1

Some Remarks on the Theory of Optimal Monetary Policy Marc Giannoni Columbia Business School CEPR, CIRANO, NBER HEC Montréal October 20, 2007

2

Background: Fascinating developments over last 15 years Practice: Early 1990s: introduction of inflation targeting (IT) in several countries (NZ, Canada, Sweden, UK…) Wonderful results: Inflation at low, stable levels; inflation expectations well anchored, economic activity fairly stable Macroeconomic Theory: Early 1990s: little role for monetary policy Principles for conduct monetary policy based on IT (Bernanke, Laubach, Mishkin, Posen, 1999) Mid-1990s: DSGE models usable for monetary policy analysis (Goodfriend-King, Rotemberg-Woodford, Clarida-Gali-Gertler, Svensson) Optimal monetary policy expressed as flexible inflation target Progress on analysis of model uncertainty Practice and Theory meet again!

3

IT in practiceSet instrument (interest rate) so as to achieve long-term constant inflation target (e.g. 2%) Flexibility in short-term (flexible-IT) Extensive explanation to public (monetary policy reports…) Transparency of decision process, objectives But typically, forecasts of monetary policy instrument not announced

4

Optimal IT in theory (Svensson, Giannoni-Woodford)Macroeconomic model (DSGE, estimated) Used to compute optimal path of all variables of interest (inflation, output gap, interest rates...) Implementation: inflation forecast targeting (IFT) Fixed long-run inflation target But in short term, CB commits to adjust instrument (interest rate) as required for a projection of future path of economy (inflation, output gap…) to satisfy a target criterion Target criterion specifies optimal short-term deviations form long-run inflation target Can be robust to model misspecifications e.g. Norges Bank

5

Optimal IT in theory (2) (Svensson, Giannoni-Woodford) CB commits to publish projections and explain policy decisions May include interest-rate projections: entire path of expected short-term interest rates matters Helps manage expectations E.g. RBNZ, Norges Bank, Riksbank… Possible that limits on transparency are desirable

6

Advantages of IFTTransparency (of monetary policy goals, decisions…) Anchor medium-term inflation expectations Optimal short-run responses to shocks Can be robust to nature of disturbances

7

Features of optimal IT criterionTarget for adjusted inflation projection (flexible IT) output gap… also matter (includes all of CB's stabilization goals) Optimal target value should vary over time (history-dependent) even though long-term inflation pinned down at a constant level helps for expectations management Use all available information about current state of the economy to make projections (may include judgment) Not a mechanical reaction function

8

Features of optimal IT criterion (2) Projections under forecasted future policy not constant interest rate or market expectations Forecasts at relatively short horizons horizons at which CB actions begin to affect target variables

9

Inflation or Price-Level Targeting?Path of log price level with no shockQuartersDlog(P) = 2%

10

Inflation or Price-Level Targeting?Effect of a cost-push shock (date 0)QuartersOptimal path under ITDlog(P) = 2%Dlog(P) = 2%

11

Inflation or Price-Level Targeting?Effect of a cost-push shock (date 0)QuartersOptimal path under ITOptimal path under PLT Commitment to bringing P to trend stabilizes prices

12

Inflation or Price-Level Targeting?Effect of a cost-push shock (date 0) PLT close to optimal QuartersOptimal path under PLTOptimal path under ITOptimal policy

13

Inflation or Price-Level Targeting?Effect of a demand or productivity shock (date 0) PLT: close to optimal (?) QuartersOptimal path under ITOptimal policyOptimal path under PLT

14

Inflation or Price-Level Targeting?IT: “bygones are bygones” PLT: Commitment to bringing prices to target limits incentive to raise prices too much Error correction desirable given model misspecification Optimal policy: desirability of “history dependence” Effective management of expectations Not exactly PLT, but close

15

Forecasts: Key for implementation of IFTNeed to be reliable But “Assessing the economy's true state remains a formidable challenge” (Bernanke, 10/19/07) productivity? potential output? inflation?....

16

Inflation: Which series?US quarterly inflation (demeaned)

17

Assessing state of the economyIntuitive approach Greenspan: large set of detailed statistics.... (sales of men’s underwear…) Complement: Formal statistical analysis Exploiting systematically information from in data-rich environments (Stock-Watson, Bernanke-Boivin-Eliasz, Reichlin et al., Boivin-Giannoni…) Can be combined in fully consistent way with theoretical models Successful for forecasting, estimating state of the economy

18

Estimated Inflation in US: Close to core PCEUS quarterly inflation (demeaned)Inflation estimated using large data setData: GDP deflatorSource: Boivin-Giannoni (2006)

19

ConclusionGreat progress has been achieved in: Conduct of monetary policy Monetary theory Estimation of models It is now possible to: Estimate accurately state of economy using large information sets Use estimated models as effective guides for conduct of monetary policy Some lessons from the theory: Optimal policy can be implemented by flexible IT History dependence desirable PLT may be good approximation to optimal policy / robust Progress on analysis of model uncertainty, but more work to be done