"Monetary Policy, Asset Prices and Misspecification: The robust approach to bubbles with model uncer

1
"Monetary Policy, Asset Prices and
Misspecification The robust approach to bubbles
with model uncertainty"by Robert J. Tetlow
Comments by Klaus Schmidt-Hebbel Central Bank of
Chile Bank of Canada Conference on Issue in
Inflation Targeting Ottawa, 28 and 29 April 2005
2
Outline

• 1. The Paper
• 2. Comments
• 3. A digression from the South

3
1. The paper
4
The paper

• This paper pushes the frontier of analytics and
  evaluation of monetary policy response to asset
  price bubbles
• Extends and uses BGG in innovative directions
• Solves for Taylor-rule response coefficients and
  CB quadratic loss for base case and conducts
  robustness analysis in several dimensions CB
  preferences, Taylor rule specification,
  correlation of shocks, bubble features, and model
  uncertainty
• Along the BG CGLW line, results put this paper
  very close to the former position most of the
  time, refrain from responding to bubbles.

5
2. Comments
6
Taylor rule specification

• Interest rate does not converge to neutral rate
  r because the third determinant is actual stock
  price change (s-s(-1)) instead of deviation from
  fundamentals (s-q), like in BGG. Paper justifies
  former choice because of difficulty of measuring
  q but then its implicit assumption is that the
  steady-state s is constant, hence any deviation
  from it is a bubble
• Interest rate adjustment is instantaneous
  although it is unlikely empirically relevant.
  Under policy sluggishness I would expect that the
  gains from responding to bubbles (highly
  temporary and high-variance bubbles) would be
  even smaller than those reported in the paper.

7
Coefficient Estimates

• Optimal response coefficients are huge
  inflation, output-gap, and bubble response
  coefficients range from 2 to 35 across tables 1-5
  (with an average of 4 or 5)
• Take for instance the coefficients in line 3,
  Table 2, on average close to 3. Assume a
  demand-driven shock of 1 to inflation forecast,
  1 of the output gap, and 1 in stock prices
  the interest rate should be raised by roughly 9?
• These values seem to be well above those reported
  in much of the empirical literature, suggesting a
  possible problem in model specification or
  large irrationality by the central bank.

8
Coefficient Estimates and Methodology

• An extreme case is line 4 in Table 4 when the
  bubble response is shut down, optimal response
  coefficients are 11 (inflation) and 35 (output
  gap)! for a very reasonable model
  specification.
• Incidentally, the latter result shows that not
  responding to bubbles (line 4) is loss-superior
  to responding, contrary to the claim that a rule
  based on a broader set of variables must
  outperform one based on a smaller set (p. 16,
  para. 2).

9
Model Stability

• The regions of stability, instability and
  indeterminacy in Fig. 4 are somewhat surprising
• Instability arises under combinations of the
  inflation response coefficient satisfying the
  Taylor principle (say, 1.1) and the bubble
  response coefficient is still large (say, 2.0)
• The model is stable under negative reactions to
  inflation shocks and bubbles?
• In my experience of dynamic rational-expectations
  models, large regions of model instability or
  indeterminacy are often a symptom of certain lack
  of fundamentals in parts of the model. Another
  possible symptom are the large derived
  optimal-response coefficients.

10
(No Transcript)
11
Risks of responding to a Wrong Bubble

• A nice robustness test is performed in the paper
  by deriving the optimal response to bubbles under
  uncertainty about the degree of bubble
  irrationality, based on a minimax strategy of
  quadratic losses
• A related question could be posed about the
  bubble itself. Paper assumes full certainty about
  a bubble when it arises (whenever a positive
  stock price change is observed). But if the
  alternative BGG modeling were adopted (specifying
  policy response to deviations of the s from its
  fundamental q), one could address two questions
• A robustness test for the optimal response under
  uncertainty about the fundamental value (q),
  based on the authors minimax strategy
• Simulation of the costs of responding wrongly to
  a perceived bubble, as a result of
  underestimating q. This would help to provide
  more perspective in assessing the mostly small
  but positive gains from reacting to bubbles under
  certainty of bubbles.

12
3. A digression from the South
13
Monetary policy response to other prices

• As discussed in the paper, the same logic of
  responding to stock prices could be applied to
  exchange-rate, commodity-price, and real-estate
  price bubbles.
• Among the latter variables, a major question that
  arises in developing countries is if monetary
  policy reacts to an exchange-rate devaluation,
  over and above its effects on inflation and
  output
• Null hypothesis on fear of floating
  emerging-country central banks react to the
  exchange rate out of concerns of direct inflation
  effects (high pass-through), financial
  instability (balance-sheet effects), and
  traded-goods production competitiveness.

14
Does MP react strongly to the ER?

• General results for selected Latin American
  inflation targeters (Schmidt-Hebbel and Werner
  2002, Schmidt-Hebbel and Tapia 2004)
• Relative exchange-rate to reserve volatility is
  high similar or higher than in Australia,
  Canada, and New Zealand
• Devaluation-inflation pass-through coefficients
  are much lower after floating and achievement of
  low stationary inflation is lower (e.g.,
  0.15-0.20 in Chile)
• MP does not react generally to the ER over the
  long term but it has reacted at times of
  significant stress (1998-2001)
• To test for the latter, the ER is included as an
  argument in a Taylor rule

15
Estimation of Taylor Rules for Brazil, Chile,
Mexico
Source Schmidt-Hebbel and Werner (2002)
16
MP response to the ER (rolling coefficients)
Source Schmidt-Hebbel and Werner (2002)
17
References

• Bernanke, B. and M. Gertler (2001) Should
  Central Banks Respond to Movements
• in Asset Prices? 91, American Economic Review
  Papers and Proceedings (May), pp.253-257.
• Schmidt-Hebbel, K. and M. Tapia (2004).
  Inflation Targeting and Floating in Chile.
  Mimeo, Banco Central de Chile.
• Schmidt-Hebbel, K. and A. Werner (2002).
  Inflation Targeting in Brazil, Chile and Mexico
  Performance, Credibility, and the Exchange Rate,
  Economia, 2 (2) 30-89.

18
"Monetary Policy, Asset Prices and
Misspecification The robust approach to bubbles
with model uncertainty"by Robert J. Tetlow
Comments by Klaus Schmidt-Hebbel Central Bank of
Chile Bank of Canada Conference on Issue in
Inflation Targeting Ottawa, 28 and 29 April 2005