Applied Economerics and Inernaional Developmen. AEID.Vol. 5-3 (5) EFFECT OF OIL PRICE SHOCKS IN THE U.S. FOR 1985-4 USING VAR, MIXED DYNAMIC AND GRANGER CAUSALITY APPROACHES AL-RJOUB, Samer AM * Absrac This paper uses bivariae VAR, Mixed Dynamic and Granger Causaliy Approaches, o analyze he news effec of oil prices on he sock marke index in he U.S during he recen oil price hikes from he lae eighies up o dae. All used models show similar evidence. They sugges ha oil shock negaively affec he sock marke reurns in he U.S. Oil prices granger cause movemens in he sock marke index. The Sock marke index will absorb he informaion of oil price shocks and incorporae i ino he sock price insananeously. Oil price shocks have an immediae negaive effec on he U.S sock marke. JEL classificaions: Q43;E44 Keywords: Oil shocks, Sock marke reacion. 1. Inroducion: Oil prices have risen very sharply afer he year 1991 up o dae o regiser a hisory record of 6 $ a barrel during June 5. Given his recen behavior of oil prices world markes will be severely affeced hrough negaive oupu reacion, slowed growh of oil imporing counries, and high inflaion. This sudy examined wheher he U.S. sock marke reac o he oil shocks, a big imporer of crude oil. Darby, 198; Hamilon, 1983; Burbridge and Harrison, 1984; Gisser and Goodwin, 1986; Mork, 1989; Ferdered, 1996 and Jones and Kaul,1996 among ohers examined he inerrelaion beween oil price changes, economic aciviy and sock marke for mos indusrialized counries. Mos of he exising work concenraed is effors on esing for hese inerrelaions in he pos World War II era * Samer AM AL-Rjoub, PhD.Deparmen of Banking and Finance The Hashemie Universiy Zarqa- 13133. Jordan 69
Applied Economerics and Inernaional Developmen. AEID.Vol. 5-3 (5) (Hamilon 1983). The assumpion was hen ha oil price changes or loosely oil price shocks may have several consequences on he real economic aciviies and sock marke changes. Alhough he bulk of he empirical research has sudied he consequences of oil price changes, lile research has been conduced on he recen oil price change caused by he hird gulf war on beginning of he year 3. Besides, few sudies have examined he effecs of oil shocks on he sock marke and economic aciviy for he Unied Saes (U.S), a big imporer of crude oil. This paper sudies he dynamic ineracions among oil prices, and sock reurns uilizing a mulivariae vecor auoregressive model (VAR), Mixed Dynamic and Granger casualy approaches for he U.S. The reminder of his aricle is organized as follows. Secion presen a review lieraure. Secion 3 explains he mehodological issue, Secion 4 concludes.. Lieraure review: Sudies ha deals wih oil-macroeconomy inerrelaionship are many (see for example Hamilon, 1983; Hamilon, 1988a, b; Mork, 1989; Hooker, 1997, 1999; Hamilon, 1996; Hooker, 1996a, b), bu he sudies ha examine aspecs of he impac of oil price on he sock marke are rare. Jones and Kaul (1996) es he possible effec of inernaional oil price shock on he curren and fuure cash flows and expeced reurns. He argues ha, in he poswar period, he reacion of only he U.S. and Canadian (his sample also includes unied Kingdome and Japan) sock prices o oil price shocks can be compleely accouned for by he impac of hese shocks on real cash flow alone. Recenly, Papaperou (1) sudied he oil priceeconomic aciviy-sock marke relaionship in a mulivariae VAR framework in Greece. She concludes ha oil price changes affec indusrial producion index (as a measure of real economic aciviy) and employmen. Also she concludes ha oil price shocks are imporan in explaining sock price movemens. Sadorsky (1999) esimaes VARs of oil prices, indusrial producion, hree-monh T- bill rae, and real sock reurns. He disinguishes beween posiive and negaive oil price shocks. Oil shocks have negaive impac on 7
Al-Rjoub, Samer AM Effec of oil price shocks in he U.S. for 1985-4 sock marke reurns ha lass for 3 mohs. Before 1986, posiive oil price shocks explain more of he forecas variance in real ock reurns han do negaive shocks. Afer 1986, posiive oil price volailiy shocks explain more of he sock marke reurns forecas errors han do ineres raes. More recen, Cunado and Garcia (3) esed for he oil price macroeconmy relaionship in a 16 indusrialized European counries. Their main resul indicaes ha oil price shocks has shor run and ling run effecs on producion growh raes. 3. Mehodological issue and daa: In he empirical analysis we perform a VAR analysis, Mixed Dynamic model and Granger Causaliy approach o explain oil price changes and is effec on sock reurns. All mehods of analysis allows use o es for he endogeneiy of all variables in he model and he response of sock reurns o oil prices shocks. The daa abou crude oil impors came from he monhly daa series of he Federal Reserve Banks of he U.S. The S&P 5 excess reurn is used o represen he sock marke aciviy. Real sock marke aciviy is measured as he difference beween he coninuously compounded reurn on he S&P 5 and he inflaion rae calculaed using he consumer price index. The choice of oil price variables is difficul, and some sudies show ha i is also relevan (See Cunado and Garcia, 3). We know ha oil prices have been influenced by producion quoas, price-conrols, high and varying axes on peroleum producs, poliical insabiliies and naional price variaions. Such consideraions jusify he use of he world price of crude oil in dollars. We have o noe ha we will no face he problem of convering currency since we are alking abou U.S dollars. Sudy daa covers he period from 1985 o 4 in order o examine he effec oil price shock on he U.S sock marke during which he Gulf war ook place and he crude oil prices reached is highs. 1 The evoluion of world oil prices and sock marke reurns are 1 The monhly oil price chronology up o he curren days is published by he Deparmen of Energy's Office of he Sraegic Peroleum Reserve. 71
Applied Economerics and Inernaional Developmen. AEID.Vol. 5-3 (5) shown in Plo 1. Plo 1 show ha here is an upward rend on oil prices and especially afer he beginning of he nineies. Plo 1 Evolusion of Oil Prices and Sock Marke Reurns 6 5 4 3 1 86 88 9 9 94 96 98 4 Oil Prices 1-1 - -3 86 88 9 9 94 96 98 4 Sock Marke Reurns In wha follows hree differen models are suggesed o es for he effec of oil price shock on he sock marke aciviy. We specify a simple model for he relaion beween oil price changes and he sock marke ha capures oil innovaion effec. The only exogenous variable is nominal oil price shocks. The equaion o be esed is n n = R + i Oil i + β i R i i = 1 i = 1 R α + e (1) Updaes for 1995-3 are from he Energy Informaion Adminisraion (hp://www.eia.doe.gov). 7
Al-Rjoub, Samer AM Effec of oil price shocks in he U.S. for 1985-4 Where R sands for he S&P 5 reurns index. R i is he R a differen lag lengh. Oil is he inernaional crude oil price impored in he U.S i By consrucion e is uncorrelaed wih all he pas R (s). If his is combined wih he fac ha e is also a linear combinaion of curren and pas R, e is serially uncorrelaed. Tes crierion s o deermine appropriae lag lenghs and/or seasonaliy are he mulivariae generalizaion of he Akaike informaion crierion (AIC) and he Schwarz crierion (SBC). I appears from running a simple VAR wih -lag lengh ha he appropriae lag selecion is. The second model, he mixed dynamic model can be wrien as follows: R = b1 D( Oil ) + br 1 + u where R sands for he S&P 5 reurns index. R 1 is he lag reurn. D(Oil) is he firs difference in he inernaional crude oil price impored in he U.S In he mixed dynamic model R is a funcion of is own lagged value and he firs differences of OIL. The mixed dynamic models according o Guisan (3) perform usually beer han firs differences models, because he laer is a paricular case of he former for he case when he coefficien of he lagged variable is exacly equal o one bu In he real world i happens very ofen ha he coefficien of he lagged endogenous variable is saisically differen of one, and in hose cases mixed dynamic model usually perform beer han he specificaion in firs differences. 73
Applied Economerics and Inernaional Developmen. AEID.Vol. 5-3 (5) Correlaion does no necessarily imply causaion in any meaningful sense of ha word he relaion can be simply spurious or meaningless. In order o es for ha, we recrui Granger Causaliy (Granger, 1969) es o deermine wheher any of he variables granger cause each oher. We es for granger causaliy beween oil shocks and he sock marke. We will deermine wheher oil shock causes he sock marke o reac i.e. o see how much of he curren oil shock can be explained by pas values of oil shocks and hen o see wheher adding lagged values of sock marke reurns can improve he explanaion. 3.1 VAR resuls VAR resuls are shown using he resuled impulse response funcion which races he effec of a one sandard deviaion shock o one of he innovaions on curren and fuure values of he endogenous variables. sandard errors used in he derivaion of he impulse response funcions are calculaed using asympoic analyical formula (see Hamilon, p.339). The resuled impulse-response funcions are represened in figure 1 and below. Figure 1 presens responses o a one sandard oil shock. A proraced decline in sock reurns follows an expeced raise in oil. An increase in oil is followed by a decline in oil and says for he year. The sock marke responses are proraced, u-shaped and small. Sock reurns roughs monhs afer he oil shock and dies ou afer four monhs, hen say sable afer ha. Sock reurns index decline by abou.5% following a one sandard deviaion shock. To be specific when combining oil and R, focus on he Oil Oil and Oil R responses ploed in figure if oil shock can affec sock Calculaing he sandard errors using Mone Carlo mehods using deferen scenarios of he number of repeiions gave similar resuls, giving he assumpion ha errors are normally disruped. 74
Al-Rjoub, Samer AM Effec of oil price shocks in he U.S. for 1985-4 marke, he sock marke response does measure he dynamic effec of he iniial oil shock. Figure 1: The Responses o Oil Shock. Horizonal axis in monhs, verical axis is response. The righ column represens a response o oil shocks. Response o One S.D. Innovaions ± S.E. 6 Response of RTR o RTR 6 Response of RTR o OIL 4 4-1 3 4 5 6 7 8 9 1 11 1-1 3 4 5 6 7 8 9 1 11 1 3 Response of OIL o RTR 3 Response of OIL o OIL 1 1-1 -1-1 3 4 5 6 7 8 9 1 11 1-1 3 4 5 6 7 8 9 1 11 1 Resuls sugges ha a posiive oil price shock depresses real sock reurns in he firs upcoming few monhs. Undersanding he properies of he forecas errors is exceedingly helpful in uncovering inerrelaionships among he variables in he sysem. So, in order o ge informaion abou he relaive imporance of he random innovaion, variance decomposiion 1 periods is run, and he resuls are in Table 1. As expeced, each ime series explains he preponderance of is own pas values; R explains over 97 % of is forecas error variance in 6 and 1 monh period, whereas Oil explains 95 % of is forecas error variance in one year period. I is 75
Applied Economerics and Inernaional Developmen. AEID.Vol. 5-3 (5) ineresing ha oil prices explain.47% and.53% of he forecas error variance of U.S sock marke index ( R ), whereas he sock marke index explains 4.13% and 4.7% of he forecas error variance of oil prices. Figure : Oil and Sock Marke Responses. Horizonal axis in monhs, verical axis is response. Response funcion is combined. Response of RTR o One S.D. Innovaions 5 4 3 1-1 1 3 4 5 6 7 8 9 1 11 1 RTR OIL Response of OIL o One S.D. Innovaions.5. 1.5 1..5. -.5-1. 1 3 4 5 6 7 8 9 1 11 1 RTR OIL Table 1 Variance Decomposiion Percenage of 4-monh Error Variance Variance Decomposiion of R: Period S.E. R OIL 1 4.49849 1.. 4.55699 97.67893.3169 3 4.566696 97.56383.436171 4 4.56775 97.55485.445145 5 4.567448 97.5389.46176 76
Al-Rjoub, Samer AM Effec of oil price shocks in he U.S. for 1985-4 6 4.567773 97.5546.474544 7 4.56884 97.5181.487195 8 4.568371 97.518.4989 9 4.56863 97.4944.59558 1 4.568865 97.4884.519163 11 4.56977 97.4716.57838 1 4.56969 97.46433.53567 Variance Decomposiion of OIL: Period S.E. R OIL 1 1.871877.348336 97.65166.99847 4.111911 95.8889 3 3.83967 4.4433 95.75577 4 4.46356 4.18877 95.8117 5 4.9743 4.1519 95.84998 6 5.394995 4.1734 95.8766 7 5.748774 4.111879 95.8881 8 6.565 4.163 95.89937 9 6.31837 4.9176 95.978 1 6.536987 4.85634 95.91437 11 6.73475 4.8446 95.91955 1 6.98463 4.765 95.9375 Ordering: R OIL Noe: 1- he firs column is he forecas error of he variable for differen forecas horizons. -he remaining columns give he percenage of he variance due o specific innovaions. One period ahead, all of he variaion in a variable comes from is own innovaion, so he firs number is always 1 percen. To examine wheher he resuls are sensiive o he ordering of variables, normally he choice is arbirary. This ordering issue becomes serious when here is a conemporaneous correlaion among he innovaions. When exiss, i can make significance difference for he variance decomposiion; se for example Gordon and King (198), Cooly and LeRoy (1985), and Leamer (1985). In order o examine he poenial sensiiviy of he innovaion accouning resul o ordering, anoher variable order is examined. 77
Applied Economerics and Inernaional Developmen. AEID.Vol. 5-3 (5) The assumpion here is ha, for he resuls o be conclusive, hey mus be robus o ordering. The new order is: oil is firs wih R second. Resuls in Table () and figure (3) shows ha he same resuls sill hold even when he orderings are changed; a suppor for he robusness of our resuls. Figure 3: The Responses o Moneary Shock. Horizonal axis in quarers, verical axis is response. Variables are in logs. The righ column represens a response o sock marke reurns shocks. Response o One S.D. Innovaions ± S.E. 3 Response of OIL o OIL 3 Response of OIL o RTR 1 1-1 1 3 4 5 6 7 8 9 1 11 1-1 1 3 4 5 6 7 8 9 1 11 1 6 Response of RTR o OIL 6 Response of RTR o RTR 4 4-1 3 4 5 6 7 8 9 1 11 1-1 3 4 5 6 7 8 9 1 11 1 78
Al-Rjoub, Samer AM Effec of oil price shocks in he U.S. for 1985-4 Table Variance Decomposiion Percenage of 4-monh Error Variance Variance Decomposiion of OIL: Period S.E. R OIL 1 4.49849. 1. 4.55699.3655 99.63745 3 4.566696.35576 99.6449 4 4.56775.3375 99.6767 5 4.567448.33853 99.69615 6 4.567773.969 99.7793 7 4.56884.8467 99.71593 8 4.568371.7885 99.717 9 4.56863.73946 99.765 1 4.568865.7588 99.7941 11 4.56977.6795 99.738 1 4.56969.65771 99.7343 Variance Decomposiion of R: Period S.E. R OIL 1 1.871877 97.65166.348336.99847 95.1835 4.781651 3 3.83967 95.168 4.831799 4 4.46356 95.15719 4.84814 5 4.9743 95.14198 4.85819 6 5.394995 95.1844 4.87156 7 5.748774 95.11555 4.884448 8 6.565 95.1364 4.89636 9 6.31837 95.985 4.97151 1 6.536987 95.831 4.916896 11 6.73475 95.743 4.95698 1 6.98463 95.6635 4.933647 Ordering: OIL R Noe: 1- he firs column is he forecas error of he variable for differen forecas horizons. -he remaining columns give he percenage of he variance due o specific innovaions. One period ahead, all of he variaion in a variable comes from is own innovaion, so he firs number is always 1 percen. 79
Applied Economerics and Inernaional Developmen. AEID.Vol. 5-3 (5) 3. Mixed Dynamic Model resuls Regressing curren reurns on oil shock and pas reurns afer adjusing Whie Heeroskedasiciy-Consisen Sandard Errors & Covariance give us he following resuls: RT = -.364 D (OIL) +.457RT (-1) -sa (-.33) ** (.638) The oil shock coefficien are negaive and significan a he 5% level indicaing ha unexpeced oil price increase have a negaive impac on he sock marke of 36%. Lagged sock marke reurn doesn convey any informaion abou he fuure movemens in he marke. 3.3 Granger Causaliy es resuls The granger causaliy resuls are shown in Table 3 below. The resuls are so ineresing we rejec he null and accep ha boh oil shock and sock marke reurns are granger cause each oher as can be see from he F-saisic and he probabiliy. The change in oil prices send negaive signals abou he performance of he economy and will be considered as a bad news where boh he oil shocks and sock marke reurns granger cause each oher as can be seen from he F-saisics, 1.57 and.6 respecively. Table 3: Pairwise Granger Causaliy Tess Lags:, number of observaions:37 Null Hypohesis: F-Saisic Probabiliy Reurns does no Granger Cause OIL 1.5736.976 OIL does no Granger Cause Reurns.655.167 4. Conclusion Oil prices changes do maer. Innovaions in he oil prices cause innovaions he sock marke. Using a bivariae VAR model, Mixed Dynamic approach, and Granger es of causaliy his paper examines 8
Al-Rjoub, Samer AM Effec of oil price shocks in he U.S. for 1985-4 he dynamic relaionship beween oil prices and sock marke reacion in he U.S during he oil price hikes in he lae eighies up o dae. The empirical evidence from VAR suggess ha oil shock affec he sock marke reurns in he U.S. Oil prices are imporan in explaining he sock marke reacions. Oil price shocks have an immediae negaive effec on he sock marke. Impulse response funcions show ha oil prices are also imporan in explaining sock price movemens. Mixed dynamic model show similar resuls of an immediae negaive impac of oil price shock on he sock marke wih significan coefficien of 36%. Granger es of causaliy rejec he null and accep ha boh oil shock and sock marke reurns are granger cause each oher as can be see from he F-saisic and he probabiliy. The change in oil prices send negaive signals abou he performance of he economy and will be considered as a bad news where boh he oil shocks and sock marke reurns granger cause each oher as can be seen from he F-saisics, 1.57 and.6 respecively.for all models specificaions he resuls sugges ha a negaive oil price shock depresses real sock reurns in he U.S sock marke. Reference Burbridge, J., Harrison, A., 1984. Tesing for he effecs of oil-price rises using vecor auoregressions, Inernaional Economic Review. 5 (1), 459-484. Cunado, J., Garcia. F.P., 3. Do oil price shocks maer? Evidence for some European counries, Energy Economics 5,137-154 Cooly, T., and S. LeRoy, 1985, A heoreical Macroeconomics: A Criique, Journal of Moneary Economics, 16, 83-38. Darby, M.R., 198.The price of oil and world inflaion in recession, American Economic Review 7,738-751. Ferdered, J.P., 1996.Oil price volailiy and he macroeconomy: A soluion o he asymmery puzzle, Journal of Macroeconomic 18,1-16. Gisser, M., Goodwin, T.H., 1986. Crude oil and he macroeconomy: Tess of some populal noions, Journal of Money Credi and Banking 18 (1), 95-13. Granger, C. W.J. 1969, Invesigaing causal relaions by economeric models and cross-specral mehods, Economerica 37, 44 438. 81
Applied Economerics and Inernaional Developmen. AEID.Vol. 5-3 (5) Guisan M.C., 1, Causaliy and coinegraion beween consumpion and GDP in 5 OECD counries: Limiaions of coinegraion approach, Applied Economerics Inernaional Developmen 1-1, pp 39-61. Guisan M.C., 3, Causaliy Tess, Inerdependence and Model Selecion: Applicaion o OECD counries, Universiy of Saniago de Composela, Working Paper Series Economic Developmen. Hamilon, J.D., 1983. Oil and he macroeconomy since World War II, Journal Poliical Economics 9(), 8-48. Hamilon, J.D., 1988. A neoclassical model of unemploymen and he business cycle, Journal Poliical Economics 96,593-617. Hamilon, J.D., 1996. This is wha happened o he oil-pricemacroeconomy relaionship, Journal Moneary Economics 38,15-. Hooker, M., 1996. Wha happened o he oil-price-macroeconomy relaionship?, Journal Moneary Economics 38,195-13. Hooker, M., 1999a. Are oil shocks inflaionary? Asymmeric and nonlinear specificaions versus change in regime, Federal Reserve Board, Mimeo. Jones, C.M., Kaul, G., 1996.Oil and sock marke, Journal of Finance 51(),463-491. King, R, and Charles Plosser, 1984, Money, Credi, and Prices in a Real Business Cycle, American Economic Review 74(June), 363-38. Koop,G.,Pesaran,M.H.,Poer,S.M.,1996.Impulse response analysis in nonlinear mulivariae models, Journal of Economerics 74,119-147. Leamer, E.E. 1985, Vecor auoregression for casual inference? Carneiqu-Rocheser Series on Public Policy, 55-33. Mork, K.A., 1989.Oil and he macroeconomy when prices go up and down: An exension of he Hamilon s resuls, Journal of Poliical Economics 97(3), 74-744. Papaperou, E., 1, Oil price shocks, sock marke, economic aciviy and employmen in Greece, Energy Economics 3,511-53. Sadorsky, P., 1999, Oil price shocks and sock marke aciviy, Energy Economics 1,449-469. Journal published by he Euro-American Associaion of Economic Developmen. hp://www.usc.es/econome/eaa.hm 8