EXCHANGE RATE AND GOLD PRICE: EVIDENCE FROM MALAYSIA. JEL Classification: F31, G11, G15 Keywords: Exchange rate, gold price, cointegration

Transcription

1 Labuan Bullein OF INTERNATIONAL BUSINESS & FINANCE Volume 12, 2014 ISSN EXCHANGE RATE AND GOLD PRICE: EVIDENCE FROM MALAYSIA Wong Hock Tsen 1 School of Business and Economics, Universii Malaysia Sabah Absrac This sudy examines he relaionship of exchange rae and gold price in Malaysia. The auoregressive disribued lag (ARDL) approach shows ha here is long-run relaionship beween exchange rae and is deerminans, includes gold price. An increase in gold price will lead o a depreciaion of he US dollar. There is a negaive relaionship beween exchange rae and gold price. Gold price is found o have a significan impac on exchange rae in he shor run. The generalized forecas error variance decomposiions demonsrae ha changes of gold price influence he forecas error variance of changes of exchange rae. There is a link beween he gold marke and he exchange rae marke. JEL Classificaion: F31, G11, G15 Keywords: Exchange rae, gold price, coinegraion 1. Inroducion Gold is a precious medal, which has inrinsic value. Gold is used widely as a sore of wealh. Gold can be used as agains inflaion or depreciaion of he Unied Saes (US) dollar. This is because gold is priced in he US dollar. When he US dollar falls, he nominal price of gold in he US dollar will rise o preserve he gold value. Gold can be as agains he US dollar exchange rae risk (Adibe and Fei, 2009: 3; Baur and McDermo, 2010). However, he role of gold as agains he US dollar varies from ime o ime (Capie, Mills and Wood, 2005: 352). Gold can be as agains currency as i is a homogeneous commodiy and easily raded in he well-organised spo and fuure markes. Gold canno be produced by he auhoriies ha produce currencies. Thus hose who can increase he money supply and herefore weaken is value canno do 1 Corresponding auhor: Locked Bag No. 2073, Koa Kinabalu, Sabah, Malaysia. address: hwong@ums.edu.my 1

2 so o weaken he gold value (Capie, Mills and Wood, 2005: 351; Sjaasad, 2008: 119). Since he breakdown of he Breon Woods in he year 1973, gold can be publicly raded wih lile governmen inervenion in many counries. Gold is no direcly linked wih he value of currency and also he moneary policy (Adibe and Fei, 2009: 5). The world winesses an aggressive growh in gold price in recen years (Figure 1). For he period, quarer 1, 1976-quarer 1, 2012, he mean of he gold price in he US dollar was per roy ounce. The sandard deviaion was The skewness and kurosis were and , respecively. The world gold price was volaile. The role of gold as an invesmen has drawn more aenion during global economic crisis in he year 2008 (Adibe and Fei, 2009: 3). In he ime of uncerainy, when he values of financial asses become ambiguous due o unwillingness of invesors o rade, he araciveness of gold increase due o he relaive simpliciy of he gold marke (Baur and McDermo, 2010). This sudy examines he relaionship of exchange rae and gold price in Malaysia. Gold ends o hold is value wih he depreciaion of exchange rae. Hence he relaionship beween exchange rae and gold price is negaive (Adibe and Fei, 2009: 10). The Cenral Bank of Malaysia does no regulae sricly he ransacion of he gold in Malaysia. Malaysia adops a managed floaing wih no pre-deermined pah for exchange rae mos of he ime (IMF, 2009). The ime series evidence of exchange rae and gold price is relaively limied. The previous sudies mainly focus he link such as beween gold price and oil price (Narayan, Narayan and Zheng, 2010; Zhang and Wei, 2010), beween gold price and inflaion (Sjaasad, 2008), beween gold price and sock price (Baur and McDermo, 2010; Basher, Haug and Sadorsky, 2012) and beween gold price and he exchange rae volailiy (Capie, Mills and Wood, 2005). Moreover, he previous sudies mosly examine he link for developed counries (Capie, Mills and Wood, 2005; Baur and McDermo, 2010; Wang and Lee, 2011). Joy (2011) examines he dynamic condiional correlaion beween gold price and he US dollar exchange rae, ha is, an increase in gold price ends o be lead o a decrease in he value of he US dollar. Wang and Lee (2011) invesigae he causaliy beween exchange rae and gold price in a non-linear hreshold vecor auoregressive model for Japan wih he assumpion he marke is imperfecly compeiive. On he oher hand, his sudy examines he relaionship beween exchange rae and gold price in a linear framework for a small and a relaively fas developing economy in Asia, namely Malaysia. Moreover, he relaive imporance of exchange rae and gold price in his sudy is examined using he generalized forecas error variance decomposiion (Koop, Pesaran and Poer, 1996; Pesaran and Shin, 1998), which solves he orhogonalised problem of he forecas error variance decomposiions of Sims (1980). Thus he 2

3 resul obained is invarian wih he posiion of he variables in he esimaion. Pukhuanhong and Roll (2011) invesigae he causaliy beween he gold reurn and he exchange rae change in a mulicurrency framework. However, he sudy did no examine he relaive imporance of he variables. The auoregressive disribued lag (ARDL) approach of Pesaran, Shin and Smih (2001) is used. The approach is suiable regardless all variables examined are I(1), I(0) or he mixure of I(1) and I(0) variables. On he oher hand, he Johansen coinegraion mehod is more relevan for all variables inegraed in he same order, namely I(1). Furhermore, his sudy includes he impac of he Asian financial crisis, in invesigaing he relaionship beween exchange rae and gold price. This sudy is srucured as follows. Secion 2 provides a lieraure review of exchange rae and gold price. Secion 3 illusraes he daa and mehodology and secion 4 provides empirical resuls and discussions. The las secion is concluding remarks. 2. Lieraure review Gold can be used as agains he value of currency. Capie, Mills and Wood (2005) invesigae he role of gold as agains he US dollar. The finding amongs ohers is ha gold serves as agains changes in he value of he US dollar. However, is role varies depending on unpredicable poliical aiudes and evens. Reboredo (2013) invesigaes he link beween he price of gold and exchange raes using differen copula funcions. The sudy uses weekly daa for he period from January 2000 o Sepember The resuls show significan posiive relaionship beween gold and he US dollar depreciaion agains differen currencies. This implies ha gold can be used as agains he US dollar movemens. The sudy concludes he usefulness of gold in he risk-managemen of a currency porfolio. Ciner, Gurdgiev and Lucey (2013) examine he correlaions beween oil, gold, currency, he bond and sock markes in he US and he Unied Kingdom (UK) using he quanile regression mehods. Gold acs as a safe haven when exchange raes drop significanly in boh he US and he UK cases. Currency depreciaion would lead an increase in he gold reurn. Pukhuanhong and Roll (2011) es he relaionship beween he gold reurn and exchange raes of he US dollar agains euro, Japanese yen and he UK pounds serling, respecively. The period of sudy is from 2 January 1971 o 10 December The resuls of Granger causaliy show ha lagged values of exchange rae changes influence he fuure gold reurns. Moreover, exchange raes of he US dollar agains euro, Japanese yen and he UK pounds serling, respecively exhibi he negaive relaionship wih he gold reurn. An increase in he gold reurn can be linked wih currency depreciaion. 3

4 Conversely, Sjaasad (2008) examines he heoreical and empirical relaionships beween exchange raes and gold price using he forecas error daa. The period of sudy is from 1991 o Floaing exchange raes is a main source of price insabiliy in he world gold marke and he world gold marke is dominaed by he US dollar bloc. Appreciaions or depreciaions of he US dollar would have significan impacs on gold price in oher currencies. Gold is no longer o be a sore of value agains world inflaion. Joy (2011) assesses he role of gold as a safe haven wih respec o he US dollar using a mulivariae generalized auoregressive condiional heeroskedasiciy (GARCH) model of dynamic condiional correlaions for 16 dollar-paired exchange raes (euro, Japanese yen, Indian rupee, Taiwan dollar, Ausralian dollar, Canadian dollar, Danish krone, Israeli shekel, Malese lira, New Zealand dollar, Norwegian krone, Singapore dollar, Souh African, Swedish krona, Swiss franc and he UK pounds serling). The daa are weekly from 10 January 1986 o 29 Augus The resuls show ha changes in he price of gold and changes in exchange raes are negaives. The negaive relaionships are increasingly in he year Quanile correlaions show gold does no ac as an effecive safe haven from marke sress. Gold has aced agains he US dollar. However, gold has been a poor safe haven. Gold is a safe haven for some sock markes bu no for ohers. Baur and McDermo (2010) analyse he role of gold as a safe haven agains socks in emerging and developing counries using daily, weekly and monhly daa for he period from 2 March 1979 o 2 March The resuls show ha gold is a safe haven for major European sock markes and he US bu no for Ausralia, Canada, Japan and large emerging markes such as Brazil, Russia, India and China. Moreover, he sudy repors ha gold can be as a sabilising force for he financial sysem by reducing losses in he exreme negaive marke shocks. Gold is a weak safe haven for some emerging markes. However, gold is a safe haven for mos developed markes during he financial crisis. Gold moves agains currency bu depends on he degree of he exchange rae change. Wang and Lee (2011) invesigae he causaliy beween he gold reurn and Japanese yen depreciaion rae in a non-linear hreshold vecor auoregressive model. The daa are monhly for he period from April 1986 o March The resuls show ha differen levels of Japanese yen flucuaion have differen effecs on he effeciveness of gold agains he exchange rae depreciaion. The effeciveness of gold as agains Japanese yen depends on he depreciaion rae of Japanese yen. More specifically, he resul shows ha holding gold can avoid lose when Japanese yen depreciaes agains he US dollar by more han he 2.62 percen else gold does no have a role agains he depreciaion of Japanese yen. Wang and Chueh (2013) examine he shor-run and long-run dynamic relaionships among 4

5 ineres raes, oil prices, gold prices and he US dollar using daily daa from 2 January 1989 o 20 December Gold prices, crude oil prices and ineres raes influence each oher in he shor run. In some longrun ranges, ineres raes lead gold prices. Changes in ineres raes can affec invesors expecaions of he US dollar, which will be ranslaed ino exchange rae. The US dollar will depreciae. Invesors will move heir capial o he gold marke for speculaion or capial preservaion, resuling in flucuaions in gold prices. Gold prices wen up because gold preserve values and provides hedging effecs. Generally, here is a negaive relaionship beween exchange rae and gold price (Capie, Mills and Wood, 2005; Joy, 2011). However, he role of gold as agains exchange rae varies across counries and ime (Baur and McDermo, 2010). There are sudies indicae ha here is a weak relaionship beween exchange rae and gold price. The empirical findings of he relaionship beween exchange rae and gold price are mixed. 3. Daa and Mehodology Exchange rae is Malaysian ringgi agains he US dollar exchange rae. Thus an increase in exchange rae implies depreciaion of Malaysian ringgi. The relaive money supply (RMS) is expressed as MS m, RMS, where MS is he money supply M2 (Malaysian ringgi MSus, in million and he US dollar in million, respecively), he subscrip m denoes Malaysia and he subscrip us denoes he US. The relaive IPI m, demand (RD) is expressed as RD, where IPI is he indusrial IPI us, producion index (2000 = 100). The ineres rae differenial (ID) is expressed as ID im, ius,, where i is he reasury bills rae. Gold price (GP) is expressed as GP GPw, ER, where GP w, is he world gold price in he US dollar per roy ounce (2000 = 100) and ER is Malaysian ringgi agains he US dollar. The sample period is from quarer 1, 1976 o quarer 1, All daa were seasonal unadjused. 2 Also, all daa were ransformed ino he naural logarihms before esimaion, excep he ineres rae differenial. All he daa were obained from Inernaional Financial Saisics, he Inernaional Moneary Fund (IFS, IMF). 2 This sudy has ried o include he seasonal dummy variables o capure he influence of seasonaliy in he esimaions, which resuls are no repored. However, he conclusions wih he seasonal dummy variables are he same as he esimaions wihou including he seasonal dummy variables. 5

6 Figure 1 displays he plos of he naural logarihms of exchange rae and gold price. Generally, here is no specific paern beween exchange rae and gold price. The correlaion beween he naural logarihms of exchange rae and gold price is 0.51***, which is significance a 1 percen level. The descripive saisics of he daa in his sudy are given in Table 1. The Jarque-Bera es saisics are found o be saisically significan. Figure 1 The Plos of exchange rae and gold price, 1976:Q1-2012:Q1 2,000 1,600 1, GP (he US dollar) log ER log GP log GP log ER Source: IFS, IMF. Noes: ER denoes exchange rae and GP denoes gold price. 6

7 Table 1 Descripive Saisics, 1976:Q1-2012:Q1 log ER log GP log RMS log RD ID Mean Median Maximum Minimum SD JB *** ** *** *** *** Correlaion Marices, 1976:Q1-2012:Q1 log ER log GP log RMS log RD log ER log GP 0.51*** log RMS 0.8*** 0.8*** 1 - log RD 0.79*** 0.71*** 0.94*** 1 ID 0.49*** 0.33*** 0.6*** 0.65*** Source: IFS, IMF. Noe: SD denoes sandard deviaion. JB denoes he Jarque-Bera es saisic. *** (**) denoes significance a he 1% (5%) level. The ess of correlaion are a 2-ailed es. The Zivo and Andrews (1992) (ZA), Lee and Srazicich (2004) (LS) and Perron (1997) (P) uni roo es saisics are used o examine he saionary of he daa. The power o rejec he uni roo null hypohesis declines if here is a srucural break in daa ha is ignored. The ZA uni roo es saisics are an augmened Dickey-Fuller ype endogenous break uni roo es. The LS uni roo es saisics are an endogenous uni roo es for one or wo srucural breaks ha is unaffeced by srucural breaks under he null and alernaive hypoheses and hus spurious rejecion will no occur. The LS uni roo es saisics are based on Lagrange Muliplier es. The P uni roo es saisics consider he dae of possible change is no fixed a priori bu is considered as unknown. The Pesaran, Shin and Smih (2001) (PSS) bounds esing approach is used o examine he long-run relaionship of he variables. The PSS bounds esing approach does no impose resricive assumpion ha all he variables are o be inegraed of he same order. The PSS bounds esing approach is said o be robus for finie samples even in he presence of regime shifs. The regime shifs can be modelled by using he dummy variables. The PSS bounds esing approach allows conclusion abou coinegraion among variables even wih he dummy variables, ha is, he asympoic heory developed in he PSS bounds esing approach is no influenced by he inclusion of he dummy variables. Moreover, he PSS bounds esing approach allows a differen number of opimal lags o be handled (Fuinhas and Marques, 2012). 7

8 The unresriced error correcion model for exchange rae is specified as follows: 3 log ER r i0 p q 10 11Trend 12D 13 i log RMS i 14 i i0 i0 s v 15iID i 16 ilog GP i 17 ilog ER i 18 log i0 i1 19 log RD 1 110ID 1 111log GP 1 112log ER 1 u 1, log RD RMS (1) 1 i where is he firs difference operaor, log is he naural logarihm, ER is exchange rae, Trend is he ime rend, D is he dummy variable o capure he influence of he Asian financial crisis, , RMS is he relaive money supply, RD is he relaive demand, ID is he ineres rae differenial, GP is gold price and u1, is a disurbance erm. The Wald or F-saisic is compued o es he null hypohesis, H0: 18 = 19 = 110 = 111 = 112 = 0 agains he alernaive hypohesis, Ha: If he Wald or F-saisic falls ouside he upper bound, he null hypohesis of no coinegraion is rejeced or log ER and is deerminans are said o be coinegraed. However, no conclusive inference can be made for he Wald or F-saisic falls inside he criical bounds. If he Wald or F-saisic falls below he lower bound, he null hypohesis of no coinegraion canno be rejeced. If here is evidence of coinegraion, he long-run model of he ARDL approach for exchange rae can be esimaed as follows: 4 log ER p q 20Trend 21D 22 i log RMS i 23 i log i0 i0 RD i r i0 24i ID i s i0 v 25i log GP i 26 i log ER i u2, i1 (2) where u2, is a disurbance erm. The orders of he lags in he ARDL model can be seleced by he Akaike Informaion Crierion (AIC). The model is esimaed by using he ordinary leas squares (OLS) esimaor. 3 The esimaed model is based on a model of he moneary approach of he exchange rae deerminaion and adding he gold price variable (Rapach and Wohar, 2002). 4 Pukhuanhong and Roll (2011: ) provide an explanaion of he correlaion beween exchange rae and gold price. 8

9 The error correcion model of he ARDL approach can be esimaed as follows: log ER p q i log RMS i 32 i log i0 i0 s i0 RD i r i0 ID 33i v 34i log GP i 35 i log ER i 36EC 1 u3, i1 (3) i where EC-1 is he one period lag of he error correcion erm and u3, is a disurbance erm. The error correcion erm is obained from he error erm of equaion (2). The general-o-specific modelling sraegy is used o esimae he error correcion model. Consider he vecor auoregressive (VAR) as follows: x p x i1 i i w, = 1, 2,, T (4) i where x is a (m 1) vecor of joinly deermined dependen variables, w is a (q 1) vecor of deerminisic and or exogenous variables, is a (m m) coefficien marix, is a (m q) coefficien marix, i is a vecor of (p 1) vecor of disurbance erms and T is he sample size. The equaion (4) can be rewrien as he infinie moving average represenaion as follows: x A G w, = 1, 2,, T (5) i i i0 i0 i i where Ai = 1Ai-1 + 2Ai pai-p and Gi = A. An impulse response funcion measures he ime profile of he effec of shocks a a ime on he fuure values of variables in a dynamic sysem. The generalized impulse response funcion of x a horizon n is defined as follows: GI x( 1 n 1 n 1 n,, ) E( x, E( x ) (6) Using equaion (6) in equaion (2), GI ( n,, 1 ) A, which is independence of -1 bu depends on he shocks defined as. x n 9

10 The scaled generalized impulse response funcion can be rewrien as follows: 1 2 j n) jj n ( A e, n = 0, 1, 2, (7) j where ej is an (m 1) vecor wih uniy as is j-h elemen and zeros elsewhere. Finally, he generalized forecas error variance decomposiion can be wrien as follows (Pesaran and Shin, 1998; Lee and Chien, 2010: 570): n 1 ii l0 n ' ei l0 ' 2 ( ei Al e j ) ij ( n), i, j = 1,, m (8) ' A A e l l i The generalized forecas error variance decomposiion examines he proporion of forecas error variance in one variable caused by he innovaions in he oher variables. Thus he generalized forecas error variance decomposiion assesses which variables are he mos endogenous and also assesses he relaive imporance of is own shock and shocks of oher variable. The main advanage of he generalized forecas error variance decomposiion is ha he esimaed resul is invarian o he posiion of he variables enered in he VAR (Koop, Pesaran and Poer, 1996; Pesaran and Shin, 1998). Therefore i does no pre-assume any ordering ha has heoreical implicaion. On he oher hand, he approach of Sims (1980) is sensiive o he posiion of he variables in he VAR. 4. Empirical Resuls and Discussions The ZA, LS and P uni roo es saisics are repored in Table 2. The lag lenghs used o esimae he ZA and LS uni roo es saisics are based on he -saisic, ha is, he number of lags for which he las included lag has a marginal significance level less han he cuoff given by he 10 percen level. The fracion of enries on each end of daa o exclude as he breaks and minimum gap beween breaks is he 10 percen level. For he P uni roo es saisics, he lag lenghs used are based on he minimum of he -saisic among he maximum of welve lags. The resuls of he ZA, LS and P uni roo es saisics show ha he variables are mosly no rejeced in heir levels and also afer aking he firs differences, excep he ZA uni roo es saisics (crash and break) show exchange rae is a saionary variable and he relaive money supply is a non-saionary variable, he LS uni roo es saisic (crash) 10

11 shows ha he relaive demand is a non-saionary variable, he P uni roo es saisic (crash) shows ha ineres rae differenial is a saionary variable and he P uni roo es saisics (break) show ha exchange rae, ineres rae differenial and gold price is a saionary variable. Thus he variables examined are he mixure of I(1) and I(0) variables and he use of he ARDL approach of PSS is suiable for he esimaion in his sudy. Table 2 The Resuls of he Zivo and Andrews (1992) (ZA), Lee and Srazicich (2004) (LS) and Perron (1997) (P) Uni Roo Tes Saisics ZA - Crash ZA - Break LS - Crash LS - Break P - Crash P - Break log ER -4.85** (1997:3) -7.83*** (1997:3) (1998:4) (1998:4) (1997:1) -7.84*** (1997:1) log ER -8.48*** (1998:4) -8.62*** (1997:2) -7.81*** (1997:1) -7.90*** (1996:2) *** (1997:3) *** (1997:3) log RMS (1992:2) (1991:4) (2007:1) (1993:2) (1993:1) (1993:1) log RMS (1998:1) (1998:1) -8.71*** (1989:2) -8.70*** (1988:3) -9.32*** (1997:3) -9.32*** (1997:3) log RD (1987:3) (1988:1) (1984:1) (1997:2) (1987:1) (1987:3) log RD *** (1990:1) *** (2008:1) (1982:4) -7.83*** (2008:3) -5.81*** (2011:2) -5.44* (2011:2) ID (1998:3) (1998:3) (1983:4) (1991:3) -5.46** (1998:1) -5.59** (1998:1) ID -6.40*** (1980:2) -6.46*** (1980:2) -5.47** (1991:4) -5.62** (2008:4) -5.84*** (1981:2) -5.84** (1981:2) log GP (2005:4) (1998:2) (1982:2) (1995:3) (2006:4) 6.57*** (1999:4) log GP *** (1980:2) *** (1980:2) -8.99*** (2000:2) -9.02*** (1992:1) *** (1980:4) *** (1979:3) Noes: Crash denoes he ZA, LS or P uni roo es saisic for esing an abrup change in level bu no change in he rend rae. Break denoes he ZA, LS or P uni roo es saisic for esing an abrup change in level and a change in he rend rae. Values in parenheses are he breaks. The criical values for he ZA, LS or P uni roo es saisics can be obained from Zivo and Andrews (1992), Lee and Srazicich (2004) and Perron (1997), respecively. *** (**) denoes significance a he 1% (5%) level. The PSS bounds esing approach is repored in Table 3. The Waldsaisic is found o be saisically significance a he 1 percen level. Hence here is a long-run relaionship beween exchange rae and is deerminans. In oher words, hose variables are moving ogeher and would no move oo far from each oher in he long run. Table 3 The Resuls of Bounds Tesing Approach for Coinegraion Wald-Saisic *** Noe: *** denoes significance a he 1% level. 11

12 The long run coefficiens of he ARDL approach are given in Table 4. The model fulfils he condiions of homoscedasiciy of error erm and no-funcional form bu i does no fulfil he condiions of normaliy of error erm and no-auocorrelaion. The condiion of no-auocorrelaion is found only o be significan a he 10 percen level. The coefficiens of all he variables are found o be saisically significan a he 1 percen level, excep he relaive demand. One possible explanaion for he insignifican of he variable is ha exchange rae could be acively managed raher han mainly refleced by he fundamenals. In he long run, an increase in he relaive money supply, he ineres rae differenial and gold price will lead o an appreciaion of he Malaysia ringgi. The negaive coefficien of he relaive money supply could be he resuls of heavily used of he moneary policy o achieve he argeed value of Malaysian ringgi. The posiive coefficien of he dummy variable which is used o capure he Asian financial crisis is found o be posiive. This implies ha he crisis had led o depreciaion of Malaysian ringgi. In he crisis, Malaysian ringgi was fixed o RM3.80 per one US dollar wih he aim o sabilise Malaysian ringgi and he economy as well. The fixed currency policy lased unil Table 4 The Long Run Coefficiens of he ARDL Approach (4,0,4,0,0) log ER = Trend*** D *** log RMS *** log RD ID *** log GP *** + u 2, Diagnosic ess: Adj. R 2 = Equaion Log-likelihood = AIC = LM = *. Rese = Normaliy = ***. Heero = Noes: Adj. R 2 is he adjused R 2. LM is he Lagrange Muliplier es of disurbance serial correlaion. Rese is he es of funcional form. Normal is he es of he normaliy of disurbance. Heero is he es of heeroscedasiciy.***(*) denoes significance a he 1% (10%) level. The error correcion model is repored in Table 5. The adjused coefficien of deerminaion (R 2 ) is The model fulfils he condiion of no-auocorrelaion bu does no fulfill he condiions of homoscedasiciy of error erm, no-funcional form and normaliy of error erm. Thus he model is esimaed by using he OLS wih he Whie's Heeroscedasiciy adjused sandard errors. The plos of cumulaive sum of recursive residuals and cumulaive sum of squares of recursive residuals indicae ha he model is relaively sable a he 5 percen level (Figure 2). The one-lagged error correcion erm is found o have he expeced negaive sign and saisically significan a he 1 percen level. This implies he validiy of an equilibrium relaionship among he variables in he esimaed model. There is some evidence ha gold price is found o have a significan impac on exchange rae. 12

13 The ineres rae differenial is also found o have a significan impac on exchange rae in he shor run. Table 5 The Error Correcion Model log ER = *** log RMS log RMS log RD ID *** ID -1** ID ID -3*** log GP log GP -2* log ER -1*** log ER -2** log ER EC -1*** + u 3, Diagnosic ess: Adj. R 2 = Equaion Log-likelihood = AIC = LM = Rese = ***. Normaliy = ***. Heero = ***. Noes: The esimaion is based on Whie's Heeroscedasiciy adjused sandard errors. *** (**, *) denoes significance a he 1% (5%, 10%) level. Figure 2 The Plos of Cumulaive Sum of Recursive Residuals (a) and Cumulaive Sum of Squares of Recursive Residuals (b) (a) (b) Noe: The sraigh lines represen criical bounds a 5% significance level. The generalized forecas error variance decomposiions are repored in Table 6. 5 The choice of he lag lengh used in he esimaions of he generalized forecas error variance decomposiions are based on he AIC. The resuls of he generalized forecas error variance decomposiions, which are repored, are based on he 1-5, 10, 15 and 20 horizon periods. On he whole, changes of exchange rae conribue o he forecas error variance of changes of gold price. Gold price accouns for abou 1.1 percen of he forecas error variances of exchange rae. Conversely, exchange rae accouns for abou 2.6 percen of he forecas error variances of gold price. This provides some evidence ha he wo markes are linked ogeher. 5 All variables are in he firs difference of he naural logarihm. 13

14 Table 6 The Generalized Forecas Error Variance Decomposiions log ER Horizon log ER log RMS ID log RD log GP log RMS Horizon log ER log RMS ID log RD log GP ID Horizon log ER log RMS ID log RD log GP log RD Horizon log ER log RMS ID log RD log GP

15 Table 6 The Generalized Forecas Error Variance Decomposiions (coninued) log GP Horizon log ER log RMS ID log RD log GP Noe: The VAR = 3 is used in he esimaion. The relaionship beween exchange rae and gold price is found o be negaive. This implies ha increase in gold price will lead o depreciaion of he US dollar. This sudy finds ha he negaive relaionship beween exchange rae and gold price in a linear approach whereas Wang and Lee (2011) repor he negaive relaionship beween exchange rae and gold price in a non-linear model. Therefore he negaive relaionship beween exchange rae and gold price can happen in he linear fashion. Malaysian ringgi seems o be relaively appreciaed agains he US dollar. This does no imply ha Malaysian ringgi appreciae agains gold price. Malaysian ringgi and he US dollar could be boh depreciaed bu relaively he US dollar depreciaed more han Malaysian ringgi. Pukhuanhong and Roll (2011) show ha increase in gold price is linked wih currency depreciaion in every counry. I is no rue ha only he US dollar depreciaed and he oher currency appreciaed. Thus gold can be used as agains for he US dollar or currency in general. In oher words, gold can be as agains inflaion or purchasing power of currency. The negaive relaionship beween exchange rae and gold price also implies ha exchange rae depreciaion will increase gold demand and herefore will increase gold price. Invesors can consider gold as par of heir porfolios in invesmen. Gold has he funcion of money. Gold is a good preserving value asse and hus inflaion resisance. The gold value is relaive sable compared wih many oher financial asses such as socks and bonds especially in he ime of uncerainy. In such as way holding gold means ha one has cerain level of purchasing power. A good porfolio can reduce risk and reain a beer rae of reurn. Moreover, he inensiy of gold holding gold is srongly correlaed wih global power (Aizenman and Inoue, 2013). 15

16 The Asian financial crisis, is found o have a significan negaive impac on Malaysian ringgi. Under financial crisis or uncerainy, holding commodiies such as gold can reduce he risk of lower purchasing power of currency. Thus gold as an alernaive o currency becomes more imporan. Capie, Mills and Wood (2005) amongs ohers show gold serves as agains he value of he US dollar. Conversely, Sjaasad (2008) demonsraes gold is no longer o be a sore of value agains world inflaion. Baur and McDermo (2010) also documen ha gold is a safe haven for mos developed markes during he financial crisis. Joy (2011) finds ha changes in he price of gold and changes in exchange rae is negaive and he negaive relaionship is increasingly during he global economic crisis in he year of Gold can be an alernaive of holding he US dollar especially when he US dollar urns o be weak. Wang and Lee (2011) demonsrae ha gold acs as agains exchange rae depreciaion. However, he effeciveness of gold as agains Japanese yen depends on he depreciaion rae of Japanese yen. There is a significan impac of gold price o exchange rae in he shor run. This means ha he informaion in he gold marke can be used o predic he exchange rae marke. Noneheless, he power of he predicion of he gold marke on he exchange rae marke could be small. One possible explanaion of he low power of he predicion is ha Malaysia adops a managed floaing exchange rae regime and he Cenral Bank of Malaysia could be acively inervene Malaysian ringgi in he marke and hus he value of Malaysian ringgi migh no closely refleced wih he fundamenals. The exchange rae and gold markes are linked. In oher words, here is a link beween he financial marke and he real secor. Any shock occurred in he real secor can be refleced in he financial marke. Pukhuanhong and Roll (2011) repor ha lagged values of exchange rae changes are influenial for fuure gold reurns. 5. Concluding Remarks This sudy has examined exchange rae and gold price in Malaysia. The resuls of he ARDL approach show ha here is long-run relaionship beween exchange rae and gold price. Moreover, here is a significan impac from gold price o exchange rae in he shor run. The negaive relaionship beween exchange rae and gold price implies ha gold can be an alernaive o holding currency. This can be more imporan especially during he financial crisis as he value of currency depreciaed grealy. The exchange rae marke and he gold marke are conneced. Consequenly informaion in he gold marke can be used o predic he exchange rae marke. There is sill a link beween exchange rae and gold price even afer he breakdown of he Breon Woods sysem. 16

17 References Aizenman, J. and Inoue, K. (2013). Cenral banks and gold puzzles. Journal of he Japanese and Inernaional Economies, 28, Basher, S.A., Haug, A.A. and Sadorsky, P. (2012). Oil prices, exchange raes and emerging sock markes. Energy Economics, 34(1), Baur, D.G. and McDermo, T.K. (2010). Is gold a safe haven? Inernaional evidence. Journal of Banking and Finance, 34(8), Capie, F., Mills, T.C. and Wood, G. (2005). Gold as a hedge agains he dollar. Journal of Inernaional Financial Markes, Insiuions and Money, 15(4), Ciner, C., Gurdgiev, C. and Lucey, B.M. (2013). Hedges and safe havens: An examinaion of socks, bonds, gold, oil and exchange raes. Inernaional Review of Financial Analysis, 29, Fuinhas, J.A. and Marques, A.C. (2012). Energy consumpion and economic growh nexus in Porugal, Ialy, Greece, Spain and Turkey: An ARDL bounds es approach ( ). Energy Economics, 34(2), Inernaional Moneary Fund (IMF). (2009). Classificaion of exchange rae arrangemens and moneary frameworks. hp:// Joy, M. (2011). Gold and he US dollar: Hedge or haven. Finance Research Leers, 8(3), Koop, G., Pesaran, M.H. and Poer, S.M. (1996). Impulse response analysis in nonlinear mulivariae models. Journal of Economerics, 74(1), Lee, C.C. and Chien. M.S. (2010). Dynamic modelling of energy consumpion, capial sock, and real income in G-7 counries. Energy Economics, 32(3), Lee, J., and Srazicich, M.C. (2004). Minimum LM uni roo es wih one srucural break. Appalachian Sae Universiy Working Paper, hp://econ.appsae.edu/repec/pdf/wp0417.pdf Narayan, P.K., Narayan, S. and Zheng, X. (2010). Gold and oil fuures markes: Are markes efficien? Applied Energy, 87(10), Perron, P. (1997). Furher evidence on breaking rend funcions in macroeconomic variables. Journal of Economerics, 80(2), Pesaran, M.H. and Shin, Y. (1998). Generalized impulse response analysis in linear mulivariae models. Economics Leers, 58(1), Pesaran, M.H., Shin, Y, and Smih, R.J. (2001). Bounds esing approaches o he analysis of level relaionships, Journal of Applied Economerics, 16(3),

18 Pukhuanhong, K. and Roll, R. (2011). Gold and he dollar (and he Euro, Pound, and Yen). Journal of Banking and Finance, 35(8), Rapach, D.E. and Wohar, M.E. (2002). Tesing he moneary model of exchange rae deerminaion: new evidence from a cenury of daa, Journal of Inernaional Economics, 58(2), Reboredo, J.C. (2013). Is gold a safe haven or a hedge for he US dollar? Implicaions for risk managemen. Journal of Banking and Finance, 37(8), Sims, C. (1980). Macroeconomics and realiy. Economerica, 48(1), Sjaasad, L.A. (2008). The price of gold and he exchange raes: Once again. Resources Policy, 33(2), Wang, K.M. and Lee, Y.M. (2011). The Yen for gold. Resources Policy, 36(1), Wang, Y.S. and Chueh, Y.L. (2013). Dynamic ransmission effecs beween he ineres rae, he US dollar, and gold and crude oil prices. Economic Modelling, 30, Zhang, Y.J. and Wei, Y.M. (2010). The crude oil marke and he gold marke: Evidence for coinegraion, causaliy and price discovery. Resources Policy, 35(3), Zivo, E. and Andrews, D. (1992). Furher evidence of he grea crash, he oil-price shock and he uni roo hypohesis. Journal of Business and Economic Saisics, 10(3),