Regional President Speech

Disclaimer The views expressed in this talk are my own. • They may not be shared by others in the Federal Reserve System ... • Especially my colleagues on the Federal Open Market Committee (FOMC). • Acknowledgements I thank Manuel Amador, Cristina Arellano, Kei-Mu Yi, and Samuel Schulhofer- Wohl for helpful comments. Classic Question Should the central bank (CB) be required to follow a pre-specified rule Classic Question Should the central bank (CB) be required to follow a pre-specified rule when setting the level of monetary accommodation? Classic Question, cont’d Or should the central bank (CB) have discretion Classic Question, cont’d Or should the central bank (CB) have discretion to choose accommodation as it deems necessary to achieve its goals? Motivation for the Question Many observers speak highly of interest rate targeting rules as a constraint • on CB choices. In fact: Congress is considering legislation that would enshrine the Taylor • Rule as a reference interest rate rule for the FOMC. The Argument in This Talk Key trade-off: • — a rule is good if CB has inflationary bias (from time consistency or other factors). — discretion is good if CB’s information and analysis are hard to quantify. In the US: little evidence of an inflationary bias by the CB. • Also: FOMC relies, in a complex way, on many indicators of inflationary • pressures. Conclusion: In the US, discretion is better than any rule. Formal Basis for My Analysis: Three Assumptions ASSUMPTION 1: CBs may or may not have different objective from public. ASSUMPTION 2: CBs may have relevant non-ruleable information . ASSUMPTION 3: CB compensation does not vary with outcomes. Assumption 1: Possibility of Different Objectives My analysis allows for the possibility that the CB has an inflationary bias. • I allow for such bias for two reasons: • — Oft-heard concerns that FOMC has a pro-inflation bias. — Relatedly: inflationary bias is the focus of academic literature on time consistency. BUT: I will argue that there is little evidence to support the existence of • such bias. Assumption 2: Non-Ruleable Information CBs use lots of information to forecast inflation and output. • As well: CBs analyze that information in many ways. • Both the information and its analysis are hard to quantify. • Example: CB models of inflationary pressures include latent variables like: • — potential output — natural real rate of interest Assumption 3: Rigid Compensation (In US), CB real compensation is largely independent of outcomes. • Implication: can’t use incentive compensation as a tool for CBs. • — rules out elegant approach of Walsh (1993; AER). Delegation Framework Given the three assumptions: There is a tension between: • — CB’s possible bias — CB’s non-ruleable information And compensation can’t be used to resolve this tension. • This creates a delegation problem for society with CB. • — See Holmstrom (1984; volume). Rules versus discretion trade-off becomes: • How much flexibility should society (delegator) give to CB (delegee)? This Talk Is Not About Simple vs. Complex Rules Some observers argue that simple reaction functions will give rise to better • outcomes. — Simple reaction functions may be near-optimal in many models. — Complicated reaction functions may be highly suboptimal for some model. Conclusion: simple rules are good because they protect against very bad • outcomes. But I’m not talking about that. • Some Other References Amador and Bagwell (2011, Econometrica). • Athey, Atkeson, and Kehoe (2005, Econometrica). • Svensson (2003, JEL). • Canzoneri (1985, AER) - especially section III. • BASIC FRAMEWORK Objectives Society has objective: • 2 (  )  − − CB has objective: • 2 (  )  − − Possible CB bias:   .   • ≥ Comment on Objectives Formally, I only model choice at a single date. • — Political economy factors could generate possible bias. But this static choice can be embedded in a dynamic model. • — Time inconsistency could generate possible bias. Timing Stage 1: inflationary pressures  are publicly observed. • Stage 2: central bank (CB) observes inflationary pressures  • —  may or may not be public. — Key:  is not ruleable. Stage 3: CB chooses accommodation  from set Φ() • — Φ() is endogenous - described later. Comment on  It may seem to be easy to encode (scalar)  into a rule. • But I think about  as being the dot product of  : •  =   0 where  and  are both random vectors. May be hard to describe . • As well,  may be the product of complicated analysis. • Inflation Inflation is determined by: •  =  +  +  +  (  ) are all mean zero and mutually independent. •  has unbounded support. • Institutional Design Before stage 1, society chooses a correspondence Φ. • The correspondence maps to an action set Φ() for the CB. • Society chooses Φ so as to maximize expectation of its objective. • Some Terminology CB has a rule if Φ() is a singleton for all  • CB has discretion if Φ() is the entire real line for all  • CB has constrained discretion if Φ has any other form. • FIVE CASES Five Cases 1. Unbiased CB and no non-ruleable information. 2. Biased CB and no non-ruleable information. 3. Unbiased CB with non-ruleable information. 4. Slightly biased CB with non-ruleable information. 5. Highly biased CB with non-ruleable information. Case 1: Unbiased CB and No Non-Ruleable Information Case 1:  () = (  ) = 0   • − Both (good!) rules and discretion are optimal. • If Φ() =   for all  then  =  +    • { − } If Φ() = (  ) for all  then  =  +   • −∞ ∞ — CB offsets inflationary pressures  ( =  )  − Case 2: Biased CB with No Non-Ruleable Information Case 2:  =  and  () = 0   • 6 In this case: a good rule dominates discretion. • If Φ() =   for all  then  =  +    • { − } — this rule forces CB to choose socially optimal accommodation. If, for some  ( ) Φ() then  =  +    • − ∈ — discretion allows CB to choose suboptimal accommodation. Case 3: Unbiased CB Has Non-Ruleable Information Case 3:  =  and  ()  0   • In this case: discretion dominates any rule. • The best rule isΦ () =   for all .  • { − } Under this rule,  =  +  +   • If Φ() = (  ) for all , then  =  +   • −∞ ∞ Discretion allows CB to offset  shocks. • Discretion therefore reduces variance of  • Case 4: Slightly Biased CB with Non-Ruleable Information. 2 Case 4:  ()  (  )   • − Again, discretion dominates any rule. • The best rule isΦ () =   for all   • { − } Under this rule,  =  +  +   • If Φ() = (  ) for all  then  =  +   • −∞ ∞ Discretion reduces variance and increases bias. • 2 Overall, discretion reduces (  ) by:  • − 2  () (  )   − − Intuition for Case 4 A rule is good because it eliminates any bias in CB’s choices. • Discretion is good because it allows CB to offset non-ruleable inflationary • pressures. 2 If  ()  (  ) , then the second force dominates the first.   • − Discretion wins. Case 5: Highly Biased CB 2 Case 5:  ()  (  )   • − Now, a (good) rule dominates discretion. • The best rule isΦ () =   for all   • { − } Under this rule,  =  +  +   • If Φ() = (  ) for all  then  =  +   • −∞ ∞ 2 Compared to the best rule, discretion increases (  ) by:  • − 2 (  )  ()   − − But rules are still not best ... • Case 5, Continued: Optimality of Constrained Discretion Suppose    (biased central bank).   • Then: incentives are aligned if  is sufficiently positive. • So: let Φ() =  + (   ].  • − −∞ This constrained discretion is better than the best rule. • It allows CB to offset sufficiently inflationary shocks. • Summary of Results 1. Unbiased CB and no non-ruleable information: Rule OR Discretion. 2. Biased CB and no non-ruleable information: Rule. 3. Unbiased CB with non-ruleable information: Discretion. 4. Slightly biased CB with non-ruleable information: Discretion. 5. Highly biased CB with non-ruleable information: Constrained Discretion. COMMENTS Comment 1: Optimality of Discretion Most macroeconomic models assume no non-ruleable information. • CB typically has inflationary bias (for time consistency reasons). • As in cases 1 and 2: rules are at least weakly optimal, if not strictly so. • Empirically: little recent evidence to suggest an inflationary bias. • Optimal inflation ( ) is 2% per year.  • — It is the long-run target in the reference policy rule before Congress. — FOMC has adopted 2% as its long-run target. PCE inflation has averaged: • — 1.9% over past 20 years. — 1.9% over past 10 years. — 1.6% over past five years. At the same time, FOMC relies on many indicators of inflationary pressures. • Hence, in the US: • — Little, if any, CB inflationary bias. — CB has a lot of non-ruleable information. The relevant cases are case 3 (maybe 4), not cases 1, 2, or 5. • (Unconstrained) discretion dominates rules. • Comment 2: Discretion Always Beats Many Rules In cases 1 and 2, a given rule dominates discretion. • But discretion still dominates many rules (all but one in case 1). • Key: rule has to exactly offset undue and public inflationary pressures () • Comment 3: Communication Suppose that the CB has discretion. • The public observes  and  • In the model, the public can then infer  via: •  =     − − In reality: public is typically uncertain about    • Then, the CB’s choice of  does not imply  • CB should reveal its information  • Equivalently: it should explain the choice of given  and   • CONCLUSION Classic Argument Central bank is tempted to over-inflate relative to long-run societal goal. • Best to eliminate central banks and replace them with computers. • In This Talk ... I argue that: • — the inflationary bias of the FOMC is negligible. — the FOMC has many non-quantifiable sources of information. — FOMC compensation is ineffective at providing incentives. Conclusion: discretion is better than any rule. Preferences Are Key (Rogoff Redux) Appointed central bankers must have little inflationary bias. • It takes the right organizational culture and appointment procedures to • deliver this outcome. In my view, we have achieved this desirable goal in the US. •

Narayana Kocherlakota (2015, January 3). Regional President Speech. Speeches, Federal Reserve. https://whenthefedspeaks.com/doc/regional_speeche_20150104_narayana_kocherlakota

@misc{wtfs_regional_speeche_20150104_narayana_kocherlakota,
  author = {Narayana Kocherlakota},
  title = {Regional President Speech},
  year = {2015},
  month = {Jan},
  howpublished = {Speeches, Federal Reserve},
  url = {https://whenthefedspeaks.com/doc/regional_speeche_20150104_narayana_kocherlakota},
  note = {Retrieved via When the Fed Speaks corpus}
}