Inflation Persistence

1
Inflation Persistence

• Comments by Sharon Kozicki on
• Factor-Market Structure, Shifting Inflation
  Targets and the New Keynesian Phillips Curve by
  Amano and Murchison
• The New-Keynesian Phillips Curve When Inflation
  is Non-Stationary The Case of Canada by Barkbu
  and Batini

2
Inflation Persistence

• Sources of inflation persistence
• Superneutrality
• Empirical regularities
• Empirical irregularities and data
• The structural model and deep structural
  parameters

3
1. In the NKPC, inflation depends on

• Steady-state inflation
• Lagged and expected inflation gap
• Marginal Cost
• Euler equation residual
• Expectations formation / Learning
• Each may be a source of inflation persistence

4
1A. Steady-state Inflation

• Steady-state inflation is typically linked to the
  actual or perceived inflation target.
• Shifts in the target contribute to persistence
  (Amano-Murchison, Barkbu-Batini, Kozicki-Tinsley,
  Levin-Piger)
• Inflation persistence can be held down by
  monetary policy that provides a credible constant
  inflation target.

5
1B. Lagged and expected inflation gap

• Most studies find lags and expected inflation
  gaps significant
• Many theories offer explanations for lags
• Indexation by non-optimizing firms
• Rule-of-thumb firms that set prices in a
  backward-looking way
• Staggered contracts (Taylor, Fuhrer-Moore)
• Frictions on price adjustment (Kozicki-Tinsley)
• Difficult to determine whether policy can reduce
  such sources of persistence.

6
1C. Real marginal cost gap

• The NKPC suggests persistence in rmc will be
  magnified and reflected in inflation
• A puzzle
• Since 1991, rmc has been considerably more
  persistent than inflation
• Are we sure we are measuring rmc correctly?
• Is rmc I(0)? I(1)?
• What is the steady-state of rmc?

7
1D. Euler equation residual

• When all else fails to explain inflation
  persistence, economists appeal to stories of
  persistent shocks
• Justification of persistent shocks is no less ad
  hoc than stories motivating inclusion of the
  lagged inflation gap

8
1E. Expectations and Learning

• Expectations model-consistent, rational, less
  restrictive
• A-M provide two pre-estimates of nominal anchor
• MEP market view SEP BoC view
• MEP-SEP gap prior to inflation targeting suggests
  imperfect policy credibility a historical
  justification for learning
• Short-horizon learning not in A-M or B-B
• Milani suggests persistence is US inflation is
  mainly from expectations and learning

9
2. Superneutrality in the NKPC

• The NKPC is a model of the deviation of
  inflation from steady state

• If the steady-state moves with market
  expectations of the target (K-T), then
  superneutrality does not require the sum of
  coeffs on leads and lags of inflation to equal 1.
• If the sum of coeffs doesnt equal 1, then
  steady-state inflation is an unobserved variable
  in the NKPC, cointegrated with I(1) inflation
• ?B-B estimates may be biased

10
3. Empirical regularities

• Over a longer history, inflation appears
  non-stationary (A-M, B-B, K-T, L-P) suggesting
  steady-state inflation wasnt constant
• Both forward- and backward-looking terms appear
  to be important in determining inflation
  persistence in Canada
• ?b 0.3
• ?f 0.7

11
4. Empirical irregularities

• Same model, different results

12
4. Empirical irregularities

• Same model, but different data and different
  scaling
• Measurement of rmc not yet fully resolved

13
4. Data Real Marginal Cost
14
4. Data Inflation
15
5. Deep structural parameters

• As noted by B-B, A multitude of models lead to
  the same general form of the NKPC
• Estimates of deep structural parameters in A-M
  are conditioned on calibrated values of other
  parameters and a model specification
• Standard errors of deep structural parameters are
  probably larger than reported
• A-M choose a model that gives sensible estimates
  of deep structural parameters

16
5. Deep structural parameters
17
5. Deep structural parameters

• A model of pricing decisions by optimizing firms
  links inflation to real marginal cost
• In A-M the coeff on rmc is tiny
• But, this coefficient is used to estimate the
  average duration between re-optimizations.
• Its as if the lack of a relationship b/w
  inflation and rmc is used to estimate average
  duration
• Firm-specific capital adjustment is necessary for
  sensible parameter estimates

18
Conclusions

• Inflation persistence can be held down a credible
  constant inflation target.
• Other sources of persistence remain.
• More measurement before more theory
• Real marginal cost and its steady state?
• What measure of inflation?
• Expectations?
• Premature to claim we have a good structural
  model of inflation dynamics or good estimates of
  deep structural parameters
• Are we there yet? No, but were getting closer!