Inflation Targeting and Optimal Monetary Policy Michael Woodford Princeton University
Intro Inflation targeting an increasingly popular approach to conduct of monetary policy worldwide associated with reductions in both average level and volatility of inflation Largely developed by central banks with little guidance from academic literature on monetary policy rules How does it compare with the ideal monetary policy that would be recommended on the basis of current theory?
Overview of conclusions: IT incorporates a number of important features of optimal policy: an explicit target for policy, announced in advance publication of projections on which policy decisions are based greater (though not exclusive) emphasis on price stability Yet current practices also fall short of the theoretical ideal in several respects
Features of IT In its most developed form: inflation-forecast targeting A publicly stated target for some measure of inflation Commitment to a decision procedure under which CB will adjust its instrument as necessary in order to ensure that economy s projected evolution satisfies a target criterion example of what Svensson calls a targeting rule in practice: inflation to equal the target value at a certain horizon Regular publication of Inflation Reports that justify recent policy decisions in the light of these projections
Advantages of an Explicit Target The most important achievement of IFT: commitment to an explicit target as opposed to the precise nature of the target A traditional prejudice of central bankers: unwise to commit, in order to preserve flexibility to respond to unanticipated developments Isn t this obviously prudent, if one can appoint someone competent to the job? No, because forward-looking private-sector behavior makes it important for central bank to affect expectations about future policy
Effective central banking depends on management of expectations Current level of funds rate of little importance: policy effective by changing expected future path of overnight rates When Fed affects expectations in the way desired, bond market does its work for it Efforts to signal future policy especially important when zero lower bound is a relevant constraint (Japan, US now?) Eggertsson and Woodford (2003): advantages of a price-level target as a way of creating the expectations that ward off a deflationary contraction
Explicit targets useful approach to communication about future policy under normal circumstances as well Inflation Reports an especially effective means of clarification of policy commitments can clarify nature of state-contingent policy by showing how decisions have been made in particular cases that have arisen allows verification of CB s commitment to its target projected future path of economy is what private sector most needs to know
Not enough simply to follow a rule, without having also to explain it? If people have RE, explanation does no harm; so a more robust approach A better outcome if people must learn: analysis of Orphanides and Williams (2003) see better output/inflation variability tradeoff available if private sector knows the long-run average inflation rate, and only must estimate dynamics of transitory fluctuations around it
Advance commitment also necessary in order to avoid the suboptimality of sequential (discretionary) optimization Policy tradeoffs affected by expectations; but ex post, no continuing motive to create the outcome that it was earlier desirable for people to expect Inflationary bias of discretionary policy: Kydland- Prescott, Barro-Gordon But also results in sub-optimal dynamic responses to shocks, even if average inflation is optimal optimal policy generally history-dependent, whereas discretionary policy purely forwardlooking
Example: optimal policy when the natural rate of interest is temporarily negative Discretionary optimization: return to pursuit of the optimal (zero) rate of inflation as soon as consistent with zero bound on interest rates But this implies that any price level decline while zero bound prevents achievement of zero inflation will be subsequently locked in Optimal policy (E-W) instead involves a commitment to subsequently undo any unwanted deflation: if expected, less deflation occurs Simply assigning a different inflation target, different output-gap target, or different relative weights on the objectives will not solve this problem, as discretionary policy will still be purely forwardlooking why commitment to a target criterion is superior to mere commitment to a loss function
6 (a) interest rate 4 2 0 5 0 5 10 15 20 25 (b) inflation 0 5 10 5 0 5 10 15 20 25 (c) output gap 0 5 10 optimal π*=0 15 5 0 5 10 15 20 25 Figure 5: Comparison of the state-contingent paths under optimal policy [solid line] and under the zero inflation target [dashed line], in the case that the natural rate of interest is negative for 15 quarters.
Advantages of a targeting rule as way of specifying a policy commitment can improve upon simple rules by allowing optimal responses to shocks, in addition to anchoring long-run inflation expectations More practical than attempting to explicitly commit oneself to an optimal state-contingent instrument path Giannoni and Woodford (2002) show the possibility of deriving a robustly optimal target criterion commitment to same target criterion is optimal, regardless of the statistical properties of disturbances necessary for commitment to be appealing in practice see also Svensson and Woodford (2003), Svensson (2003)
The Importance of Price Stability Another characteristic feature of IT: particular (even exclusive?) emphasis on inflation stabilization Clearly a particularly important aspect of privatesector expectations to influence through policy commitment because of particular importance of inflation expectations in determining effects of current policy this means desirable to be clear about implications of policy commitments for future inflation, even if the inflation target is ideally state-contingent
The case for complete stabilization of prices is more robust than previously believed Under certain circumstances, complete price stability is optimal despite allowance for a wide range of types of real disturbances (affecting both supply and demand) Key condition needed: despite these disturbances, equilibrium with fully flexible prices would be optimal Then a policy that stabilizes prices creates an environment in which allocation of resources is same as with fully flexible prices, and hence optimal (Goodfriend and King, 1997) While special assumptions needed for this to be exactly true, may be a useful first approximation
Nonetheless, a number of important reasons for strict inflation targeting not to be an optimal policy Complete stability of (low) inflation may not be feasible, e.g., due to zero bound Monetary frictions imply that high nominal interest rates increase distortions but minimization of these conflicts with inflation stabilization Wages as well as prices sticky cannot fully stabilize both similarly, cannot generally stabilize all prices at once
Cost-push shocks: disturbances that shift flex-price equilibrium output and efficient output to different extents variations in taxes or market power virtually any disturbances, if steady state is sufficiently distorted Nonetheless, calibrated examples suggest that optimal policy should involve a great deal of stability of inflation rate Thus appropriate to emphasize that mediumterm outlook for inflation should vary little Even when other stabilization goals are taken into account, makes sense to refer to the optimal regime as flexible inflation targeting
Improving Upon Current Practice An optimal target criterion would involve more than just inflation Numerous examples: Svensson and Woodford (2003), Giannoni and Woodford (2003) Other variables that plausibly enter: output gap wage as well as price inflation (or real wage growth) other relative prices (or sectoral, regional inflation) nominal interest rate Optimal target criteria for an empirical model: Giannoni and Woodford (2003)
Estimated Model: Robustly Optimal Policy Rule A three-stage criterion for optimal policy 1. Bring about i t equal to i t,t 1 announced at t 1 2. Short-term criterion : Choose i t+1,t to announce, so as to ensure satisfaction of target criterion F t (π)+φ w [F t (w) w t ]= π t given target π t chosen at date t 1 3. Long-term criterion : Choose target π t+1 for next quarter, consistent with projections in which criterion F τ (π)+φ wf τ (w)+φ xf τ (x) =π τ is expected to be satisfied at all dates τ t.
Notation for projections: F t (z) X k=1 α z k E tz t+j, X k=1 α z k =1 History-dependent optimal target values π t E t 1 {F t (π)+φ w [F t (w) w t ]} π t (1 θ π) π + θ πft 1 1 (π) +θ wft 1 1 (w)+θ xft 1 1 (x)
Short-term criterion F t (π)+0.57 [F t (w) w t ]= π t π t E t 1 {F t (π)+0.57 [F t (w) w t ]} 0.25 α π k 1.2 α w k 0.2 1 0.8 0.15 0.6 0.1 0.4 0.05 0.2 0 0 1 2 3 4 5 6-0.2 1 2 3 4 5 6 Relative weights on projections
Long-term criterion F t (π)+0.26f t (w)+0.13f t (x) =π t π t 0.42π +0.58F 1 t 1 (π)+0.25f 1 t 1 (w)+0.12f 1 t 1 (x) 0.4 α *π k 1.5 α *w k 1.5 α *x k 0.3 1 1 0.2 0.5 0.5 0.1 0 0 0 0.6 2 4 6 α π1 k -0.5 1.5 2 4 6 α w1 k -0.5 1.5 2 4 6 α x1 k 0.4 1 1 0.2 0.5 0.5 0 0 0-0.2-0.5-0.5 2 4 6 2 4 6 2 4 6 Relative weights on projections
Explicit commitment needed to a near-term target criterion, not just medium-term target Need to say by what path it is appropriate to return to the medium-term target following a disturbance Otherwise, commitment is vague at best, as to how policy will actually be conducted Example: Giannoni-Woodford rule: target criteria place greatest weight on projections for 1 or 2 quarters in future Does not mean that optimal response to disturbance may not involve only gradual return to steady state, over period of years the point is to state the criterion that must hold along the transition path, which determines both the degree of permissible temporary departure from the medium-term target, and the appropriate speed of return
The fact that target criterion is only a mediumterm target explains the sole emphasis on inflation in the explicit target Other goals presumably determine the choice of transition path But arguments for explicit commitment apply here as well Optimal target criteria do imply a fixed target for the long-run rate of inflation but commitment to that alone would be an incomplete policy commitment, that fails to uniquely determine REE hence would not even determine what CB must do to comply
CBs may believe that commitment cannot be made about transition path, because it will depend on the nature of the individual disturbance But Giannoni and Woodford (2002) show precisely that it is possible to state a near-term target criterion that identifies the optimal transition path regardless of the nature of the disturbance
Optimal target criteria involve projections conditional on future policy consistent with the rule, and not constant-interest-rate projections Why are CIR projections used by several CBs? Probably to solve the problem of the indeterminacy of the transition path consistent with the medium-term target Problem should instead be solved by commitment to a near-term target
Why not CIR projections? Optimizing models often imply REE indeterminate under this assumption thus no unique projection obtained Backward-looking models instead often imply unstable (Wicksellian) dynamics BOE model? then highly arbitrary to be content with hitting target at one date only Not the CB s own forecast of future policy so either policy is not really expected to satisfy target criterion or else is not really based on the published projections in either event, bad for transparency, accountability
Conclusions IFT an important advance in the conduct of monetary policy big step toward more rule-based policy great improvement in transparency both very important given forward-looking behavior stability of inflation expectations valuable Yet room for improvement: avoidance of commitment to a near-term target criterion a fundamental weakness of current practice
The appropriate near-term criterion obviously much more controversial an urgent topic for further research requires progress in modeling economy