Volatility spillovers between
agricultural commodity and
financial asset markets
ZEF Volatility Workshop, 1 February 2013
Stephanie Grosche
Stephanie.grosche@ilr.uni-bonn.de
Growing importance of commodities as portfolio assets
More investment vehicles available
Global financial crisis intensify
Growth in Commodity ETP assets
2002-11, bn USD
2007-2012 global financial crisis
 Subprime mortgage crisis
50
45
40
35
30
25
20
15
10
5
0
(2007/2008)
 Sovereign debt crisis
(from ~2010)
2002
2003
2004
2005
2006
2007
2008
2009
2010
Use of agricultural commodities
as portfolio diversifiers facilitated
2011
Higher importance of agricultural
commodities refuge assets
Source: BlackRock (2011); Conover et al. 2010, Chong and Miffre 2010, Gorton and Rouwenhorst 2006, Ankrim and Hensel 1993
1 February 2013
Stephanie Grosche, ILR, University of Bonn
2
Development of trading volume in asset markets
NYMEX WTI Crude oil, 1st gen.
CBOT C, S, W, 1st gen.
400.000
350.000
300.000
250.000
200.000
150.000
100.000
50.000
0
Jun-11
Jun-10
Jun-08
Jun-08
Jun-09
Jun-07
Jun-06
Jun-05
Jun-04
Jun-03
Jun-02
Jun-01
Jun-07
DJ Equity REIT Index
Jun-00
Jun-99
Jun-98
Jun-11
Jun-10
Jun-09
Jun-08
Jun-07
Jun-06
Jun-05
Jun-04
Jun-03
Jun-02
Jun-01
Jun-00
Jun-99
Jun-98
800.000
700.000
600.000
500.000
400.000
300.000
200.000
100.000
0
S&P 500 Index
140.000.000
3.500.000.000
120.000.000
Sharp
decrease,
not zero
100.000.000
80.000.000
60.000.000
40.000.000
3.000.000.000
2.500.000.000
2.000.000.000
1.500.000.000
1.000.000.000
20.000.000
500.000.000
Jun-11
Jun-10
Jun-09
Jun-06
Jun-05
Jun-04
Jun-03
Jun-02
Jun-01
Jun-00
Jun-99
0
Jun-98
Jun-11
Jun-10
Jun-09
Jun-08
Jun-07
Jun-06
Jun-05
Jun-04
Jun-03
Jun-02
Jun-01
Jun-00
Jun-99
Jun-98
0
Significant increase in commodity trading volume after 2006
Source: Bloomberg
1 February 2013
Stephanie Grosche, ILR, University of Bonn
3
Research objective
Investigate whether market interdependence and volatility
transmission between agricultural commodity markets and
financial asset markets increases…
A
In normal markets:
As a result of portfolio rebalancing and asset weight
adjustments.
B
In crisis markets:
As s a result of real asset substitution and use of
agricultural commodities as refuge assets.
1 February 2013
Stephanie Grosche, ILR, University of Bonn
4
Methodology
Selection criteria
 Multivariate
(~8 variables)
 Link to economic
theory
 Account for
potential regimeswitches
Methodology
 Structural VAR
(rolling estimation)
 Generalized
Forecast Error
Variance
Decompositions
(Pesaran and Shin,
1998)
Application examples
 Diebold and Yilmaz
(2009, 2012)
 Dimpfl and Jung
(2012)
 “Volatility spillover
indices”
1 February 2013
Stephanie Grosche, ILR, University of Bonn
5
Modeling steps
I
II
 Selection of included financial and commodity assets
 Data gathering
 Computation of volatility proxies
III
 Estimation of rolling VAR models
 Generalized forecast error variance decompositions (FEVDs)
IV
 Calculation of volatility spillover indices
1 February 2013
Stephanie Grosche, ILR, University of Bonn
6
I
Assets included in the analysis
Commodity
Financial
Agricultural
Equity
 S&P 500 Index (SPX)
 Corn, CBOT (C1)*
 Wheat, CBOT (W1)*
 Soybeans, CBOT (S1)*
Real estate
 DJ Equity All REIT Index (REIT)
Energy
 WTI Crude oil , NYMEX (CL1)*
3 June 98 –
30 Mar 2012/
Daily data
Fixed income
 10-y-U.S. Treasury, CBOT (TY1)*
Foreign exchange
 ICE Futures U.S. Dollar Index (DXY)
* Future contracts, 1st generic (Bloomberg), rolling “relative to expiration”, contracts rolled after last trading day of front month
1 February 2013
Stephanie Grosche, ILR, University of Bonn
7
II
Volatility proxies used in the models
Focus
Range-based volatility*
ˆ Range,it
1   PitHigh 

ln  Low 
4ln 2   Pit 
Pro:
 Captures intraday movements
2
Con:
 May show high volatility in times
of a persistent trend in returns
 May be inflated due to intraday
periods of low trading volume
Return-based volatility
1 m
ˆ Re turn,it (m) 
( Rit  n  Ri (m)) 2

m  1 n 1
with
and
 P Closeit 
Rit  ln  Close 
 P it 1 
m = 5, 30, 90, 180
Pro:
 Captures trends
Con:
 Neglects intraday movements
 Sensitive to included no. of
observations/ time period of
investigation
* based on Parkinson (1980)
1 February 2013
Stephanie Grosche, ILR, University of Bonn
8
II
Asset volatility profiles(Range-based, Annualized*)
* Multiplied by 252 0.5
Source: Own calculations
1 February 2013
Stephanie Grosche, ILR, University of Bonn
9
III
Estimation of VARs – Rolling regression
Model
Specification*
Included observations (No. per variable = T)
w = 252
No. of windows
(T-w + 1)
log ˆ Range
VAR(4)
06/03/98
…
03/30/12
(3,488)
3,237
log ˆ Return (5)
VAR(1)
06/10/98
…
03/30/12
(3,483)
3,232
log ˆ Return (90)
VAR(1)
10/09/98
…
03/30/12
(3,398)
3,147
* Lag length selected with SBC, VAR models for 30 and 180 day return-based volatilities estimated, results not reported
1 February 2013
Stephanie Grosche, ILR, University of Bonn
10
Generalized vs. Orthogonalized impulse responses
IV

MA representation:
yt      h ut  h
h 0
 j (h)
IR function:
Restriction required
Focus
Orthogonalized
Generalized (Pesaran and Shin 1998)
 Response to specific shock in
 Response to typical composite
one variable (equation j, c.p.)
shock (equation j and others)
 Via Cholesky decomposition of
 Via information on history
covariance matrix  = PP’:
contained in estimated :
 (h)  h Pe j
g
j
 Sensitive to ordering
 Theory required
1
2
 Not sensitive to ordering
 Sensitive to past behavior
jj = variance of errror term for jth equation
ej = selection vector (1 as jth element, 0 otherwise)
1 February 2013

 (h)   jj  h e j
o
j
h = time period for forecast
Stephanie Grosche, ILR, University of Bonn
11
IV
Variance decompositions and volatility spillovers
Generalized FEVDs
 Own variance shares:
fraction of H-step
ahead FEVs for one
asset class (i) that are
due to shocks to this
asset class (i).
 Spillovers (cross
variance shares):
fraction of H-step
ahead FEVs for one
asset class (i) that are
due to shocks to
another asset class (j).
Spillover indices (Diebold and Yilmaz 2009, 2012)
 Total spillovers (H)
= sum of spillovers across all asset classes in
relation to the total forecast error variance.
 Directional spillovers FROM (H)
= spillovers received by asset i from all other assets j =
1,…,N , ji in relation to the total forecast error
variance.
 Directional spillovers TO (H)
= spillovers transmitted by asset i to all other assets j
= 1,…,N , ji in relation to the total forecast error
variance.
 Net (pair wise) spillovers (H)
= spillovers transmitted by asset i to all other assets j
= 1,…,N , ji (one asset j) – spillovers received by
asset i from all other assets j = 1,…,N , ji (one asset
j) in relation to the total forecast error variance.
1 February 2013
Stephanie Grosche, ILR, University of Bonn
12
Own variance share
IV
Index calculations based on FEVDs
C1
S1
W1
CL1
SPX
REI
TY1
Cross variance shares
(spillovers)
DXY (j) *
C1
1
S1
1
W1
1
CL1
1
SPX
1
REI
1
TY1
1
DXY
(i)
1
Matrix with FEVDs for a given
forecast horizon H
* Entries have been normalized
with row sum
N
A
Total spillover index =
Sum of cross-variance shares rows 1:N / Sum of all variance shares rows 1:N (=N) * 100
B
Spillover index FROM all j to i =
Sum of cross variance shares in row (i)/ sum of all variance shares in rows 1:N (= N) *100
C
Spillover index from i TO all j =
Sum of cross variance shares in column (i)/ sum of all variance shares in columns 1:N (= N) * 100
D
Net (pairwise) spillover index i=
Spillover index from i TO all j (one j) – spillover index FROM all j (one j) to i
Source: Diebold and Yilmaz (2012, 2009)
1 February 2013
Stephanie Grosche, ILR, University of Bonn
13
IV
Results – Total volatility spillover index, H = 10
 Nasdaq crash, end
of dot.com bubble
(03/03)
50
40
Percent
 Stock market
downturn of 2002
60
 Low real GDP
growth in EU 27 and
US
 Growth in imports
from China
(soybeans) and India
 Aftermath of
Afghanistan/ Iraq
wars
0
50
45
40
35
30
25
20
15
10
5
0
 12 successive
decreases of interest
rates by Fed b/w
Aug 07 and Dec 08
 Biofuel mandates in
EU and US
 Further growth in
imports from China
and India
 Low stock levels
Time (end of regression window)
log ˆ Range,it
1 February 2013
 Subprime crisis/
Sovereign Bond
Crisis
 Low /negative real
GDP growth in in
EU27 and US
30
10
Percent
 Continued reduction
of EU buffer stocks
Late 2000s recession
(01/07 – …)
20
 September 11
 Beginning of war in
Afghanistan,
Invasion in Iraq
Early 2000s
recession
(03/00 – 12/03)
log ˆ Return,it (5)
log ˆ Return,it (90)
Stephanie Grosche, ILR, University of Bonn
 Commodity index
fund trading volume
growth
14
IV
Net directional spillovers* (Range-based)
* > 0 = net transmitter
< 0 = net receiver
1 February 2013
Stephanie Grosche, ILR, University of Bonn
15
IV
Pairwise analysis* (Range-based): Corn
* > 0 = net transmitter
< 0 = net receiver
1 February 2013
Stephanie Grosche, ILR, University of Bonn
16
IV
Pairwise analysis* (Range-based): Wheat
* > 0 = net transmitter
< 0 = net receiver
1 February 2013
Stephanie Grosche, ILR, University of Bonn
17
IV
Pairwise analysis* (Range-based): Soybeans
* > 0 = net transmitter
< 0 = net receiver
1 February 2013
Stephanie Grosche, ILR, University of Bonn
18
IV
Pairwise analysis* (Range-based): Crude oil
* > 0 = net transmitter
< 0 = net receiver
1 February 2013
Stephanie Grosche, ILR, University of Bonn
19
First insights and preliminary conclusions
 Total volatility spillovers generally increase during times of financial crises
 Net volatility spillovers from equity and real estate markets reached high levels during
and after subprime crisis
 Commodities (except soybeans) mostly net receivers of volatility spillovers during and
after subprime crisis  crude oil net transmitter of volatility during early 2000 crisis
 Most effects more pronounced in the short-term (range-based / 5D return-based)
 No general evidence on effects of financial crises on intra-commodity market spillovers
 Some evidence for closer integration of commodity and financial asset markets
during times of crises
 Some evidence for a structural change in volatility spillovers in soybean-corn
and soybean-wheat market pairs, soybean market net volatility transmitter
1 February 2013
Stephanie Grosche, ILR, University of Bonn
20
Robustness checks and possible extensions
Robustness checks
 Sensitivity analysis (e.g. different lag lengths (HQ, AIC criteria), different forecast horizons,
different window size)
 Check for whiteness of residuals for each window (Ljung Box Test, Breusch-Godfrey LM Test)
 Check for structural breaks within the windows
Planned extensions
 Use of index composed of wheat, corn, soybeans (weight e.g. trading volume?)
 Check for structural breaks within volatility spillover indices
 Complementary structural analysis (e.g. Impulse responses, Granger Causality Analysis)
 Inclusion of metal markets
 Introduction of seasonality (e.g. harvest dummies)
 Comparison with conditional volatility model (M-GARCH)
 Use of implied volatility
1 February 2013
Stephanie Grosche, ILR, University of Bonn
21
Sources (1/2)
Ankrim, E.M. and Hensel, C.R. (1993): Commodities in Asset Allocation: A Real-Asset Alternative to Real Estate?, Financial Analysts
Journal, Vol. 49, No. 3, pp. 20–29.
Aulerich, N.M. et al. (2010): The price impact of index funds in commodity futures markets: Evidence from the CFTC's daily large
reporting system, Working paper, Department of Agricultural and Consumer Economics, University of Illinois.
http://www.farmdoc.illinois.edu/irwin/research/PriceeImpactIndexFund,%20Jan%202010.pdf, last accessed 13.07.2011.
BlackRock (2011): ETP Landscape: Global Handbook.
http://www.blackrockinternational.com/content/groups/internationalsite/documents/literature/etfl_globalhandbook_q411.pdf, last
accessed 03.07.2012.
Chan, K.F. et al. (2011): Asset market linkages: Evidence from financial, commodity and real estate assets, Journal of Banking & Finance,
Vol. 35, No. 6, pp. 1415–1426.
Cheung, Y. and Ng, L.K. (1996): A causality-in-variance test and its application to financial market prices, Journal of Econometrics, Vol.
72, No. 1-2, pp. 33–48.
Chiang, M.-H. and Wang, L.-M. (2011): Volatility contagion: A range-based volatility approach, Journal of Econometrics, Vol. 165, No.
2, pp. 175–189.
Chong, J. and Miffre, J. (2010): Conditional Correlation and Volatility in Commodity Futures and Traditional Asset Markets, The Journal
of Alternative Investments, Vol. 12, No. 3, pp. 61–75.
Conover, C.M. et al. (2010): Is Now the Time to Add Commodities to Your Portfolio?, The Journal of Investing, Vol. 19, No. 3, pp. 10–19.
Diebold, F.X. and Yilmaz, K. (2012): Better to give than to receive: Predictive directional measurement of volatility spillovers, International
Journal of Forecasting, Vol. 28, No. 1, pp. 57–66.
Diebold, F.X. and Yilmaz, K. (2009): Measuring Financial Asset Return and Volatility Spillovers, with Application to Global Equity Markets, The
Economic Journal, Vol. 119, No. 534, pp. 158–171.
Dimpfl, T. and Jung, R.C. (2012): Financial market spillovers around the globe, Applied Financial Economics, Vol. 22, No. 1, pp. 45–57.
1 February 2013
Stephanie Grosche, ILR, University of Bonn
22
Sources (2/2)
Du Xiaodong et al. (2011): Speculation and volatility spillover in the crude oil and agricultural commodity markets: A Bayesian
analysis, Energy Economics, Vol. 33, pp. 497–503.
Edwards, S. and Susmel, R. (2001): Volatility dependence and contagion in emerging equity markets, Journal of Development Economics,
Vol. 66, No. 2, pp. 505–532.
Enders, W. (2010): Applied econometric time series, 3rd, Hoboken, N.J, Wiley.
Engle, R.F. et al. (2012): Volatility Spillovers in East Asian Financial Markets: A Mem-Based Approach, Review of Economics and
Statistics, Vol. 94, No. 1, pp. 222–223.
Federal Reserve Bank of New York (Fed) (n.d.): Historical Changes of the Target Federal Funds and Discount Rate.
http://www.newyorkfed.org/markets/statistics/dlyrates/fedrate.html, last accessed 20.09.2012.
Gorton, G. and Rouwenhorst, G. (2006): Facts and Fantasies about Commodity Futures, Financial Analysts Journal, Vol. 62, No. 2.
Koop, G. et al. (1996): Impulse response analysis in nonlinear multivariate models, Journal of Econometrics, Vol. 74, No. 1, pp. 119–147.
Irwin, S.H. and Sanders, D.R. (2010): The Impact of Index and Swap Funds in Commodity Futures Markets: Preliminary Results, OECD Food,
Agriculture and Fisheries Working Papers, No. 27. OECD Publishing. http://dx.doi.org/10.1787/5kmd40wl1t5f-en, last accessed 10.12.2010.
Hamilton, J.D. (1994): Time series analysis, Princeton, NJ,Princeton Univ. Press.
Hong, Y. (2001): A test for volatility spillover with application to exchange rates. Studies in estimation and testing, Journal of
Econometrics, Vol. 103, No. 1-2, pp. 183–224.
Parkinson, M. (1980): The extreme value method for estimating the variance of the rate of return,
Journal of Business, Vol. 53, pp. 61–65.
Pesaran, H.H. and Shin, Y. (1998): Generalized impulse response analysis in linear multivariate models, Economics Letters, Vol. 58, No. 1,
pp. 17–29.
Trujillo-Barrera, A. et al. (2011): Volatility spillovers in the U.S. crude oil corn and ethanol markets, Proceedings of the NCCC-134
Conference on Applied Commodity Price Analysis, Forecasting and Market Risk Management. St. Louis, MO.
http://www.farmdoc.illinois.edu/nccc134/conf_2011/pdf/confp14-11.pdf, last accessed 27.03.2012.
1 February 2013
Stephanie Grosche, ILR, University of Bonn
23