On the correlation between commodity and equity returns

University of Macedonia

School of economic and regional studies

Department of Economics

Bachelor Thesis

ON THE CORRELATION BETWEEN COMMODITY AND EQUITY

RETURNS

Of

Soupionis Georgios

Supervisor: Panagiotidis Theodoros

June 2019

Thessaloniki

1

CONTENTS

ABSTRACT ... 2 INTRODUCTION ... 3 2.LITERATURE REVIEW... 4 3.DATA ... 18 3.1 DATA EXPLANATION ... 19 3.2 GRAPHS ... 20 3.3 DESCRIPTIVE STATISTICS ... 28 4.METHODOLOGY ... 29 4.1 ADF TEST ... 29 4.2 GARCH (1,1) ... 30 4.3 DCC MODEL ... 30 4.4 DCC VS DUMMIES MODEL ... 31 5.RESULTS ... 32

5.1 UNIT ROOT TEST ADF ... 32

5.2 AUTOCORRELATION TEST ... 33 5.3 CORRELATION MATRIX ... 41 5.4 DCC GARCH 1,1 RESULTS ... 42 5.5 DCC MODELS ... 43 5.6 DCC GRAPHS ... 44 5.7 DCC VS DUMMIES ... 54 6.DISCUSSION ... 55 7.CONCLUSION ... 55 REFERENCES ... 57

2

ABSTRACT

In this paper, we investigate the correlation between various commodity and equity markets from different countries during the period of the 2000s. We used weekly data to avoid the problem with the differenced time between the markets. We applied the dynamic conditional correlation methodology and ran an OLS estimation with two shock dummies in order to examine the hypothesis that the correlation must be negative during shock periods, as long as the safe heaven hypothesis. Our main findings supports that the correlation between commodity and equity markets tend to be negative during the two crisis, but after the financial crisis of 2008 the correlation seems to be positive due to the financialization of the economy.

3

INTRODUCTION

The objective of each successful financial investor is to accomplish high returns with low risk. The financial theory suggests, that the best way to do this, is by diversifying portfolios with financial assets that have negative correlation. In the past , the most popular method to create one diversified portfolio was the usage of equities and commodities assets(Gorton and Rouwenhorst, 2006). After the dot-com bubble, at the begging of this century , the investing in commodity markets became more and more popular to the investors, due to the high returns same to those ones from the stock market and also, their negative correlation with the equity markets, which lowered the total risk. Moreover, the safe heaven hypothesis of gold gave an extra motive for investing in commodities such as gold or silver. Although, nowadays their correlation seems to have been changed after the recent financial crisis of 2007-2009 and tend to be positive (Büyükşahin, Haigh and Robe , 2008). All in all, the correlation between commodities and equities doesn’t have a constant pattern, it changes through the years, thus more research is needed in order to determine the real relationship between commodity and equity assets.

In this thesis we try to analyze the correlation between various commodity and equity asset, during a period from January 2000 to December 2018. We chose this period of time in order to focus our analysis to the doc-com bubble and the real estate shock in USA , which led to the worldwide financial crisis of 2007-2009. We used the Dynamic Conditional Correlation (DCC) approach established by Engle (2002) which is an extension of the GARCH models to extracted the correlation and then a simple OLS regression between the DCC and the two shocks in order to examine the correlation in that period.

The paper is organized as follows. In section 2 we present the literature review tables, that summarize the previous analysis, that have been made for commodity and equity assets correlation. Section 3 contains the data that we used. Section 4 is devoted to the methodology of the paper. In section 5 one can observe the results from our empirical analysis and finally section 6 concludes and summarizes our main findings.

4

2.

LITE

R

ATUR

E RE

VIE

W

The m e T hi s arti c le a na ly s es the c o -m ov em e nts be twe en t he U S s toc k ma rk et a nd s ev era l c om mo di ty f utu res be twe en 19 9 8 a n d 20 11 . It c o mp ut es a DCC mo de l fo r ( i) 1 -ho ur, ( ii) 5 -m in ute , ( iii ) 10 -s ec on d, an d (i v ) 1 -s ec on d f req ue nc ies a nd do c um e nts a s y nc hroni z ed s tr uc tura l break , c ha rac ter iz ed by c orr el ati on s th at h av e s ign ifi c an tl y d ep art ed fr om z ero t o p os iti v e terr itori es , s inc e lat e S ep tem be r 2 00 8. A ls o the wr iters s up p ort th e ide a t h at h igh freq ue nc y tr ad in g a nd al g orit hm ic s tr ate gi es ha v e an eff ec t o n t h e b eh av ior of c om mo di ty pr ic es . Dat a T he da ta us e d a re , ti c k -lev el da ta f ro m J an u ary 19 98 to D ec em be r 20 11 an d t he y are ob tai ne d fr om T ho ms on Reu ters Tic k Hi s tory ( T RT H) da tab as e. T he y us ed da ta f or E -mi ni S & P 5 00 f or s toc k ma rk et, lig ht c rud e o il( W T I) (NY M E X ), c orn (C B O T ), whea t ( CB O T ), s ug ar #1 1(IC E U S ), s oy be a ns (CB O T ) an d liv e c att le (CME ) fo r t he c o mm o di ty ma rk et. Th e f req ue nc ies of the d ata are 1 -h ou r, 5 -mi n ute , 10 -s ec on d a n d 1 -s ec on d. M e th o d o log y Dy na mi c Con di ti on al Cor rel at io n (DC C) G A RCH mo d el propos e d b y E ng e l (20 02 ) an d a m ov in g -wi nd ow c orr el ati on s wi th wi n do w wi d ths s et to 5 or 15 ob s erv ati on s for eac h c o ns id ered fr eq ue nc y . A u th o r B ic c he tt i a nd Ma y s tr e (201 3) Titl e T he s y nc hroni z e d a nd lon g -l as ti ng s tr uc tur al c ha ng e on c o mm od ity ma rk ets : E v ide nc e fr om hi g h f req ue nc y d ata (1)

5 The m e T hi s pa pe r i nv es ti g ate s the l ink s b etwe en pr ic e retu rns fo r 2 5 co mm o di ti es a nd s toc k s by pa y in g p arti c u lar att en ti o n t o en ergy r aw ma ter ial s du ri ng t he fi na nc ia l c ri s is of 20 07 -20 08 . T hi s pa pe r i nv es ti g ate s the inc rea s e of th e c orr el ati on be tw ee n eq u ity an d c om mo d ity retu rns , b lam in g f or t ha t s urgi ng i nv es tme nt i n c om mo di ty -r el ate d produc ts . M oreov er, t he au th ors us e v ario us me as ures o f c orr el a ti o n an d as s es s wha t a re th e imp lic at io ns fo r as s et al loc ati on be tw ee n the two m ark ets . T he y a ls o us e a f orec as ti n g m od e l an d ru n a n a lloc a ti o n ex erc is e t o v eri fy th e ir res ul ts . Dat a T he da ta ar e da ily s p ot pr ic e s erie s fo r a larg e s am pl e o f c om mo di ti es s uc h a s en erg y , prec iou s me ta ls , a gric ul tura l, non -ferr o us me tal s , foo d, ol e ag ino us , ex oti c an d Li v es toc k ov er th e p erio d J an ua ry 3, 20 0 1 –N ov em b e r 28 2 01 1 (s o urc e: Data S tr ea m, T h om s on F ina nc ia l) . T he y a ls o u s e d da ta fr om C om mo d ity Res ea rc h B urea u (CRB) i n de x a nd S & P 5 00 ind ex fo r th e s toc k ma rk et. T he da ta us e d a re , we ek ly retu rns G en erate d b y t he Mo rga n S tan ley C ap ita l In ternat io n al gl o ba l E qu ity i nd ex ( M S CI) an d th e S tan da rd & P oo r's G o ldm a n S ac hs c om mo d ity i nd ex (S P G S CI) , s tarti ng fro m t he fi rs t wee k o f J an u ary 19 8 0 un ti l t he l as t we ek of A u gu s t 20 15 . A dd iti on a lly i n t h e forec as ti n g m od e l th ey us e d da ta fr om S P 50 0 , B re nt, B CO M an d W T I. M e th o d o log y Dy na mi c Con di ti on al Cor rel at io n (DC C) G A RCH mo d el propos e d b y E ng e l (20 02 ) T ime -v ary in g B ay es ia n Dy na mi c Con di ti on al Cor rel at io n m od e l b as ed on E ng e l’s ( 20 02 ) DCC me th od , ro lli n g wi nd ows for me an , v ari an c e an d c orr el ati on . A ls o f or t he forec as ti n g e x erc is e the y us ed mo de ls s uc h as Ran d om W a lk , A R mo de l, B ay es ia n V A R an d B ay es ian D CC. T h e c ri teri on s us ed ar e, MS P E , K L IC a nd L S . A u th o r Cr eti , Jo ëts and Mi g no n (20 13 ) Lo mb ardi and Rav az z ol o (20 16 ) Titl e O n t he l ink s be twe en s toc k an d c o mm o di ty ma rk ets ' v ol at ili ty O n t he c orr el ati on be twe en c om mo di ty a nd eq u ity r etu rns : Imp lic at ion s f or po rtfo lio al loc ati on (2) (3)

6 The m e T hi s arti c le s tu di es t he tem p oral v ar iat ion s i n t he c on di ti o na l r etu rn c orr el ati on s b etwe en c om mo di ty f utu res a nd tr ad it ion al as s et c las s es , it a ls o a na ly s es th e l ink be twe en T re as ury -bi ll po rtfo lios an d c om m od ity ma rk ets . T hi s pa pe r i nv es ti g ate s the r et urn l ink s an d v ol at ili ty trans m is s ion be twe en t he S & P 50 0 an d c o mm o di ty pric e ind ic es fo r e ne rgy , f oo d, go ld a n d b ev erages ov er the t urbul en t pe ri od fr om 2000 to 20 1 1. A ls o ana ly s es th e o pti ma l wei g hts an d h ed g e rat ios for c om m od iti es /S & P 50 0 po rtfo lio h ol d in gs . Dat a T he da ta us e d a re , we ek ly c om pris e ret urns an d c los in g pric es on t he ne arby an d s ec on d n ea rby c on tr ac ts o f twen ty -fi v e c o mm od iti es (el ev en ag ri c u ltur al f utu res , f iv e en ergy , f ou r l iv es toc k fu tures an d f iv e m eta l fu tures ) an d thi rte en tra di ti o na l as s et c las s es ( fou r from the US , three f ro m g lo ba l ma rk ets a nd s ix i nte rn at ion al bo n d i nd ic es fr om J P . Mo rg an ). Th e d a ta s et s pa ns th e pe ri od Dec em be r 12 , 19 80 to D ec em be r 27 , 2 00 6. T he da ta us e d a re , da ily c los e retu rns from the S & P 50 0 (be v erage pric e ), w he a t, g o ld pric e i nd ex es an d two c rud e o il be nc hm ark s : Cus hi ng W es t Tex as I nte rm ed iate ( W T I) an d E uro pe ’s B ren t from J an ua ry 3, 20 0 0 t o Dec em be r 31 , 2 01 1 . T he s o urc es of t h e da ta wer e the U S E ne rgy Inf orm ati on A dm in is tr ati on , Int erna ti o na l G rai ns Cou nc il, an d S & P 5 00 webs ites . M e th o d o log y G A RCH ( 1, 1) ( Hans e n an d Lu n de 2 00 5), de v el op e d b y B ol lers lev (19 86 ), D CC G A RCH (1, 1) mo de l pro po s ed by E ng el ( 20 0 2) an d ma x im um l ik e liho o d proc ed ure us in g t h e al g orit h m o f B F G S . V A R ( 1) – G A RCH ( 1, 1) propos e d b y L ing an d Mc A Le er ( 20 0 3) wh ic h inc lu de s th e m ul ti v ari ate CCC -G A RCH of B ol lers lev ( 19 9 0 ).A ls o the y us ed qu as i -ma x im um l ik e liho o d (Q ML ) to es ti ma te the pa ram ete rs of t he mo de l. A u th o r Chon g a n d Mi ffre (20 08 ) Me ns i, B el jid, B ou ba k er and Ma na gi (20 13 ) Titl e Cond iti on a l Retu rn Cor rel at io ns be twe en Comm od ity F utu res an d T rad it ion al A s s ets Cor rel at io ns an d v o lat ili ty s pi llov ers ac ros s c om mo di ty a nd s toc k ma rk ets : Li nk ing en erg ies , fo od , an d go ld (4) (5)

7 The m e T hi s pa pe r a na ly s es th e c o -mo v em en t be tw ee n c om mo di ty a nd eq u ity ma rk ets fo r a pe ri od of 1 7 y ea rs by us ing Dy na m ic Cond iti on a l C orr el a ti o n a nd rec urs iv e t ec hn iqu es . They fi nd th a t th e c orr el a ti o n a nd the co -m ov em e nt b etwe en the r et urns on c o mm od ity and eq u ity i nv es tme nts du ri n g p erio ds of ex tr e me retu rns ha v e no t c h an g ed s ign ifi c an tly thr ou g h t h is pe ri o d o f ti me . T hi s pa pe r ex a mi ne s th e rel at ion s h ip be tw ee n th e en ergy a n d eq u ity ma rk ets by es ti m ati ng v o lat ili ty imp ul s e res p on s e f un c ti on s fr om a mu lti v ar iate B E K K mo de l. T he ir an a ly s is s ug ge s ts th at th e e ne rgy ind ex i s ge ne ra lly a po or h ed g in g i ns tr um e nt. Dat a T he da ta us e d a re , d a ily , week ly a nd mo nt hl y r et urns fr om J an ua ry 19 91 t hroug h Ma y 20 08 , fr om S & P 5 00 I n de x (S tan da rd a nd P o or' s ), G S CI Ind ex ( G ol dm an S ac hs Comm od ity In d ex ), DJ IA I n de x an d DJ -A IG In d ex ( Dow -J on es 's ). T he s ou rc e o f th e da ta was B lo om b erg. T he da ta us e d f or thi s pa p e r are week ly r etu rns fro m S & P 50 0 an d G ol d ma n S ac k s E ne rgy E x c es s Retu rn I nd e x , fr om J an ua ry 1 , 1 98 5 t o A p ri l 24 , 2 01 3 . T he s o urc e o f the da ta was Tho ms on ’s Data S tr ea m. A ls o t he y us e d we ek ly retu rn i n t h e Tr ad e W ei gh te d U.S . do llar ind ex : M aj or c urr en c ies from the S t. Lo ui s F red da tab as e . M e th o d o log y Rol lin g h is tori c a l c orr el ati on , ex po n en ti a l s mo oth er, an d DCC by log -l ik e lih oo d fo r m ea n -rev erti ng mo de l es ti ma ti o n p ro po s e d b y E ng el ( 2 00 2). A ug me n ted Di c k ey F ul ler un it ro ot tes t an d rec urs iv e c oi nt eg rat ion tec hn iqu es (J oh an s en , 19 98 , 19 91 ; J o ha ns en an d J us e lius ,19 9 0) V A R mo de l wi th o ne lag , mu lti v aria te GAR CH mo de l ( B E K K ’s s pe c ifi c at io n), he dg e rati o f or th e S & P 50 0 an d ea c h c o mm o di ty ind ex us in g t h e es ti ma ted c o nd iti on a l c o v aria nc es a nd v ari an c e imp ul s e res p on s e fun c ti on s pro po s ed by Hafn er an d Her war tz ’s (20 06 ). A u th o r B üy ük şah in , Hai gh an d Robe (20 08 ) O ls on , V iv ian a nd W oh ar (20 14 ) Titl e Comm od iti es an d E qu it ies : A “Ma rk et of O ne ”? T he r el ati on s h ip be twe en en ergy a nd eq u ity ma rk ets : ev ide nc e f rom v ol at ili ty imp ul s e res po ns e fun c ti on s (6 ) (7 )

8 The m e T hi s pa pe r ex a mi ne s th e lon g -r un tre nd s an d t h e s ho rt -r un fl uc tua ti o ns of t he c om mo di ty -eq u ity c orr el ati on , an d i t do es s o to ind ic e f ro m 4 5 eq ui ty ma rk ets . They f ind th a t c om mo di ty i nv es tm en ts ha v e m ore d iv ers ifi c a ti o n be ne fi ts i n t he l on g run wor ldw id e. T hi s arti c le ai ms at inv es ti ga ti ng the ti me v ary ing r el ati on s h ip be twe en Is lam ic eq ui ty an d c om mo di ty r et urns i n order to ex am in e h ow c om bi na ti on of Is lam ic eq u iti es an d c om m od iti es c on tr ibu te to t h e b en ef its of po rtfo lio i nv es tors an d ma na ge r. Dat a T he da ta us e d a re , d a ily pri c es fr om S & P 50 0 Go ldm an S a c k s Comm od ity In d ex ( G S CI) a nd Mo rga n S tan ley Ca pi tal Int erna ti o na l A ll Co un tr y W orld Inv es tab le I nd ex ( M S CI A C W I) , fr om J an ua ry 1, 2 00 0 t o Dec em b er 31, 20 10 . T he s ou rc e o f th e d ata was Data S tr ea m. T he da ta us e d a re da ily s p ot pric e s erie s ex tr ac ted fro m Data S tr ea m f or t he di ff eren t c om mo di ti es c ov eri ng v ari o us s ec tors ov er the J an ua ry 3, 20 01 Ma rc h 28 , 20 1 3 p er iod . T he y i nv es ti ga te 5 di ffe ren t c om mo di ty gr ou ps ; en ergy , prec iou s me ta ls , a gric ul tura l, no nf err ou s me tal s an d s o ft’ s group. A n ag grega te c om mo di ty pr ic e i nd ex , the DJ S po t c om m od ity In de x , i s al s o c on s ide re d. R eg ard in g t h e eq u ity ma rk et, th ey us ed D ow J on es Is la mi c i nd ex . M e th o d o log y DCC mo de l pro po s ed by E ng el ( 20 0 2), l og lik e lih oo d r ati o t es t, mu lt ip le -bre ak un it ro ot tes t fo r tre nd s tat io na ry wi th the us e o f M o nte Car lo S imu lat ion s . Dy na mi c Con di ti on al Cor rel at io ns G A RCH mo de l pr op os e d b y E ng le an d S h ep p ard (20 01 ) a nd E n gl e (20 0 2) . A u th o r Li , Z h an g and Du (20 11 ) K ha n, K ab ir , B as ha r an d Ma s ih (20 15 ) Titl e Cor rel at io n i n c om mo di ty fut ures a nd eq u ity ma rk ets aroun d t h e wor ld: Long -r u n t ren d an d s h ort -r un fl uc tu ati on T ime v ary ing c orr el ati on be twe en Is lam ic eq u ity an d c om mo di ty retu rns : imp lic at io ns fo r po rtfo lio di v ers ifi c ati on (8 ) (9)

9 The m e T hi s pa pe r m od e ls ti me -v ary ing c orr el ati on s be twee n c om mo di ty a nd s toc k m ark ets to un c ov er t he dy na m ic na ture of c orr e lat ion s d urin g the f ina nc ia liz at io n of c om mo di ty m ark ets an d i n t he aft erm ath of th e rec en t fi na nc ia l c ri s is . Th ei r r es u lt s s ho w th at c om mo d it ies wor k wel l fo r d iv ers if ic ati on i n t he c al m p erio ds du e t o the ir he tero ge n eo us s tr uc ture . Dat a T he da ta us e d a re , da ily pri c e s erie s of a gric u ltu ra l c om m od ity an d prec io us me tal s u b -i n di c es of S & P G S CI an d S & P 50 0 ind ex fr om J a nu ar y 4, 19 90 to Dec em b er 20, 20 12 .Th e d a ta are ob ta ine d f ro m Glo ba l Fi na n c ial Data . W e ek ly r etu rn r ate s c al c ul a ted by l o g d iffe ren c ing T hu rs da y 12 c los in g p ri c es are us ed i n e s ti ma ti o ns . The ex tr em e ret urns wh ic h a re ou ts id e t h e f o ur s tan da rd de v iat io ns c on fi de nc e i nte rv al aroun d t h e m e an are r ep la c ed wi th the ir bo un d ary v al u es . M e th o d o log y S mo oth tr an s it io n c on di ti o na l c orr el a ti o n (S T CC) an d d o ub le s mo oth tr an s it io n c on di ti o na l c orr el a ti o n (DS T CC) mo de ls of S ilv en n oi ne n an d T eräs vi rta (20 0 5, 2009) . A u th o r Ö ztek and Öca l ( 20 1 7 ) Titl e F ina nc ia l c ri s es an d the n atu re of c orr el ati on be tw ee n c om mo di ty a nd s toc k ma rk ets (1 0 )

10 The m e T hi s pa pe r us es V A R MA -A G A RC H a nd DCC AG A RCH mo de ls to mo de l v ol ati liti es an d c on di ti o na l c orr el a ti o ns be twe en em erg in g ma rk et s toc k pri c es , c op pe r pr ic es , o il pr ic es an d w he a t p ri c es . A ls o i t us es th e D CC m od e l t o c al c ul a te the h ed g e rati os a nd t he op ti ma l po rtfo lio we igh ts i n order to an al y z e t he l ink be twe en c o mm o di ty an d eq u ity ma rk et. T hi s pa pe r i nv es ti g ate s the c orr el ati on be tw ee n eq u ity an d c om mo d ity retu rns no t on ly i n US ma rk ets , b u t a ls o o n th e em erg ing ma rk ets o f A s ia, E uro pe an d La ti n A me ri c a. A dd iti on a lly , i t pres en ts th e d iffe re nc es be twe en em erg in g a nd de v el op e d m ark ets c o -mo v em en ts w ith c om mo di ti es . Dat a T he da ta us e d a re , d a ily da ta fr om J an ua ry 3, 2 0 0 0 t o J un e 2 9, 2 01 2 an d th ey are ob ta ine d f ro m Da ta S tr ea m Int erna ti o na l. For t he s toc k ma rk et, t he y us ed th e M S C I E me rg in g M ark ets In d ex an d for the c o mm od ity ma rk et the y us ed W es t Tex as Int erm ed iate Cr ud e oi l, CO ME X c op p er c on tr ac t pric es an d I nt ernat ion al G rai ns Co un c il wh ea t pric e in d ex . T he da ta us e d a re , d a ily da ta fr om J an ua ry 20 06 t o J an ua ry 20 16 an d the y ar e ob ta ine d f ro m Da taS tr ea m . T he c om mo d ity i nd ic es us e d are S & P G S CI In de x a nd t hree s u b -ind ic es : S & P G S CI E n ergy , A gric u lture , a n d P rec iou s Me ta ls . F or th e s toc k ma rk et, the y u s ed th e M S CI E me rg in g m ark ets i n de x an d al s o t hree re gi o na l ind ic es : MS CI E me rgi ng Ma rk ets E urop e, A s ia an d La ti n A me ri c a. M e th o d o log y V A R MA -A G A RC H mo de l o f M c A le er et al . (20 09 ) a nd t he D CC -A G A RC H mo de l of E ng le (2 00 2). A ls o, th ey c on s tr uc ted h ed g e rati os ( K ro ne r a nd S ul tan 19 93 ) an d op ti ma l p ortfol io we igh ts s ub jec t t o a no s h orti n g c on s tr ai n t (K ron er an d Ng, 1 99 8) f or the po rtfo lios . Dy na mi c Con di ti on al Cor rel at io n (DC C) G A RCH mo d el propos e d b y E ng e l (20 02 ) A u th o r S ad ors k y (20 14 ) D e B oy ri e and Pav lov a (20 16 ) Titl e Mo de ling v ol at ili ty an d c orr el ati on s be twe en em erg ing ma rk et s toc k pric es an d t h e pric es of c op pe r, o il an d whea t Li nk a ge s be twe en E q ui ty an d C om mo d it y Ma rk ets : A re E me rg in g Ma rk ets Di ffe re nt? (11 ) (12 )

11 The m e T he pu rpo s e o f th is arti c le is to an al y z e t h e rel at ion s h ip be tw ee n c om mo di ty a nd eq u ity ma rk ets i n th e A s ian Ma rk et b efo re an d du ri ng the f ina nc ia l c ri s is of 2007 -20 0 9. A ls o, it inv es ti ga tes t he s p ill ov er eff ec ts from eq ui ty ma rk ets to c o mm o di ty fut ures . T hi s arti c le a na ly s es th e s toc k -c om mo di ty c orr el ati on du ri ng 1 96 0 to 20 12 an d proofs t he ex is ten c e of a bu s ine s s c y c le c om po ne nt i n th e c orr el ati on . A ls o, it rep orts th a t th e bu s ine s s c y c le e ffe c t c an ex pl ai n the s p ik es i n the s toc k -c om mo di ty c orr el ati on i n the e arly 19 80 s a nd t he late 20 00 s . Dat a T he da ta us e d a re , d a ily da ta fr om J ul y 5 , 2 0 05 to Dec em b er 14, 20 11 a nd th ey are ob tai ne d from B loo mb e rg Data b as e. A ls o, the y s p lit t he s am pl e i n t w o p er iod s , on e be fore the c ri s is of 2 00 7 -20 09 an d o n e d urin g t h e c ri s is . Fo r th e s toc k m ark et the y us ed i n di c es f or 14 A s ian c ou ntr ies a nd f or the c om mo di ty m ark et th ey us e d go ld a n d c rud e oi l. T he da ta us e d a re , d a ily an d week ly d ata fr om J an ua ry 19 60 to D ec em be r 20 12 an d the y are ob ta in ed fro m Comm od ity Res ea rc h B urea u (CR B ), B loo mb erg, Lo nd on Me tal s E x c ha ng e, U.S . D ep art me n t o f La bo r - B urea u o f La bo r S tat is ti c s an d B oa rd o f G ov ernor s of the Fe de ra l Res erv e S y s te m. T he y us ed t he D ow J o ne s -UB S Com mo di ty I nd ex a nd the S & P -G S CI ,S & P 50 0 ind ex , rea l G D P gro wth , inf lat ion an d th e de fau lt S prea d f or U S A . M e th o d o log y B E K K -G A RCH ( 1, 1) mo de l pr op os e d b y E ng le an d K ro ne r (19 95 ). R ealiz ed c o rrela tio n and co vari an ce p ro p o sed by An d erse n e t al. ( 2 0 0 1 ) an d Dy na mi c Cond iti on a l C orr el a ti o n (DCC) G A RCH mo d el propos e d b y E ng e l (20 02 ). A u th o r T hu rai s a m y , S ha rm a and A h me d (20 13 ) B ha rd waj and Duns by (20 12 ) Titl e T he r el ati on s h ip be twe en A s ia n eq u ity an d c om mo di ty fut ures m ark ets T he B us in es s Cy c le a n d t h e Cor rel at io n be twe en S toc k s and Comm od iti es (1 3 ) (14 )

12 The m e T hi s pa pe r i nv es ti g ate s the l ink be tw ee n c om mo di ty a nd eq u ity ma rk ets i n S a ud i A rab ia and draws i mp lic at ion s for portf ol io ri s k ma na ge me nt . T hi s pa pe r fo c us es on c om mo di ty m ark ets an d the ir r e lat io n wi th Is lam ic s toc k ma rk ets du ri n g t h e rec en t fi na nc ia l c ri s is a nd ex am ine s th e a ltern ati v e inv es tme nt o pp ortun iti es for the i nv es tors d urin g tha t pe ri o d. Dat a T he da ta us e d a re , d a ily da ta fr om J un e 1 , 20 0 5 t o A ug u s t 13 , 2 01 4 an d t h ey are ob ta ine d f ro m M organ S tan ley C ap ita l In ternat io n al . F or the s toc k m ark et th ey c ho s e Tada wu l A ll-S ha re Ind ex ( T A S I) fo r S a ud i A rab ia. T h ey us ed W es t T ex as In term ed iate ( W T I) c rud e o il, prec io us me tal s (go ld an d s ilv er) a nd c ere al (whea t, c or n a n d ri c e) f or th e c om mo di ty m ark et. T he da ta us e d a re , da ily s p ot pric e s erie s ex tr ac ted fro m Data S tr ea m f or t he di ff eren t s ec tor c om mo di ti es ov er t h e J an ua ry 3, 20 0 1 - M arc h 2 8 , 20 13 pe ri od . They us e d 5 di ff erent c o mm o di ti es s ec tors : e ne rgy , p rec iou s me ta ls , a gr ic ul tural , n o n -ferr ou s me tal s a nd s o ft’ s s po t ind ex . A s for t he s toc k ma rk et, t he s e lec ted th e Do w J on es Is la mi c In de x . M e th o d o log y T he b iv aria te DCC – F IA P A RCH mo de l, whi c h inc lud es t he F IG A RCH ( p, d, q) mo de l o f B ai lli e et a l. (19 96 ) a nd t he Dy n am ic Cond iti on a l C orr el a ti o n (DCC) G A RCH mo d el propos e d b y E ng e l (20 02 ). A ls o, the y us ed a m o di fi e d v ers ion of the I nc lan an d Ti ao (19 94 ) IC S S te s t de v el op e d b y S an s o et al . (2 00 4) to d ete c t t he s tr uc tural bre ak s i n t he un c on d it ion al v o lat ili ty . Dy na mi c Con di ti on al Cor rel at io n (DC C) G A RCH mo d el propos e d b y E ng e l (20 02 ) A u th o r Me ns i, Hamm ou d e h a n d K an g (20 15 ) K ha n an d Ma s ih (20 14 ) Titl e P rec io us me ta ls , c erea l, oi l an d s toc k ma rk et link a ge s an d po rtfo lio ri s k ma na ge me nt : E v ide nc e f ro m S au di A rab ia Cor rel at io n be twe en Is lam ic s toc k an d Comm od ity ma rk ets : A n inv es ti ga ti on into the i m pa c t of fi na nc ial c ri s is an d fi na nc ia liz at io n of c om mo d ity ma rk ets (1 5 ) (16 )

13 The m e T hi s pa pe r i nv es ti g ate s the l ink s b etwe en t he S & P 50 0 i nd ex an d S & P c om mo di ti es i nd ex es ba s ed on t he DD C G A RCH mo d el , in order to ex am in e t h e h ed g e hy po th es is be twee n them. Thi s arti c le a na ly s es th e link s b etwe en Is lam ic c ap ita l m ark ets an d c om mo di ti es . Th ei r s tu dy foc us es on ga s an d c rud e o il an d the ir rel at ion s h ip wi th t h e Q E A l R ay an Is lam ic In de x . Dat a T he da ta us e d, ar e d a ily retu rns of t he S & P 5 00 , S & P -G S CI, S & P -G S CI A gric u lture , S & P G S CI E ne rgy , S & P -G S C I In du s tr ial Me ta ls , S & P -G S CI Li v es toc k an d S & P -G S CI P rec iou s Me ta ls , from J an u ary 19 9 9 to J un e 2 01 7. The da ta we re prov ide d b y S & P Do w J on e s Ind ex es . T he da ta us e d a re , d a ily s p ot pric e f rom 15 Ma rc h 20 1 1 t o 25 Dec e mb er 201 4 a nd t he y are ob tai ne d from Q ata r E x c ha ng e . Th e s erie s as s oc iate d wi th t w o s tr ate gi c c om mo di ti es , n am el y c rud e oi l an d g as . T he s toc k i nd e x c on s ide re d for t hi s s tu dy i s the Q E A l Ray a n I s lam ic Ind ex . M e th o d o log y Dy na mi c Con di ti on al Cor rel at io n (DC C) G A RCH mo d el propos e d b y E ng e l (20 02 ) Dy na mi c Con di ti on al Cor rel at io n (DC C) G A RCH mo d el propos e d b y E ng e l (20 02 ) A u th o r Daum as , A iub e an d B a idy a (20 17 ) Cheb bi an d Der ba li ( 2 01 5) Titl e Hedg ing s toc k s throug h c om mo di ty ind ex es : a DCC -G A RCH ap pro ac h T he dy na mi c c orr el ati on be twe en en ergy c om mo di ti es an d I s lam ic s toc k ma rk et: A na ly s is an d forec as ti n g (1 7 ) (18 )

14 The m e T hi s pa pe r, u s ing J ap an es e ma rk et da ta, pres en ts th at th e c orr el ati on be tw ee n eq u ity ma rk ets a nd c om mo di ty m ark et s ha v e c ha ng e d s ig ni fi c an tl y throug h t h e fi n a nc ia l c ri s is i n A utu mn of 2 00 8 an d an a ly s es th e res u lts of th is c ha n ge . T hi s pa pe r, a na ly z es ti me -v ary ing c orr el ati on s be twe en c o mm o di ty ma rk ets an d S & P 50 0 ind ex , em pl oy ing a rec en t a n d n ov el tec hn iqu e : a s y mm etri c dy na m ic c on d iti on a l c orr el ati on ( A DCC) mo de l, i n order t o ex am ine t he i mp ac ts of fi na nc ia l c ri s es on th e c on di ti o na l c orr el a ti o ns . Dat a T he da ta us e d a re , d a ily pric es from Ma y 3 1, 1 98 6 to F eb rua ry 28 , 20 07 an d the y are ob tai ne d from T ok y o Comm od ity E x c ha ng e (T O CO M). T ok y o Co mm od ity E x c ha ng e (T O CO M) us ed as a n ind ic ato r of th e c om mo d ity ma rk et a nd t he y c om pa re d it wi th the T ok y o S toc k E x c ha ng e S toc k P ri c e Ind e x (T O P IX ). T he da ta us e d a re , we ek ly da ta of c om mo di ty i nd ex (Dow J on es -U B S ), i ts s ub -ind ic es an d S & P 50 0 s toc k ind ex from J a nu ary 3, 19 92 to Dec e mb er 27 , 20 1 3 a nd the y are ob ta in ed fro m B lo om be rg. M e th o d o log y A s y mm etr ic dy na m ic c on di ti o na l c orr el ati on (A DCC) m od e l propos e d b y Capp ie llo et al . (20 06 ). A DCC mo de l is a ge ne ral iz ati on o f DCC -G A RCH mo de l o f E ng le (20 02 ). A u th o r Y am ori (20 10 ) Demi ral ay and Ul us oy ( 20 14 ) Titl e Co -mo v e me nt be twe en Comm od ity Ma rk et a nd E qu ity M ark et: Does Comm od ity Ma rk et Chan ge ? Li nk s B etwe en Comm od ity F utu res A nd S toc k Ma rk et: Di v ers ifi c a ti o n B en efi ts , F ina nc ia liz at ion a nd F ina nc ia l Cr is es (1 9 ) (20 )

15 The m e T hi s pa pe r ex a mi ne s th e pric e v ol ati lity an d he dg in g b e ha v ior of c om mo di ty f utu res ind ic es an d s toc k ma rk et ind ic es . They i nv es ti ga te the w ee k ly h ed g in g s tr ate gi es ge ne rate d b y retu rn -ba s e d a n d ran ge -ba s ed as y mm etr ic dy na m ic c on d iti on a l c orr el ati on ( DC C) p roc es s es . T hi s pa pe r ex a mi ne s th e rel at ion s h ip be tw ee n c rud e o il pric es , e q ui ty pric es , a n d c om mo d ity ma rk ets by ad di ng n ew mo d ifi c at io n, s uc h as ex pa nd in g t h e m od e l to inc lu de t he eq u ity pri c e for an oi l pro du c in g f ir m, Cono c o P h ill ips , wh ic h am e liorates o mi tted v aria b le b ias a nd es ti ma ti n g t h e e x p an d ed mo de l us ing th e K a lm an F ilter, wh ic h red uc es un c ertai nty as s oc ia ted wi th O LS es ti m ate s fro m rol ling wi nd ows . Dat a T he da ta us e d a re , we ek ly da ta fr om J a nu ary 1, 19 93 t o Dec em b er 25, 20 09 a nd th ey are ob ta ine d from Data S tr ea m. T he da ta c on s is ted of S & P 50 0, DA X , a nd N ik k ei 22 5, two p rec iou s m eta l c om mo di ti es pr ic es ( go ld an d s ilv er) , a n d o ne c rud e oi l pric e (W T I) . T he da ta us e d a re , da ily pric es from O c to be r 7 , 19 9 6 to A ug us t 21 , 20 1 1, for th e W es t Tex as I nte rm ed iate (W T I) on t he N ew Y ork Me rc an ti le E x c ha ng e, th e pric e o f Con oc oP hi lli ps s toc k on th e New Y ork S to c k E x c ha ng e (CO P ), an d the S tan da rd an d P oo r's 50 0 i nd ex ( S P 50 0) an d t he y are ob tai ne d fr om the Th om s on -R eu ters Tic k Hi s tory da tab as e . M e th o d o log y Retu rn -b as ed DC C -G A RCH a nd r a ng e -ba s ed D CC -CA R R mo de ls . O LS fo r es ti m ati ng mo de ls from r ol ling wi nd ows an d th e K al ma n Fi lter t o es ti ma te ti m e -v ary ing c oe ffi c ien ts fro m t h e en ti re s am p le . A u th o r La h ian i a nd G ue s mi (20 14 ) K ol od z iej , K au fma nn , K ul ati lak a , B ic c he tt i and Ma y s tr e (20 14 ) Titl e Comm od ity P ri c e Cor rel at io n a n d T ime V ary in g He dg e Rati os Cr ud e o il: Comm od ity or fi na nc ia l as s et? (21 ) (22 )

16 The m e T hi s pa pe r i nv es ti g ate s the r o le o f g o ld an d o the r prec iou s me ta ls r el ati v e to v ol ati lity of s toc k ma rk ets an d V IX as a he dg e a nd s af e h av en us ing da ta f rom th e US s toc k ma rk et. A ls o, the y ex am ine a p ortf ol io an a ly s is wi th i n teres ti ng res ul ts . T hi s arti c le i nv es ti ga te s the a uto -c orr el ati on s a nd c ros s -c or rel ati on s of the v ol at ili ty ti me s eri es i n the B raz ili a n s toc k an d c om mo di ty m ark et, us ing the r ec en tl y i ntrod uc ed Detrend ed Cr os s -Cor rel at io n A na ly s is . Dat a T he da ta us e d a re , da ily c los ing s po t pric es fo r g ol d, s ilv er , p lat inu m, S & P 5 00 Ind ex a nd V IX . A ll th e da ta us ed i n t h e p ap er w as ob ta ine d fr o m B lo om b erg. T he da ta c ov ers a pe ri od fr om Nov e mb er 30, 19 9 5 t o Nov em b er 30 , 2 01 0 . T he da ta us e d, ar e d a ily pric es from A u gu s t 1 0, 2 00 0 to A pril 30 , 20 08 ,f or c ott o n ,s ug ar, s oy be an , c o ffe e , catt le, B raz il’ s T E L E CO M S .A . ( brto4), Us ina s S id d e MG S .A . (us im5 ), G ás de S ão P a ul o (c ga s5) an d I tau s Inv es ti m en tos S .A . (i ts a4 ).The s o urc es of t he da ta wer e c e pe a es al q u s p an d B O V E S P A . M e th o d o log y R eg res s io n m o de l propos e d by B au r an d Mc Der m ott (20 10 ) Detrend ed F luc tu ati on A na ly s is ( DF A ) int rod uc ed by P e ng (19 94 ) a nd D etren de d Cr os s -Cor rel ati on A na ly s is ( DCCA) introd uc ed by P o do bn ik an d S ta nl ey ( 20 08 ). A u th o r Hood an d Ma lik (20 13 ) S iq ue ir a J r, T .S tos ic , B ej an , B .S tos ic (20 10 ) Titl e Is G ol d t he B es t Hedg e a n d a S afe Ha v en Unde r Chan gi n g S toc k Ma rk et V ol ati lity Cor rel at io ns an d c ros s -c orr el a ti on s i n the B raz ili a n ag raria n c om mo di ti es an d s toc k s (23 ) (24 )

17 The m e T hi s arti c le inv es ti ga tes the p ote nti al f or portf ol io di v ers ifi c ati on b en e fi ts of c om mo d ity fu tures i n the I nd ian c o nte x t. Mo reo v er, t he y c al c ul a ted the DCC c orr el ati on be tw ee n c om mo di ti es a nd s toc k s , b on d a nd T rea s ure b ill s be fore an d du ri ng t he f ina nc ia l c ri s is of 20 07 -20 09 an d fou n d i nte res ti ng res ul ts . Dat a T he da ta us e d a re , da ily da ta f ro m J un e 7, 20 05 t o S ep tem b er 30 , 2 01 1 an d t h ey are ob ta ine d f ro m M C X Ind ia, C lea ri ng Cor po rat ion of I nd ia Li m ite d a nd t he Nati on a l S toc k E x c ha ng e o f Ind ia. T he c om mo d it ies us ed are, CO MD E X , a n ind ex r ep res en ti n g a ll c om mo di ti es , MCX ME T A L, MCX E N E RG Y an d MCX A G RI. A s a as prox y of th e In d ia n s toc k ma rk et, th ey s el ec ted th e S & P CNX Ni fty In de x . A ls o t he y us ed th e l on g -t erm bo nd i n de x a nd T rea s ury bi ll ind ex . T he y s pl it th e s am pl e into two pe ri od s o ne , be fore the c ri s is an d o ne du ri ng t he c ri s is of 2007 -20 0 9. M e th o d o log y Dy na mi c Cond iti on a l Cor rel at io n (DCC) G A RCH mo de l pr op os e d by E ng el ( 20 0 2) A u th o r La ge s h , K as im an d P au l (20 14 ) Titl e Comm od ity Futures Ind ic es and T rad it ion al A s s et Ma rk ets i n In d ia: DCC Ev ide nc e f or P ortfo lio Di v ers ifi c a ti o n B en efi ts (25 )

18

3.DATA

We analyzed the correlation between different stock markets and several commodity markets between January 3 2000 and December 28 2018, using weekly data from Quantl, Fred St Louis, London Stock Exchange, Gold World Council and Yahoo Finance. More specifically, we used weekly closing prices for S&P 500 INDEX (USA), DAX INDEX (GERMANY), NIKKEI 225 INDEX (JAPAN) and FTSE 100 INDEX (UK) and weekly spot prices for WTI OIL, BRENT CRUDE OIL, GOLD and SILVER.

We used the logarithmic values for each variable, in order to achieve better economic results and then calculate the log returns.

Rt = Log (Xt) – Log (Xt-1)

where

Xt: variables price at time t

Xt-1: variables price at time t-1

DATA SUMMARY

Firstly, we plotted the log variables and their log returns and then we presented the descriptive statistics for each log return.

19

3.1 DATA EXPLANATION TABLE 1

VARIABLES

DEFINITION

SOURCE

LOGBRENT LOG (BRENT_PRICES) Fred St Louis

BRENT_R

LOG(BRENT_PRICES)-LOG(BRENT_PRICES(-1))

LOGDAX LOG (DAX_PRICES) Yahoo Finance

DAX_R

LOG(DAX_PRICES)-LOG(DAX_PRICES(-1))

LOGFTSE100 LOG (FTSE100_PRICES) London Stock

Exchange

FTSE100_R

LOG(FTSE100_PRICES)-LOG(FTSE100_PRICES(-1))

LOGGOLD LOG (GOLD_PRICES) Gold World Council

GOLD_R

LOG(GOLD_PRICES)-LOG(GOLD_PRICES(-1))

LOGNIKKEI LOG (NIKKEI_PRICES) Yahoo Finance

NIKKEI_R

LOG(NIKKEI_PRICES)-LOG(NIKKEI_PRICES(-1))

LOGSILVER LOG (SILVER_PRICES) Quantl

SILVER_R

LOG(SILVER_PRICES)-LOG(SILVER_PRICES(-1))

LOGSP500 LOG (SP500_PRICES) Yahoo Finance

SP500_R

LOG(SP500_PRICES)-LOG(SP500_PRICES(-1))

LOGWTI LOG (WTI_PRICES) Fred St Louis

WTI_R

20 3.2 GRAPHS

BRENT

2.8 3.2 3.6 4.0 4.4 4.8 5.2 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 LOGBRENT Figure 1 -.3 -.2 -.1 .0 .1 .2 .3 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 BRENT_R Figure 2

Figure 1 illustrates the natural logarithm of BRENT line graph and we can see that the index has a trend. Figure 2 represents the BRENT’s returns.

21

DAX

7.6 8.0 8.4 8.8 9.2 9.6 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 LOGDAX Figure 3 -.16 -.12 -.08 -.04 .00 .04 .08 .12 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 DAX_R Figure 4

Figure 3 illustrates the natural logarithm of DAX line graph and we can see that the index has a trend. Figure 4 represents the DAX’s returns.

22

FTSE100

8.1 8.2 8.3 8.4 8.5 8.6 8.7 8.8 8.9 9.0 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 LOGFTSE100 Figure 5 -.12 -.08 -.04 .00 .04 .08 .12 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 FTSE100_R Figure 6

Figure 5 illustrates the natural logarithm of FTSE100 line graph and we can see that the index has a trend. Figure 6 represents the FTSE100’s returns.

23

GOLD

5.2 5.6 6.0 6.4 6.8 7.2 7.6 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 LOGGOLD Figure 7 -.12 -.08 -.04 .00 .04 .08 .12 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 GOLD_R Figure 8

Figure 7 illustrates the natural logarithm of GOLD line graph and we can see that the index has a trend. Figure 8 represents the GOLD’s returns.

24

NIKKEI

8.8 9.0 9.2 9.4 9.6 9.8 10.0 10.2 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 LOGNIKKEI Figure 9 -.20 -.15 -.10 -.05 .00 .05 .10 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 NIKKEI_R Figure 10

Figure 9 illustrates the natural logarithm of NIKKEI line graph and we can see that the index has a trend. Figure 10 represents the NIKKEI’s returns.

25

SILVER

1.0 1.5 2.0 2.5 3.0 3.5 4.0 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 LOGSILVER Figure 11 -.25 -.20 -.15 -.10 -.05 .00 .05 .10 .15 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 SILVER_R Figure 12

Figure 11 illustrates the natural logarithm of SILVER line graph and we can see that the index has a trend. Figure 12 represents the SILVER’s returns.

26

SP500

6.4 6.6 6.8 7.0 7.2 7.4 7.6 7.8 8.0 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 LOGSP500 Figure 13 -.20 -.15 -.10 -.05 .00 .05 .10 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 SP500_R Figure 14

Figure 13 illustrates the natural logarithm of SP500 line graph and we can see that the index has a trend. Figure 14 represents the SP500’s returns.

27

WTI

2.8 3.2 3.6 4.0 4.4 4.8 5.2 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 LOGWTI Figure 15 -.2 -.1 .0 .1 .2 .3 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 WTI_R Figure 16

Figure 15 illustrates the natural logarithm of WTI line graph and we can see that the index has a trend. Figure 16 represents the WTI’s returns.

28

3.3 DESCRIPTIVE STATISTICS Table 2

The above table summarizes the descriptive statistics for each variable. In order to have normal distribution the skewness must be about zero and the kurtosis must be about 3. As we can see the skewness is near zero in most of the variables, but the kurtosis is bigger than 3 in all of them, so there is not enough strong evidence for normal distribution.

29

4.METHODOLOGY

The basic methodology that we use in our analysis is a multivariate GARCH model with time varying conditional correlation. To begin with we applied the unit root test (ADF) in the log returns and we tested for autocorrelation, by using correlogram. The results from the previous analysis can be seen in the results section. Then, we applied the DCC model proposed by Engle (2002). The DCC model is estimated in two steps. In the first step we estimate the GARCH parameters and in the second we estimate the correlation. For the estimation we used the maximum likelihood method and tried to maximize the likelihood function. Also, in the estimation of the quasicorrelation matrix Qt we

utilized a mean-reverting model which is also a GARCH (1, 1) model Engle (2002). Finally, with the results from the DCC correlations we ran an estimation with two dummy variables, bubble dummy and estate dummy. These dummies represented the two shocks occurred during the period we investigated. Bubble dummy represented the dot.com shock in 2000-2002 and estate dummy represented the financial crisis in 2007-2009.

4.1 ADF TEST

Augmented Dickey Fuller tests the null hypothesis that a unit root is present in a time series sample. The alternative hypothesis is stationarity but in our analysis is trend-stationarity. The test continues as follows :

Δy

t

= α + βt + γy

t-1

+ δ

1

Δy

t-1

+ … + δ

p-1

Δy

t-p+1

+ ε

t

,

where α is a constant, β the coefficient on a time trend and p the lag order of the autoregressive process. The unit root test is carried out under the null hypothesis γ = 0 against the alternative hypothesis of γ < 0 . εt is the error

term. We run the estimation and then we save the γ^.Then we calculate the DF stat :

t

If |DFτ| > t-values , we reject the null hypothesis and we accept the alternative

one. Also, if the p-value of the test is smaller than 0,05 ,we also accept the alternative hypothesis.

If |DFτ| < t-values , we reject the alternative hypothesis and we accept the null

hypothesis. Also, if the p-value of the test is more than 0,05 ,we also accept the null hypothesis.

30

4.2 GARCH (1,1)

The GARCH (1,1) model can be written as follows:

σ

t2

= α

0

+ α

1

ε

2t-1

+ β

1

σ

2t-1

where

σt2: is the variance in time t

εt-1: is the error term of previous period

σ2

t-1: is the variance of the previous period

α0: is a constant term

α1: is a coefficient

β1: is a coefficient

Also , these conditions must be valid α0 > 0, α1 > 0, β1 > 0 and α1 + β1 < 1.

4.3 DCC MODEL

The DCC GARCH model can be written as follows:

H

t

= D

t

R

t

D

t

where

Ht: is the conditional covariance matrix

Rt: is the conditional correlation matrix

Dt: is a diagonal matrix with time varying standard deviations on the diagonal

Rt = diag(Qt)-1/2 Qt diag(Qt)-1/2

Qt: is symmetric positive definite matrix

Qt = (1- θ1 – θ2) Ǭ + θ1 et-1 e΄t-1 + θ2 Qt-1

Ǭ: is the unconditional correlation matrix of standardized residuals eit

θ1>0, θ2>0 and θ1 + θ2 < 1

The correlation estimator is:

=

31

4.4 DCC VS DUMMIES MODEL

The equation we used in order to investigate the correlation between commodities and equities during the shocks in 2000-2002 and 2007-2009 can be written as follows:

RHO_12_i

t

= c + d

1

buble_dummy + d

2

estate_dummy + u

t

where

RHO_12_it : is the DCC correlation result for each relationship

c : is a constant term d1: is a coefficient

d1: is a coefficient

ut : is the error term.

We constructed the correlation matrix for the variables in Eviews and then we applied the DCC GARCH (1, 1) methodology in Eviews. Finally we ran the equation with the two shock dummies with OLS estimation. You can see results from our analysis in the results section.

32

5.RESULTS

5.1 UNIT ROOT TEST ADF TABLE 1

Variables

Probability

BRENT_R 0,0000 DAX_R 0,0000 FTSE100_R 0,0000 GOLD_R 0,0000 NIKKEI_R 0,0000 SILVER_R 0,0000 SP500_R 0,0000 WTI_R 0,0000

After the ADF test for the first log difference variables, we found that all are stationary, thus we accepted the alternative hypothesis.

If probability is smaller than 0,05 then we must accept the H1: Variable is stationary.

If the returns are stationary then the indexes have a unit root. Our results follow the theory and proof that commodity and equity prices are I(1) variables.

33

5.2 AUTOCORRELATION TEST TABLE 2

BRENT_R CORRELOGRAM

The evidence of autocorrelation is strong in our variable , because all the

p-values is smaller than 0,05 and we accept the H1, that variable has

34

TABLE 3

DAX_R CORRELOGRAM

The evidence of autocorrelation is strong in our variable , because all the

p-values is smaller than 0,05 and we accept the H1 that, variable has

35

TABLE 4

FTSE 100_R CORRELOGRAM

The evidence of autocorrelation is strong in our variable , because all the

p-values is smaller than 0,05 and we accept the H1 that, variable has

36

TABLE 5

GOLD_R CORRELOGRAM

The evidence of autocorrelation is strong in our variable , because all the

p-values is smaller than 0,05 and we accept the H1, that variable has

37

TABLE 6

NIKKEI_R CORRELOGRAM

The evidence of autocorrelation is strong in our variable , because all the

p-values is smaller than 0,05 and we accept the H1, that variable has

38

TABLE 7

SILVER_R CORRELOGRAM

The evidence of autocorrelation is strong in our variable , because all the

p-values is smaller than 0,05 and we accept the H1, that variable has

39

TABLE 8

SP500_R CORRELOGRAM

The evidence of autocorrelation is strong in our variable , because all the

p-values is smaller than 0,05 and we accept the H1, that variable has

40

TABLE 9

WTI_R CORRELOGRAM

The evidence of autocorrelation is strong in our variable , because all the

p-values is smaller than 0,05 and we accept the H1, that variable has

41

5.3 CORRELATION MATRIX TABLE 10

Covariance analysis and correlation Observations 990

DAX_R BRENT_R GOLD_R SILVER_R WTI_R

Covariance 0,000211 -1,18E-05 0,000124 0,000204 Correlation 0,192272 -0,022926 0,141909 0,186288 t-Statistic 6,158479 -0,720808 4,506139 5,959809

Probability 0,0000 0,4712 0,0000 0,0000

FTSE100_R BRENT_R GOLD_R SILVER_R WTI_R

Covariance 0,000212 1,07E-05 0,000123 0,000212 Correlation 0,266019 0,028443 0,193903 0,265439 t-Statistic 8,674179 0,894387 6,212762 8,653831

Probability 0,0000 0,3713 0,0000 0,0000

NIKKEI_R BRENT_R GOLD_R SILVER_R WTI_R

Covariance 0,000248 -1,50E-05 0,000126 0,000226 Correlation 0,233398 -0,029891 0,148643 0,212768 t-Statistic 7,544659 -0,939971 4,724690 6,844537

Probability 0,0000 0,3475 0,0000 0,0000

SP500_R BRENT_R GOLD_R SILVER_R WTI_R

Covariance 0,000206 1,55E05 0,000118 0,000200 Correlation 0,254731 0,040691 0,183518 0,245870 t-Statistic 8,279970 1,280090 5,868100 7,973049

Probability 0,0000 0,2008 0,0000 0,0000

Brent has a positive correlation with Dax, Ftse100, Nikkei and Sp500, because t-stat is bigger than 1,96 , thus we accept the H1, that we do have correlation and it is positive.

Gold has no correlation with Dax, Ftse100, Nikkei and Sp500, because t-stat is smaller than 1,96 , thus we accept the H0, that we don not have correlation.

Silver has a positive correlation with Dax, Ftse100, Nikkei and Sp500, because t-stat is bigger than 1,96, therefor we accept the H1, that we have correlation and it is positive.

Wti has a positive correlation with Dax, Ftse100, Nikkei and Sp500, because

t-stat is bigger than 1,96 , thus we accept the H1, that we have correlation and

42

5.4 DCC GARCH 1,1 RESULTS TABLE 11

DCC GARCH 1,1 THETAS COEFFICIENTS PROBABILITIES

BRENT_R-DAX_R Theta(1) 0,065661 0,0003

BRENT_R-DAX_R Theta(2) 0,911633 0,0000

STABILITY CONDITION: Theta (1) + Theta (2) < 1 is met

BRENT_R-FTSE100_R Theta(1) 0,036784 0,0091 BRENT_R-FTSE100_R Theta(2) 0,955200 0,0000

STABILITY CONDITION: Theta (1) + Theta (2) < 1 is met

BRENT_R-NIKKEI_R Theta(1) 0,088671 0,0002 BRENT_R-NIKKEI_R Theta(2) 0,834204 0,0000

STABILITY CONDITION: Theta (1) + Theta (2) < 1 is met

BRENT_R-SP500_R Theta(1) 0,052812 0,0000 BRENT_R-SP500_R Theta(2) 0,937725 0,0000

STABILITY CONDITION: Theta (1) + Theta (2) < 1 is met

GOLD_R-DAX_R Theta(1) 0,057868 0,0217

GOLD_R-DAX_R Theta(2) 0,901174 0,0000

STABILITY CONDITION: Theta (1) + Theta (2) < 1 is met

GOLD_R-FTSE100_R Theta(1) 0,025875 0,3566 GOLD_R-FTSE100_R Theta(2) 0,926091 0,0000

STABILITY CONDITION: Theta (1) + Theta (2) < 1 is met

GOLD_R-NIKKEI_R Theta(1) 0,007241 NA

GOLD_R-NIKKEI_R Theta(2) 1,037751 NA

STABILITY CONDITION: Theta (1) + Theta (2) < 1 is not met

GOLD_R-SP500_R Theta(1) 0,050797 0,0210

GOLD_R-SP500_R Theta(2) 0,895809 0,0000

STABILITY CONDITION: Theta (1) + Theta (2) < 1 is met

SILVER_R-DAX_R Theta(1) 0,047128 0,1143

SILVER_R-DAX_R Theta(2) 0,903350 0,0000

STABILITY CONDITION: Theta (1) + Theta (2) < 1 is met

SILVER_R-FTSE100_R Theta(1) 0,032552 0,0262 SILVER_R-FTSE100_R Theta(2) 0,941242 0,0000

STABILITY CONDITION: Theta (1) + Theta (2) < 1 is met

SILVER_R-NIKKEI_R Theta(1) 0,083427 0,0002 SILVER_R-NIKKEI_R Theta(2) 0,862257 0,0000

STABILITY CONDITION: Theta (1) + Theta (2) < 1 is met

SILVER_R-SP500_R Theta(1) 0,061175 0,0050 SILVER-R-SP500_R Theta(2) 0,864954 0,0000

STABILITY CONDITION: Theta (1) + Theta (2) < 1 is met

WTI_R-DAX_R Theta(1) 0,078660 0,0001

WTI_R-DAX_R Theta(2) 0,876851 0,0000

STABILITY CONDITION: Theta (1) + Theta (2) < 1 is met

WTI_R-FTSE100_R Theta(1) 0,035288 0,0248 WTI_R-FTSE100_R Theta(2) 0,957321 0,0000

STABILITY CONDITION: Theta (1) + Theta (2) < 1 is met

WTI_R-NIKKEI_R Theta(1) 0,065068 0,0075

WTI_R-NIKKEI_R Theta(2) 0,864627 0,0000

STABILITY CONDITION: Theta (1) + Theta (2) < 1 is met

WTI_R-SP500_R Theta(1) 0,050411 0,0021

WTI_R-SP500_R Theta(2) 0,938262 0,0000

43

5.5 DCC MODELS

Based on the above parameters is possible to build different models: 1) Qi,j,t = ci,j + 0,065661 εi,t-1 εj,t-1 + 0.911633 Qi,j,t-1

Equation 1 BRENT_R – DAX_R DCC

2) Qi,j,t = ci,j + 0.036784 εi,t-1 εj,t-1 + 0.955200 Qi,j,t-1

Equation 2 BRENT_R – FTSE100_R DCC

3) Qi,j,t = ci,j + 0.088671 εi,t-1 εj,t-1 + 0.834204 Qi,j,t-1

Equation 3 BRENT_R – NIKKEI_R DCC

4) Qi,j,t = ci,j + 0.052812 εi,t-1 εj,t-1 + 0.937725 Qi,j,t-1

Equation 4 BRENT_R – SP500_R DCC

5) Qi,j,t = ci,j + 0.057868 εi,t-1 εj,t-1 + 0.901174 Qi,j,t-1

Equation 5 GOLD_R-DAX_R DCC

6) Qi,j,t = ci,j + 0.025875 εi,t-1 εj,t-1 + 0.926091 Qi,j,t-1

Equation 6 GOLD_R-FTSE100_R DCC

7) Qi,j,t = ci,j + 0.007241 εi,t-1 εj,t-1 + 1.037751 Qi,j,t-1

Equation 7 GOLD_R-NIKKEI_R DCC

8) Qi,j,t = ci,j + 0.050797 εi,t-1 εj,t-1 + 0.895809 Qi,j,t-1

Equation 8 GOLD_R-SP500_R DCC

9) Qi,j,t = ci,j + 0.047128 εi,t-1 εj,t-1 + 0.903350 Qi,j,t-1

Equation 9 SILVER_R-DAX_R DCC

10) Qi,j,t = ci,j + 0.032552 εi,t-1 εj,t-1 + 0.941242 Qi,j,t-1

Equation 10 SILVER_R-FTSE100_R DCC

11) Qi,j,t = ci,j + 0.083427 εi,t-1 εj,t-1 + 0.862257 Qi,j,t-1

Equation 11 SILVER_R-NIKKEI _R DCC

12) Qi,j,t = ci,j + 0.061175 εi,t-1 εj,t-1 + 0.864954 Qi,j,t-1

Equation 12 SILVER_R-SP500 _R DCC

44

Equation 13 WTI_R-DAX_R DCC

14 ) Qi,j,t = ci,j + 0.035288 εi,t-1 εj,t-1 + 0.957321 Qi,j,t-1

Equation 14 WTI_R-FTSE100_R DCC

15) Qi,j,t = ci,j + 0.065068 εi,t-1 εj,t-1 + 0.864627 Qi,j,t-1

Equation 15 WTI_R-NIKKEI_R DCC

16) Qi,j,t = ci,j + 0.050411 εi,t-1 εj,t-1 + 0.938262 Qi,j,t-1

Equation 16 WTI_R-SP500_R DCC

As Bampinas, Ladopoulos & Panagiotidis (2018) mentioned in their paper, the conditional variance has to be stationary and the estimators of the GARCH(1,1) models has to be significant in order to have significant DCC GARCH(1,1) results. Also, the stability condition of the model Theta (1) + Theta (2) < 1 has to be met. We can observe that almost all thetas are significant. The theta1 and 2 in equation 7 are not significant, the stability condition is not met and this dcc analysis is false. However, all though theta 1 in equation 6 and theta 1 in equation 9 are not significant, the stability condition is met, so we do not encounter a problem in the results.

5.6 DCC GRAPHS

Figure 1 BRENT_R –DAX_R DCC

-.6 -.4 -.2 .0 .2 .4 .6 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 RHO_12_01

We notice that, the correlation started to be negative during the dot-com bubble and it took its minimum value -0.4 during 2003-2004. Then an upward trend was noticed until 2008 and the financial crisis, which caused a new

45

negative correlation. After the 2007-2009 shock we can observe a positive correlation, which might be occurred by the financialization of the market.

Figure 2 BRENT_R –FTSE100_R DCC

-.4 -.2 .0 .2 .4 .6 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 RHO_12_02

We notice that, the correlation started to be negative during 2003-2004. Then had an upward trend until 2008 and the financial crisis, which caused a new negative correlation. After the 2007-2009 shock we can observe a positive correlation ,which might be occurred by the financialization of the market.

Figure 3 BRENT_R –NIKKEI_R DCC

-.6 -.4 -.2 .0 .2 .4 .6 .8 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 RHO_12_03

46

We notice that, the correlation started to be negative during the dot-com bubble and continued until 2004, after that we can see an upward trend until the financial crisis , which brought back negative correlation. After the 2007-2009 shock we can observe a positive correlation, which might be occurred by the financialization of the market.

Figure 4 BRENT_R –SP500_R DCC -.6 -.4 -.2 .0 .2 .4 .6 .8 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 RHO_12_04

We notice, that the correlation started to be negative during the dot-com bubble, and continued through until 2004. Then the financial crisis changed the correlation, but after 2008 we have positive correlation.

47 -.5 -.4 -.3 -.2 -.1 .0 .1 .2 .3 .4 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 RHO_12_05

We can observe that the correlation has a negative trend. During the two shocks it took its minimum values, thus we have an evidence of the safe heaven hypothesis of gold.

Figure 6 GOLD_R –FTSE100_R DCC

-.3 -.2 -.1 .0 .1 .2 .3 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 RHO_12_06

We can observe that the correlation has a negative trend. During the two shocks it took negative values, thus we have an evidence of the safe heaven hypothesis of gold.

48 -0.4 0.0 0.4 0.8 1.2 1.6 2.0 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 RHO_12_07

The stability condition of our analysis is not met.

Figure 8 GOLD_R –SP500_R DCC -.4 -.3 -.2 -.1 .0 .1 .2 .3 .4 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 RHO_12_08

We can observe that the correlation has a negative trend. During the two shocks it took negative values, thus we have an evidence of the safe heaven hypothesis of gold.

49 -.2 -.1 .0 .1 .2 .3 .4 .5 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 RHO_12_09

We can observe that the correlation has a negative trend. During the two shocks it took negative values, thus we have an evidence of the safe heaven hypothesis of silver.

Figure 10 SILVER_R –FTSE100_R DCC

-.2 -.1 .0 .1 .2 .3 .4 .5 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 RHO_12_10

We can observe that the correlation has a negative trend. During the two shocks it took negative values , thus we have an evidence of the safe heaven hypothesis of silver. After 2008 we can see a positive correlation, which proves the financialization of the market.

50

Figure 11 SILVER_R –NIKKEI_R DCC

-.6 -.4 -.2 .0 .2 .4 .6 .8 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 RHO_12_11

We can observe that the correlation has a negative trend. During the two shocks it took negative values, thus we have an evidence of the safe heaven hypothesis of silver. Figure 12 SILVER_R –SP500_R DCC -.6 -.4 -.2 .0 .2 .4 .6 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 RHO_12_12

We can observe that the correlation has a negative trend. During the two shocks it took negative values, thus we have an evidence of the safe heaven hypothesis of silver.

51 -.6 -.4 -.2 .0 .2 .4 .6 .8 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 RHO_12_13

We notice that the correlation has a negative trend. During the two shocks it took negative values. After 2008 we have positive correlation but then the trend seems negative.

Figure 14 WTI_R –FTSE100_R DCC

-.4 -.2 .0 .2 .4 .6 .8 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 RHO_12_14

We notice that the correlation has a negative trend. During the two shocks it took negative values. After 2008 we have positive correlation, the cause of this change might be the financialization.

52 -.4 -.2 .0 .2 .4 .6 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 RHO_12_15

We notice that the correlation has a positive trend. Although, during the two shocks it took negative values.

Figure 16 WTI_R –SP500_R DCC. -.6 -.4 -.2 .0 .2 .4 .6 .8 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 RHO_12_16

We notice that the correlation has a negative trend. During the two shocks it took negative values. After 2008 we have positive correlation, the cause of this change might be the financialization.

To sum up the two shocks changed the level of the correlation between commodity and equity markets. The investors tried to add commodities in their

53

portfolios in order to lowered their loses from the downward trend of the equity market. The gold seemed to be the most popular option , because of the safe heaven hypothesis that was valid during that, troubled period.

54

5.7 DCC VS DUMMIES TABLE 12

Dependent Variables Constant Term Bubble Dummy Estate Dummy

RHO_12_01 0,148914 (Prob.=0,0000) 0,018441 (Prob.=0,4085) -0,000715 (Prob.=0,9805) RHO_12_02 0,222423 (Prob.=0,0000) -0,024263 (Prob.=0,2299) -0,039838 (Prob.=0,1333) RHO_12_03 0,188335 (Prob.=0,0000) -0,054246 (Prob.=0,0020) -0,002280 (Prob.=0,9210) RHO_12_04 0,182191 (Prob.=0,0000) -0,046135 (Prob.=0,0737) -0,109808 (Prob.=0,0012) RHO_12_05 -0,027552 (Prob.=0,0000) -0,072436 (Prob.=0,0000) -0,046147 (Prob.=0,0244) RHO_12_06 0,010380 (Prob.=0,0003) -0,058266 (Prob.=0,0000) -0,056061 (Prob.=0,0000) RHO_12_08 0,038710 (Prob.=0,0000) -0,087800 (Prob.=0,0000) -0,124896 (Prob.=0,0000) RHO_12_09 0,130726 (Prob.=0,0000) -0,028426 (Prob.=0,0128) -0,042047 (Prob.=0,0051) RHO_12_10 0,161668 (Prob.=0,0000) -0,080761 (Prob.=0,0000) -0,079428 (Prob.=0,0000) RHO_12_11 0,105080 (Prob.=0,0000) 0,023735 (Prob.=0,1911) -0,043375 (Prob.=0,0690) RHO_12_12 0,161987 (Prob.=0,0000) -0,077245 (Prob.=0,0000) -0,105515 (Prob.=0,0000) RHO_12_13 0,140754 (Prob.=0,0000) -0,001063 (Prob.=0,9603) -0,003296 (Prob.=0,9064) RHO_12_14 0,234371 (Prob.=0,0000) -0,101900 (Prob.=0,0000) -0,093033 (Prob.=0,0004) RHO_12_15 0,173953 (Prob.=0,0000) -0,068923 (Prob.=0,0000) 0,002113 (Prob.=0,9106) RHO_12_16 0,185803 (Prob.=0,0000) -0,093259 (Prob.=0,0001) -0,127790 (Prob.=0,0001)

We ran the OLS estimation and we want to exploit the significance of our coefficients. We tested the H0 non-significant over the H1 significant , by using the probability. If probability is less than 0.05 then, we accept the H1 and reject the H0, otherwise we accept H0. The constant term is significant in all dynamic correlations. The bubble dummy estimators ,which referred to the dot-com-bubble are significant to almost every correlation except the RHO_12_01, RHO_12_02,RHO_12_04, RHO-12_11 and RHO_12_13.The estate dummy, which are referred to the financial crisis of 2007-2009 are significant to almost every correlation except the RHO_12_01, RHO-_12_02,RHO_12_03, RHO-12_11, RHO_12_12 and RHO_12_13.

From our estimation, it is noticeable that the two shocks have a negative impact in the dynamic correlation between the commodity and equity

55

products. All the significant estimations of the dummies have a negative value, thus the correlation during the crisis tend to be negative, this is a fact that proves the theory.

6.DISCUSSION

In the beginning the commodities and the stocks were two separate assets for different kinds of investors. Moreover, stocks tend to have very high returns in the early 2000s and commodities were more stable, we can observe this by looking at all the figures in the data section. After the first shock occurred by the dot-com bubble the loses of the investors guided them to add commodities and equities in their portfolios in order to diversify their positions and lower their management risk. As the commodity investment began to be more and more popular, their returns started to increase and when the market healed from the first shock the investors had big profits, due to the negative correlation between the two financial products, as we can see from the DCC correlation results. All this changed suddenly, after the 2008 financial crisis, for the first time we observed a strong positive correlation between commodities and equities markets. This was the evidence, that we needed to support that, the shock was too strong, that created spillover effects to the commodity investing. That was also proof of the financialization of the economy. The prices of all the assets started to get higher and higher and then the collapse came with huge loses and the investors in order to take cover they invested in commodity market making the commodity prices very high. As Bampinas, Panagiotidis and Rouska (2018) mentioned in their paper the excess demand of noise traders for oil and gold reduced volatility, especially in the presence of a negative shocks, when there is an intense need for alternative investment vehicles or ‘safe havens’, this also proves the theory that commodity investing is more intense during the turmoil periods. Nowadays, the correlation tends to be negative again as the economy starts to recover from the financial crisis, but more research is needed in order to understand the patterns of the correlation between these two assets, mainly during the crisis period when the correlation tends to be negative.

7.CONCLUSION

Our analysis revealed that, the correlation between commodity and equity markets during the dot-com bubble and the financial crisis of 2007-2009 , was negative. This is the evidence that we needed to support that the investors could cover their loses during the crisis by investing into commodity products. In addition, the safe heaven hypothesis of gold was also valid, because this

56

specific correlation had a continuous negative trend. On the other hand the financial crisis has changed the image of the market. Now we see that the correlation tends to be positive, due to the popularity of the commodity investing. This is the result from financialization in the economy. To sum up, we find that the commodity futures could be used as a hedge during the turmoil periods of the economy. However correlation is similar to a living organism, which always develops, therefore more investigation is needed, in order to have a clear perspective of the relationship between commodity and equity markets.

57

REFERENCES

Bampinas, G., Panagiotidis, T., Rouska, C., 2019. Volatility persistence and asymmetry under the microscope: the role of information demand for gold and oil. Scottish Journal of Political Economy 66, 180–197.

Bampinas, G., Ladopoulos, K., & Panagiotidis, T. (2018). A note on the estimated GARCH coefficients from the S&P1500 universe. Applied

Economics: Vol 50, No 34-35.

Bhardwaj, G., & Dunsby, A. (2013). The Business Cycle and the Correlation between Stocks and Commodities (SSRN Scholarly Paper No. ID 2371355).

Bicchetti, D., & Maystre, N. (2013). The synchronized and long-lasting

structural change on commodity markets: Evidence from high frequency data.

Algorithmic Finance, 2(3–4), 233–239.

de Boyrie, M.E., Pavlova, I., (2016). Linkages between Equity and Commodity Markets: Are Emerging Markets Different? 31.

Buyuksahin, B., Haigh, M.S., Robe, M.A. (2008). Commodities and Equities: “A Market of One”? (SSRN Scholarly Paper No. ID 1069862).

Chebbi, T., & Derbali, A. (2015). The dynamic correlation between energy commodities and Islamic stock market: Analysis and forecasting. International

Journal of Trade and Global Markets, 8(2), 112.

Chong, J., & Miffre, J. (2008). Conditional Return Correlations between Commodity Futures and Traditional Assets. 17.

Creti, A., Joëts, M., & Mignon, V. (2013). On the links between stock and commodity markets’ volatility. Energy Economics, 37, 16–28.

Daumas, L. D., Aiube, F. A. L., & Baidya, T. K. N. (2017). Hedging stocks through commodity indexes: a DCC-GARCH approach. 20.

58

Demiralay, S. & Ulusoy V. (2014). Links Between Commodity Futures And Stock Market: Diversification Benefits, Financialization And Financial Crises

[MPRA Paper].

Engle, R. (2002). Dynamic Conditional Correlation: A Simple Class of Multivariate Generalized Autoregressive Conditional Heteroskedasticity Models. Journal of Business & Economic Statistics, 20(3), 339–350.

Gorton,G., & Rouwenhorst, K. G.(2006). Facts and Fantasies about Commodity Futures. NBER Working Paper No. 10595.

Hood, M., & Malik, F. (2013). Is gold the best hedge and a safe haven under changing stock market volatility? Review of Financial Economics, 22(2), 47– 52.

Khan, A., & Masih, M. (2014). Correlation between Islamic stock and Commodity markets: An investigation into the impact of financial crisis and financialization of commodity markets [MPRA Paper].

Khan, A. P., Kabir, S. H., Bashar, O. K. M. R., & Masih, A. M. M. (2015). Time Varying Correlation Between Islamic Equity and Commodity Returns:

Implications for Portfolio Diversification. The Journal of Developing Areas,

49(5), 115–128.

Kolodziej, M., Kaufmann, R. K., Kulatilaka, N., Bicchetti, D., & Maystre, N. (2014). Crude oil: Commodity or financial asset? Energy Economics, 46, 216– 223.

Lagesh, M. A., Kasim C., M., & Paul, S. (2014). Commodity Futures Indices and Traditional Asset Markets in India: DCC Evidence for Portfolio