Monetary Policy, Asset Prices and Misspecification

1
Monetary Policy, Asset Prices and Misspecification

• The robust approach to bubbles with model
  uncertainty

Robert J. Tetlow Federal Reserve
Board www.roberttetlow.com
2
Disclaimer!

• The views expressed in this presentation and in
  the paper this presentation is based upon are
  those of the author alone and are not necessarily
  shared by any member of the Board of Governors or
  its staff.

3
Motivation

• Observation bubbles are often followed by
  economic underperformance
• U.S. stock market boom and bust (1929-33)
• Japan real estate bubble (1986-1995)
• U.S. tech stock boom and bust (1997- 2002)

4
(No Transcript)
5
In theory

• if bubbles matter'...
• Control theory says feed back on everything
• Uncertainty is just a signal-extraction problem.
• But
• Inflation forecasts can encompass a lot of
  information, so reacting to asset prices might
  not be necessary.
• Non-linear, non-Gaussian processes could point in
  either direction, particularly given uncertainty
• Historical experience argue for humility.

6
The debate

• The majority view is no
• a policy maker that looks ahead, responding to
  the implications everything for future inflation,
  but should place no added emphasis on bubbles.
  (Bernanke-Gertler)
• The minority view is yes
• policy makers can, should and do deal with
  bubbles, not by targeting asset prices as such,
  but by responding to them in reflection of their
  episodic role in transmitting crises. (Cecchetti
  et al.)

7
Partial list of references

• policy references
• Batini and Nelson (2000), Bernanke and Gertler
  (1999,2001), Bullard and Schaling (2002),
  Cecchetti et al. (2000, 2002), Cochrane (2002),
  Cogley (1999), Filardo (2001), Gilchrist and
  Leahy (2002), Kent and Lowe (1997), Mishkin and
  White (2002), Reinhart (1998).
• robust control references
• Giannoni (2002), Hansen and Sargent (various),
  Onatski (2001), Onatski and Stock (1999), Rustem
  (1988), Rustem et al. (2002), Stock (1999),
  Tetlow-von zur Muehlen (2001,2002)

8
Methodology

• Take a model that is conventional, except that
  financial markets matter
• Subject it to a process of bubbles in stock
  prices
• Compute the optimal response in
  inflation-forecast based (IFB) rules, and
  outcome-based (Taylor-type) policy rules, with
  and without a stock-price term.
• Then consider the uncertainty-averse robust
  response to possible misspecification
• Examine the costs of protecting against
  uncertainty

9
The model

• Begin a version of BGG (1997), a standard NKB
  model except
• It has capital
• There are financial frictions
• Hybrid NK price equation
• Consumption depends on observed stock prices
• A bubble process for stock prices

10
The model (continued)

• And we go further. We also add
• Habit persistence in consumption
• Generalized adjustment costs in investment
• Investment that depends on observed stock prices

11
Modeling bubbles
12
(No Transcript)
13
(No Transcript)
14
(No Transcript)
15
Methodology

• Formulate the state-space form

• Loss function

16
Methodology (continued)

• Simple rules

• Note that rule can be outcome based (j0), or
  forecast based (j1).
• Note also that the policy rule has no error term.
• Constants are normalized to zero.

17
Robust policy

• Takes uncertainty seriously
• Non-parametric in nature
• Three different variants exist.
• We use structured perturbation method in this
  paper.
• A malevolent nature is assumed to choose either
  a or p within some bound to maximize the loss
  function

18
Robust policy (continued)

• More generally, some element of either C or B in
  the state-space form, call it bij is chosen

• Subject to the model, the rule and any
  restrictions on its specification and the
  variance-covariance matrix of shocks.

19
Results 1 the basics

20
Results 1 the basics (continued)

21
Results 2 correlated shocks

22
Results 3 continuation probabilities

23
Results 4 bubble magnitudes

24
Results 5 Robust results

25
Results 6 Robust results

26
Concluding remarks

• What we did
• Examined the role for monetary policy in
  responding directly to stock market bubbles
• Did so working in three directions
• The breadth of the BG argument
• Broadened the experiments
• Looked at Knightian uncertainty

27
Concluding remarks (continued)

• What we found
• Results are mostly supportive of the majority
  view that policy should not respond directly to
  bubbles.
• However there are calibrations for which the case
  for direct response is stronger.