Uncovering equity market contagion among BRICS countries: an application of the multivariate GARCH model

Transcription

1 MPRA Munich Personal RePEc Archive Uncovering equiy marke conagion among BRICS counries: an applicaion of he mulivariae GARCH model Lumengo Bonga-Bonga Universiy of Johannesburg 24. Augus 2015 Online a hps://mpra.ub.uni-muenchen.de/66262/ MPRA Paper No , posed 25. Augus :21 UTC

2 Uncovering equiy marke conagion among BRICS counries: an applicaion of he mulivariae GARCH model by Lumengo Bonga-Bonga Absrac This paper assesses he exen of he ransmission of financial shocks beween Souh Africa and oher members of he BRICS grouping in order o infer he degree of conagion during he period The paper makes use of a mulivariae VAR-DCC-GARCH model for his end. The paper finds evidence of cross-ransmission and dependence beween Souh Africa and Brazil. However, he empirical resuls show ha Souh Africa is more affeced by crises originaing from China, India and Russia han hese counries are by crises originaing from Souh Africa. The findings of his paper should be of ineres o policy makers in he BRICS grouping should hey be considering he possibiliy of full capial marke liberalizaion and o he inernaional invesor who is looking a diversifying porfolios in he BRICS grouping..

3 1. Inroducion 1 In he pas wo o hree decades, various counries have been bese by severe financial crises: he Mexican peso collapse of 1994, he Eas Asian crisis of 1997, he Russian collapse of 1998, he Argeninean crisis of 2002, he US (Unied Saes) subprime, also referred o as he housing marke crisis of 2007, and he European sovereign deb crisis of 2010, jus o name a few. Alhough hese financial crises sared in a specific counry and region of he globe, heir effecs spread o oher counries and regions. For example, he Eas Asian currency crisis ha sared in Thailand spread wihin a shor period of ime o Indonesia, Malaysia, Korea, Taiwan, and he Philippines (Chancharoenchai and Dibooglu, 2006). Such ransmission of shocks is dubbed conagion in he financial economics lieraure. The erm conagion generally refers o he inernaional ransmission of shocks during financial crises. Alhough here is no concise definiion of he concep, financial economiss noneheless widely use he erm o describe he exen and magniude of he ransmission of shocks from one region or marke o ohers. For example, Bekaer, Harvey, and Ng (2005) refer o conagion as he excess correlaion beween markes over and above wha one would expec from economic fundamenals. Dornhbusch, Park and Claessens (2000) define conagion as a significan increase in cross-marke linkages afer a shock o an individual or group of counries. The lieraure divides he concep of conagion ino wo broad caegories (Dornbusch e al., 2000; Forbes & Rigobon, 2001; Masson, 1998), namely, fundamenal-based and invesor-behaviour conagions. Fundamenal-based conagion refers o he ransmission of shocks ha is due o real and financial linkages or fundamenal relaionship of any kind, such as rade or macroeconomic policy, beween counries. Invesor-behaviour conagion refers o a change in invesor behavior which alers he flow of inernaional porfolio invesmens in such a manner ha i canno be explained by economic fundamenals. For example, a crisis in one emerging marke counry can rigger invesors o wihdraw funds from many emerging markes wihou aking ino accoun he fundamenal economic differences beween hem. 1 We acknowledge he conribuion of Naacha Brink.

4 Sudying he effec of conagion of financial crises beween BRICS counries is imporan given he magniude of ineracion beween member counries and wha he BRICS counries represen globally. The BRICS counries, consising of Brazil, Russia, India, China, and since December 2010 Souh Africa (SA), represen he world s leading emerging marke economies (EME), disinguished by heir large, fas-growing economies. The growh poenials in hose culurally and geographically disparae counries are based on diverse aribues. Brazil is a resource-rich counry, wih resources such as coffee, soybean, sugar cane, iron ore and crude oil. Russia is well known for is massive deposis of oil, naural gas and minerals. India has a rising manufacuring base and is a srong service provider. China has a highly skilled workforce a low wage cos and is seen as he manufacuring workshop of he world. SA, he smalles of he five BRICS counries by land mass and world GDP conribuion, is he world s larges producer of plainum and chromium, and holds he world s larges known reserves of manganese, plainum group meals, chromium, vanadium and alumino-silicaes (New Delhi, 2012). BRICS financial indicaors are ousanding in ha equiy indices more han doubled beween 1999 and 2009, and BRICS marke capialisaion in equiy markes grew from US$1.2 rillion o US$6.4 rillion beween 2000 and 2010 (New Delhi, 2012). Noneheless, in erms of muual influence and ineracion beween BRICS member counries, a number of auhors have quesioned he imporance and influence of Souh Africa (SA) wihin his prospecively powerful grouping. For example, Naidoo (2012) conends ha SA does no fi ino BRICS given he size of is economy. The auhor sees he presence of SA as weakening he group for hree reasons. Firsly, because of SA s GDP growh lags compared o he res of he BRICS counries and oher EMEs. Secondly, SA doesn feaure wihin he op 20 larges world economies in US dollar erms. Thirdly, SA has a populaion of 50 million compared o he second smalles BRICS counry (Russia) wih a populaion of 140 million, and is herefore a small counry in comparison. The quesioning by criics of he imporance and influence of SA wihin he BRICS grouping promped his paper, which endeavours o assess he exen of Souh Africa s financial influence on oher BRICS counries, and also he degree and magniude of he ransmission of financial shocks beween Souh Africa and each of he oher BRICS counries during periods of financial crises. In oher words, he paper endeavours o assess he exen of conagion beween Souh Africa and each of he BRICS counries during he period

5 The financial influence of Souh Africa on emerging-marke economies is well documened. For example, Flvin and O connor (2010) show ha Souh Africa has one of he mos liberalized sock exchange and financial sysems among emerging-marke economies. However, he exen of is financial influence in he BRICS grouping is a maer of empirical analysis. The hypohesis of his paper is ha if i can be found ha a crisis ha originaes in SA spreads o oher BRICS counries o he same exen as shocks from oher BRICS counries ransmi o Souh Africa, hen one could infer he possibiliy of muual financial inerdependence beween Souh Africa and oher BRICS economies, proving wrong he view ha Souh Africa is of lile financial influence in he BRICS grouping. The paper assesses he ransmission of shocks in he conex of he equiy marke given is imporance as a significan financial secor in BRICS counries. A number of sudies have made use of sock exchange daa o assess he degree of financial dependence and inegraion of counries (Bonga-Bonga, 2009; Singh, 1997). While oher sudies have assessed conagion beween BRICS and oher developed economics (Nikkinen, e al., 2013; Berikos, 2014; Morales and Gassie, 2011; Sheu and Liao, 2011), o he bes of our knowledge here is no sudy ha assesses conagion wihin BRICS counries, especially since he ime of Souh Africa s inclusion in he BRICS grouping. The finding of his paper should inform policy makers in BRICS counries on he benefi ha each member can derive from furher liberalizing is capial markes. I is imporan o noe ha capial marke liberalizaion in he presence of asymmeric conagion may lead o porfolio re-allocaion and capial fligh a he derimen of he mos vulnerable or relian counry, especially during he periods of financial crisis (Sigliz, 2004; Borjas and Ramy, 1995). Thus, he finding of his paper should also be of grea ineres o inernaional invesors and asse managers. In order o assess he exen of conagion beween Souh Africa and each of he BRICS counries, his paper idenifies periods of major crises in each of he BRICS counries and assesses how condiional correlaion of equiy marke reurns beween Souh Africa and each of he BRICS counry fared during hese periods. For example, he dynamic condiional correlaion of equiy marke reurns beween Souh Africa and each of he BRIC counries will be assessed during he 2001 Souh African currency crisis. I is imporan o noe ha during he 2001 currency crisis in Souh Africa, he nominal rand depreciaed 26% agains he US dollar, especially beween Sepember 2001 and December 2001 (Bhundia & Ricci, 2005). Bhundai and Goschalk (2003) as well as Preorius and de Beer (2004)

6 aribue he sharp depreciaion of he rand during hose periods o he nominal disurbance ha originaed from he US, he Sepember 11, 2001 aack and he poliical unres in Zimbabwe. Moreover, he impac of crises emanaing from oher BRICS counries on he Souh African economy will also be assessed. The empirical lieraure on conagion is vas, mosly promped by he aemp of a number of sudies o undersand he widespread effecs of he financial crises in he 1990s. Differen empirical approaches emerged, which could be classified in four differen caegories (Forbes & Rigobon, 2001): he analysis of cross-marke correlaion coefficiens; GARCH frameworks; coinegraion and probi models. The cross-marke correlaion es measures he correlaion in reurns beween wo markes a wo disincly differen ime periods, he ranquil and urmoil periods. A significan increase in he correlaion coefficien during he urmoil period would sugges a ransmission mechanism or he occurrence of conagion (King and Wadhwani, 1990; Kim, 1993; Rengasamy, 2012). Noneheless, cross-correlaion models for he analysis of conagion have been criicised for heir inabiliy o accoun for heeroscedasiciy in he variables used. To remedy his criicism, Forbes and Rigobon (2001) and Bouaziz, Selmi, and Boujelbene (2012) sugges he use of he generalised auoregressive condiional heeroscedasiciy (GARCH) model as per Bollerslev (1986). The GARCH frameworks use high-frequency daa o assess he ransmission of volailiy across markes (Kalkberg, Liu, & Pasquariello 2005). Morales and Gassie (2011) use a sandard univariae GARCH, TGARCH and srucural breakpoin idenificaion algorihms o analyse he co-movemen of he BRICS and US energy markes (oil, naural gas and elecriciy). Moreover, in he conex of mulivariae GARCH, Bouaziz e al. (2012) used Markov Swiching in conjuncion wih a DCC- GARCH model o deermine he worldwide conagion effec of he US subprime crisis of A number of sudies make use of he coinegraion echnique o es for conagion by deermining he long-run relaionship beween markes in he presence of financial crises (Longin and Solnik, 1995; Fahami, 2011). For example, Fahami (2011) used his mehod o es he srucure of linkages and he causal relaionships beween BRIC and oher developed counries during he 2007 US subprime crisis. The auhor shows ha BRIC equiy markes correlaed more closely wih he US equiy marke han wih UK and Japanese equiy markes. Gupa (2011) underook similar ess on equiy markes by comparing BRIC counries inerdependence during he US subprime crisis of 2007 and he European

7 sovereign deb crisis of The auhor found long-erm correlaion beween he BRIC counries, and ha bi-direcional causaliy exiss beween China, India and Russia. The fourh group of empirical analysis assesses financial conagion by making use of exogenous evens and microeconomic raher han macroeconomic daa (Forbes and Rigobon, 2001). The advanage of microeconomic daa is ha i provides a more concise and clear idenificaion of he channels hrough which conagion can occur. For example, Forbes and Rigobon (2001) examined how differen ypes of firms were globally affeced by he Russian and Asian Crises and how hese crises affeced oher firms worldwide. The auhors showed ha firms ha ransac wih counries ha are affeced by economic crisis are also significanly affeced; his herefore suggess ha rade channels are imporan in ransmiing conagion. In order o assess he magniude of he ransmission of financial shocks in he conex of conagion, his paper applies he GARCH framework by making use of a mulivariae vecor auoregressive dynamic condiional correlaion GARCH (VAR-DCC GARCH) model, whereby aenion will be given o he ransmission of equiy marke volailiy shocks and ime-varying condiional correlaion o assess he evoluion of he correlaion beween he Souh African and oher BRICS equiy markes. Conrary o sudies ha made use of he DCC GARCH model in assessing conagion beween differen counries (Celik, 2012; Chao and Parhizgani, 2008), his paper makes use of he VAR framework in he mean equaion o accoun for possible endogeneiy and inerdependence of equiy reurns of BRICS economies. In addiion, as saed earlier, his paper is he firs o deal wih he issue of conagion among he BRICS counries. The remainder of his paper is organised as follows: Secion 2 deals wih he mehodology of mulivariae VAR-DCC GARCH model, he resuls will be discussed in Secion 3. This paper concludes wih a presenaion of he findings in Secion Mehodology In order o examine financial conagion beween SA and is BRICS counerpars during he differen crisis periods a VAR DCC GARCH model is esimaed. The esimaion of he VAR DCC GARCH model is broken down ino hree sages. In he firs sage, a vecor auoregressive (VAR) model is esimaed as he mean equaion. This esimaion informs of he ineracion beween sock reurns of

8 BRICS counries and brings up o dae he possible spillover beween he sock exchanges of hose counries. In he second sage, he residuals obained from he firs sage are used o model he GARCH equaions. In his paper use of he GARCH (1,1) model is made, which is suiable for equiy reurns (Engle & Paon, 2001). 2 Lasly, he covariance marix obained in he second sage is used o calculae he ime-varying correlaion marix. The mean equaion is represened by he following VAR equaion of order n: Y n 0 iy Z i1 i + (1) wih and Y is a 5-variable vecor conaining he following equiy marke variables in order: Russia, Souh Africa, India, Brazil and China. Z. represens he vecors of deerminisic and exogenous variables and he residual combines he whie noise process and he heeroscedasic componen. Parameers 0, i and need o be esimaed. The advanage of using he VAR framework in he mean equaion is o accoun for he inerdependence of reurns beween BRICS counries and he influence of he deerminisic and/or exogenous variable Z (here we accoun for he influence of he S$P 500 reurns on BRICS equiy reurns). The second sage uses he residuals obained from Equaion 1 in he firs sage o inpu hem ino he univariae condiional-variance model specified for each BRICS equiy reurn. To accoun for equiymarke asymmery, we use he Glosen, Jagannahan and Runkle (GJR) (1993) GARCH model, which accouns for he asymmeric effec of equiy-marke reurns. The GJR GARCH (1,1) 3 model is represened as follows: d I 0) (2) ( 1 2 Mis-specificaion ess are conduced o ascerain he validiy of he model used. 3 The order of he EGARCH is deermined by he log likelihood of he model esimaion.

9 where he parameer ω refers o he long-erm condiional variance and α is he lag coefficien. I( ) is an indicaor variable ha akes he value of 1 when if 1 0 and zero oherwise. Thus, he impac of on is d for negaive shocks and for posiive shocks. The las sage in a DCC GARCH model consiss of deermining he ime-varian condiional correlaion marix from he condiional variance expressed as: H D R D (3) R D =diag h (4) ( 1 a b) R a 1 br 1 h 1/ / 2 11 nn Where D is he diagonal marix of condiional variances such as. R is a posiive definie N x N correlaion marix and is defined as follows: Where a, b >0 and a b 1. R is a scalar for consan condiional correlaion in ha R= R if a=b=0. 1 is expressed as: ij, 1 u m1 M M ( u m1 1, m 2 i, m u j, m M )( u h1 2 j, h ) (5) and u / i i h ii The logarihm of he likelihood funcion of he DCC GARCH model is represened as: T 1 ln L ln(2 ) (6) 2 2 T ' 1 ln D R D ln R ( R ) 1 3. Daa, esimaions, resuls and discussion The paper makes use of weekly daa ha covers he period December 1996 o May The iniial period corresponds wih he liberalizaion of a number of BRICS equiy markes. BRICS equiy reurns are compued from he following equiy indices: he Johannesburg Sock Exchange (JSE) All Share Index for Souh Africa, he Bovespa Index for Brazil, Shanghai A Share Index for China, he RTS

10 Index for Russia and he S&P CNX500 Index for India. The S&P 500 reurns are used as an exogenous variable in he VAR model o conrol for he influence of he US on he BRICS equiy markes. Table 1 repors he summary saisics for he weekly equiy reurns of he five BRICS counries. The mean reurns range from 0.27 % for Brazil o 0.08% for China. Russia has he highes sandard deviaion for he full sample observaion followed by Brazil and India. The high kurosis and negaive skewness for all he BRICS counries indicae ha heir equiy reurns are characerized by fa ails and exremely negaive reurns, respecively. This migh explain he vulnerabiliy of BRICS counries o global crises. The Jarque-Bera saisics show ha BRICS reurns exhibi subsanial non-normaliy; hus quasi-maximum likelihood is considered for GARCH esimaion. Table 1. BRICS reurns descripive saisics Brazil China India Souh Africa Russia Mean Median Sd. Dev Skewness Kurosis Jarque-Bera Probabiliy Observaions Figures 1 o 5 display he equiy reurns of he five BRICS counries superimposed on periods of major financial and economic crises. Figure 1 shows ha he Souh African equiy reurns, he JSE All Share Index reurns, were highly volaile during 1997, 1998, 2001, and 2002 and during he 2007 o 2009 period (as indicaed by he shaded areas), which correspond o he Asian crisis in 1997, he Russian crisis in 1998, he Argeninean crisis in 2002, and he US subprime crisis and European sovereign deb crises during 2007 and 2009, respecively. This suggess ha Souh Africa has been vulnerable o conagion from boh emerging and developed markes. In addiion, high volailiy clusering hough o a lesser exen is also visible during 2000, 2004 and I is worh noing ha he observed high volailiy in 2006 was due o a balance of paymens problem in emerging markes riggered by a srong signal by he US Federal Reserve Bank ha

11 here would be a hike in he Fed Fund rae in May 2006, which led o massive capial flows from emerging markes (Bonga-Bonga, 2014). Figure 1. Reurns on he Souh African JSE All Share Index A similar picure is eviden for Brazil s Bovespa Index reurns. Figure 2 shows ha Brazils Bovespa Index reurns experienced high levels of volailiy during 1997, 1998, 2000, 2001, 2002, and beween 2008 and 2009, jus like he Souh African equiy reurns. Noneheless, he Brazilian equiy reurns were more volaile during he crisis periods han Souh Africa s were, as evidenced by high reurns spikes during hose periods. Figure 3 displays he equiy reurns of Russia s RTS index. Russian equiy reurns porray a differen picure from hose of SA and Brazil. Periods of high volailiy are few and far beween, bu much more pronounced han hose of boh SA and Brazil. Periods in which high volailiy is eviden are beween 1997 and 1998, he laer par of 2008, he beginning of 2009 and during This explains Russia s suscepibiliy o he Asian crisis and is own (Russian) crisis as well as o he US subprime crisis and he European sovereign deb crisis. India s S&P CNX500 reurns, displayed in Figure 4, resembles SA s more closely. However, high volailiy clusering is more frequen compared o periods of low volailiy. Periods of high volailiy in India s S&P CNX500 Index reurns are during 1998 o 2001, he beginning of 2004 and 2006, and beween 2007 and High volailiy during 2001, however, can be explained by eiher he currency crisis ha originaed in SA or he 9/11 error aacks in he US. Similarly, he high volailiy eviden in

12 boh 2004 and 2006 also does no coincide wih periods of known crises. The period from 2007 o 2009 corresponds o boh he US subprime crisis and he European sovereign deb crisis.

13 Figure 2. Reurns on Brazil s Bovespa Index Figure 3. Reurns on Russia s RTS Index Figure 4. Reurns on India s S&P CNX

14 Figure 5 shows ha he volailiy in China s Shanghai A share is moderae, wih he mos significan deviaions observed during he period Figure 5. Reurns on China s Shanghai A Share Index In order o obain he condiional correlaion esimaes of he Souh African equiy marke and each of he oher BRICS equiy markes, we firs make use he mean equaion approximaed by a VAR model wih one lag, 4 where he endogenous variables consis of equiy reurns from he differen BRICS counries. In addiion, we conrol exogenously for he influence of he S&P 500 equiy reurns on BRICS counries. Given ha here is evidence ha he series co-breaks, he VAR model did no include specific dummy variables. Secondly, he residuals obained from he VAR esimaion are used o model he GJR-GARCH(1,1) from he differen counries. In he hird sep, he likelihood funcion in Equaion 6 is used o obain he parameers of he VAR-DCC-GARCH model. Table 1 presens he resuls of he esimaion of he models represened by Equaions 1 o 5. The resuls repored in Table 2 show ha on average he S&P 500 equiy index reurns have a posiive impac on BRICS equiy reurns, wih he impac being saisically significan for all he BRICS counries. While he Brazilian equiy marke seems o be he mos influenced by he US equiy marke, wih he coefficien equals , he Chinese equiy marke is he leas influenced by he US equiy marke among BRICS counries. I is imporan o noe ha he posiive influence of he US equiy marke on emerging marke equiy reurns is well documened (see Bonga-Bonga and Mwamba, 2015). 4 The choice is based on informaion crieria such as he Akaike Informaion Crieria and he Bayesian Informaion Crieria

15 Moreover, he resuls repored in Table 2 show ha he asymmeric effec is saisically significan in he Souh African, Brazilian and Russian equiy markes and ha he sum of coefficien a and b is less han uniy, which jusifies he sabiliy of he volailiy model used. Table 2. VAR DCC GARCH esimaion of BRICS equiy markes Brazil Souh Africa India Russia China *** *** *** *** *** *** *** *** *** *** * ** * ** *** *** ** *** *** *** *** ** ** *** *** *** *** *** *** *** *** * *** *** d *** *** *** DCC Coefficiens coefficiens sandard error -sa Probabiliy a b ***, ** and * denoes rejecion of he null hypohesis a 1%, 5% and 10% respecively. The order of counries is 1 for Russia, 2 for Souh Africa, 3 for India, 4 for Brazil and 5 for China. The Q-saisics and he LM ARCH ess in Table 3 confirm ha he null hypohesis of no serial correlaion and no ARCH effec is no rejeced for he esimaed VAR DCC-GARCH(1,1) model. This confirms he validiy of he model used, from which he dynamic condiional correlaion graphs displayed in Figures 6 o 9 are obained.

16 Table 3. GARCH Q es coefficiens Equiy Q Probabiliy ARCH Probabiliy reurns Russia SA India Brazil China As he focus of his paper is on he dynamic condiional correlaion obained in he hird sep of he VAR-DCC-GARCH model, Figures 6 o 9 display he dynamic condiional correlaion beween Souh Africa and each of he oher BRICS counries from 1996 o From hese figures, conagion is inferred when here is evidence of an increasing correlaion during paricular crisis periods. The periods of crisis idenified in his paper are mainly specific BRICS counries crises, such as he Souh African currency crisis of 2001, he Brazilian currency crisis of 2002, and he Russian currency crisis of Because no paricular crisis emanaed from India and China during he sample period of our sudy, we will use he Asian crisis as he originaing crisis for he wo counries. Our approach o assessing he financial influence of Souh Africa on he BRICS grouping is o compare he magniude of he correlaion beween he Souh African equiy marke reurns and each of he BRICS equiy marke reurns during he crisis emanaing from Souh Africa and he crises emanaing from each of he oher BRICS counries. The idenified crisis periods during which possible conagion is assessed include he monh of he beginning of he crisis in each BRICS counry and he following monh in order o accoun for conemporaneous and possible lag effecs in he ransmission of shocks. Thus, he idenified ime period for he Souh African currency crisis is from December 2001 o January The period for he Russian financial crisis is from Augus o Sepember The Brazilian currency crisis is idenified from July o Augus 2002 and he Asian financial crisis from November o December These crisis periods are indicaed by he dark shades in Figures 6 o 9 wih he addiion of he subprime crisis. Figure 6 shows he dynamic correlaion beween Souh Africa and Brazil. The display in Figure 6 shows ha here is clear evidence of an increasing correlaion beween Souh Africa and Brazil during

17 he Brazilian currency crisis, bu no such clear evidence during he Souh African currency crisis. 5 This should indicae ha crises from Brazil spill over o Souh Africa and no he opposie. Anoher observaion from Figure 6 is ha he correlaion beween equiy markes in Brazil and Souh Africa increases during crisis periods emanaing from oher counries and regions, such as he Russian, Asian and subprime crises. This indicaes ha he synchronizaion of he wo equiy markes is also riggered by exernal shocks. Figure 6. Dynamic correlaion beween Souh Africa and Brazil Figure 7 shows he dynamic correlaion beween Souh Africa and Russia. While here is a clear evidence of an increase in he correlaion beween he equiy marke reurns of he wo counries during he Russian crisis, here is no such srong evidence during he Souh African currency crisis. This evidence should indicae ha he Souh African currency crisis had a negligible influence on Russia and, hus, he absence of conagion of he Souh African equiy marke o he Russian equiy marke. As in he case of Brazil, he wo equiy markes commove o differen exernal shocks. Figure 8 shows he dynamic correlaion beween he Souh African and Indian equiy marke reurns. The increasing correlaion beween he wo equiy markes during he Asian financial crisis and he negligible increasing correlaion during he Souh African currency crisis indicae unidirecional conagion from India o Souh Africa. 5 The observed spike in 2001 corresponds o he 11 Sepember even in he Unied Saes

18 Figure 7. Dynamic correlaion beween Souh Africa and Russia Figure 8: Dynamic correlaion beween Souh Africa and India Figure 9 displays he dynamic correlaion beween he Souh African and Chinese equiy marke reurns. The correlaion beween he Souh African and Chinese equiy markes is lower han hose beween he Souh African and oher BRICS equiy markes. Moreover, conrary o oher equiy markes where he correlaion wih he Souh African equiy marke has remained posiive during he sample periods, he correlaion beween he Souh African and Chinese equiy markes is characerised by periods of negaive correlaion, indicaing ha he wo markes occasionally decouple.

19 Figure 9. Dynamic correlaion beween Souh Africa and China The resuls displayed in Figures 6, 7, 8, and 9 show ha here seems o be weak evidence of an increasing correlaion beween Souh Africa and each of he BRICS counries during a period of crisis ha sems from Souh Africa, which may lead o he conclusion ha Souh Africa is a receiver raher han ransmier of financial shocks o oher BRICS counries during periods of financial crisis. Such a conclusion will no be robus wihou assessing wheher he difference in he means dynamic correlaion observed during periods of crisis emanaing from Souh Africa and oher BRICS counries is saisically differen. We use he -saisics es of means difference o his end whereby he null and alernaive hypoheses for he -saisics es of means difference are defined as: SAcrisis BRICS crisis Ho, Ha SAcrisis BRICS crisis SAcrisis BRICS crisis Where and are he means of he condiional correlaion coefficiens during he periods of crisis emanaing from Souh Africa and each of he BRICS counries, respecively. The - saisics are calculaed as follows: _ SAcrisis ( s ij 2 SAcrisis n _ BRICS crisis ij s 2 BRICS crisis n )

20 Where _ SAcrisis ij denoes he mean of dynamic correlaion coefficiens beween Souh Africa () and each of he BRICS counry (j) during he crisis emanaing from souh Africa. s 2 SAcrisis is he variance of hese coefficiens esimaed as: s 2 _ 2 1 n ij ij 1 n 1 Table 4 presens he DCC mean values, he -saisics for means difference and he oucome of he es of he means difference. While here is a rejecion of equal magniude of conagion beween Souh Africa-China, Souh Africa-India and Souh Africa-Russia during he periods of crisis emanaing from each of hese counries, here is evidence of equal magniude of conagion beween Souh Africa and Brazil during periods of crisis semming from each of he wo counries. These resuls show ha Souh Africa is more affeced by crises originaing from China, India and Russia han he hree counries are by crises originaing from Souh Africa. China seems o be decoupled from crises originaing from Souh Africa, wih a slighly negaive condiional correlaion during period of crisis originaing from Souh Africa. Brazil and Souh Africa are equally affeced during crises originaing from heir respecive counries. Table 4. DCC mean values and -saisics for means difference SAcrisis BRICS crisis -saisics oucome Souh Africa - Brazil H0 no rejeced Souh Africa - India * H0 rejeced Souh Africa - Russia * H0 rejeced Souh Africa - China * H0 rejeced * denoes rejecion a 1% level These findings should be of ineres o policy makers in BRICS as well inernaional invesors and porfolio managers who inend o inves in BRICS. Policy makers in Souh Africa, in paricular, should be cauious in aemping o pursue he agenda of full capial marke liberalizaion wih oher BRICS counries wih he possibiliy of scrapping he exising exchange conrol. Such a move may resul in capial fligh from Souh Africa o oher BRICS counries, especially during periods of financial insabiliy in Souh Africa. However, Souh Africa may aemp he full capial marke liberalizaion

21 embarked on by Brazil, which is already Souh Africa s mos imporan rade parner in he BRICS grouping. The wo counries are also boh members of he IBSA (India, Brazil and Souh Africa) grouping. For inernaional invesors and porfolio managers, hese findings should inform on he possibiliy of porfolio diversificaion and equiy and opion pricings when invesing in he BRICS bloc. Souh Africa is shown o be far from a safe haven during crises originaing from Russia, India and China. To deermine he validiy of our findings, we conduced a robusness es by making use of uncondiional correlaion measures obained from adjusing he DCC for heeroscedasiciy. Forbes and Rigobon (2002) sugges he use uncondiional correlaion when inferring for conagion. The auhors shows ha formal ess for conagion based on condiional correlaion may be biased if he laer is no adjused o uncondiional correlaion (see discussion in Forbes and Rogobon, 2002). We adjused he condiional correlaion o uncondiional correlaion by making use of he relaive increase in he variance of he Souh African reurns before and during he 2001 currency crises. Table 5 presens he mean values, he -saisics for means difference and he oucome of he es of he means difference during he crisis originaed from Souh Africa and he one originaed from specific BRICS counry by making use of uncondiional correlaion measures. Table 5 Uncondiional correlaion mean values and -saisics for means difference SAcrisis BRICS crisis -saisics oucome Souh Africa - Brazil H0 no rejeced Souh Africa - India * H0 rejeced Souh Africa- Russia * H0 rejeced Souh Africa - China * H0 rejeced * denoes rejecion a 1% level The resuls repored in Table 5 show ha alhough he magniude of means of uncondiional correlaion is less han he mean of he condiional correlaion across all he idenified crises, he oucome of he es of he means difference is idenical o ha repored for condiional correlaion in Table 4. This indicaes ha he inference drawn for he exen of conagion beween Souh Africa and oher BRICS counries holds across differen measures of correlaion. Souh Africa is more conaminaed by crises originaed from China, India and Russia han hose counries are by crises

22 originaing from Souh Africa. Noneheless, here is evidence of inerdependence beween Souh Africa and Brazil. 4. Conclusion This paper assesses he exen of financial conagion beween Souh Africa and oher BRICS counries by using he VAR-DCC-GARCH model. The magniude of he correlaion beween Souh Africa and oher BRICS counries is analysed during BRICS-specific and global financial crises, such as he 1998 Russian currency crisis, he 2001 Souh African currency crisis, he 2002 Brazilian currency crisis and he 1997 Asian financial crisis. The findings of he paper indicae ha Brazil and Souh Africa are equally affeced during crises emanaing from heir respecive counries. These findings indicae ha here is inerdependence beween Souh Africa and Brazil. However, he empirical resuls show ha Souh Africa is more affeced by crises originaing from China, India and Russia han hose counries are by crises originaing from Souh Africa. The findings of his paper should be of ineres o policy makers in he BRICS grouping when hey consider he possibiliy of full capial marke liberalizaion. Moreover, he findings of his paper should inform inernaional invesors and porfolio managers on he possibiliy of porfolio diversificaion and equiy and opion pricings when invesing in BRICS. For furher research we sugges ha oher volailiy measures be considered when analysing he possibiliy of conagion wihin he BRICS grouping.

23 References Bekaer, G., Harvey, C.R., & Ng, A. (2005). Marke inegraion and conagion. The Journal of Business, 78(1): Berikos, S.D. (2014). Conagion, decoupling and he spillover effecs of he US financial crisis: Evidence from he BRIC marke. Inernaional Review of Financial Analysis, 33(1): Borjas, G.J., Ramey V.A., Foreign Compeiion, Marke Power and Wage Inequaliy. Quarerly Journal of Economics, 110(4): Bhundai, A. J. & Goschalk, J. (2003). Sources of nominal exchange rae flucuaions in SA. IMF Working Paper. WP/03/252 Bhundai, A. J. & Ricci, L. A. (2005) The Rand crises of 1998 and 2001: Wha have we learned? In Nowak, M. and Ricci, L. A. (Eds.). Pos-aparheid Souh Africa: The firs en years Washingon, DC: Inernaional Moneary Fund. Bollerslev, T. (1986). Generalized auoregressive condiional heeroskedasiciy. Journal of Economerics, 31(3): Bonga-Bonga, L. (2009). An Assessmen of he Degree of Souh Africa's Financial Inegraion ino he World Economy. The African Finance Journal, 11(2): Bonga-Bonga, L. ( 2014). Assessing he readiness of BRICS grouping for muually beneficial financial inegraion,. MPRA Paper 60701, Universiy Library of Munich, Germany. Bonga-Bonga, L. & Mwamba, M. (2015). A mulivariae model for he predicion of sock reurns in an emerging marke: A comparison of parameric and non-parameric models. MPRA Paper 62028, Universiy Library of Munich, Germany. Bouaziz, M. C., Selmi, N., & Boujelbene, Y. (2012). Conagion Effec of he Subprime Financial Crisis: Evidence of DCC Mulivariae GARCH Models. European Journal of Economics, Finance and Adminisraive Sciences. 44: Celik, S. (2012). The more conagion effec on emerging markes: he evidence of DCC-GARCH model. Economic Modelling, 29(5):

24 Chancharoenchai, K. and Dibooglu, S. (2006). Volailiy spillovers and conagion during he Asian crisis: evidence from six Souheas Asia sock markes. Emerging markes Finance and Trade 42(2): Cho, J.H. and Parhizgani, A.M. (2008). Eas Asia financial conagion under DCC-GARCH. The Inernaional Journal of Banking and Finance, 6(1): Dornbusch, R., Park, Y. C.,& Claessens, S. (2000). Conagion: Undersanding how i spreads. World Bank Research Observer, 15(2): Engle, R. F. & Paon, A. (2001). Wha good is a volailiy model?. Quaniaive Finance, 1: Fahami, N. A. (2011). The srucure of linkages and causal relaionship beween BRIC and developed equiy markes Inernaional Conference on Informaion and Finance. IPEDR Vol. 21. IACSIT Press, Singapore. Flavin, T. and O'Connor, T. (2010). The sequencing of sock marke liberalisaion daes and corporae financing decisions. Emerging Markes Review, 11(3): Forbes, K.J. and Rigobon, R. (2002). No conagion, only inerdependence: measuring sock marke comovemens. Journal of Finance, 57(5), Forbes, K. & Rigobon, R. (2001). Measuring conagion: Concepual and empirical issues. In Sijn Claessens & Krisin Forbes (Eds.), Inernaional Financial Conagion. Boson: Kluwer Academic Publishers, pp Forbes, K.J. and Rigobon, R. (2002). No conagion, only inerdependence: measuring sock marke comovemens. Journal of Finance, 57(5), Glosen, L. R., R. Jagannahan, and D. E. Runkle, On The Relaion beween The Expeced Value and The Volailiy of Nominal Excess Reurn on socks. Journal of Finance 48: Gupa, S. (2011). Sudy of BRIC counries in he financial urmoil. Inernaional Affairs and Global Sraegy, 1(1): 1-16 Kalkberg, J. G., Liu, C. H., & Pasquariello, P. (2005). An examinaion of he Asian crisis: Regime shifs in currency and equiy. The Journal of Business, 78(1):

25 King, M. & Wadhwani, S. (1990). Transmission of volailiy beween sock markes. Review of Financial Sudies, 3(1): Lee, S. B. & Kim, K. J. (1993). Does he Ocober 1987 crash srenghen he co-movemens among naional sock markes? Review of Financial Economics, 3(1): Longin, F.& Solnik, B. (1995). Is he correlaion in inernaional equiy reurns consan: ? Journal of Inernaional Money and Finance, 14: Masson, P. (1998). Conagion: Monsoonal effecs, spillover, and jumps beween muliple equilibrium. IMF Working Paper. No. 98/142. Morales, L. & Gassie, E. (2011). Srucural breaks and financial volailiy: Lessons from BRIC counries. IAMO Forum 2011, No. 13, hp:// hdl.handle.ne/10419/ Accessed 12 January Naidoo, S. (2012). Souh Africa s presence drags down BRICS. (Mar, 23). The Mail & Guardian. hp://mg.co.za/aricle/ sa-presence-drags-down-brics. Accessed 12 February New Delhi. (2012). The BRICS Repor. Oxford Universiy Press. Nikkinen, J. ; Saleem, K. and Marikainen, M. (2013). Transmission of he subprime crisis: evidence from indusrial and financial secors of he BRICS counries. Journal of Applied Business Research, 29(5): Preorius, A. & De Beer, J. (2004) Conagion in Africa: Souh Africa and a roubled neighbour, Zimbabwe. Economic Modeling, 21(4), Rengasamy, E. (2012). Sovereign deb crisis in he Euro zone and is impac on he BRICS s sock index reurns and volailiy. Economics and Finance Review. 2(2): Sheu, H. J. & Liao, C. H. (2011). Dynamics of sock marke inegraion beween he US and he BRIC. African Journal of Business Managemen. Vol. 5 (9):

26 Singh, A.(1997). Financial liberalizaion, sock marke and economic developmen. The Economic Journal, 107: Sigliz, J.E. (2004). Capial-marke kiberalisaion, globalizaion and he IMF. Oxford Review of Economic Policy, 20(1): 57-71