A Structural Break in the Effects of Foreign Exchange Intervention on YenDollar Exchange Rate Volati

1
A Structural Break in the Effects of
Foreign Exchange Intervention on Yen/Dollar
Exchange Rate VolatilityEric Hillebrand,
Louisiana State University Gunther Schnabl,
European Central Bank and Tübingen University
March 2005
2
Outline

• Setting the stage Yen/Dollar and Interventions
• The measure of success
• The empirical literature Is (Japanese) foreign
  exchange intervention effective?
• Data and reaction function for the Bank of Japan
• Global GARCH, Local GARCH
• Interpretation of the findings
• Conclusion

3
1. The Syndrome of the Ever-Higher Yen
??
observation period
4
Japanese Foreign Exchange Intervention
5
US Foreign Exchange Intervention
foreign reserves in Dec. 2004 Japan 820 billion
dollars (20 GDP) US 43 billion dollars (0.5
GDP)
6
2. The Measure of Success

• Returns cause Intervention, Intervention causes
  returns?
• Endogeneity bias!
• Two strands of literature 1) Intervention as a
  problem of optimal control (Mundaca/Oksendal
  1998, Cadenillas/Zapatero 2000)
• 2) Intervention as empirical problem GARCH plus
  intervention as exogenous variable
• Successful intervention Reduces volatility. No
  statements possible about direction of exchange
  rate movement.
• Is there an endogeneity problem with volatility?

7
Exchange Rate Level and Intervention
8
3. The empirical literature

• Observation consensus" among central bankers
• Sterilized interventions are ineffective.
• Foreign exchange intervention increases
  volatility because of higher uncertainty.
• Only interest rate changes (unsterilized
  interventions) trigger adjustments of
  international capital flows and thereby exchange
  rate changes.
• sterilized intervention is effective
• portfolio-balance models (Branson 1983)
• signalling effects (Mussa 1981)
• international coordination (with US) (Bryant 1995)

9
3. The empirical literature

• Japanese foreign exchange intervention is not
  effective
• Schwartz 1996 exercise in futility causing
  higher volatility
• Bonser-Neal/Tanner 1996 higher volatility
  (85-91)
• Dominguez 1998 higher volatility (77-94)
• Galati and Melick 1999 higher uncertainty
  (93-96)
• Japanese foreign exchange intervention is
  effective
• Ramaswamy/Samiei 2000 at least partially
  effective (90-98)
• Ito 2002 effective since 1995 (91-01)
• Beine/Szafarz 2003 effective when coordinated
  with US (91-01)
• Fatum/Hutchison 2003 successful (91-00)
• Nagayasu 2004 effective when coordinated with US
  (91-01)

10
4. Data

• actual intervention data (day, volume, currency)
• Japanese Ministry of Finance (MOF)
• Federal Reserve Board
• observation period April 1991 Oct. 2004
  (updated)
• daily exchange rate data provided by Bloomberg
  Datastream
• Tokyo 5 p.m.
• New York noon (London 5 p.m., Tokyo 2 a.m. t1)
• New York 5 p.m. (Tokyo 7 a.m. t1)
• number of observations
• 3542 trading days
• 344 intervention days (9.71) (US 22 / 1.01)

11
Bank of Japan Reaction Function

• Possible determinants of intervention
• percentage exchange rate changes of the previous
  day rt-1
• deviation of exchange rate from a target value
• exchange rate volatility (rt-1)2
• intervention of the previous day
• binary probit model (ID0 or ID 1)

12
Results for the Reaction Function

• Japanese foreign exchange intervention is
  triggered by the exchange rate but not exchange
  rate volatility.
• The exchange rate term in the mean estimation can
  be subject to endogeneity bias.

13
5. Global GARCH Estimation 1991-2004
14
Local GARCH Estimation

• All previous estimations (Ramaswamy/Samiei 2000,
  Ito 2002, Beine/Szafarz 2003, Fatum/Hutchison
  2003, Nagayasu 2004) were global.
• Global approaches assume that the effects of
  Japanese foreign exchange intervention are
  unchanged during the whole observation period.
• Structural breakswhich may distort the
  resultsare ignored.
• To identify structural breaks we re-estimate the
  model
• year by year (arbitrary),
• intervention clusters (arbitrary),
• periods built by change point detection (non-
  arbitrary).

15
Yearly Results 1991-2004

• Intervention reduces exchange rate volatility
  starting in 1999.
• Before volatility seems to increase.

16
Results for Intervention Clusters
17
Change Point Detector

• Change Point Detector for ARCH models
  (Kokoszka/Leipus 1999) identifies
• periods with different volatility
• non-arbitrary segmentation
• specification
• Detected change points
• 7. May 1997
• 3. April 2000

18
Results for Change Point Detection

• Structural break after 1999?

19
Rolling GARCH(1,1) Coefficients

• Window size 500, 750, 1000, 1250, 1500

20
Rolling GARCH(1,1) Coefficients

• Window size 500, 750, 1000, 1250, 1500

structural break around 1999/2000
21
6. Interpretation of the Findings

• Empirical Plausibility
• In line with recent financial press reports.
• Horst Köhler (August 2003) Tokyos
  unprecedented foreign exchange intervention is
  pragmatic and helps stabilizing the financial
  system.
• In line with reduced low-frequency yen/dollar
  exchange rate volatility (McKinnon/Schnabl
  2004a).
• Theoretical Plausibility
• If Japanese foreign exchange intervention remains
  sterilized the finding is surprising.
• If in the liquidity trap foreign exchange
  intervention is regarded as unsterilized, the
  finding is in line with the conventional wisdom
  on the efficacy of foreign exchange intervention.

22
7. Conclusion

• Marked change in the influence of Japanese
  interventions on the volatility around 1999/2000.
• Possible reasons changed intervention
  credibility, liquidity trap
• Challenges
• Endogeneity problem Success in influencing the
  direction of FX movements?