www.ceep.cas.cn BingYue Liu Department of Statistics and Finance University of Science and Technology of China June 18-21, Singapore Interval Tests for Structural Breaks in the Dependence: Empirical Evidence of Oil and Gold Markets 40 th IAEE International Conference
19
Embed
Interval Tests for Structural Breaks in the Dependence ... 1 Liu.pdf Interval Tests for Structural Breaks Introduction Crude Oil •Important commodity •Large trading volume •Low
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
www.ceep.cas.cn
BingYue Liu
Department of Statistics and FinanceUniversity of Science and Technology of China
June 18-21, Singapore
Interval Tests for Structural Breaks in the Dependence:
Empirical Evidence of Oil and Gold Markets
40th IAEE International Conference
www.ceep.cas.cn
Interval Tests for Structural Breaks
Introduction
Crude Oil
•Important commodity
•Large trading volume
•Low price
Gold•Important precious metal
•Hedge heaven
Co-movement•Positive
•Dynamic
Guidance•Investment portfolio
•Risk management
2
www.ceep.cas.cn
Literature
Baffes (2007), Hammoudeh and Yuan (2008), Soytas et al.
(2009), Sari et al. (2010), Narayan et al. (2010), Zhang and Wei
(2010), Reboredo (2013)•Method
Static relationship
Linear relationship
•Conclusion
Short-term or long-term relationship
Causal relationship
But•Market volatility
•Market mechanism change under extreme market shock
Structural (Large) change in dependence
3
Interval Tests for Structural Breaks
www.ceep.cas.cn
Contribution
Apply TVCopula to analyze dependence of oil & gold•Nonlinear
•Dynamic
•Small change in dependence
Change point test to analyze dependence of oil & gold•Large change in dependence
•But
Extreme shocks last not long
Hardly capture the dependent features in extreme risk period
Propose change interval test•Capture the dependent features in an extreme period
•Test the effect of extreme shocks on the co-movement across markets
•Empirical evidence of crude oil and gold markets
4
Interval Tests for Structural Breaks
www.ceep.cas.cn 5
Interval Tests for Structural Breaks
0 t1 t2 t3 t4 t5 T
Dependent
Rela
tionship
Comparision between Change Point and Change Interval
www.ceep.cas.cn 6
Interval Tests for Structural Breaks
www.ceep.cas.cn 7
Interval Tests for Structural Breaks
, , 1
1
ˆ arg max log | ;i
T
i i t i t t i
t
f x
F , 1, 2i , and then 1, 2, 1
1
ˆ ˆ ˆarg max log , | ;T
t t t t
t
c u u
F
1 2ˆ ˆ ˆ ˆ, ,IFM
1
0 0ˆ 0,IFMT N G
www.ceep.cas.cn
Methodology
Change point test (Dias and Embrechts, 2009)•Random vector U = (U1, U2), joint distribution
•Test
Likelihood ratio statistic
Note , then referring to Csörgő and Horváth (1997)
8
Interval Tests for Structural Breaks
;Ct
u θ
HT: H0: 1 2 Tθ θ θ vs. HA:
1 k k+1 Tθ θ θ θ
, 1
1
sup ; ;
sup ;
t k k t T
t T
c c
kc
t tθ θ
tθ
u θ u θ
u θ
1max 2logT
k TZ k
2
1 2
2 2 2 4
exp 2 1 1 1 1 4 1Pr log log
2 2
p
T p
x x h l h lpZ x O
p hl x hl x x
, as x
www.ceep.cas.cn
Methodology
Change interval test•Define parameter space
•Construct hypothesis test
9
Interval Tests for Structural Breaks
1 1 2 21 1, , : , ,
1 T 1 k k k k Tθ θ θ θ θ θ θ θQ ,
11
0 , , : 1 T 1 2 Tθ θ θ θ θQ ,
1 1
12
0 1, , : , 1 T 1 k k Tθ θ θ θ θ θQ ,
2 2
13
0 1, , : , 1 T 1 k k Tθ θ θ θ θ θQ .
HT11: H0: 11
0, , 1 Tθ θ Q vs. HA: 11
0, , 1 Tθ θ Q Q ,
HT12: H0: 12
0, , 1 Tθ θ Q vs. HA: 12
0, , 1 Tθ θ Q Q ,
HT13: H0: 13
0, , 1 Tθ θ Q vs. HA: 13
0, , 1 Tθ θ Q Q .
Null hypothesis
HT11: No change point
HT12: No change point or
only one change point in k1
HT13: No change point or
only one change point in k2
www.ceep.cas.cn 10
Interval Tests for Structural Breaks
www.ceep.cas.cn
Empirical Analysis
11
Interval Tests for Structural Breaks
Estimates for marginal distribution models
Oil
Model: t tr a , t t ta , 2 2 2
1 1 1t t t tI a , 0,1t t
1.07e-04 0.008* a 0.012** 0.961*** 0.050*** 8.465***
Gold
Model: t tr a , t t ta , 2 2 2
1 1t t ta , 0,1t t
0.049** 0.013*** 0.038*** 0.954*** 5.247***
Note: a *, **, *** denote the significant levels of 10%, 5% and 1%, respectively.
Oil & Gold: volatility clustering
Oil: volatility asymmetry
Oil & Gold: innovation symmetric fat-tailed distribution
www.ceep.cas.cn
Empirical Analysis
12
Interval Tests for Structural Breaks
Change point test over the whole sample period
Period n T 1 2
Tz 1 2 1 2
T TP Z z H0 (0.95) Time of Change
Normal
1/4/2006–4/29/2016 1 2546 5.542 4.40e-06 R. a 10/9/2013
1/4/2006–10/9/2013 1 1913 1.851 1 -- --
10/10/2013–4/29/2016 1 633 2.765 0.165 -- --
t
1/4/2006–4/29/2016 2 2546 5.720 1.16e-05 R. 10/9/2013
1/4/2006–10/9/2013 2 1913 2.758 0.608 -- --
10/10/2013–4/29/2016 2 633 2.739 0.516 -- --
Clayton
1/4/2006–4/29/2016 1 2546 4.749 2.22e-04 R. 11/7/2013
1/4/2006–11/7/2013 1 1934 2.347 0.484 -- --
11/8/2013–4/29/2016 1 612 2.183 0.525 -- --
Note: a R. denotes the null hypothesis is rejected at the 5% level.
Change point test can only check out one change point over the whole sample period.
www.ceep.cas.cn
Empirical Analysis
13
Interval Tests for Structural Breaks
Change interval test for financial crisis over the whole sample period
Copula n
Initial Period: 7/1/2008–6/30/2009
Period of Change Change Interval Test Statistics
1k 2k HT11 HT12 HT13
Normal 1 9/10/2008 4/23/2009 13.677** a 6.828** 13.140**