a monetary authority and a …scal authority. Government consumption is de…ned as Gt = (1 1=gt ) Yt , with the variable gt following a stationary …rst-order autoregressive process: ln gt = 1 where gt g ln g + g ln gt 1 + g gt ; gt N (0; 1) ; (2) is an i.i.d. government expenditure shock. The …scal authority always follows a Ricardian …scal policy and collects a lump-sum tax. The aggregate resource constraint reads Yt = Ct + Gt : 10 The monetary authority sets the nominal interest rate Rt according to the Taylor rule p r; t Rt = Rt where t = Pt =Pt The variable rt 1 1 h p ; t t= (Yt = (zYt 1 )) p y; t i1 p r; t e r rt ; (3) N (0; 1) ; rt denotes the gross in‡ation rate and Yt is aggregate output in period t. captures nonsystematic exogenous deviations of the nominal interest rate p t Rt from the rule. The variable controls the policy regime that determines the policy coe¢ cients of the rule re‡ecting the emphasis of the central bank on in‡ation stabilization relative to output gap stabilization. 2.1 Policy Regimes We model changes in the central bank’s emphasis on in‡ation and output stabilization by introducing a three-regime Markov-switching process p t that evolves according to this matrix: 2 p12 p13 6 p11 6 Pp = 6 6 1 p22 p22 0 4 1 p33 0 p33 3 7 7 7: 7 5 (4) The realized regime determines the monetary policy parameters of the central bank’s reaction function. In symbols, R ( p t = j) ; ( p t = j) ; p t ( R y ( = j) ; p t ( = j) = p t = j) ; P P R; y ; ( p t P y , if j = 2 or j = 3: Under Regime = j) = A R; A ; A y , if j = 1 and 1 (the active regime), the central bank’s main emphasis is on stabilizing in‡ation and the Taylor principle is satis…ed: ( p t = 1) = A 1. Under Regime 2 (the short-lasting passive regime), the central bank de-emphasizes in‡ation stabilization, but only for short periods of time (on average). The same parameter combination also characterizes Regime 3 (the long-lasting passive regime). Therefore, ( p t = 1) = A P = ( p t = 2) = ( p t = 3). However, under Regime 3, deviations are generally more prolonged. In other words, Regime 2 is less persistent than Regime 3: p22 < p33 . Therefore, the two passive regimes do not 11 di¤er in terms of response to in‡ation P and the output gap P y, but only in terms of their relative persistence. The three policy regimes are meant to capture the recurrent changes in the Federal Reserve’s attitude toward in‡ation and output stabilization in the postwar period. A number of empirical works (Clarida, Galí, and Gertler 2000 Lubik and Schorfheide 2004) have documented that the Federal Reserve de-emphasized in‡ation stabilization for prolonged periods of time in the 1970s. Furthermore, as argued by Bernanke (2003), while the Federal Reserve has been mostly focused on actively stabilizing in‡ation starting from the early 1980s, it has also occasionally engaged in short-lasting policies whose objective was not to stabilize in‡ation in the short run. This monetary policy approach has been dubbed constrained discretion. We introduce this three-regime structure so as to give the model enough ‡exibility to explain both the long-lasting passive monetary policy of the 1970s, as well as the recurrent and short-lasting passive policies of the post-1970s. The probabilities p11 , p12 , p22 govern the evolution of monetary policy when the central bank follows constrained discretion. The larger p12 is vis-a-vis p11 , the more frequent the short-lasting deviations are. The larger p22 is, the more persistent the short-lasting deviations are. The probability p13 controls how likely it is that constrained discretion is abandoned in favor of a prolonged deviation from the active regime. The ratio p12 = (1 p11 ) captures the relative probability of a short-lasting deviation conditional on having deviated to passive regimes and can be interpreted as a measure of central bank’s long-run reputation. This is because this composite parameter controls how likely it is that the central bank will abandon constrained discretion the moment it starts deviating from the active regime. As it will become clear later on, central bank’s long-run reputation has deep implications for the general equilibrium properties of the macroeconomy. This is because agents are fully rational and form expectations while taking into account the possibility of regime changes, implying that their beliefs matter for the way shocks propagate through the economy. Therefore, the proposed de…nition of central bank reputation has the important advantage of being 12 measurable in the data, even over a relatively short period of time. 2.2 Communication Strategies It can be shown that regime changes do not a¤ect the steady-state equilibrium, but only the way the economy propagates around it. Since technology Zt follows a random walk, we normalize all the nonstationary real variable by the level of technology. We then log-linearize the model around the steady-state equilibrium in which the steady-state in‡ation does not have to be zero.5 Once log-linearized around the steady state, the imperfect information model can be solved under di¤erent assumptions on what the central bank communicates about the future monetary policy course. The central bank’s communication a¤ects agents’information set Ft . We consider two cases: no transparency and transparency. If the central bank is not transparent, it never announces the duration of passive policies. We call this approach no transparency. We make a minimal departure from the assumption of perfect information by assuming that agents can observe the history of all the endogenous variables and the history of the structural shocks but not the policy regimes p t. It should be noted that agents are always able to infer if monetary policy is currently active or passive. However, when monetary policy is passive, agents cannot immediately …gure out whether (short-lasting) Regime 2 or (long-lasting) Regime 3 is in place. To see why, recall that the two passive regimes are observationally equivalent to agents, given that p and p y are the same across the two regimes. Therefore, agents conduct Bayesian learning in order to infer which one of the two regimes is in place. In the next section we will discuss how agents’beliefs evolve as agents observe more and more deviations from the active regime. Under transparency all the information held by the central bank is communicated to agents. We assume that the central bank knows for how long it will be deviating from active monetary policy when conducting short-lasting deviations. Long-lasting deviations 5 The log-linearized equations and a detailed discussion on how nonzero steady-state in‡ation a¤ects the agents’behaviors in the model is in the online appendix. 13 are intended to capture structural changes in the way monetary policy is conducted (e.g., the type of a newly appointed central banker). Therefore, their duration is always unknown to the central bank and, hence, cannot be announced. Notice that under transparency, rational agents immediately infer when such a structural change in the conduct of monetary policy has occurred. If a transparent central bank starts deviating from active policy without announcing the duration of such a passive policy, this deviation must be a long lasting one.6 A transparent central bank announces the duration of short-lasting passive policies, revealing to agents exactly when monetary policy will switch back to the active regime. Agents form their beliefs by taking into account that the central bank will systematically announce the duration of every short-lasting passive policy. We assume that the central bank’s announcements are truthful and are believed as such by rational agents. In Section 6, we will consider the case in which the announcements made by the central bank are not always truthful. In Section 7.2, we will study the case in which the central bank can only announce the likely duration of passive policies 3 that is, the type of passive regime. Beliefs Dynamics and Model Solution Here we provide a brief discussion of how to solve the model under the two di¤erent communication strategies. More details are provided in the online appendix. No Transparency: To solve the model under no transparency we use the methods developed in Bianchi and Melosi (2016a). Denote the number of consecutive deviations from the active regime at time t as t 2 f0; 1; :::g, where at time t. Conditional on having observed t t = 0 means that monetary policy is active 1 consecutive deviations from the active regime at time t, agents believe that the central bank will keep deviating in the next period, 6 Our results still hold if one allows the central bank to announce the duration of the long-lasting deviations as well. 14 t + 1; with probability: prob f t+1 6= 0j Equation (5) makes it clear that p22 (p12 =p13 ) (p22 =p33 ) t + p33 : (p12 =p13 ) (p22 =p33 ) t + 1 t 6= 0g = (5) t is a su¢ cient statistic for the probability of being in the passive regime next period. This equation captures the dynamics of agents’ beliefs about observing yet another period of passive policy in the next period, which is the key state variable we use to solve the model under no transparency. It should be also observed that equation (5) has a number of properties that are quite insightful to the key mechanism of the model at hand. The probability of observing yet another period of passive policy in the next period is a weighted average of the probabilities p22 and p33 , with weights that vary with the number of consecutive periods of passive policy t. ( t When agents observe the central bank deviating from the active regime for the …rst time = 1), the weights for the probabilities p22 and p33 are p12 = (1 p11 ) and p13 = (1 p11 ), respectively. These weights re‡ect the central bank’s long-run reputation. When its long-run reputation is high, it is very unlikely that the central bank engages in a long-lasting passive policy. Therefore, as the …rst period of passive policy is observed, agents are con…dent that the economy has entered the short-lasting passive regime (Regime 2). If the central bank keeps deviating from the active regime, agents will eventually become convinced of being in the long-lasting passive regime (Regime 3). After a su¢ ciently long-lasting passive policy, the probability of observing an additional deviation in the next period degenerates to the persistence of the long-lasting passive regime (Regime 3). Hence, p33 is the upper bound for the probability that agents attach to staying in the passive regime next period. It follows that for any e > 0, there exists an integer for any t > such that p33 prob f t+1 6= 0j t = g < e: Therefore, , agents’beliefs can be e¤ectively approximated using the properties of the long-lasting passive regime. Endowed with these results, we can solve the model under no transparency by expanding 15 the number of regimes in order to take into account the evolution of agents’ beliefs. Now each regime is characterized by the central bank’s behavior and the number of observed consecutive deviations from active policy at any time t; set of regimes indexed by t 2 f0; 1; :::; t: The transition matrix for this new g can be derived by equation (5), as shown in the online appendix. Now regimes are de…ned in terms of the observed consecutive durations, t, which, unlike the primitive set of policy regime p t 2 f1; 2; 3g, belongs to the agents’ information set Ft . Hence, we can solve this model by applying any of the methods developed to solve Markov-switching rational expectations models with perfect information, such as Davig and Leeper (2007); Farmer, Waggoner and Zha (2011); and Foerster et al. (2013). We use Farmer, Waggoner, and Zha (2011). It is worth emphasizing that this way of recasting the learning process allows us to tractably model the behavior of agents that know that they do not know. In other words, agents are aware of the fact that their beliefs will change in the future according to what they observe in the economy. This represents a substantial di¤erence from the anticipated utility approach, in which agents form expectations without taking into account that their beliefs about the economy will change over time. Furthermore, our approach di¤ers from the one traditionally used in the learning literature in which agents form expectations according to a reduced-form law of motion that is updated recursively (for example, using discounted least squares regressions). The advantage of adaptive learning is the extreme ‡exibility given that, at least in principle, no restrictions need to be imposed on the type of parameter instability characterizing the model. However, such ‡exibility does not come without a cost, given that agents are not really aware of the model they live in. Transparency: When the central bank is transparent, the exact duration of every shortlasting deviation from active policy is truthfully announced. In this model the number of announced short-lasting deviations from active policy yet to be carried out a t is a su¢ cient statistic that captures the dynamics of beliefs after an announcement. Since the exact 16 duration of long-lasting passive policies is not announced, we also have to keep the longlasting passive regime as one of the possible regimes. Regimes are ordered from the smallest number of announced deviations (zero or the active policy) to the largest one ( a ). The long-lasting passive regime, whose conditional persistence is p33 , is ordered as the last regime. eA . The online appendix The evolution of the regimes is controlled by the transition matrix P explains how to build such a transition matrix. As in the case of no transparency, we recast the MS-DSGE model under transparency as a Markov-switching rational expectations model with perfect information, in which the short-lasting passive regime is rede…ned in terms of the number of announced deviations from the active regimes yet to be carried out, a t. This rede…ned set of regimes belongs to the agents’information set Ft under transparency. This result allows us to solve the model under transparency by applying any of the methods developed to solve Markov-switching rational expectations models of perfect information. 4 Empirical Analysis In order to put discipline on the parameter values, the model under no transparency is …tted to U.S. data. We believe that the model with a non-transparent central bank is better suited to capture the Federal Reserve communication strategy in our sample that ranges from the mid-1950s to just prior to the Great Recession. We then use the results to quantify the Federal Reserve’s reputation and the potential gains from making the Federal Reserve’s monetary policy more transparent. 4.1 Data and Estimation For observables, we use three series of U.S. quarterly data: the annualized Gross Domestic Product (GDP) growth rate, the annualized quarterly in‡ation (GDP de‡ator), and the federal funds rate (FFR). The sample spans from 1954:Q4 through 2009:Q3. Table 1 reports the prior and the posterior distribution of model parameters. The model is estimated by using 17 Name A A y A R P P y P R p11 p22 =p33 p33 p12 =(1 p11 ) " ! g z 100 ln z 100 ln 100 g 100 m 100 z 100 r Median Posterior 5% 95% Type Prior Mean Std. 3.0993 0.6947 0.6864 1.3365 0.4034 0.6925 0.9497 0.7070 0.9689 0.9536 0.9998 7.7764 0.6569 0.9189 0.9974 0.9496 0.3690 0.4808 0.7433 2.5499 3.1215 1.4827 0.6009 2.5299 0.5368 0.5186 1.0426 0.2541 0.5864 0.8955 0.5114 0.9403 0.9003 0.8469 4.3259 0.3562 0.8734 0.9955 0.9260 0.1435 0.3144 0.5571 1.8268 1.7974 1.0278 0.4531 3.7031 0.9126 0.7778 1.6032 0.6384 0.7774 0.9776 0.8915 0.9868 0.9883 1.1788 13.1161 0.8799 0.9496 0.9986 0.9682 0.5959 0.6368 0.9115 3.8237 5.8893 2.2091 08799 N G B G G B B B B B G G B B B B B N N IG IG IG IG 2.5 0.25 0.5 0.9 0.25 0.5 0.9 0.8 0.95 0.95 1 8 0.5 0.5 0.99 0.5 0.5 0.4 0.5 2 2 2 0.5 0.5 0.15 0.2 0.3 0.15 0.2 0.05 0.1 0.025 0.025 0.1 3 0.2 0.2 0.005 0.2 0.2 0.125 0.125 1 1 1 0.2 Table 1: Posterior modes, means, and 90% error bands of the model parameters. Type N, G, B, and IG stand for Normal, Gamma, Beta, and Inversed Gamma density, respectively. Dir stands for the Dirichelet distribution a Gibbs sampling algorithm in which both the regime sequence and the model parameters are sampled. The algorithm is similar to the one used in Bianchi (2013). Convergence is checked by using the Brooks-Gelman-Rubin potential reduction scale factor. The …ve chains consist of 270; 000 draws each and 1 of every 1; 000 draws is saved. The parameter values are quite standard with the central bank responding fairly aggressively to in‡ation when monetary policy is active. The central bank is also responding more aggressively to output under active policy. The response of the FFR to in‡ation in the passive regimes is estimated to be around 1.33, with 90% error bands spanning the interval between 1.04 and 1.60. This implies that many draws for the passive regime are well above 1, which is the threshold that is generally associated with the Taylor principle and active monetary policy. However, in a model like the one considered in this paper, the threshold for determinacy 18 is a¤ected by the absence of full indexation to trend in‡ation, as pointed out by Coibion and Gorodnichenko (2011). Once the determinacy region is properly adjusted, around 5% of the draws associated with the passive monetary policy rule fall in fact into the passive region and 30% of them are within 0.25 from the passive region. We still refer to this rule as passive to the extent that in‡ation stabilization is de-emphasized. Furthermore, other studies that use richer models instead of the prototypical …xed-parameter three-equation new-Keynesian model also …nd a sizable probability that the response of monetary policy to in‡ation was not violating the Taylor principle in the 1970s (see, for example, Bianchi (2013) and Davig and Doh (2014b)). The posterior median of the elasticity of substitution " implies a net markup equal to approximately 13%. The Calvo parameter implies a fairly large degree of nominal rigidities as is common when small-scale models are estimated. The Frisch elasticity of labor supply is close to one. The probability of being in the short-lasting passive regime conditional on having switched to passive policies, p12 = (1 p11 ), plays a critical role in the model. As noticed in Section 2, this parameter value relates to the strength of the Federal Reserve’s long-run reputation. This parameter is found to be fairly close to one, con…rming that the Federal Reserve has a strong reputation. This number means that as agents observe a deviation from the active regime, they expect that the Federal Reserve is conducting a short-lasting passive policy with a probability of 95:36%. Recall that in the estimated model, regimes are indexed with respect to the number of consecutive periods of passive policy, t. We have a total of + 1 regimes, where depends on the speed of learning and can be larger than 100. Reporting the regime probabilities for such a large number of regimes is not practical. A most e¤ective approach is to report the estimated expected number of consecutive deviations from active policy over the sample. As explained above, the higher the number of expected consecutive deviations, the larger is the posterior probability mass associated with the long-lasting passive regime. Furthermore, this statistic re‡ects agents’beliefs, and it is, therefore, critical to understand the e¤ects of 19 Expected Number of Consecutive Deviations 20 10 0 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 Figure 1: The gray shaded areas mark periods of passive monetary policy based on the regime sequence associated with the posterior mode based on the Gibbs sampling algorithm. The blue solid line reports the corresponding expected number of consecutive deviations from the active regime. central bank communication on social welfare, as we will show later. The shaded areas in Figure 1 show the periods of passive monetary policy based on the regime sequence associated with the posterior mode. The solid line reports the corresponding expected number of consecutive deviations from the active regime. This can be considered a measure of agents’pessimism because, as we will show later in the paper, a larger number of expected consecutive deviations determines an increase in uncertainty and, as a result, a decline in agents’welfare. The …gure highlights that short-lasting deviations from active policy only imply a modest increase in this statistic. In contrast, at the end of the 1970s and early 1980s the number of expected consecutive deviations approaches its highest value, (1 p33 ) 1 , re‡ecting the fact that most of the posterior probability is shifted toward regimes associated with passive policies of fairly long duration. The expected duration of passive policy grows gradually throughout the 1970s and reaches relatively high levels at the end of this decade. This suggests that agents slowly changed their expectations about future policy as they observed more and more periods of passive policy in the 1970s. After the 1970s, a large posterior probability is attributed to either the active regime or passive policies of very short realized duration. This is captured by the number of expected deviations from active policy being either close to zero, when the active regime prevails, or else slightly positive, 20 No Transparency Transparency 50 40 Pessimism 40 30 30 20 20 10 10 10 20 30 40 50 10 Consecutive Periods of Passive Policy 20 30 40 50 Consecutive Period of Passive Policy Figure 2: Pessimism on the vertical axis is measured as the expected number of consecutive deviations. 1 On the left plot the two horizontal lines denote the smallest lower bound (1 p22 ) and upper bound of 1 pessimism (1 p33 ) . These statistics are computed at the posterior mode. but below 10 quarters, i.e., 2 years and a half, when short-lasting deviations occur during the 2001 recession and in correspondence with the most recent recession. This is the essence of constrained discretion we want to study in this paper. Finally, we want to evaluate whether there is empirical support for our benchmark model with no transparency. To this end, we estimate an alternative model in which parameters are not allowed to change and then compare the two models by using Bayesian model comparison. We …nd that the data strongly favor the Markov-switching speci…cation, despite the larger number of parameters. In fact, the model with …xed parameters can attain a higher posterior probability only if one attaches extremely low prior probabilities (< 1:39E 11) to this model. 4.2 Communication and Beliefs Dynamics Regime changes in monetary policy and communication strategies critically a¤ect social welfare and the macroeconomic equilibrium by in‡uencing agents’pessimism about future monetary policy. In this paper, we use the word pessimism to precisely mean agents’expec- 21 tations about the duration of an observed passive policy. A high level of pessimism means that agents expect an observed passive policy to last for fairly long expected duration of the long-lasting passive regime: (1 that is, close to the p33 ) 1 . While expecting a longer lasting deviation from the active regime is not necessarily welfare decreasing, we will show that expecting a prolonged period of passive policy impairs social welfare in the estimated model. We measure pessimism by computing the number of expected consecutive periods of passive monetary policy conditional on the observed duration of passive policy 0. The evolution of this variable is tightly linked to the estimated transition matrix, that in turn captures the central bank’s long-run reputation. Let us consider the case in which the central bank decides to engage in passive policies lasting 50 consecutive periods. While such a long deviation from the active regime is not so likely, this example illustrates how transparency a¤ects pessimism relative to no transparency. Figure 2 reports the evolution of pessimism under no transparency (left graph) and under transparency (right graph) at the posterior mode. The two horizontal lines mark the smallest lower bound and upper bound for pessimism. The former is given by the expected duration of the short-lasting passive Regime (1 p22 ) 1 . The smallest lower bound is attained at the …rst period of passive policy only if the conditional probability of a short-lasting deviation is one: p12 = (1 p11 ) = 1. The left graph shows that the intercept of the solid line is quite close to the bottom dashed line, implying that agents expect that the Federal Reserve is engaging in a short-lasting deviation as the …rst period of passive policy is observed. This result is due to the fact that the Federal Reserve’s reputation is estimated to be fairly high (p12 = (1 p11 ) = 0:9536). The upper bound for pessimism is given by the expected duration of the long-lasting passive policy (1 p33 ) 1 and is attained only after a very large number of consecutive deviations from the active regime. Such a gradual increase in pessimism suggests that the Federal Reserve can enjoy a great deal of leeway in deviating from active monetary policy in order to stabilize alternative short-lasting objectives. This result is again due to the strong 22 reputation of the Federal Reserve. If the reputation coe¢ cient p12 = (1 p11 ) were close to zero, then the expected number of consecutive deviations would experience a larger jump and, hence, the convergence to the upper bound would be faster. As shown in the right graph, pessimism follows an inverse path under transparency. Unlike the case of no transparency, agents’ pessimism is very high at the initial stages of the deviation from active policy, but it decreases as the time goes by. This result comes from assuming that agents are fully rational and the announcement is truthful. As the 50 periods of passive monetary policy are announced (t = 0), an immediate rise in pessimism occurs. As the number of periods of passive policy yet to be carried out decreases, agents’ pessimism declines accordingly. At the end of the policy (t = 50), pessimism reaches its lowest level, with agents expecting to return to the active regime with a probability of one in the following period. It should be noted that at the end of the announced deviation, transparency allows the central bank to lower agents’ pessimism below the smallest lower bound attainable under no transparency: This result emerges because the central bank is able to inform agents about the exact period in which passive policy will be terminated. This assumption will be relaxed in Section 7.2. To sum up, Figure 2 allows us to isolate two important e¤ects of transparency on agents’ pessimism about future monetary policy: (i) transparency raises pessimism at the beginning of the policy and (ii) transparency anchors down pessimism at the end of the policy. As we shall show, these two e¤ects play a critical role for the welfare implications of transparency. 5 Welfare Implications of Transparency In this section, we assess the welfare implications of introducing transparency. Before proceeding, it is worth emphasizing that the regime changes considered in this paper do not a¤ect the steady state, but only the way the economy ‡uctuates around the steady state. The period welfare function can then be obtained by taking a log-quadratic approximation 23 of the representative household’s utility function around the deterministic steady state: P1 Wi (st (i)) = h=1 h [ 0 + 1 vari (^ yt+h jst (i)) + 2 vari (^ t+h jst (i))] ; (6) where vari ( ) with i 2 fT; N g stands for the stochastic variance associated with agents’ forecasts of in‡ation conditional on transparency (T ) or no transparency (N ) and the output gap at horizon h. The coe¢ cients i, i 2 f0; 1; 2g are functions of the model’s parameters and are de…ned in the online appendix. The subscript i refers to the communication strategy: i = N stands for the case of no transparency, while i = T denotes transparency. Finally, st (i) denotes the policy regime: st (i = N ) 2 f0; 1; :::; g= t and st (i = T ) 2 f0; 1; :::; a + 1g = a 7 t. The term 0 captures the steady-state e¤ects from positive trend in‡ation. These e¤ects stem from positive trend in‡ation raising cross-sectional steady-state dispersion in prices that in turn leads to ine¢ cient allocations of resources across industries (Coibion, Gorodnichenko, and Wieland, 2012). These steady-state e¤ects are eliminated if price indexation is perfect (! = 1). The term 1 is directly related to the increasing disutility of labor supply. Since households’costs of supplying labor are convex, the expected disutility from labor rises with the volatility of output around its steady state. As discussed in Coibion, Gorodnichenko, and Wieland, 2012, the magnitude of this coe¢ cient is invariant to the level of trend in‡ation The term 2 . captures the e¤ects of price dispersion on social welfare. Positive trend in‡ation generates some price dispersion. The increased price dispersion following an in‡ationary shock becomes now more costly because of the higher initial price dispersion due to positive trend in‡ation. Higher nominal rigidities ( ) lead to stronger e¤ects of price dispersion on welfare ( 2 ). It should be noted that zero trend in‡ation ( = 1) or positive trend in‡ation with perfect indexation (! = 1) would imply that the steady-state costs of positive trend in‡ation go to zero ( 7 Recall st (i = T ) = announced. a 0 = 0). A detailed derivation of the welfare function can be found + 1 denotes the long-lasting passive regime, whose exact realized duration is not 24 in Coibion, Gorodnichenko, and Wieland, 2012. These welfare coe¢ cients 0, 1, and 2 depends on the government-purchase-to-output ratio in steady state, which we assume to be equal to 22%. It can be shown that conditional on a price markup shock, the active regime is associated with a lower volatility of in‡ation but a higher volatility of the output gap compared with deviating to passive policies. This result captures the monetary policy trade-o¤ due to these ine¢ cient shocks, which is a well-known feature in the context of linear DSGE models. However, conditional on the other three shocks (i.e., the discount factor shock technology shock zt , and the monetary shock rt ), g;t , the active policy always leads to a lower level of both volatilities and, hence, to an unambiguously higher welfare. Equation (6) makes it explicit that social welfare depends on agents’uncertainty about future in‡ation and future output gaps. It should be noted that agents’uncertainty in any given period captures the macroeconomic risk associated with the observed policy regime and communication strategy, st (i). Unlike standard New Keynesian models with …xed parameters, where welfare is merely a function of the unconditional variance of in‡ation and the output gap, our model allows us to study the dynamic e¤ects of policy actions and forwardlooking communication on welfare. To the best of our knowledge, this is the …rst paper that studies this feature using a structural model. Furthermore, the learning mechanism plays an important role in our welfare analysis by linking the concept of a central bank’s long-run reputation to a central bank’s ability to control the dynamics of the macroeconomic risk associated with policy actions. This last point will be the focus of the next session. To assess the desirability of transparency, we compute the model predicted welfare gains/losses from transparency as follows: We = P a +1 a =0 p T ( a ) WT ( a ) P =0 pN ( ) WN ( ) (7) where pT ( a ) stands for the vector of ergodic probabilities of a passive policy of announced 25 duration a and pN ( ) stands for the vector of ergodic probabilities of a passive policy of observed duration . It is important to emphasize that welfare gains from transparency are not conditioned on a particular shock or policy path. Instead, the welfare gain is measured by the unconditional long-run change in welfare that arises if the central bank systematically announces the duration of any short-lasting deviation from active monetary policy. Uncertainty about future output gaps turn out to play only a minor role for social welfare, since the estimated value of the slope of the Phillips curve is very small and the elasticity of substitution among goods " is quite large. Such a ‡at Phillips curve is a standard …nding when DSGE models are estimated using U.S. data and the estimated value of the elasticity of substitution is in line with the results of previous studies and with micro data on U.S. …rms’ average pro…tability. The estimated value of these two parameters causes the estimated coe¢ cient for the in‡ation risk in the welfare function ( coe¢ cients ( 0 and 1) 2) to be bigger than the other two by several orders of magnitude. Therefore, welfare turns out to be tightly related to agents’uncertainty about future in‡ation, which, as we shall show, depends on the time-varying level of pessimism about observing a future switch to active monetary policy. For the sake of brevity, in what follows we do not discuss the evolution of uncertainty about output gaps. 5.1 Evolution of Uncertainty We have shown that agents’uncertainty about future in‡ation crucially a¤ects social welfare in the estimated model. In this section, we will show how uncertainty is tightly linked to agents’pessimism about observing active monetary policy in the future. As shown in Section 4.2, transparency has two e¤ects on pessimism: (i) pessimism rises at the beginning of the policy (henceforth, the short-run e¤ect of transparency on pessimism) and (ii) pessimism is anchored down at the end of the policy (henceforth, the anchoring e¤ect of transparency on pessimism). As we shall show, these two e¤ects play a critical role for the welfare implications of enhancing a central bank’s transparency. 26 To illustrate how uncertainty responds to pessimism under the two communication strategies, we consider the case in which the Federal Reserve conducts a 40-quarter-long deviation from active monetary policy.8 While such a long-lasting realization of the short-lasting regime is implausible, this example allows us to highlight the key implications of the two communication strategies on welfare. The upper panel of Figure 3 shows the evolution of uncertainty about in‡ation at di¤erent horizons h under no transparency (left panel) and under transparency (right panel). At each point in time, the evolution of agents’uncertainty is measured by the h-period ahead standard deviation of in‡ation given the communication i hp p vari ( t+h jst (i)) vari ( t+h jst (i) = 0) , where strategy that is, sdi ( t+h j t ) = 100 i 2 fN; T g captures the communication strategy.9 We analytically compute the conditional standard deviations taking into account regime uncertainty by using the methods described in Bianchi (2016). As shown in the upper left graph, when the central bank does not announce its policy course beforehand, uncertainty about future in‡ation is fairly low at the beginning of the policy because agents interpret the …rst deviations from active policy as short-lasting. As more and more periods of passive policy are observed, agents become progressively more convinced that the observed deviation may have a long-lasting nature and uncertainty about future in‡ation gradually takes o¤. Uncertainty rises because expecting a longer spell of passive policies raises concerns about the central bank’s ability to control the in‡ationary consequences of future shocks. Note that the increase in uncertainty occurs at every horizon because agents expect passive monetary policy to prevail for many periods ahead. It is worth emphasizing that the pattern of agents’uncertainty over time mimics the evolution of pessimism depicted in Figure 2. Since higher uncertainty leads to bigger welfare losses, the progressive disanchoring of uncertainty about future in‡ation is a reason of concern for 8 The analysis is conducted for an economy at the steady-state and, hence, without conditioning on a particular shock. The exercise is conditioned only on the policy path and intends to facilitate the exposition of the welfare implications of transparency in the next section. 9 The graphs plot the results for h from 1 to 60: At horizon h = 0, uncertainty is zero as agents observe current in‡ation. 27 Transparency Uncertainty Uncertainty No Transparency 0.6 0.4 0.2 60 40 20 10 Horizon 20 40 30 0.4 0.2 60 40 20 10 Horizon Time 20 40 30 Time 0.6 Uncertainty 0.6 Uncertainty 0.6 0.5 0.4 Perf.Inf.Bounds 10 deviations 20 deviations 40 deviations 0.3 0.2 0.5 0.4 Perf.Inf.Bounds 10 deviations 20 deviations 40 deviations 0.3 0.2 10 20 30 40 50 60 10 Horizon 20 30 40 50 60 Horizon Figure 3: Upper graphs: Evolution of uncertainty about in‡ation at di¤erent horizons (h) over 40 periods of passive policy (time) under no transparency (left graph) and under transparency (right graph). The vertical axis reports the standard deviations in percentage points at the posterior mode. Lower graphs: Dynamics of uncertainty across horizons after having observed (left plot) or announced (right plot) 10, 20, and 40 consecutive quarters of passive policy. The gray areas denote the bounds for the dynamics of uncertainty about in‡ation across horizons when agents know the nature of the observed passive policy. The upper (lower) bound is when the passive policy is of the short-lasting (long-lasting) type. Parameter values are set at the posterior mode. a nontransparent central bank. The lower left panel shows the dynamics of uncertainty across di¤erent horizons when 10, 20, and 40 quarters of deviations are observed under no transparency. The gray area captures the dynamics of uncertainty across horizons in the case of perfect information that is, the case in which agents know whether the passive regime in place is short- or longlasting. Thus, the lower- (upper-) bound of the gray area captures the uncertainty when agents know for certain that the short- (long-) lasting passive regime is in place. After 40 consecutive deviations from active policy have been observed, uncertainty evolves as if agents knew with certainty that the central bank is conducting a long-lasting policy (the 28 black dashed-dotted line). After observing so many deviations, agents are certain that this is a long-lasting passive policy and cannot become more uncertain about future in‡ation. Therefore, the dynamics of uncertainty when agents perfectly know that the nature of passive policy is long-lasting represents an upper bound for agents’uncertainty. The dynamics of uncertainty conditional on a short-lasting passive policy under perfect information constitute a lower bound for uncertainty under no-transparency. Once the central bank starts deviating, the higher the central bank’s long-run reputation p12 = (1 p11 ) is, the closer the dynamics of uncertainty to this lower bound are. The lower left graph shows that the evolution of uncertainty remains close to the lower bound even after ten consecutive periods of passive policy. This result re‡ects the high reputation of the Federal Reserve. The right upper graph of Figure 3 shows the dynamics of uncertainty about future in‡ation in the case of transparency. Comparing the upper graphs (the scale of the z-axes are identical) illustrates that uncertainty is higher under transparency at the beginning of a 40-period long passive policy. This is captured by the pronounced hump-shaped dynamics of short- and medium-horizon uncertainty. This result is driven by the short-run e¤ect of transparency on pessimism. The announcement commits the central bank to follow a passive policy for the next 40 periods, causing agents to expect larger consequences from the shocks that will materialize during the implementation of the announced policy path. The lower right graph compares the dynamics of uncertainty after announcing the passive policy of increasing durations (10, 20, and 40 quarters) with the upper and lower bounds for the case of no transparency (the gray area). After announcing 40 quarters of passive policy, uncertainty is above the gray area at short and medium horizons, implying that uncertainty becomes higher than the upper bound for the case of no transparency. This overreaction of short-run uncertainty is driven by the short-run e¤ect of transparency on pessimism and contributes to lowering the welfare gains from transparency. Compared with uncertainty in the case of no transparency, uncertainty in the case of transparency is larger at the beginning of the policy at both short and medium horizons. 29 However, 40-quarter-ahead in‡ation uncertainty appears to be smaller in the case of transparency. This result is due to the anchoring e¤ect of transparency on pessimism. While agents know that monetary policy will be passive for 40 quarters, they also take into account that there will be a switch to the active regime in 40 quarters. Announcing the timing of the return to active monetary policy determines a fall in uncertainty in correspondence with the horizons that coincide with the announced date. In the upper right graph of Figure 3, such a decline in uncertainty shows up as a valley in the surface representing the level of uncertainty. As we shall show, this feature of transparency has the e¤ect of raising social welfare by systematically anchoring agents’uncertainty at the end of the announced deviations from the active regime. Furthermore, at long horizons, uncertainty is always lower under transparency. In fact, the lower right graph shows that in the case of transparency long-horizon uncertainty is lower than the lower bound for the no-transparency case even when very persistent passive policies are announced. This result is due to the anchoring e¤ect of transparency on pessimism and contributes to raising welfare gains from transparency. To sum up, under the no transparency, uncertainty increases across all horizons as the policy is implemented while under transparency, uncertainty decreases over time because agents are aware that the end of the prolonged period of passive monetary policy is approaching. These opposite patterns for uncertainty under the two communication strategies are due to the anchoring e¤ect of transparency on pessimism. It should be noted that the evolution of uncertainty conditional on being in the active regime is not the same across the two alternative communication strategies. This is because transparency determines an overall reduction in uncertainty that manifests itself also under the active regime, even if under the active regime no announcement is made. A transparent central bank enjoys lower uncertainty even when monetary policy is active because agents understand that should a short-lasting passive policy of any duration be implemented in the future, the central bank will announce its duration beforehand. As it will soon become clear, such a communication strategy is e¤ective in reducing uncertainty by removing the fear of 30 Welfare Gains From Transparency Ergodic prob (left scale) Welfare gains from transparency (right scale) 0.6 3.75 3.7 0.4 3.65 0.2 3.6 0 0 5 10 15 20 25 30 35 Observed Periods of Passive Policy Figure 4: The solid line shows the dynamics of the welfare gains from transparency as a function of the observed periods of passive policy ( t ). The bars show the ergodic probability of observing the periods of passive policy on the x-axis ( t ). Parameter values are set at the posterior mode. a long-lasting deviation for the frequent short-lasting deviations and creating an anchoring e¤ect for the sporadic long-lasting deviations. Since the active regime occurs often, its weight for the welfare calculation in equation (7) is rather large, implying that welfare gains conditional on being in the active regime will critically a¤ect the welfare-based ranking of the two alternative communication strategies. 5.2 Welfare Gains from Transparency To assess the overall welfare gains from transparency, we use equation (7), which combines the welfare associated with the policy regimes ( t for the case of no transparency and a t for the case of transparency) and their ergodic probabilities.10 To facilitate the comparison, we rede…ne the regimes under transparency a t in terms of observed periods of passive policy t and recompute welfare in the case of transparency associated with these new set of regimes. The line in Figure 4 shows the welfare gains from transparency associated with having 10 This is a long-run welfare measure. The online appendix shows the evolution of welfare under transparency and no transparency as passive policies of di¤erent length are implemented. 31 observed passive policies for t periods based on the posterior mode estimates. The bars report the ergodic probabilities of regimes t. Only short deviations from active policy are plausible for the U.S. The line shows that for passive policies of plausible durations, transparency raises welfare, implying that the model predicted welfare gains from transparency We in equation (7) are positive. Interestingly, the welfare gains from transparency for observed deviations t gradually decline as the number of observed deviations increases. This result stems from the fact that announcing longer and longer deviations progressively strengthens the short-run e¤ect of transparency on pessimism. This, in turn, raises the risk of macroeconomic instability, as shown in Figure 3. We …nd that the gains from transparency are roughly 3.74% of steady-state consumption for the U.S. economy, with a 70% posterior credible interval covering the range 1:74% 5:30%. . This result implies that the anchoring e¤ect of transparency dominates the short-run e¤ects. In other words, transparency is welfare improving because it allows the central bank to e¤ectively sweep away the fear of a return to the 1970s-type of passive policies. This explains why when the central bank conducts an active policy ( t = 0), the welfare gains from transparency are not zero. They are, in fact, positive, capturing the welfare gains from expecting that the central bank will systematically and truthfully announce the duration of any future short-lasting passive policy. 5.3 In‡ation Uncertainty in the Data One property of the estimated model is that beliefs change gradually as more and more periods of passive policy are observed. If we assume that for an alternative model, agents perfectly know the realization of policy regimes (perfect information), their beliefs would respond abruptly as the central bank changes its attitude toward in‡ation stabilization. In this section, we want to test the diverging predictions of these two models on the dynamics of in‡ation uncertainty in the 1970s, which both models identify as a period in which a 32 Inflation Uncertainty 1 Data: D'Amico Orphanides (2014) Model with Learning Model with Perfect Information 0.9 0.8 0.7 0.6 0.5 1970 1972 1974 1976 1978 1980 Figure 5: Long-Run Trend of Uncertainty about Future In‡ation Predicted by the Model with Learning and the Perfect Information Model. The dynamics of in‡ation uncertainty resulting only from policy actions. The dynamics of uncertainty predicted by the two models is rescaled so that in 1968:Q4, in‡ation uncertainty is equal to the least square constant estimated using uncertainty in the data. long-lasting passive policy was implemented.11 This test is intended to empirically validate the learning mechanism put forward in the paper.12 We focus on uncertainty about in‡ation because this variable is the key driver of social welfare in the estimated model. Figure 5 compares the dynamics of one-year-ahead in‡ation uncertainty measured by D’Amico and Orphanides (2014) from the Survey of Professional Forecasters, with the trend in the one-year-ahead in‡ation uncertainty predicted by the estimated Markov-switching model with learning (no transparency) and by the estimated Markov-switching model with 11 We focus on the 1970s for two reasons. First, our model does not have interesting predictions for uncertainty during periods characterized by active monetary policy because the central bank’s reputation is immediately rebuilt once monetary policy moves back to the active regime. As a result, the dynamics of uncertainty is ‡at, like in a perfect information model. Second, the dynamics of beliefs during periods of active policy or short-lasting passive policy are very similar to the ones that would arise under perfect information, making it hard to distinguish our benchmark model from the alternative model with perfect information over sample periods that do not include the 1970s. 12 Comparing the marginal likelihood of the estimated model with that of the same model with perfect information leads to inconclusive results. It is likely that estimating the models with a data set that includes in‡ation expectations may help us select one of the two models. However, reliable survey-based expectations data are available only from the early 1970s onward. Yet, using the entire sample is key to estimating the properties of the active regime and those of the transition matrix. Using uncertainty as an observable variable is computationally very challanging as uncertainty follows a nonlinear law of motion, requiring MC …ltering to evaluate the likelihood. 33 perfect information. In the latter model, agents perfectly know which type of policy regime is in place. The model with perfect information predicts a sharp rise in in‡ation uncertainty as monetary policy switches to the long-lasting passive policy. After the switch to passive policy, the perfect information model predicts that long-run uncertainty stays put at this higher level throughout the 1970s. In contrast, the model with learning predicts a smaller rise in uncertainty as monetary policy becomes passive in the late 1960s and a gradual run-up in in‡ation uncertainty in the subsequent years. This gradual increase in uncertainty is due to the prolonged period of passive policy that caused agents to become progressively more convinced that this policy had a long-lasting nature. The data (the dotted line) suggest that in‡ation uncertainty grew slowly in the 1970s, favoring the dynamics predicted by the model with learning. Since uncertainty is not an observable in our estimation, the comparison in Figure 5 constitutes an external validation exercise. 6 Imperfectly Credible Announcements In this section we study the consequences of imperfectly credible announcements. Let us consider the case in which the central bank systematically announces the duration of shortlasting deviations, but it lies if the duration is longer than . To model this idea, we assume that the duration of passive policies is drawn accordingly with the Markov-switching process implied by the primitive matrix P, de…ned in Section 3. If the drawn duration of the passive policy is smaller than or equal to , the central bank announces drawn duration of the passive policy a number of consecutive deviations is discovered after a a = . If the is larger than , the central bank lies and announces a between 1 and . Note that the central bank’s lie + 1 periods, and at this point, rational agents know that the policy will stay passive for sure until a. For any period of the short-lasting passive policy observed after , agents have to learn the persistence of the regime in place as they do in the no-transparency case. 34 We also assume that if the central bank lies, it is more likely to announce a fairly large number of deviations in order to deceive the lie for longer. This property also implies that the longer the announced deviation, the more likely it is that the central bank has lied and will keep deviating from active policy at the end of the announced passive policy. Thus, the probability of returning to an active policy declines as the horizon of the announced policy increases. The central bank does not announce long-lasting deviations, which is consistent with how we have de…ned transparency throughout this paper. A detailed description of how to specify the transition matrix for the regimes in the case of imperfectly credible announcements is provided in the online appendix. In what follows, we assume that the central bank always lies if the number of deviations is larger than = 4: We then assume a mapping f ( a ) to control the probability that the central bank announces a periods of passive policy conditional on having lied ( > ): f ( a ) = f:1; :2; :3; :4g for a = f1; 2; 3; 4g, respectively. This mapping implies that if the central bank lies, it is four times more likely that a four-quarter deviation from active policy is announced relative to a one-quarter deviation. This last assumption causes the probability of returning to an active policy to decline as the horizon of the announced policy increases. It turns out that this speci…cation does not substantially change the main results of the paper. Even if the central bank can lie, transparency is still welfare improving. In fact, limiting the number of periods of passive policy the central bank truthfully announces, leads to a slight increase in the welfare gains from transparency. For the sake of illustrating the key mechanism, Figure 6 shows the dynamics of welfare as a 20-period passive policy is implemented under three di¤erent announcements. In all cases, the announcements are made by the central bank at the beginning of the period of passive policy.13 The solid blue line corresponds to the benchmark model and captures the dynamics of welfare in the 13 This example is about a very longed passive policy ( = 20 > ) and hence it is very unrealistic for countries such as the U.S, in which the average duration of a short-lasting passive policy is three quarters. Nevertheless, this numerical example allows us to isolate two key e¤ects of lying about passive policies with duration > on the welfare gains from transparency. It can be shown that for a given transition matrix P, the magnitude of these two e¤ects becomes smaller and smaller as grows. 35 Transparency: 20-Quarter Passive Policy Truthful Announcement Untruthful Announcement: 1 Quarter Untruthful Announcement: 4 Quarters Welfare -102 -104 -106 -108 -110 2 4 6 8 10 12 14 16 18 20 Time Figure 6: Welfare evolution as a 20-quarter deviation from the active policy is implemented under imperfectly credible announcements. case in which the central bank truthfully announces the duration of the passive policy. The dot-dashed line and the solid line with circles capture the dynamics of welfare in the case in which the central bank announces that the passive policy will last one quarter and four quarters, respectively. This plot illustrates that lying about the duration of passive policies raises welfare in the short run. These gains stem from preventing the sudden rise in pessimism that occurs at the beginning of an announced passive policy. Nevertheless, lying eventually back…res. Once agents realize that the announcement was in fact a lie, welfare experiences a discrete drop. After that, it starts following a pattern similar to the one prevailing under no transparency. Of course, relaxing the assumption that announcements are fully credible has also a third e¤ect on the welfare gains from transparency. This has to do with the fact that agents question the veracity of central bank’s announcements even when the central bank is, in fact, telling the truth. This e¤ect contributes to lowering the welfare gains from transparency. For the U.S., the positive short-run e¤ects on the welfare gains from transparency dominate the other two negative e¤ects. Consequently, our results suggest that even if the central bank is 36 Welfare gains from transparency 6.5 6 5.5 5 4.5 4 2 4 6 8 10 12 14 16 18 20 Realized Duration of Passive Policy Figure 7: Average per-period welfare gains from transparency associated with passive policies of given realized durations. Parameter values are set at their posterior mode. unable to make perfectly credible announcements, transparency still determines an increase in welfare and this gain might be even larger than under perfectly credible announcements. 7 Robustness In this section we conduct a series of robustness exercises. In Section 7.1 we investigate whether transparency is welfare increasing for passive policies of every plausible duration. In Section 7.2, we relax the assumption that the central bank knows exactly the realized duration of the ongoing passive policy. 7.1 Short-Run Bene…ts from Transparency In the previous sections, we have showed that enhancing a central bank’s transparency would raise welfare. The computation of expected welfare gains from transparency is obtained using the ergodic distribution of the policy regimes and, hence, captures the long-run gains. However, it remains to be seen if the central bank is better o¤ following transparency for any possible duration of the short-lasting deviations. In other words, are there short-lasting 37 deviations for which the central bank would rather be nontransparent? We …nd that the positive gains from transparency occur for every plausible duration of the passive policy. To see this, the solid line of Figure 7 shows the dynamics of the average per-period welfare gains from transparency associated with the short-lasting passive policy of various durations. The important result is that welfare gains are positive for all plausible durations of short-lasting passive policies. This …nding suggests that the central bank is better o¤ by announcing passive policies of every plausible duration. Quite interestingly, this plot suggests that the central bank is better o¤ even if it has to announce passive policies of fairly long durations. This is an important result that implies that the overall reduction in uncertainty that occurs owing to introducing transparency overcomes any short-run loss associated with announcing a prolonged deviation from active policy. It should be noted the di¤erence between the welfare gains from transparency in Figure 4 and those of Figure 7. Figure 4 reports the welfare gains from having announced the duration of the ongoing passive policy when t deviations out of the announced a t t have been observed. Figure 7 shows the average per-period welfare gains, should the Federal Reserve decide to announce a passive policy of a certain duration. The latter measure evaluates the average welfare gains from transparency across periods of policy implementation, whereas the former measure, combined with the ergodic probability distribution of the deviations, captures the expected bene…t from being transparent over the long run. 7.2 Limited Information We have modeled transparency as a communication strategy in which the central bank shares all the information about the policy regime to households and …rms. Since we assume that the central bank knows the exact duration of its short-lasting passive policies, transparency implies that such information is shared with the public. In this section, we relax the assumption that the central bank knows the exact duration of passive policies. Rather, we assume that the central bank knows only the expected duration of the deviations from the 38 active regime that is, the bank perfectly knows only if the passive policy is short-lasting or long-lasting. Thus, now under transparency the central bank truthfully announces whether it will be conducting a short-lasting or a long-lasting passive policy.14 We …nd that under limited information, welfare gains from transparency are always positive for policies of any plausible duration.15 Compared with the case in which the central bank announces the exact duration of the short-lasting passive policies, the magnitude of the welfare gains from transparency is smaller as the central bank knows less about the duration of the policy it is implementing. In fact, the model predicted welfare gains from transparency amount to 0.54% of steady-state consumption. Thus, our analysis suggests that the welfare gains from transparency are positive and are quanti…ed to range from 0.54% to 3.74% depending on how much informed the central bank is regarding the duration of passive policies. 8 Concluding Remarks We have developed a general equilibrium model in which the central bank can deviate from active in‡ation stabilization. Agents observe when monetary policy becomes passive, but they face uncertainty regarding the nature of these deviations. Importantly, when observing passive policy, they cannot rule out the possibility of a return to the 1970s-type of passive policies. The longer the deviation from active policy is, the more pessimistic about the evolution of future monetary policy agents become. This implies that as the central bank keeps deviating, uncertainty increases and welfare deteriorates. When the model is …tted to U.S. data, we …nd that the Federal Reserve bene…ts from a strong reputation. As a result, policymakers can deviate for a prolonged period of time from active monetary policy before losing control over agents’ uncertainty about future 14 Therefore, the model boils down to a Markov-switching model with perfect information given that now the history of policy regimes pt 2 f1; 2; 3g belongs to the agents’information set. 15 We plot the welfare gains from transparency conditional on the observed periods of passive policy as well as the distribution of these periods in the online appendix. 39 in‡ation. Nevertheless, increasing transparency about the central bank’s future behaviors would improve welfare by anchoring agents’ pessimism following exceptionally prolonged periods of passive monetary policy and by removing the fear of a return to the 1970s following the frequent short-lasting deviations. A limitation of the current analysis is that agents learn only the persistence of passive policies, while the active regime is fully revealing. This implies that agents’expectations are completely revised as soon as the central bank returns to the active regime. In Bianchi and Melosi (2016a) we develop a more general methodology that could be used to study a model in which agents have to learn about the likely duration of both passive and active policies. This extension would add further realism to the model because it would make the cost of losing reputation more persistent. While we regard the estimation of a model with richer learning dynamics as an important direction for future research, at this stage estimating a model of this type is computationally challenging. Furthermore, we believe that such an extension is unlikely to a¤ect the main conclusions of this paper. This is because announcing the return to a long-lasting period of active monetary policy would still have the e¤ect of anchoring agents’pessimism and uncertainty. References Barro, Robert J., and David B. Gordon. 1983. “Rules, discretion and reputation in a model of monetary policy.”Journal of Monetary Economics, 12(1): 101–121. Bernanke, Ben S. 2003. “Constrained Discretion" and Monetary Policy.” Remarks by Governor Ben S. Bernanke. Bernanke, Ben S, and Frederic S Mishkin. 1997. “In‡ation Targeting: A New Framework for Monetary Policy?” Journal of Economic Perspectives, 11(2): 97–116. 40 Bernanke, Ben S, Thomas Laubach, Frederic S. Mishkin, and Adam S. Posen. 1999. In‡ation Targeting: Lessons from the International Experience. Princeton University Press. Bianchi, Francesco. 2013. “Regime Switches, Agents’Beliefs, and Post-World War II U.S. Macroeconomic Dynamics.”Review of Economic Studies, 80(2): 463–490. Bianchi, Francesco. 2016. “Methods for Measuring Expectations and Uncertainty in Markov-switching Models.”Journal of Econometrics, 190(1): 79 –99. Bianchi, Francesco, and Cosmin Ilut. 2013. “Monetary/Fiscal Policy Mix and Agents’ Beliefs.”C.E.P.R. Discussion Papers CEPR Discussion Papers 9645. Bianchi, Francesco, and Leonardo Melosi. 2014. “Dormant Shocks and Fiscal Virtue.” In NBER Macroeconomics Annual 2013. Vol. 28 of NBER Chapters, , ed. Jonathan Parker and Michael Woodford, 1–46. National Bureau of Economic Research, Inc. Bianchi, Francesco, and Leonardo Melosi. 2016a. “Modeling the Evolution of Expectations and Uncertainty in General Equilibrium.” The International Economic Review, 57(2): 717–756. Bianchi, Francesco, and Leonardo Melosi. 2016b. “Modeling the Evolution of Expectations and Uncertainty in General Equilibrium.” The International Economic Review, 57(2): 717–756. Campbell, Je¤rey R., Charles L. Evans, Jonas D.M. Fisher, and Alejandro Justiniano. 2012. “Macroeconomic E¤ects of Federal Reserve Forward Guidance.”Brookings Papers on Economic Activity, 44(1): 1–80. Campbell, Je¤rey R., Jonas D.M. Fisher, Alejandro Justiniano, and Leonardo Melosi. Forthcoming. “Forward Guidance and Macroeconomic Outcomes Since the Financial Crisis.”NBER Macroeconomics Annual 2016. 41 Cho, In-Koo, Noah Williams, and Thomas J. Sargent. 2002. “Escaping Nash In‡ation.”Review of Economic Studies, 69(1): 1–40. Christiano, Lawrence J., Martin Eichenbaum, and Charles L. Evans. 2005. “Nominal Rigidities and the Dynamic E¤ects of a Shock to Monetary Policy.”Journal of Political Economy, 113(1): 1–45. Clarida, Richard, Jordi Galí, and Mark Gertler. 2000. “Monetary Policy Rules And Macroeconomic Stability: Evidence And Some Theory.” The Quarterly Journal of Economics, 115(1): 147–180. Cogley, Timothy, and Thomas J. Sargent. 2005. “Drift and Volatilities: Monetary Policies and Outcomes in the Post WWII U.S.”Review of Economic Dynamics, 8(2): 262– 302. Cogley, Timothy, Christian Matthes, and Argia M. Sbordone. 2011. “Optimal disin‡ation under learning.”Federal Reserve Bank of New York Sta¤ Reports 524. Coibion, Olivier, and Yuriy Gorodnichenko. 2011. “Monetary Policy, Trend In‡ation, and the Great Moderation: An Alternative Interpretation.” American Economic Review, 101(1): 341–70. Coibion, Olivier, Yuriy Gorodnichenko, and Johannes Wieland. 2012. “The Optimal In‡ation Rate in New Keynesian Models: Should Central Banks Raise Their In‡ation Targets in Light of the Zero Lower Bound?” Review of Economic Studies, 79(4): 1371– 1406. D’Amico, Stefania, and Athanasios Orphanides. 2014. “In‡ation Uncertainty and Disagreement in Bond Risk Premia.” Federal Reserve Bank of Chicago Working Paper Series WP-2014-24. 42 Davig, Troy, and Eric M. Leeper. 2007. “Generalizing the Taylor Principle.”American Economic Review, 97(3): 607–635. Davig, Troy, and Taeyoung Doh. 2014a. “Monetary Policy Regime Shifts and In‡ation Persistence.”The Review of Economics and Statistics, 96(5): 862–875. Davig, Troy, and Taeyoung Doh. 2014b. “Monetary Policy Regime Shifts and In‡ation Persistence.”The Review of Economics and Statistics, 96(5): 862–875. Del Negro, Marco, Marc Giannoni, and Christina Patterson. 2012. “The Forward Guidance Puzzle.”Federal Reserve Bank of New York Sta¤ Reports 574. Eusepi, Stefano, and Bruce Preston. 2010. “Central Bank Communication and Expectations Stabilization.”American Economic Journal: Macroeconomics, 2(3): 235–71. Evans, George W., and Seppo Honkapohja. 2003. “Expectations and the Stability Problem for Optimal Monetary Policies.”Review of Economic Studies, 70(4): 807–824. Evans, G.W., and S. Honkapohja. 2001. Learning and Expectations In Macroeconomics. Princeton:Princeton University Press. Farmer, Roger E.A., Daniel F. Waggoner, and Tao Zha. 2011. “Minimal State Variable Solutions to Markov-Switching Rational Expectations Models.”Journal of Economic Dynamics and Control, 35(12): 2150–2166. Faust, Jon, and Lars E O Svensson. 2001. “Transparency and Credibility: Monetary Policy with Unobservable Goals.”International Economic Review, 42(2): 369–97. Fernandez-Villaverde, Jesus, and Juan F. Rubio-Ramirez. 2008. “How Structural Are Structural Parameters?” In NBER Macroeconomics Annual 2007, Volume 22. NBER Chapters, 83–137. National Bureau of Economic Research, Inc. Foerster, Andrew, Juan Rubio-Ramirez, Dan Waggoner, and Tao Zha. 2013. “Perturbation methods for Markov-switching DSGE model.” 43 Galí, Jordi, and Mark Gertler. 2007. “Macroeconomic Modeling for Monetary Policy Evaluation.”Journal of Economic Perspectives, 21(4): 25–46. Gorodnichenko, Yuriy. 2008. “Endogenous information, menu costs and in‡ation persistence.”National Bureau of Economic Research, Inc NBER Working Papers 14184. Justiniano, Alejandro, and Giorgio E. Primiceri. 2008. “The Time-Varying Volatility of Macroeconomic Fluctuations.”American Economic Review, 98(3): 604–41. Kydland, Finn E, and Edward C Prescott. 1977. “Rules Rather Than Discretion: The Inconsistency of Optimal Plans.”Journal of Political Economy, 85(3): 473–91. Liu, Z., D. Waggoner, and T. Zha. 2011. “Sources of the Great Moderation: A RegimeSwitching DSGE Approach.”Quantitative Economics, 2(2): 251–301. Lubik, Thomas A., and Frank Schorfheide. 2004. “Testing for Indeterminacy: An Application to U.S. Monetary Policy.”American Economic Review, 94(1): 190–217. Mackowiak, Bartosz, and Mirko Wiederholt. 2009. “Optimal Sticky Prices under Rational Inattention.”American Economic Review, 99(3): 769–803. Mankiw, N. Gregory, and Ricardo Reis. 2006. “Pervasive Stickiness.” American Economic Review, 96(2): 164–169. Marcet, Albert, and Thomas J Sargent. 1989a. “Convergence of Least-Squares Learning in Environments with Hidden State Variables and Private Information.” Journal of Political Economy, 97(6): 1306–22. Marcet, Albert, and Thomas J. Sargent. 1989b. “Convergence of least squares learning mechanisms in self-referential linear stochastic models.”Journal of Economic Theory, 48(2): 337–368. Melosi, Leonardo. 2014. “Estimating Models with Dispersed Information.”American Economic Journal: Macroeconomics, 6(1): 1–31. 44 Melosi, Leonardo. Forthcoming. “Signaling E¤ects of Monetary Policy.” Review of Economic Studies. Mester, Loretta J. 2014. “Forward Guidance and Communications in U.S. Monetary Policy.”Speech at the Imperial Business Insights Series, Imperial College, London, U.K. Mishkin, F. S. 2002. “In‡ation Targeting.” In An encyclopedia of macroeconomics. Elgar Original Reference Series, , ed. B. Snowdon and H.R. Vane. Edward Elgar. Nimark, Kristo¤er. 2008. “Dynamic Pricing and Imperfect Common Knowledge.”Journal of Monetary Economics, 55(8): 365–382. Plosser, Charles I. 2012. “Economic Outlook and Monetary Policy.”Policy Speech at the CFA Society of Philadelphia/The Bond Club of Philadelphia, September 25. Primiceri, Giorgio E. 2005. “Time Varying Structural Vector Autoregressions and Monetary Policy.”Review of Economic Studies, 72(3): 821–852. Schorfheide, Frank. 2005. “Learning and Monetary Policy Shifts.” Review of Economic Dynamics, 8(2): 392–419. Sims, Christopher A., and Tao Zha. 2006. “Were There Regime Switches in U.S. Monetary Policy?” American Economic Review, 96(1): 54–81. Smets, Frank, and Rafael Wouters. 2007. “Shocks and Frictions in US Business Cycles: A Bayeasian DSGE Approach.”American Economic Review, 97(3): 586–606. 45 Online Appendix Not For Publication 46 A The Log-Linearized Model Since technology Zt follows a random walk, we normalize all the nonstationary real variable by the level of technology. We then log-linearize the model around the steady-state equilibrium in which the steady-state in‡ation does not have to be zero. Let us denote logdeviations of the detrended variable xt from its own steady-state value x with x bt ln (xt =x). The log-linearized model can be expressed as follows:16 y^t = E (^ yt+1 jFt ) r^t Et (^ t+1 jFt ) ^t zz + 1 g (8) g^t 1^ bt ^t = (9) 1 ^bt = (1 + "= ) d^t = (1 1 +1 2) + 2 h h d^t Et (^ yt+1 jFt ) Et (^ yt+1 jFt ) e^t + y^t + y^t + (10) m m;t 2 " ^t zz +1 ^t R i + 1 Et (^ t+1 jFt ) + 2 Et d^t+1 jFt (11) (12) i ^ t + "Et (^ t+1 jFt ) + 1 Et (^ y^t + z z^t R et+1 jFt ) i h p ( p ^t + p y ^ + z ^ ) + r r;t t t y; t ; t R; t (13) e^t = 1 r^t = R; r^t 1 + 1 g^t = ^t 1 gg + g gt (15) z^t = ^t 1 zz + z zt ; (16) where 1 p t (! 1)(1 ") and 2 (1 !)["( 1 +1)] (14) and ^bt denotes the optimal reset price of …rms. Following Coibion, Gorodnichenko and Wieland (2012), we add an i.i.d. cost-push shock m;t to the Phillips curve. If one abstracts from imperfect information, this model is very similar to the model studied by Coibion and Gorodnichenko (2011) and Coibion, Gorodnichenko and Wieland (2012). Equation (9) suggests that in‡ation, ^ t , is less sensitive to changes in the re-optimizing price, ^bt , as steady-state in‡ation rises. Coibion, Gorodnichenko and Wieland (2012) explain that this e¤ect has to do with the fact that, with positive steady-state in‡ation, …rms that 16 The detailed derivations of these equations are in an appendix, which is available upon request. 47 reset their price have higher prices than others and receive a smaller share of expenditures, thereby reducing the sensitivity of in‡ation to these price changes. Indexation of prices tends to o¤set this e¤ect, with full indexation completely restoring the usual relationship between reset prices and in‡ation. However, equations (10)-(13) suggest that higher trend in‡ation makes …rms more forward-looking in their price-setting decisions by raising the importance of expected future marginal costs and in‡ation and by inducing them to respond to expected future output growth and interest rate. The increased coe¢ cient on expectations of future in‡ation, which re‡ects the expected future depreciation of the reset price and the associated losses, plays a critical role. Coibion, Gorodnichenko and Wieland (2012) explain that in response to an in‡ationary shock, a …rm that can reset its price will expect higher in‡ation today and in the future as other …rms update their prices in response to the shock. Given this expectation, the more forward-looking a …rm is, the greater the optimal reset price must be in anticipation of other …rms raising their prices in the future. Thus, reset prices become more responsive to current shocks with higher . Coibion, Gorodnichenko and Wieland (2012) argue that this e¤ect dominates the reduced sensitivity of in‡ation to the reset price in equation (9). B Solving the Model with No Transparency It is very important to emphasize that the evolution of agents’beliefs about the future conduct of monetary policies plays a critical role in the Markov-switching model with learning. In fact, three policy regimes p t are not a su¢ cient statistic for the dynamics of the endoge- nous variables in the model with learning. Instead, agents expect di¤erent dynamics for the next period’s endogenous variables depending on their beliefs about a return to the active regime. To account for agents learning, we expand the number of regimes and rede…ne them as a combination between the central bank’s behaviors and agents’beliefs. Bianchi and Melosi 48 (2016a) show that the Markov-switching model with learning described previously can be recast in terms of an expanded set of ( + 1) > 3 new regimes, where t t > 0 is de…ned by the following convergence theorem. For any e > 0, there exists an integer prob f p33 t+1 6= 0j t = g < e: Therefore, that for any , agents’ beliefs can be > t such that e¤ectively approximated using the properties of the long-lasting passive regime (Regime 3). These new set of regimes constitute a su¢ cient statistics for the endogenous variables in the model as they capture the evolution of agents’beliefs about observing a switch to the active regime in the next period. The [( p t = 1; t = 0) ; ( p t + 1 regimes are given by 6= 1; t = 1) ; ( p t 6= 1; t = 2) ; :::; ( and the transition matrix Pep is de…ned using equation (5) 2 p11 6 6 6 6 6 6 6 e Pp = 6 6 6 6 6 6 1 6 4 1 1 p12 p22 +p13 p33 p12 +p13 1 p12 p222 +p13 p233 p12 p22 +p13 p33 p22 (p12 =p13 )(p22 =p33 ) (p12 =p13 )(p22 =p33 ) 1 t = )] ; that is, 3 0 ::: 0 0 0 p12 p22 +p13 p33 p12 +p13 ::: 0 0 0 .. . 0 .. . ::: .. . 0 .. . 0 .. . 0 0 0 0 p22 (p12 =p13 )(p22 =p33 ) (p12 =p13 )(p22 =p33 ) 2 0 0 0 0 p22 (p12 =p13 )(p22 =p33 ) (p12 =p13 )(p22 =p33 ) 1 2 2 6= 1; p12 + p13 .. . p22 (p12 =p13 )(p22 =p33 ) (p12 =p13 )(p22 =p33 ) p t +p33 +1 1 +p33 +1 2 +p33 +1 1 +p33 +1 (17) Equation (5) measures the probability that monetary policy remains passive in period t + 1 conditional on having observed Realize that prob f t+1 t+1 6= 0j t t consecutive periods of passive policy at time t. 6= 0 can be true only if either p t+1 = 2 or p t+1 = 3. Hence, the probability, 6= 0g, in the main text, can be obtained by using the law of total probability 49 7 7 7 7 7 7 7 7: 7 7 7 7 7 7 5 as follows: prob f t+1 6= 0j t p t 6= 0g = prob f +prob f = 2j p t t 6= 0g prob = 3j p t+1 p t+1 6= 0g prob t p t = 2j = 3j = 2; p t t = 3; 6= 0 t (18) 6= 0 ; where we have used the fact that p23 = p32 = 0 to simplify the expression. Note that the Markovian property of the process implies that prob p t+1 = 2j p t = 2; t 6= 0 = p22 ; (19) prob p t+1 = 2j p t = 2; t 6= 0 = p33 : (20) (21) p13 p33t ; p12 p22t + p13 p33t (22) The Bayes’theorem allows us to write: prob f p t = 2j t p12 p22t 6 0g = = ; p12 p22t + p13 p33t prob f p t = 3j t 6= 0g = where p12 (p13 ) is the probability that switching to the passive block was originally due to a switch to the short-lasting (long-lasting) passive regime p t =2( p t = 2) and p22t (p33t ) is the probability that the short-lasting (long-lasting) passive regime lasts for t consecutive periods conditional on the original switch to the short-lasting (long-lasting) passive regime. Replacing this results into equation (18) leads to prob f t+1 6= 0j t 6= 0g = p12 p22t +1 + p13 p33t +1 : p12 p22t + p13 p33t Dividing both sides by p13 p33t delivers equation (5) in the main text. 50 (23) C Welfare Function The period welfare function can then be obtained by taking a log-quadratic approximation of the representative household’s utility function around the deterministic steady state: P1 Wi (st (i)) = h h=1 [ 0 + 1 vari (^ yt+h jst (i)) + 2 vari (^ t+h jst (i))] ; (24) where vari ( ) with i 2 fT; N g stands for the stochastic variance associated with agents’ forecasts of in‡ation conditional on transparency (T ) or no transparency (N ) and the output gap at horizon h. The coe¢ cients i, i 2 f0; 1; 2g are functions of the model’s parameters and are de…ned in the online appendix. The subscript i refers to the communication strategy: i = N stands for the case of no transparency, while i = T denotes transparency. Finally, st (i) denotes the policy regime: st (i = N ) 2 f0; 1; :::; g= t and st (i = T ) 2 f0; 1; :::; a + 1g = a 17 t. The coe¢ cients 1 0 1 1 1 2 1 g g 1 1+ 2 = 1 gy 1 2 " 2 = 3 (1 gy ) where 17 1 1 0; and 1 2 1 1+ 1+ Q0y 2 are de…ned as follows: Q0y (1 ln x Q0y + "2 (1 + 2 ) 1+ 2 (1 gy ) 1 ) 1 ln (x )2 1 Q1y " 1 1 + 1+ 1 1+ " 1 " Q0y Q1y 1+ 1 " 1 " Q1y ln x 1) ="]; the steady-state government purchase share gy is set equal to log [(" Recall st (i = T ) = announced. 1, a + 1 denotes the long-lasting passive regime, whose exact realized duration is not 51 ; 0.22; and " x Q0y h 0 1 1 1 0:5 " " 1 " 1 2 " 1 + (" 1) Q1p (1 1 1) (1 (" 1 (1 !)(" 1) 1 0:5 h 1 + 0:5 ) b + Q0p !) Q1p " 1 2 " " 1 2 " # 2 i3 (1 !) 0:5 (1 ")2 1 + 0:5 (1 ")2 1 0 ) M2 + (1 1 0:5 (1 Q0p (1 !)"( 1 +1) 1 1 Q1y i2 ; 1 + 0:5 0 3 +" (1 ") (! 1)(1 ") 1 + 0:5 (1 ") ")2 2; 1 Q1p 2 3; where the cross-sectional price dispersion in the nonstochastic steady state is given by 2 2 (1 !) 2 (1 ) = and the cross-sectional dispersion of output in the nonstochastic steady state is = "2 . The log of the optimal reset price in the nonstochastic steady state is given by 1 1 b = log D (! 1)(1 1 ) 1 and M t bt b log( ) = (" 1)(1 !) 1 (" 1)(1 !) . Transition Matrix under Transparency When the central bank is transparent, the exact duration of every short-lasting deviation from active policy is truthfully announced. In this model the number of announced shortlasting deviations from active policy yet to be carried out a t is a su¢ cient statistic that captures the dynamics of beliefs after an announcement. Since the exact duration of longlasting passive policies is not announced, we also have to keep the long-lasting passive regime as one of the possible regimes. Regimes are ordered from the smallest number of announced deviations (zero, or active policy) to the largest one ( a ). The long-lasting passive regime, whose conditional persistence is p33 , is ordered as the last regime. Notationally, regime a t = a + 1 denotes the long-lasting passive regime. Hence, we rede…ne the set of policy regimes in terms of this variable with the following mapping to the parameter values of the 52 policy rule: a t r( where a t a = j) ; ( a t = j) ; y( a t 2 A r; 6 = j) = 4 P P r ; A P y ; ; A y 3 , if j = 0 , if 1 a j 7 5 +1 (25) is a large number at which we truncate the rede…ned set of regimes. The regimes 2 f0; 1; :::; a 0 [e p10 e20 e30 e1A is a 1 A; p A; p A ] , where p a a + 1g are governed by the ( ( a + 2) ( a eA = + 2) transition matrix P + 2) vector whose j-th element is p11 if j = 1; p12 pj22 2 p21 if + 1 (the probability that the realized short-lasting passive policy will last exactly 2 j j 1 consecutive periods conditional on being in the active regime); and p13 if j = a + 2. This vector pe1A captures the probability of remaining in the active regime; switching to a short-lasting passive regime of duration 1 up to regime a; and switching to the long-lasting passive a all conditional on being currently in the active regime. The pe2A is de…ned as I a ; 0 a 2 , where I a a is a a identity matrix and 0 ( a 2 a + 1) matrix is a 2 null matrix. This submatrix captures the transition while the announced deviation from active policy is carried out. pe3A is de…ned as a 1 j = 1; zero if 2 j a ( + 1; and p33 if j = a + 2) vector whose j-th element is (1 a eA captures + 2. The last row of the matrix P p33 ) if the probability of staying in the long-lasting passive regime or switching to the active regime, conditional on being currently in the long-lasting passive regime. To ensure that the …rst row Pa 1 1 sums up to one, we set pe1A ( a ) = 1 p11 eA (j) p13 , which, e¤ectively, becomes the j=1 p probability for the central bank to announce a deviation longer than a to be large enough so that pe1A ( a ) a periods. We choose 0 and the approximation error becomes negligible.18 Let us make a simple example to illustrate how to construct the transition matrix gov- erning the evolution of the policy regimes in the case of transparency. To serve the purpose of this simple example, let us truncate the maximum number of announced deviations at a = 3 periods. We need to construct a total of 18 a + 2 = 5 regimes. The …rst regime is Since p22 < 1, it can be easily shown that the larger the truncation error. 53 a, the lower the approximation A r; active P r ; P ; P y A ; A y and all of the other regimes (from the second to the …fth) are passive . The 5 eA can be constructed as follows: 5 transition matrix P 2 6 6 6 6 6 A e =6 P 6 6 6 6 4 p11 1 p12 p21 p12 p22 p21 p12 p222 p21 p13 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 p33 3 7 7 0 7 7 7 0 7 7: 7 7 0 7 5 p33 Let assume that at time t the central bank announces it will implement a three-period passive policy (i.e., a t = 3). The system will move to the fourth regime in period t + 1, to third regime in period t + 2, to the second regime in period t + 3, and then back to the active regime (i.e., the …rst regime) in period t + 4. Similarly to the case of no transparency, we have recast the MS-DSGE model under transparency as a Markov-switching rational expectations model with perfect information, in which the short-lasting passive regime is rede…ned in terms of the number of announced deviations from the active regimes yet to be carried out, a t. This rede…ned set of regimes belongs to the agents’information set Ft in the case of transparency. This result allows us to solve the model under transparency by applying any of the methods developed to solve Markov-switching rational expectations models of perfect information. E Transformation of Regimes under Transparency In Figure 4 we express welfare under transparency in terms of number of observed deviations from the active regime. This corresponds to the de…nition of policy regime under no transparency. This is done in order to facilitate the analysis of how the welfare gains from transparency varies with passive policies of duration : First of all, the probability that periods of deviation from the active regime are due 54 to the implementation of a short-lasting passive policy (primitive Regime 2) is de…ned as follows: P( ) = p12 pi22 1 : i 1 p12 p22 + p13 pi33 1 Furthermore, we compute the probability that i consecutive periods of passive policy has been announced conditional on having observed i period of short-lasting passive policy as follows: i 1 p12 p22 p21 for any 1 p12 pj22 1 p21 j= (i) = P i a+ + a : Note that the numerator captures the probability that a deviation of duration i is realized and, hence, announced (recall all announcements are truthful). The denominator is the probability of (announcing) a short-lasting passive policy lasting the truncation periods or longer (up to ). The welfare associated with a policy that has been deviating for 1 consecutive periods under transparency is given by19 f WT ( ) = P ( ) a X (j + ) WT ( a = j) + [1 p ( )] WT ( a = a + 1) : (26) j=0 Note the di¤erence from WT ( a ) in equation (6), which is the welfare function de…ned in terms of policy regimes for the case of transparency (i.e., a the number of announced deviations yet to be carried out). WT ( a ) is the welfare under transparency associated with a announcing a periods of passive policy. f WT ( ; ) is the welfare under transparency associated with having observed consecutive periods of passive policy. We can show that this recasting of policy regimes leads to a negligible approximation error as X =0 19 The Regime a pN ( ) f WT ( ) a +1 X pT ( a ) WT ( a ) : a =0 + 1 denotes the long-lasting passive regime ( 55 p = 3). The welfare gains from transparency can be alternatively computed as follows X f We h pN ( ) =0 f WT ( ) i WN ( ) : This formula can be used to compute the welfare gains from transparency that are identical to those obtained by using the formula (7) in the main text (up to some very tiny computational h i f error). Figure 4 plots the conditional welfare gains from transparency WT ( ) WN ( ) for 1 . We analyze the welfare gains from transparency under limited information (i.e., the welfare in the perfect information case) by the central bank in Section 7.1. To do so, we …rst compute the long-run welfare under perfect information as follows f WP ( ) = ( ) WP ( p = 2) + [1 ( )] WP ( p = 3) ; for 1 where Wp denotes the welfare under perfect information and of policy regimes. The weight p12 p22 1 ; p12 p22 1 + p13 p33 1 which captures the probability that the observed stems from a short-lasting passive policy ( p t = 1 or 2 f1; 2; 3g, the primitive set is de…ned as follows: ( )= the active regime, p t ; p consecutive deviations from active policy = 2). For = 0 (i.e., conditional on being in = 0), the welfare f WP (0) = WP p =1 . This computation gives almost identical welfare gains from transparency to the alternative computation on the right-hand-side of the following expression: X =0 pN ( ) f WP ( ) X p t 2f1;2;3g 56 p ( pt ) WP ( pt ) : Four-Quarter Passive Policy Eight-Quarter Passive Policy -108 Welfare Welfare -108 -109 -110 -111 -109 -110 -111 -112 1 2 3 4 2 Time Twelve-Quarter Passive Policy 4 6 8 Time Forty-Quarters Passive Policy -108 Welfare Welfare -110 -110 -112 No transparancy Transparency -114 2 4 6 -115 -120 8 10 12 Time 10 20 30 Time Figure 8: Evolution of welfare Wi (st (i)) de…ned in equation (6) as a passive policy of duration 4 (upperleft graph), 8 (upper right graph), 12 (lower left graph), and 40 quarters (lower right graph) is implemented under no transparency (i = N ), the blue dashed line, and under transparency (i = T ), the red solid line. Parameter values are set at the posterior mode. F Welfare Dynamics: a Numerical Example For the sake of illustrating the dynamics of welfare, let us consider passive policies of duration 4, 8, 12, and 40 quarters.20 Figure 8 shows the dynamics of welfare Wi (st (i)), de…ned in equation (6), over time as these policies are implemented under the two communication schemes: no transparency i = N and transparency i = T . Welfare under transparency (red solid line) is always higher than welfare under no transparency (blue dashed line) at every time during the implementation of passive policies of four-, eight-, and twelve-quarter duration. Nonetheless, welfare under transparency is lower than welfare under no transparency at the very early stage of a 40-quarter-long passive policy. Larger gains from transparency, measured by the vertical distance between the two lines, are reaped at the end of this pro20 This is a numerical example and is made for the sake of illustrating the evolution of welfare. We pick fairly prolonged deviations from the active regime so as to make these dynamics more visible in the graphs. Such long-lasting passive policies have low probability of occuring based on our estimates. 57 Price Indexation =0.0 Price Indexation =0.5 6.55 29.6 6.5 Welfare gains from transparency Welfare gains from transparency 29.4 29.2 29 28.8 6.45 6.4 6.35 28.6 6.3 28.4 28.2 6.25 0 10 20 30 0 Observed Periods of Passive Policy 10 20 30 Observed Periods of Passive Policy Figure 9: The graphs report the dynamics of the welfare gains from transparency as a function of the observed periods of passive policy ( t ) under no price indexation (left graph) and under partial price indexation, ! = 0:5, (right graph). The other parameter values are set at the posterior mode. longed passive policy. As discussed earlier, when the announcement is made, agents become suddenly more pessimistic and, hence, being transparent may lower welfare compared to not being transparent at the beginning of the policy. However, transparency lowers pessimism as the passive policy is implemented because agents expect fewer and fewer periods of passive policy ahead. Therefore, welfare generally increases as the passive policy is implemented. In contrast welfare is downward sloping under no transparency because the central bank does not communicate the duration of passive policies; agents’pessimism gradually grows, progressively lowering welfare. G Lower Price Indexation Figure 9 shows the welfare gains from transparency conditional on observing a given number of periods of (short-lasting) passive policy. A quick comparison of these plots with Figure 58 4 shows that welfare gains from transparency are higher when price indexation is lower. Interestingly, the pattern of these gains with respect to the observed duration of passive policy is qualitatively very similar to that in the estimated model (Figure 4). H Imperfectly Credible Announcements To solve the model in which the central bank’s announcements are only partially credible, we rede…ne the structure of the three regimes (i.e., active, short-lasting passive, and longlasting passive) into a new set of 2regimes A ; A y 6 follows: ( t = i) ; y ( t = i) = 4 P ; Py set of regimes t t determining 3 the Taylor rule parameters as , if i = 1 7 5 . The evolution of the rede…ned , if i > 1 eA . To simplify the description of this is governed by the transition matrix P matrix, let us consider the case in which = 4 and 59 = 7 (i.e., the truncation for the model eA reads: with no transparency). The transition matrix P 2 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4 0 pa2 0 0 pa3 0 0 0 pa4 0 0 0 p11 0 0 0 pa1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 (1) 0 0 (1) 0 0 0 0 0 0 0 0 0 0 0 0 0 1 (2) 0 (2) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 (3) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 (4) 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 pe56 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 pe67 0 0 0 0 0 0 0 0 0 0 0 0 0 0 pe56 pe78 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 pe88 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 (3) 1 1 1 1 1 (4) 0 pe67 0 0 3 p13 7 7 0 7 7 7 7 0 7 7 7 0 7 7 7 0 7 7 7 7 0 7 7 7 0 7 7 7 0 7 7 7 7 0 7 7 7 0 7 7 7 0 7 7 7 7 0 7 7 7 0 7 7 7 0 7 7 7 7 0 7 7 7 0 7 7 7 pe78 7 7 5 pe88 where we denote the probability (conditional on being in the active regime) of announcing = i consecutive periods of passive monetary policy with pai . Note that this probability a is de…ned as paa = a + f ( a ), where i denotes the probability that the duration of the passive policy (conditional on being in the active regime) is shorter than i periods a p12 p22a 1 p21 , and that is, denotes the probability (conditional on being in the active regime) that the central bank lies when making an announcement. This probability is de…ned as 60 follows: prob ( > ) = p12 X i= i a where p12 is the probability of switching to the short-lasting passive regime conditional on being in the active regime. f ( a ) is a monotonically increasing (deterministic) function that determines the probability that the central bank announces a consecutive periods of passive policy conditional on having lied ( > ). A monotonically increasing function f captures the property that when the central bank lies, it is more likely that a a relatively longer deviation from active policy is announced. Furthermore, the probability that the announcement made turns out to be untrue after having observed the announced number of deviations a is denoted by ( a ) = = (1 p12 ) f ( a ). Recall that probabilities p~ij are the probabilities in the transition matrix in the case of no transparency. Note that after a + 1 periods the central bank’s lie is discovered and agents know that the policy will stay passive until passive policy after a. Moreover, for any period of the short-lasting agents have to learn the persistence of the regime in place as they do eA is a in the no-transparency world. This is why the lower-right submatrix of the matrix P the submatrix of the transition matrix P; which is the matrix that capture the evolution of policy regimes under no transparency. Note that in the case of an untruthful announcement agents start learning after having already observed + 1 periods of passive monetary policy. So the learning based on counting the number of consecutive periods of passive policy starts from I +1 that is, 5 periods in this example. Limited Information The upper graph of Figure 10 shows the welfare gains from transparency associated with observing di¤erent durations of passive policies. The lower graph reports the ergodic probability of observing passive policies of di¤erent durations where zero duration means active 61 Welfare gains from transparency 0.55 0.545 0.54 0 5 10 15 20 25 30 35 Ergodic Prob. Observed Periods of Passive Policy 0.6 0.4 0.2 0 0 5 10 15 20 25 30 35 Observed Periods of Passive Policy Figure 10: The upper graphs report the dynamics of the welfare gains from transparency as a function of the observed periods of passive policy ( t ) when the central bank is assumed to have limited information. The lower graph reports the ergodic probability of observing the periods of passive policy on the x-axis. Parameter values are set at their posterior mode. policy.21 The important result that emerges from this graph is that welfare gains from transparency are always positive for policies of any plausible duration. Computing the upper graph requires to transform the primitive regimes, pt 2 f1; 2; 3g ; into the set of regimes used for the case of no transparency, which are de…ned in terms of the observed durations of passive policies t . The details of this transformation are provided in Appendix E. 21 62 Working Paper Series A series of research studies on regional economic issues relating to the Seventh Federal Reserve District, and on financial and economic topics. The Urban Density Premium across Establishments R. Jason Faberman and Matthew Freedman WP-13-01 Why Do Borrowers Make Mortgage Refinancing Mistakes? Sumit Agarwal, Richard J. Rosen, and Vincent Yao WP-13-02 Bank Panics, Government Guarantees, and the Long-Run Size of the Financial Sector: Evidence from Free-Banking America Benjamin Chabot and Charles C. Moul WP-13-03 Fiscal Consequences of Paying Interest on Reserves Marco Bassetto and Todd Messer WP-13-04 Properties of the Vacancy Statistic in the Discrete Circle Covering Problem Gadi Barlevy and H. N. Nagaraja WP-13-05 Credit Crunches and Credit Allocation in a Model of Entrepreneurship Marco Bassetto, Marco Cagetti, and Mariacristina De Nardi WP-13-06 Financial Incentives and Educational Investment: The Impact of Performance-Based Scholarships on Student Time Use Lisa Barrow and Cecilia Elena Rouse WP-13-07 The Global Welfare Impact of China: Trade Integration and Technological Change Julian di Giovanni, Andrei A. Levchenko, and Jing Zhang WP-13-08 Structural Change in an Open Economy Timothy Uy, Kei-Mu Yi, and Jing Zhang WP-13-09 The Global Labor Market Impact of Emerging Giants: a Quantitative Assessment Andrei A. Levchenko and Jing Zhang WP-13-10 Size-Dependent Regulations, Firm Size Distribution, and Reallocation François Gourio and Nicolas Roys WP-13-11 Modeling the Evolution of Expectations and Uncertainty in General Equilibrium Francesco Bianchi and Leonardo Melosi WP-13-12 Rushing into the American Dream? House Prices, the Timing of Homeownership, and the Adjustment of Consumer Credit Sumit Agarwal, Luojia Hu, and Xing Huang WP-13-13 1 Working Paper Series (continued) The Earned Income Tax Credit and Food Consumption Patterns Leslie McGranahan and Diane W. Schanzenbach WP-13-14 Agglomeration in the European automobile supplier industry Thomas Klier and Dan McMillen WP-13-15 Human Capital and Long-Run Labor Income Risk Luca Benzoni and Olena Chyruk WP-13-16 The Effects of the Saving and Banking Glut on the U.S. Economy Alejandro Justiniano, Giorgio E. Primiceri, and Andrea Tambalotti WP-13-17 A Portfolio-Balance Approach to the Nominal Term Structure Thomas B. King WP-13-18 Gross Migration, Housing and Urban Population Dynamics Morris A. Davis, Jonas D.M. Fisher, and Marcelo Veracierto WP-13-19 Very Simple Markov-Perfect Industry Dynamics Jaap H. Abbring, Jeffrey R. Campbell, Jan Tilly, and Nan Yang WP-13-20 Bubbles and Leverage: A Simple and Unified Approach Robert Barsky and Theodore Bogusz WP-13-21 The scarcity value of Treasury collateral: Repo market effects of security-specific supply and demand factors Stefania D'Amico, Roger Fan, and Yuriy Kitsul Gambling for Dollars: Strategic Hedge Fund Manager Investment Dan Bernhardt and Ed Nosal Cash-in-the-Market Pricing in a Model with Money and Over-the-Counter Financial Markets Fabrizio Mattesini and Ed Nosal WP-13-22 WP-13-23 WP-13-24 An Interview with Neil Wallace David Altig and Ed Nosal WP-13-25 Firm Dynamics and the Minimum Wage: A Putty-Clay Approach Daniel Aaronson, Eric French, and Isaac Sorkin WP-13-26 Policy Intervention in Debt Renegotiation: Evidence from the Home Affordable Modification Program Sumit Agarwal, Gene Amromin, Itzhak Ben-David, Souphala Chomsisengphet, Tomasz Piskorski, and Amit Seru WP-13-27 2 Working Paper Series (continued) The Effects of the Massachusetts Health Reform on Financial Distress Bhashkar Mazumder and Sarah Miller WP-14-01 Can Intangible Capital Explain Cyclical Movements in the Labor Wedge? François Gourio and Leena Rudanko WP-14-02 Early Public Banks William Roberds and François R. Velde WP-14-03 Mandatory Disclosure and Financial Contagion Fernando Alvarez and Gadi Barlevy WP-14-04 The Stock of External Sovereign Debt: Can We Take the Data at ‘Face Value’? Daniel A. Dias, Christine Richmond, and Mark L. J. Wright WP-14-05 Interpreting the Pari Passu Clause in Sovereign Bond Contracts: It’s All Hebrew (and Aramaic) to Me Mark L. J. Wright WP-14-06 AIG in Hindsight Robert McDonald and Anna Paulson WP-14-07 On the Structural Interpretation of the Smets-Wouters “Risk Premium” Shock Jonas D.M. Fisher WP-14-08 Human Capital Risk, Contract Enforcement, and the Macroeconomy Tom Krebs, Moritz Kuhn, and Mark L. J. Wright WP-14-09 Adverse Selection, Risk Sharing and Business Cycles Marcelo Veracierto WP-14-10 Core and ‘Crust’: Consumer Prices and the Term Structure of Interest Rates Andrea Ajello, Luca Benzoni, and Olena Chyruk WP-14-11 The Evolution of Comparative Advantage: Measurement and Implications Andrei A. Levchenko and Jing Zhang WP-14-12 Saving Europe?: The Unpleasant Arithmetic of Fiscal Austerity in Integrated Economies Enrique G. Mendoza, Linda L. Tesar, and Jing Zhang WP-14-13 Liquidity Traps and Monetary Policy: Managing a Credit Crunch Francisco Buera and Juan Pablo Nicolini WP-14-14 Quantitative Easing in Joseph’s Egypt with Keynesian Producers Jeffrey R. Campbell WP-14-15 3 Working Paper Series (continued) Constrained Discretion and Central Bank Transparency Francesco Bianchi and Leonardo Melosi WP-14-16 Escaping the Great Recession Francesco Bianchi and Leonardo Melosi WP-14-17 More on Middlemen: Equilibrium Entry and Efficiency in Intermediated Markets Ed Nosal, Yuet-Yee Wong, and Randall Wright WP-14-18 Preventing Bank Runs David Andolfatto, Ed Nosal, and Bruno Sultanum WP-14-19 The Impact of Chicago’s Small High School Initiative Lisa Barrow, Diane Whitmore Schanzenbach, and Amy Claessens WP-14-20 Credit Supply and the Housing Boom Alejandro Justiniano, Giorgio E. Primiceri, and Andrea Tambalotti WP-14-21 The Effect of Vehicle Fuel Economy Standards on Technology Adoption Thomas Klier and Joshua Linn WP-14-22 What Drives Bank Funding Spreads? Thomas B. King and Kurt F. Lewis WP-14-23 Inflation Uncertainty and Disagreement in Bond Risk Premia Stefania D’Amico and Athanasios Orphanides WP-14-24 Access to Refinancing and Mortgage Interest Rates: HARPing on the Importance of Competition Gene Amromin and Caitlin Kearns WP-14-25 Private Takings Alessandro Marchesiani and Ed Nosal WP-14-26 Momentum Trading, Return Chasing, and Predictable Crashes Benjamin Chabot, Eric Ghysels, and Ravi Jagannathan WP-14-27 Early Life Environment and Racial Inequality in Education and Earnings in the United States Kenneth Y. Chay, Jonathan Guryan, and Bhashkar Mazumder WP-14-28 Poor (Wo)man’s Bootstrap Bo E. Honoré and Luojia Hu WP-15-01 Revisiting the Role of Home Production in Life-Cycle Labor Supply R. Jason Faberman WP-15-02 4 Working Paper Series (continued) Risk Management for Monetary Policy Near the Zero Lower Bound Charles Evans, Jonas Fisher, François Gourio, and Spencer Krane Estimating the Intergenerational Elasticity and Rank Association in the US: Overcoming the Current Limitations of Tax Data Bhashkar Mazumder WP-15-03 WP-15-04 External and Public Debt Crises Cristina Arellano, Andrew Atkeson, and Mark Wright WP-15-05 The Value and Risk of Human Capital Luca Benzoni and Olena Chyruk WP-15-06 Simpler Bootstrap Estimation of the Asymptotic Variance of U-statistic Based Estimators Bo E. Honoré and Luojia Hu WP-15-07 Bad Investments and Missed Opportunities? Postwar Capital Flows to Asia and Latin America Lee E. Ohanian, Paulina Restrepo-Echavarria, and Mark L. J. Wright WP-15-08 Backtesting Systemic Risk Measures During Historical Bank Runs Christian Brownlees, Ben Chabot, Eric Ghysels, and Christopher Kurz WP-15-09 What Does Anticipated Monetary Policy Do? Stefania D’Amico and Thomas B. King WP-15-10 Firm Entry and Macroeconomic Dynamics: A State-level Analysis François Gourio, Todd Messer, and Michael Siemer WP-16-01 Measuring Interest Rate Risk in the Life Insurance Sector: the U.S. and the U.K. Daniel Hartley, Anna Paulson, and Richard J. Rosen WP-16-02 Allocating Effort and Talent in Professional Labor Markets Gadi Barlevy and Derek Neal WP-16-03 The Life Insurance Industry and Systemic Risk: A Bond Market Perspective Anna Paulson and Richard Rosen WP-16-04 Forecasting Economic Activity with Mixed Frequency Bayesian VARs Scott A. Brave, R. Andrew Butters, and Alejandro Justiniano WP-16-05 Optimal Monetary Policy in an Open Emerging Market Economy Tara Iyer WP-16-06 Forward Guidance and Macroeconomic Outcomes Since the Financial Crisis Jeffrey R. Campbell, Jonas D. M. Fisher, Alejandro Justiniano, and Leonardo Melosi WP-16-07 5 Working Paper Series (continued) Insurance in Human Capital Models with Limited Enforcement Tom Krebs, Moritz Kuhn, and Mark Wright WP-16-08 Accounting for Central Neighborhood Change, 1980-2010 Nathaniel Baum-Snow and Daniel Hartley WP-16-09 The Effect of the Patient Protection and Affordable Care Act Medicaid Expansions on Financial Wellbeing Luojia Hu, Robert Kaestner, Bhashkar Mazumder, Sarah Miller, and Ashley Wong WP-16-10 The Interplay Between Financial Conditions and Monetary Policy Shock Marco Bassetto, Luca Benzoni, and Trevor Serrao WP-16-11 Tax Credits and the Debt Position of US Households Leslie McGranahan WP-16-12 The Global Diffusion of Ideas Francisco J. Buera and Ezra Oberfield WP-16-13 Signaling Effects of Monetary Policy Leonardo Melosi WP-16-14 Constrained Discretion and Central Bank Transparency Francesco Bianchi and Leonardo Melosi WP-16-15 6