oil price volatility and stock price fluctuations in an emerging market: evidence from south korea
TRANSCRIPT
Energy Economics 33 (2011) 975–986
Contents lists available at ScienceDirect
Energy Economics
j ourna l homepage: www.e lsev ie r.com/ locate /eneco
Oil price volatility and stock price fluctuations in an emerging market: Evidence fromSouth Korea
Rumi Masih a,1, Sanjay Peters b,2, Lurion De Mello c,⁎a JP Morgan Asset Management, New York, NY 10167, USAb Department of Economics, IESE Business School, 08034, Spainc Macquarie University, Sydney, Australia
⁎ Corresponding author at: Department of EconomicsNSW 2109, Australia. Tel.: +61 412560753.
E-mail addresses: [email protected] (R. M(S. Peters), [email protected], [email protected]
1 Tel.: +1 212 648 1723.2 Tel.: +34 93 253 4200.
0140-9883/$ – see front matter © 2011 Elsevier B.V. Aldoi:10.1016/j.eneco.2011.03.015
a b s t r a c t
a r t i c l e i n f oArticle history:Received 10 April 2007Received in revised form 26 March 2011Accepted 27 March 2011Available online 13 April 2011
JEL classification:F31C22C52
Keywords:Emerging marketsReal stock priceOil price shocksIndustrial productionGeneralized variance decompositionsImpulse response functions
How important are oil price fluctuations and oil price volatility on equity market performance? What are thepolicy implications if volatility turns out to be significant? We assess this issue in an economics/finance nexusfor Korea using a VEC model including interest rates, economic activity, real stock returns, real oil prices andoil price volatility. Ourmain aim is to capture the effects of crude oil prices on the Korean economy thoroughlycovering the period of the Asian Financial Crisis of 1997, which heavily affected the country, and the oil pricehikes in the early 1990s after the Gulf War. South Korea was the country most hit by the financial crisistogether with Indonesia and Thailand. Results indicate the dominance of oil price volatility on real stockreturns and emphasize how this has increased over time. Oil price volatility can have profound effect on thetime horizon of investment and firms need adjust their risk management procedures accordingly. Thisincrease in dependency has been found in other net oil importing emerging equity markets. We test therelationship between oil price movements and economic activity by using modern time series techniques in acointegrating framework. We expand the standard error correction model by examining the dynamics of outof sample causality through the generalized variance decomposition and impulse response functiontechniques. The evidence from persistence profiles also gives important guidelines based on how fast theentire system adjusts back to equilibrium. In addition, we find the cointegrating relationship to be stable andfind that the linear error correction model to be more favorable than an asymmetric 2 period Markovswitching model.
, Macquarie University, Sydney
asih), [email protected] (L. De Mello). 3 See Hamilton (1
4 See Heap (2005)5 See Stevens (200
l rights reserved.
© 2011 Elsevier B.V. All rights reserved.
1. Introduction
Since the oil price shocks of 1973–74 and 1979–80, dozens ofacademics and practitioners have explored the relationships betweenoil price shocks and the macroeconomic variables. The recessionaryimpacts of these oil prices shocks were too close for possible causallinks to be ignored, and considerable attention has been devoted tostudy the macroeconomics of these events. Policymakers have to takeserious account of the developments in the oil market, as a rise in theworld price of oil imposes macroeconomic costs in two ways. First, tothe extent that oil is both an important input to production andconsumer goods (i.e. petrol and heating oil), results in a reduction ineconomic activity as energy becomes more expensive. Second, risingoil prices contribute directly to the level of inflation, particularly in
energy dependent countries. Over time, the impact on activity andinflation will also depend on policy responses and supply-sideeffects.3
The high oil prices in 2005 and 2006 reflect the booming demandfrom Asia (especially China and India)4 and the geopolitical risk in theMiddle East5 (the “fear premium” estimated to add between $4 and $8to current prices). China and India have become two principal playerson the global energy scene. In 1990, consumption in these twocountries amounted to no less than 3.5 million barrels per day,approximately 5% of global petroleum use. In 2003, 13 years later,these two countries account for more than 10% of global oilconsumption. (BP Statistical Review of World Energy Markets, 2004).
It is difficult to distinguish temporary shocks from permanentshocks; and uncertainties related to large changes in oil prices can havesignificant effects on consumer confidence and therefore on growth.The impact of these oil price shocks is likely to be significantly greater inoil-importing countries, especially where policy frameworks are weak,
983), Bohi (1989), Bernanke et al. (1997).,Radetski (2006).5).
976 R. Masih et al. / Energy Economics 33 (2011) 975–986
foreign exchange reserves are low, and access to international capitalmarkets is limited.
Government authorities combat oil price hikes by using monetaryand fiscal policies. For example, in order to maintain high industrialproduction and exports revenues, interest rates are kept at low levels.However, it is difficult to evaluate the impact of the oil price shocks ondifferent variables of the macroeconomic environment, especially theimpact on the stock markets. It is perhaps noteworthy that it was onlyin the 1990s that researchers seriously examined the impact of oilprice shocks on stock markets.6 Macroeconomics and financialdynamics have not been captured together in one model whensubjected to oil price shocks, especially for a net oil importing andemerging economy such as South Korea.
The Republic of Korea (South Korea) is important to world energymarkets, as it happens to be the seventh largest oil consumer and thefifth largest net oil importer in the world. Korea relies entirely on oilimports as there are no oil reserves in the country or surroundingareas. Hence, being a net importer of oil, the movements orfluctuations in oil prices are of major relevance for the Koreangovernment when taking policy decisions that affect the nationaleconomy. The twomajor oil crises of the 1970s and 1980s significantlyaffected South Korea's macroeconomic performance. By usingmonetary and fiscal policies the country was able to weather thefirst oil price shock with some difficulties, but coping with the secondcrisis was much more challenging and South Korea experienced theworst stagflation between 1973 and 1990.
Korea's reliance on foreign sources for its energy requirementsdrastically increased between 1960 and 1990. For instance, Korea'stotal indigenous energy production to foreign imports continuouslydeclined from 54.77% in 1971, to 30.21% in 1980, and to 29.95% in1990. This, in turn, increased foreign energy dependence from 71.7%in 1971 to 77.95% in 1980 to 82.01% in 1990.7
More recently, petroleum accounted for 54 percent of SouthKorea's primary energy consumption in 2002. In 2004, the countryconsumed around 2.14 million barrels a day (bbl/d) of oil, all of whichwas imported.8 Glasure (2002) indicates that the real oil price is themajor determinant of real income and energy consumption in SouthKorea.
With the close dynamics between economic indicators andfinancial markets, many studies have used various proxies to illus-trate the degree and direction of causality in a cointegrating VARframework. To our knowledge these dynamics have not beenobserved together with exogenous oil price movements and oilprice volatilities.
This paper tries to answer the following questions. What is thelong-run relationship between oil price movements and stockmarkets in an emerging market like South Korea? Did the stochastictrends change between industrial production, interest rates, stockmarkets and oil price change during the financial crises and oil pricehikes in the early 1990s? What is the direction of causality betweenthese variables and what are the implications for the transmissionmechanisms of shocks? Can the domestic stock market be isolatedfrom oil price movements? Answers to these questions will haveserious fiscal and monetary policy implications not only to Korea butalso to other energy dependent countries. The negative impact of oilprice volatility on Korean industries could help the government inlooking at alternative less volatile sources of fossil fuels such asnuclear, coal and Liquefied Natural Gas (LNG) which has continued togrow in popularity especially in electricity generation.9 To ensure
6 Driesprong et al. (2003), argue that changes in oil prices strongly predict futurestock market returns in 12 out of 18 developed countries surveyed. South Korea is noton the list of the countries surveyed.
7 Glasure (2002).8 Data from Energy Information Administration (EIA), US Department of Energy.9 According to EIA estimates in 2008, South Korea primary energy consumption
constituted of 45% Petroleum, 27% Coal and 14% Natural Gas.
energy substitution, proper infrastructure needs to be in place tomake sure industries have the means to convert from oil intensive togas or coal intensive processes.
The paper is structured as follows. Section 2 provides a review ofthe literature and main debates surrounding the dynamics betweenstock markets and economic markets, together with impacts fromenergy and oil price movements. In Section 3 we describe the dataused in the analysis and briefly discuss econometric concepts andmethodology surrounding multivariate cointegration analysis and theout-of sample testing framework. The application and estimationresults are presented in Section 4 with some tables and figurespresented in Appendix A of the paper. In Section 5 we draw someimportant policy conclusions with respect to monetary policy andpolicies designed for stock markets to withstand oil price movements.
2. Literature review
James Hamilton's (1983) study of the role of oil price shocks in USbusiness cycles has had considerable influence on research on themacroeconomics of oil price shocks. As Mork et al. (1994) reviewpaper outlines, economists worked for nearly a decade on methods ofincorporating oil price shocks into macroeconomic models before asynergy developedwith real business cycle (RBC)models and oil priceshocks. This theoretical relationship betweenmacroeconomics and oilprice movements has been applied and tested using variouseconometric techniques.
Chaudhuri and Daniel (1998) use cointegration and causality todemonstrate that nonstationary behaviour of the US dollar realexchange rate is explained by nonstationary behaviour of real oilprices. The authors argue that oil price shocks can have long-runeffects on real exchange rates even if perfect markets exist in the longrun.
Greene et al. (1998) assess the impact of cartels like OPEC on theU.S. economy. They identify three main separate and additive typesof economic losses resulting from oil prices increases: the loss of thepotential to produce, macroeconomic adjustment losses and thetransfer of wealth from US oil consumers to foreign oil exporters.Whereas, Kaneko and Lee (1995) use an eight-variable VAR model totest the pricing influence of economic factors on U.S. and Japanesestock market returns and in identifying their relative importance in adynamic context. The eight variables used in this study are as follows:risk premium, term premium, growth rate in industrial production,rate of inflation, changes in terms of trade, changes in oil prices,change in exchange rates and excess stock returns. They find theaverage values of excess stock returns, rates of inflation, riskpremiums and term premiums to be higher for the United Statesthan for Japan.
Papaetrou (2001) on the other hand tests the dynamic linkagebetween crude oil price and employment in Greece using industrialproduction and industrial employment as alternative measures ofeconomic activity. His study is modelled in a cointegrated VARframework and extends out by looking at the generalized variancedecomposition and impulse response functions, which is veryencouraging as most studies have not gone beyond cointegrationand error corrections modelling.
Sadorsky's (1999) research meanwhile draws attention to anegative relationship between shocks in oil prices and real stockreturns for the US economy and a negative impact of shocks to realstock returns on interest rates and industrial production.
In a later study, Sadorsky (2001) finds a significant and positiverelationship between oil and gas equity index and the price of crudeoil in Canada. Furthermore the author indicates a positive relationshipbetween the return on the index and the return on the stockmarket asa whole. Finally a negative association is found between the stockmarket index value and both the premium on 3-month vs. 1-monthGovernment debt and the US/Canadian Dollar exchange rate.
977R. Masih et al. / Energy Economics 33 (2011) 975–986
Sadorsky's (1999) previous study on the US economy shows thatoil price volatility shocks have asymmetric effects on the economy. Byanalyzing the impulse response functions, he shows that oil pricemovements are important in explaining movements in stock returns:after 1986, furthermore, oil price movements explain a larger fractionof the forecast error variance in real stock returns than do interestrates. The results finally suggest that positive shocks to oil pricesdepress real stock returns while shocks to real stock returns havepositive impacts on interest rates and industrial production.
Faff and Brailsford (1999) study the sensitivity of the Australianindustry equity returns to an oil price factor and the sensitivity tomarket returns. Their results show a significant positive sensitivity inthe oil and gas and a negative sensitivity in paper and packaging, andtransport industries.
Faff and Chan (1998) meanwhile, empirically test the returns ongold stocks in the Australian equity market over the period 1979–1992. They find that there are just two factors explaining the returnson gold stock, namely, a market factor and a gold price factors.
Jones and Kaul (1996) however report that the reaction of theUnited States and Canadian stock markets to oil price shocks can beentirely explained on the basis of changes in the expected value offuture real cash flows. This evidence holds but is weaker for the UKand Japan. Jorion (1990) on the other hand, finds that the degree ofexposure of US multinationals to be positively and reliably correlatedwith the degree of foreign involvement.
3. Data and econometric methodology
The data were gathered from the International Financial Statisticsof the International Monetary Fund and consist of monthly observa-tions from May 1988 to January 2005 of the Korean stock marketindex, industrial production, interest rates and oil prices. The studythus thoroughly covers the period of the Asian Financial Crisis of 1997,which heavily affected the country,10 and the oil price hikes in theearly 1990s after Gulf War. The data collected allowed us to calculatethe real stock returns and the volatility of oil prices. In the results wereport industrial production as ip, real stock returns as rsr, interestrates as r, oil prices as lo, and oil price volatility as rvol.
In this paper we use a VAR model to explain the impact of oil pricechanges and volatility on real stock returns, industrial production andinterest rates. This methodological framework allows us test theendogeneity of all remaining variables when oil price shocks areintroduced as exogenous variables. We use two models to test ourdynamics. The first model uses oil prices, while the second uses theirvolatility. We test the reaction of industrial production, real stockreturns and interest rates when oil price movements and percentagegains or drops in oil price are introduced in the model. We employvarious diagnostic tests to verify linear or non-linear effects of crudeoil prices and crude oil price shocks on the stock market. One suchapproach is a bivariate Markov switching model on the residuals tosee if the speed of adjustment is asymmetric.
4. Model specification and empirical results
4.1. Unit root tests
It is standard in the literature on VARmodelling to employ a seriesof unit root tests to ensure our variables are I(1). To verify the order ofintegration of the variables we test for unit root based on Perron(1988), Phillips (1987), Phillips and Perron (1988) and Kwiatkowskiet al. (1992). The semi-parametric Phillips–Peron (PP) tests devel-
10 South Korea was the country most hit by financial crisis together with Indonesiaand Thailand. The $170.9 billion fall in 1998 was equal to 33.1% of the 1997 GDP, theexchange rate passed by 850 Won/ US$ of June 1997 to 1,290 Won/ US$ in July 1998(−34.1%).
oped by Phillips (1987), Phillips and Perron (1988), and Perron(1988) assume as its null hypothesis that a unit root exists in theautoregressive representation of the time series, while the nullhypothesis for the KPSS assumes the opposite (i.e., that a unit rootdoes not exist). The Dickey-Fuller tests (Dickey and Fuller, 1981),attempt to account for temporally dependent and heterogeneouslydistributed errors by including lagged sequences of first differences ofthe variable in its set of regressors. The PP tests try to account fordependent and IID processes through adopting a non-parametricadjustment, hence eliminating any nuisance parameters. Recentlythese tests have been shown, by Schwert (1987) and DeJong et al.(1992), to suffer from lack of power as they often tend to accept thenull of a unit root too frequently against a stationary alternative.
4.2. Modified DF-GLSτ test
In this paper, instead of the standard Augmented Dickey-Fuller testwe use themodified Dickey–Fuller test (DF-GLSτ) developed by Elliottet al. (1995). This test is conducted using the following regression:
1−Lð Þyτt−1 = a0yτt−1 + ∑
p
j=1aj 1−Lð Þyτt−j + μ t ð1Þ
where μt is a white noise error term and ytτ is locally detrended dataprocess under local alternative of α given by:
yτt = yt−Pβι′ zt ð2Þ
where zt=(1− t)' and β is the regression coefficient of yt and zt forwhich:
y1; y2;…; yTð Þ = y1; 1−αLð Þy2;…; 1−αLð ÞyTð Þ ð3Þ
z1; z2;…; zTð Þ = z1; 1−αLð Þz2;…; 1−αLð ÞzTð Þ ð4Þ
The t-test of the hypothesis H0: a0=0 against H0: a0b0 gives theADF-GLSτ test statistic. Elliott et al. (1995) recommended that theparameter c, which defines the local alternative by
α = 1 +cT
be set equal to −13.5. This test can attain a significant gain in powerover the traditional unit root tests. The critical values for Elliott et al.(1995) in Table 1 of Appendix A are estimated from Monte Carlosimulations. For finite sample correlations, Cheung and Lai (1995)provide approximate critical values. In the non-deterministic case, theuse of c = −7 is recommended where the test DF-GLSμ basicallyinvolves the same procedure as computing the DF-GLSτ test, apartfrom the exception that the locally detrended process series (ytτ) isreplaced by the locally demeaned series (ytμ) and zt=1. Theasymptotic distribution of the DF-GLSμ test is the same as that of theconventional DF test.
4.3. Confidence interval for the largest autoregressive root
ADF tests indicate the presence of a unit root in each series sincefor no series can the null of nonstationarity be rejected. To allow us tomeasure how persistent the unit root in the process is, we alsocalculate a confidence interval (CI) due to Stock (1991), who suggeststhat reporting CIs may provide useful information regarding samplinguncertainty. The confidence interval estimates tend to suggest thatthe unit root is quite persistent with all lower bounds quite clearlyabove 0.80 for both ADF (μ) and ADF (τ). We also supplement theseresults from Sims Bayesian unit root procedure which seem to besuggestive of a unit root with high value of a. Furthermore, Gewekeand Porter Hudack (GPH) tests for fractional integration, also quite
978 R. Masih et al. / Energy Economics 33 (2011) 975–986
uniformly suggest that most estimates of d fall significantly in theneighbourhood of 1. To check that these series are not integrated ofhigher orders, we also repeat these tests using first differences ofeach series. These results suggest that they are all stationary afterapplying the difference filter only once.11 Given the consistency andthe absence of ambiguity in results from all the mentioned testingmethods, we conclude that our variables are integrated at most orderone. This provides a requisite for the forthcoming multiple cointegra-tion analysis.
4.4. Multivariate cointegration analysis
As OLS estimates of cointegrating vectors, particularly in smallsamples, may be severely biased, in this analysis we employ the well-known Johansen and Juselius (JJ) procedure of testing for the presenceof multiple cointegrating vectors. It is demonstrated in Johansen(1992) that the procedure involves the identification of the rank of them by n matrix ∏ in the specification given by:
ΔXt = δ + ∑k−1
i=1ΓiΔXt−1 + ∏Xt−k + εt ð5Þ
where, Xt is a column vector of the m variables.Results of cointegration rank by the JJ procedure appear in Table 2
of Appendix A. Evidence from both traces and maximal eigenvaluetests suggest that there is at most a single cointegrating vector oranalogously 7 independent common stochastic trends within thisfour-variable system.12 This finding is consistent with studies byCorhay et al. (1993), Leachman and Francis (1995) and Jeon andChiang (1991) who, among others, find that equity markets of count-ries belonging to the G-7 countries possess at least one cointegratingvector. It is worth noting that an implicit assumption underlying thesetests is that events over this period such as the Asian financial crisesdid not significantly affect the stability of this system in terms ofaltering the number of common stochastic trends between the macroand financial variables. This issuemay now be tested using proceduresadvocated in the literature by Hansen and Johansen (1993), andQuintos and Phillips (1993). However, based on evidence usingsimilar techniques on a system of five OECD equity markets, Masihand Masih (1996c) find evidence that the crash did not affect thenumber of common stochastic trends within this particular system. Inthis study we find that the Asian financial and banking crises had verylittle effect on the stochastic trends between interest rates, industrialproduction, oil prices, oil price volatility and real stock returns.Besides observing these variables we have seen that South Korea wasone of better economies that withstood the pressures of currency
11 As a means of investigating the robustness of these results derived fromconducting tests for the total sample, we also undertake a sub-sample analysis ofthese tests taking the October 1987 crash as the break point. In order to save space,these results have not been reported for pre- and post-crash samples but availableupon request. Results, in general, indicate that the unit root approximation seems tobe quite robust to the October 1987 crash, since these sub-sample results do notchange our conclusion from conducting the tests over the full sample that thesevariables are integrated of at most order one. Once again, it is important to warnreaders that such results are very much vulnerable to low power due to poorperformance in small samples.12 Due to one of the biases of the JJ procedure being the sensitivity of cointegrationrank to the order of the lag length used in the VAR, We chose the lag subject to theAkaikes FPE criterion. In addition, results of a unique cointegrating vector wereinsensitive to slight modifications to lag length. Furthermore, there has been muchrecent work documenting the potential for severe small sample bias in Johansen tests(see Cheung and Lai (1995)).The scaling-up factor on the asymptotic critical valuessuggested by Cheung and Lai's study does not alter our conclusion of cointegrationrank. Furthermore, their study favors the trace test in that: it shows little bias in thepresence of either skewness or excess kurtosis, and is found to be more robust to bothskewness and kurtosis than the maximal eigenvalue test. (Cheung and Lai (1995,p.324)). In the light of this statement, the trace statistic of 215.64, further confirms ourinitial conclusion of r equal to at most 1.
meltdowns in Thailand and Indonesia and the severe banking crises inJapan in the early nineties.
In order to assess the relative strength of the long run relationship,Johansen and Juselius (1992) point out that larger eigenvaluesare associated with the cointegrating vector being more correlatedwith the stationary component of the process. To gain some insightinto the robustness of results for all five variables, we also conductedcointegration tests revealing r=1 at the 95% confidence level forboth models. Eigenvalues, presented in Table 2 of Appendix A are indescending order; indicate the cointegration relationship between thevariables.
Finally in order to test that each of the variables enters thecointegrating vector significantly, we test for zero restrictions uponeach of the coefficients derived by the Johansen procedure. Havingestablished the presence of a single cointegrating vector, the Johansenprocedure allows us to test several hypotheses on the coefficients byway of imposing restrictions and likelihood ratio tests which are,asymptotically, chi-square distributed with one degree of freedom.Scrutinising the cointegration vector in each model presents us witha measure of the most important component, in terms of its relativeweight, in comparison to the remaining components. Coefficientestimates and significance levels associated with the tests of zero-loading restrictions appear in Table 3 of Appendix A. Normalising oninterest rates in Korea provide evidence of each of these restrictionsbeing rejected, for most at least at the 10% level. This implies thatmostof the variables enter into the cointegrating vector at a statisticallysignificant level. However, although the weights of some of thevariables are not statistically significant, we cannot exclude this fromthe cointegrating vector as it forms a part of the long-run relationship.In general, these results indicate that almost all variables adjust in asignificant fashion to clear any short-run disequilibrium.
4.5. Short-run dynamics and long-run relations: Vector Error-CorrectionModelling (VECM)
Given the presence of a unique cointegrating vector in the nine-dimensional VAR used in the JJ cointegration tests, this then providesus with one error-correction term for constructing our models.Analogously, we may also extract (n− r) or four common trends, (forsuch an approach see Kasa (1992), Chung and Liu (1994)).
Summary results in Tables 1 and 2 offer some interesting insights.For each of the variables, at least one channel of Granger causality isactive: either the short-run through joint tests of lagged-differencesor a statistically significant Error Correction Term (ECT). This latterchannel is a novelty of the VECM formulation but it is noteworthy ofsignificance only in the rsr equation. The economic intuition arisingfrom this finding implies that when there is a deviation from theequilibrium cointegrating relationships as measured by the ECTs, it ismainly changes in the real stock returns that adjust to clear thedisequilibrium i.e. bears the brunt of short-run adjustment to long-run equilibrium. This leaves changes in interest rates, industrialproduction and oil prices, which appear to be statistically exogenousin both models and thus represents the initial receptor of anyexogenous shocks to their long-term equilibrium relationships.
Although the ECTs are not statistically significant for variablesother than real stock returns, one cannot assume that all othervariables are non-causal since the short-run channels are still active.For example, fluctuations in interest rates seem to explainmovementsin industrial production, and the exogenous oil price shocks seem tocause the biggest fluctuations in real stock returns. These short-runcausalities are explained by the significance of lagged differences inTables 1 and 2.
During the period of our study, South Korea did not have anymajorpolicy changes that had a direct impact on the variables in our model.The various financial sector reforms in the 1980s and 1990s wereacross the board. In addition to the financial crises in 1997/98 and the
Table 1VEC model estimates using real oil prices.
Equation Δrt Δrolt Δipt Δrsrt
ξ1, t−1 −0.019 −0.004 −0.000 −1.019(0.019) (0.001) (0.000) (0.160)
α 0.899 0.216 0.041 47.990(0.913) (0.068) (0.052) (7.55)
Δrt−1 0.296 0.011 0.001 −0.579(0.092) (0.007) (0.005) (0.759)
Δrt−2 −0.084 −0.003 −0.003 0.440(0.088) (0.007) (0.005) (0.729)
Δrolt−1 1.096 0.064 −0.141 15.303(1.215) (0.090) (0.069) (10.043)
Δrolt−2 -1.183 0.132 0.000 14.406(1.208) (0.090) (0.000) (9.992)
Δipt−1 −0.924 −0.005 −0.280 17.004(1.514) (0.113) (0.085) (12.516)
Δipt−2 0.643 0.171 −0.208 −4.130(1.472) (0.109) (0.083) (12.151)
Δrsrt−1 0.005 0.002 0.000 0.116(0.015) (0.001) (0.000) (0.122)
Δrsrt−2 0.012 0.000 0.000 0.098(0.011) (0.000) (0.000) (0.094)
R2
0.13 0.17 0.15 0.43σ̂ 0.661 0.052 0.042 6.760RSS 0.532 0.289 0.168 3638.812χSC2 [12] 21.421 16.56 14.666 4.819
χFF2 [1] 0.558 0.493 5.134 0.059
χNOR2 [2] 344.022 6336.890 32.301 0.714
χHET2 [1] 0.035 10.381 30.191 4.463
Notes: The underlying VAR model is of order 3 and contains unrestricted intercepts andrestricted trend coefficients. Lag order was selected by Schwarz Bayesian Criterion (SBC).Standard errors are given in parenthesis. The diagnostics are chi-squared χ2 (degrees offreedom) statistics for serial correlation (SC), functional form misspecification (FF), non-normal error terms (NOR) and heteroskedastic error variances (HET).
Table 2VEC model estimates using real oil volatility.
Equation Δrt Δrolvt Δipt Δrsrt
ξ1, t−1 −0.013 −0.001 −0.001 −0.626(0.012) (0.000) (0.001) (0.108)
α 2.035 0.212 0.239 99.474(1.983) (0.054) (0.108) (17.030)
Δrt−1 0.303 0.007 0.001 −0.388(0.094) (0.003) (0.005) (0.807)
Δrt−2 −0.110 0.000 −0.000 0.910(0.092) (0.003) (0.005) (0.791)
Δrolvt−1 −0.225 0.356 −0.503 65.539(3.195) (0.087) (0.174) (27.441)
Δrolvt−2 3.032 −0.076 0.129 10.258(3.342) (0.091) (0.182) (28.707)
Δlipt−1 −0.408 −0.029 −0.307 19.242(1.600) (0.043) (0.087) (13.747)
Δlipt−2 0.456 0.011 −0.233 −5.077(1.526) (0.041) (0.083) (13.108)
Δrsrt−1 0.002 0.001 0.001 0.055(0.014) (0.000) (0.000) (0.117)
Δrsrt−2 0.010 0.000 0.001 0.085(0.011) (0.000) (0.001) (0.095)
R2
0.12 0.32 0.21 0.39σ̂ 0.661 0.024 0.041 6.834RSS 52.231 0.043 0.148 3854.178χSC2 [12] 19.668 28.079 21.212 3.262
χFF2 [1] 0.667 2.434 4.851 1.512
χNOR2 [2] 366.331 7507.398 32.053 0.369
χHET2 [1] 0.611 1.592 25.134 1.795
Notes: The underlying VAR model is of order 3 and contains unrestricted intercepts andrestricted trend coefficients. Lag order was selected by Schwarz Bayesian Criterion (SBC).Standard errors are given in parenthesis. The diagnostics are chi-squared χ2 (degrees offreedom) statistics for serial correlation (SC), functional form misspecification (FF), non-normal error terms (NOR) and heteroskedastic error variances (HET).
979R. Masih et al. / Energy Economics 33 (2011) 975–986
high oil prices leading to the Gulf War in 2000, South Korea did suffera decline in 1989 spurred by a sharp decrease in exports and foreignorders which affected the industrial sector. Poor export performanceresulted from structural problems embedded in the nation's economy,including an overly strong won, increased wages and high labourcosts, frequent strikes, and high interest rates. The high labour costssaw industries investing heavily in automation and robotics technol-ogy which resulted in an increase in industrial products through the1990s and in early 2000 period.
It is important to test the stability of the error correction mecha-nismwhen weakly exogenous factors are present in the model. In thisstudy crude oil prices, interest rates and industrial production arefound to be weakly exogenous and the stability of these in the errorcorrection model are tested using the CUSUM and CUSUMSQ testsgiven in Figs. 1 and 2 of Appendix A. These tests are based on the nullhypothesis that the cointegrating vector is the same in every period;the alternative is simply that it (or the disturbance variance) is not.The test is quite general in that it does not require a prior specificationof when the structural change takes place. Since our study is not basedon a definitive piece of information, namely when the structuralchange takes place it is preferred to the Chow test which only worksbest when a definitive piece of information is at hand.
We test the stability of crude oil prices and interest rates afterestimating the VECMmodel of these variables as thesewere found to beexogenous in our model and shocks originating from them could havecaused some instability in the system. Both the plots of the cumulativesumof recursive residuals fall between the critical bounds and thereforesuggest that there is no cause for alarm in the structural stability of theVECM. The CUSUM plot is around zero and more importantly liesbetween the error bounds. The supporting CUSUMSQ which is deemedmore powerful than CUSUM also supports that crude oil prices andinterest rates did not cause instability in the cointegrating vector.
In addition to the above stability tests, we ran two bivariate coin-tegration tests between real stock returns and oil prices and real stock
returns and oil price volatility. Both models were found to be coin-tegrated and we tested for asymmetric effects in the error correctionmechanism by utilising a 2 regime Markov switching model. Inregime-switching models (like TAR, STAR and SETAR; see e.g. Fransesand van Dijk, 2000) the regimes and the switching mechanism areexplicitly defined through the threshold variable and the thresholdlevel. However, an upfront specification of this variable and levelis not a trivial task as the “regime switches” in our study are likelyto result from a combination of different fundamental drivers likeindustrial production, interest rates real oil price, real oil volatility. Onthe other hand, in Markov Regime Switching (MRS) models theswitching mechanism between the states is assumed to be governedby an unobserved (latent) random variable. MRS models do notrequire an upfront specification of the threshold variable and leveland, hence, are less prone to modelling risk. This gives them anadvantage in terms of parsimony.
A Markov-switching mean-adjusted auto regression can be writ-ten as:
Yt−μX Stð Þ = A + A1 Yt−1−μX Stð Þ½ � + A2 Yt−2−μX Stð Þ½ � + …
+ AN Yt−N−μX Stð Þ½ � + εt ; εteN 0;σ2� � ð6Þ
where St={1, 2} represents a two-regime state indicator variable (weallow for 2 regimes in this case). The parameter μX(St) can take twodifferent values, depending on the state which is prevailing in thesystem.
The transition among states is assumed to be governed by a first-order discrete Markov process. Denoting the pi, j the probability ofswitching from regime i to regime j, the two-regime transition matrixcontaining these transition probabilities can be written as:
P = p11 p21p12 p22
� �= p11 1−p22
1−p11 p22
� �ð7Þ
Table 3Generalized variance decompositions.
Horizon rt Rol ip rsr
Shock to interest rate (rt) explained by innovations in:1 80.55 8.30 0.83 10.316 79.27 6.34 1.01 13.3812 79.57 5.73 1.07 13.6324 79.68 5.40 1.11 13.82
Shock to economic activity (ipt) explained by innovations in:1 0.64 1.44 96.81 1.116 3.00 1.61 92.28 3.1112 3.45 1.51 91.52 3.5224 3.73 1.45 91.01 3.80
Shock to real stock return (rsrt) explained by innovation in:1 11.04 2.69 2.27 84.006 16.48 3.65 2.50 77.3712 24.09 3.67 2.32 69.9224 34.97 3.71 2.10 59.23
Table 4Generalized variance decompositions.
Horizon rt rolv ip rsr
Shock to interest rate (rt) explained by innovations in:1 82.84 6.79 1.97 8.396 78.92 7.52 2.14 11.4312 78.28 7.44 2.15 12.1324 77.89 7.41 2.17 12.53
Shock to economic activity (ipt) explained by innovations in:1 2.39 5.83 91.37 0.416 6.57 10.60 82.50 0.3312 7.34 11.30 81.07 0.2924 7.84 11.75 80.14 0.27
Shock to real stock return (rsrt) explained by innovation in:1 7.87 1.98 0.44 89.716 14.45 12.36 2.87 70.3312 19.60 17.87 2.64 59.88
980 R. Masih et al. / Energy Economics 33 (2011) 975–986
[The estimation below gives you a matrix of transition probabilitieswhich is the transpose of P above].
The state prevailing in the system in each point in time is inferredwith an efficient algorithm. Estimating the model above using amaximum likelihood procedure presents some computation compli-cations as the state and space equations for this model can only bewritten as conditional of each other. To estimate the model Hamilton(1989) proposed the use of an iterative expectation maximizationalgorithm which converges to the maximum value of the loglikeli-hood function.
We first test for asymmetry in the error correction mechanismbetween oil price and real stock returns by choosing a MSM(2)-AR(3)model, i.e. a model with 2 regimes and 3 lags. We select 2 regimes, asany higher would not be justified for the number of observations asthey increase significantly the number of parameters in the modelmaking it less parsimonious. In some cases we get a much worse AICand in other cases the algorithm fails to converge due to the lack ofobservations for each regime. The number of lags has been selected byconsidering AIC, HQ and SBC information criteria. The results arepresented in Table 4 of Appendix A.
Comparing the values of the log-likelihood for the estimatedmodel and its linear counterpart suggests that the Markov-switchingmodel does not produce any significant improvement with respect tothe linear model. Further analysis of the results suggests that there isno obvious specification error in the Markov-switching model.
Examining the regime classification in the graphs suggests thatthere is about 50% of probability of being in either regime 1 or 2. This isbecause 1 regime is associated with the positive residuals and theother one is associated with the negative values. So, after allowing themodel to “pick” any significant change in the residuals, the inferenceabout regimes classification shows no sign of structural change in thecontext of the AR model presented above. We conclude that theresiduals show no evidence of nonlinear behaviour that could beincorporated into the model. Further evidence is provided in Table 5and Fig. 3 of Appendix A.
We then estimated the same model on the bivariate relationshipbetween oil price volatility and stock returns and conclude that theerror correction mechanism is best represented by a linear model forthe period during the East Asian financial crises. A summary of theresults are given in Tables 6 and 7 and Fig. 4 of Appendix A.
24 24.43 23.32 2.45 49.81
Notes: the underlying cointegrated VAR model is of order 3, contains unrestrictedintercepts, and restricted trend coefficients. Lag order was selected using SchwarzBayesian Criterion (SBC). Standard errors generated from 10,000 replications arepresented in parenthesis. In Table 3 we cannot obtain VDCs for oil price because it isintroduced as an exogenous variable. In Table 4 we cannot obtain VDCs for oil pricevolatility as it is introduced as an exogenous variable. We do capture the out of sampledynamics in the subsequent impulse responses.
4.6. Generalized variance decomposition analysis
The relative strength of the Granger-causal chain amongst thevariables beyond the sample period is given by the Variance decom-positions (VDCs). VDCs provide a literal breakdown of the change in
value of the variable in a given period arising from changes in thesame variable in addition to other variables in previous periods. Avariable that is optimally forecast from its own lagged values will haveall its forecast error variance accounted for by its own disturbances(Sims, 1982).
The variance decompositions presented in Table 3 indicate that80.55% of shocks to interest rates are self-explained in the first month.This weighing stays at around 79% for 6, 12 and 24 months, suggestingthat other variables influence interest rates.We find that stock returnsand oil prices have a greater influence on the variance of interest ratesthan industrial production. Economic theory makes the link betweenindustrial production and interest rates through an increase ininvestment. An increase in investment results in an increase inindustrial production and then puts an increasing pressure on interestrates. In South Korea's case industrial production has minimal impacton interest rates and perhaps this indicates the country's capability toenjoy high industrial growth and not worry too much about inflation.This explains why South Korea had double digit growth in the 1980sand in early 1990s.
In Table 3 we also find that industrial production is stronglyexogenous with only small influences from interest rates and realstock returns. The brunt of the variance in endogenous real stockreturn variable is explained bymovements in interest rates. In the firstmonth, 11.04% of variance and in 6 months 16.48% variance in realstock return is explained by changes in interest rates. This influenceestablishes a link between instruments of monetary policy and thestock market. Movements in interest are indicative of the state of theeconomy and this embeds expectations among investors. An increasein variance of interest rates could suggest the direction of inflation inthe economy which in turn reflects whether economic activity haspicked up or slowed down. If interest rates are on the increase thaninvestors are likely to go easy on the stock market as the risk returntrade off in the bond market will become more attractive.
In Table 4 we get similar results except that oil price volatilityexplains a greater proportion of variation in industrial productionthan the level series of the oil price. Industrial production appears tobe less exogenous compared to Table 3 mainly because of theuncertainty caused by oil price volatility. After 6 horizons only 82.50%of shock is self-explained compared to 92.28 in Table 3. Oil price
981R. Masih et al. / Energy Economics 33 (2011) 975–986
volatility explains 10.60%, while interest rates explain 6.57% of theinnovations in industrial production. The endogeneity of the real stockreturns are illustrated by the strong causal links between interestrates and oil price volatility, with 14.45% and 12.36% of shocks in stockreturns being explained by the former and later.
13 Data from Energy Information Administration (EIA), US Department of Energy.
4.7. Impulse response functions
The information contained in the VDCs can be equivalentlyrepresented by graphs of the impulse response functions (IRFs). IRFsessentially map out the dynamic response path of a variable due to aone-period standard deviation shock to another variable. The IRFsbecome crucial in our analysis of oil price shocks, as VDCs cannot begenerated for exogenous variables.
Lee et al. (1992), Pesaran et al. (1993) and Lee and Pesaran (1993)discuss why generalized impulse responses are preferred to orthog-onalized response functions. Information from application of thesetools should provide some further evidence on the patterns of link-ages amongst stock markets and oil price shocks, as well as contributeto enhancing our insights upon how other macroeconomic variablesreact to system wide shocks and how these responses propagateover time. It is important to note, however, that although derivationof GIRFs does not suffer from the arbitrary orthogonalizations ofinnovations, GIRFs should not be strictly used to isolate responses of aparticular shock, assuming that all other shocks are not present, or notalso running in conjunction with the particular shock in question. Inthis respect, one should not attribute the shock, as in traditional IRFanalysis, to sole variables in the system, and thereby practice cautionwhen interpreting such results.
Generalized IRFs from one-standard deviation shocks to the modelusing log of oil price and oil price volatility in Korea are traced out foreach individual variable in Figs. 1–4 (including the own shock to eachmarket). In general the responses show long-lasting effects and thevariables take about 6 to 7 months to return to a new equilibriumlevel. Of all the variables interest rates seem to act positively inresponding to their own shocks and shocks to oil price and oil pricevolatility. Shocks to industrial production and real stock returns seemto have no upward pressure on interest rates. This shows that interestrates in South Korea are purely driven up by expectations embeddedin interest rates (perhaps long-term interest rates) and throughshocks in oil prices. We can see that industrial production is moresusceptible to shocks in the stockmarket and it takes about 11 monthsto reach to a new equilibrium state. The self-reactionary profile of oilprice and oil price volatility seem to settle back to their pre-shocklevels the quickest which is not surprising given the conclusions fromthe within-sample causality VECM and VAR results. Responses of thestock market are interesting when interest rates and oil prices areshocked. In both instances the stockmarket increases and then revertsback in the negative territory to its long-run level after about9 months. This shows the lag effect of interest rates and oil pricehave on stock market activity and thus shows that the Korean stockmarkets is of strong character in the short term. The identical reac-tionary profiles of real stock returns in Fig. 1 also suggest that infla-tionary expectations are evident through oil prices and throughmovements in short term interest rates.
In summary, impulse responses of the two models are very similarin nature indicate that volatility of oil price and log of oil price havethe same impact on the Korean economy. The IRFs show that theKorean economy is not affected adversely by oil price shocks anydifferently to normal oil price movements. The long-run time path ofreal stock returns in Figs. 1 and 3 when ROLV and ROL are shockedindicate a bigger impact from oil price volatility then from the levelseries. The new equilibrium for the stock market settles at a highernegative standard deviation level (−1.2) then through the impacts ofthe level of oil prices.
5. Main findings and conclusion
Few studies have thus far analysed the effects of oil pricefluctuations on the economic and financial variables of net importingcountries of oil. This study tries to capture the stochastic propertiesand long run dynamics between the macro economy, the stockmarkets, the instruments of monetary policy and the oil pricemovements.
The first result is that a long run equilibrium relationship doesexist among the four variables considered in the study. Furthermorethe system is stable throughout the entire period that is analyzed. Thismeans that the financial crisis did not affect the stability of the system,and banking has just a low effect on the stochastic trends betweeninterest rate, industrial production, oil prices and oil price volatility andreal stock returns. Interest rate and oil production appear to beexogenous variables in both our models. In other words they are thefirst receptors of any external shocks. The main conclusion of ourresearch is that oil pricemovements significantly affect the stockmarket.Our analysis indicates that real stock returns are the main channel ofshort-run adjustment to long-run equilibrium. After shocking oil pricesand oil price volatility (with a bigger effect in the second case), the stockmarket increases and then slows down, recovering to its long-runequilibrium level after a period of approximately 9 months. During thisperiod interest rates and oil prices spread their effect on the stockmarket. This conclusion confirms the linkage between real economyshocks, instruments of monetary policy and stock markets.
Results from our VECM suggest that real stock returns are themainchannel of short-run adjustment to long-run equilibrium. Oil priceshocks have two different negative effects on firm profitability. First, ithas a direct negative effect because it increases the production costs offirms. And secondly, it has an indirect negative effect because inves-tors foresee the decline in profit margins of firms and make decisionsthat affect the stock market indexes (i.e. selling shares).
For years, South Korea was one of Asia's fastest-growing economies,but the country's rapid industrialization was accompanied by acorresponding increase in its energy consumption. South Korea's rapidindustrialization process over the past several decades has resulted inthe country's industrial sector energy consumption increasing by morethan300%, from1.0quads in 1985 to 4.2 quads just over a decade later.13
Since high and volatile oil prices can have significant adversespillovers for the economy at large, there is in principle an argumentfor government intervention to reduce volatility. Three areas appearto be worthwhile for taking into consideration. First, governments ofoil-importing countries could benefit from increasing their strategicoil reserves, and thus protect themselves from the risk of supplydisruptions. Second, governments of oil-importing countries shouldconsider oil-saving measures very carefully. These measures includepolicies to improve energy efficiency, promote energy conservationand use of alternative fuels (i.e. coal, natural gas and renewableenergy). And finally, oil-importing countries should enhance dialoguewith oil-exporting countries in order to increase multilateralcooperation and to minimize shocks that have an adverse effect onthe national economy.
Acknowledgements
We would like to thank the two anonymous referees for theirpatience, comments and suggestions that greatly improved thepaper. We are very grateful for Julian Inchauspe for researchassistance.
The views expressed in this paper are those of the authors and notnecessarily shared by JP Morgan.
Generalized Impluse Response Paths when ROLV is shocked
Response of R Response of ROLV Response of LIP Response of RSR
Generalized Impluse Response Paths when R is shocked
Response of R Response of ROLV Response of LIP Response of RSR
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
0.018
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
0.00
0.00
0.00
0.00
0.00
-1.8
-1.6
-1.4
-1.2
-1.0
-0.8
-0.6
-0.4
-0.2
0.0
0.0
0.1
0.1
0.2
0.2
0.3
0.3
0.4
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
0
0.005
0.01
0.015
0.02
0.025
0.03
-1.40E-02
-1.20E-02
-1.00E-02
-8.00E-03
-6.00E-03
-4.00E-03
-2.00E-03
0.00E+00
-1.6
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
Fig. 1. Generalized impulse response function. Notes: The horizontal axis refers to months after shock. The vertical axis refers to standard deviations. Charts provide generalized impulse response functions (GIRF) of all variables in our modelwhen interest rates (R) and oil volatility (ROLV) are shocked. Dashed lines represent single standard error bounds around the point estimates. We can compare the above to Fig. 3.
982R.M
asihet
al./Energy
Economics
33(2011)
975–986
Generalized Impluse Response Paths when RSR is shocked
Response of R Response of ROLV Response of LIP Response of RSR
Generalized Impluse Response Paths when LIP is shocked
Response of R Response of ROLV Response of LIP Response of RSR
-0.18
-0.16
-0.14
-0.12
-0.1
-0.08
-0.06
-0.04
-0.02
0
-7.00E-03
-6.00E-03
-5.00E-03
-4.00E-03
-3.00E-03
-2.00E-03
-1.00E-03
0.00E+00
0.00
0.01
0.01
0.02
0.02
0.03
0.03
0.04
0.04
0.05
-1.2
-1.0
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
-0.5
-0.4
-0.4
-0.3
-0.3
-0.2
-0.2
-0.1
-0.1
0.0
-0.025
-0.02
-0.015
-0.01
-0.005
0
-0.002
-0.0015
-0.001
-0.0005
0
0.0005
0.001
0.0015
0.002
0.0025
0.003
-1
0
1
2
3
4
5
6
7
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 311 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
Fig. 2. Generalized impulse response functions (Continued). Notes: The horizontal axis refers to months after shock. The vertical axis refers to standard deviations. Charts provide generalized impulse response functions (GIRF) of all variablesin our model when log of industrial production (LIP) and real stock returns (RSR) are shocked. Dashed lines represent single standard error bounds around the point estimates. We can compare the above to Fig. 4.
983R.M
asihet
al./Energy
Economics
33(2011)
975–986
Generalized Impluse Response Paths when ROL is shocked
Generalized Impluse Response Paths when R is shocked
Response of R Response of ROL Response of LIP Response of RSR
Response of R Response of ROL Response of LIP Response of RSR
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
-0.01
-0.01
-0.01
0.00
0.00
0.00
0.00
0.00
-2.0
-1.8
-1.6
-1.4
-1.2
-1.0
-0.8
-0.6
-0.4
-0.2
0.0
0.0
0.1
0.1
0.2
0.2
0.3
0.3
0.4
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
-0.005
-0.004
-0.003
-0.002
-0.001
0
0.001
0.002
0.003
0.004
0.005
0.006
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 311 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
Fig. 3. Generalized impulse response functions (Continued). Notes: The horizontal axis refers to months after shock. The vertical axis refers to standard deviations. Charts provide generalized impulse response functions (GIRF) of all variablesin our model when interest rates (R) and oil prices (ROL) are shocked. Dashed lines represent single standard error bounds around the point estimates. We can compare the above to Fig. 1.
984R.M
asihet
al./Energy
Economics
33(2011)
975–986
1 3 5 7 9 1113151719212325272931 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 311 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
Generalized Impluse Response Paths when RSR is shocked
Generalized Impluse Response Paths when LIP is shocked
Response of R Response of LO Response of LIP Response of RSR
Response of R Response of LO Response of LIP Response of RSR
-0.12
-0.1
-0.08
-0.06
-0.04
-0.02
0
0
0.001
0.002
0.003
0.004
0.005
0.006
0.007
0.008
0.009
0.00
0.01
0.01
0.02
0.02
0.03
0.03
0.04
0.04
0.05
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
-0.5
-0.4
-0.4
-0.3
-0.3
-0.2
-0.2
-0.1
-0.1
0.0
-0.025
-0.02
-0.015
-0.01
-0.005
0
-0.002
-0.0015
-0.001
-0.0005
0
0.0005
0.001
0.0015
0.002
0.0025
0.003
-2
-1
0
1
2
3
4
5
6
7
Fig. 4. Generalized impulse response functions (Continued). Notes: The horizontal axis refers to months after shock. The vertical axis refers to standard deviations. Charts provide generalized impulse response functions (GIRF) of all variablesin our model when log of industrial production (LIP) and oil prices (LO) are shocked. Dashed lines represent single standard error bounds around the point estimates. We can compare the above to Fig. 2.
985R.M
asihet
al./Energy
Economics
33(2011)
975–986
986 R. Masih et al. / Energy Economics 33 (2011) 975–986
Appendix A. Supplementary data
Supplementary data to this article can be found online atdoi:10.1016/j.eneco.2011.03.015.
References
Bernanke, B.S., Gertler, M., Watson, M., 1997. Systematic monetary policy and theeffects of oil price shocks. Brookings Papers on Economic Activity 1997 (1), 91–142.
Bohi, D.R., 1989. Energy price shocks and macroeconomic performance. Resources forthe Future, Washington D.C.
BP, 2004. Statistical review of world energy June 2004. www.bp.com/statisticalreview2004.
Chaudhuri, K., Daniel, B.C., 1998. Long-run equilibrium real exchange rates and oilprices. Economics Letters 58, 231–238.
Cheung, Y., Lai, K., 1995. Lag order and critical values of the augmented Dickey–Fullertest. Journal of Business and Economic Statistics 13, 277–280.
Chung, P.J., Liu, J.D., 1994. Common stochastic trends in Pacific Rim stock market. TheQuarterly Review of Economics and Finance 34, 241–259.
Corhay, A., Tourani, A., Urbain, J.-P., 1993. Common stochastic trends in European stockmarkets. Economic Letters 42, 380–385.
DeJong, D., Nankervis, J., Savin, N.E., Whiteman, C.H., 1992. The power problems of unitroot tests in time series with autoregressive errors. Journal of Econometrics 53,323–343.
Dickey, D.A., Fuller, W.A., 1981. Likelihood ratio statistics for autoregressive time serieswith a Unit Root. Econometrica 49, 1057–1072.
Driesprong, G., Jacobsen, B., Maat, B., 2003. Striking oil: another puzzle. Research paperERS-2003-082-F&A. Erasmus Research Institute of Management (ERIM.
Elliott, G., Rothenberg, T.J., Stock, J.H., 1995. Efficient tests for an autoregressive unitroot, 1996. Econometrica 64, 813–836.
Faff, R., Brailsford, T.J., 1999. Oil price risk and the Australian stock market. Journal ofEnergy Finance and Development 49, 69–87.
Faff, R., Chan, H., 1998. A multifactor model of gold industry stock returns: evidencefrom the Australian equity market. Applied Financial Economics 8, 21–28.
Franses, P.H., van Dijk, D., 2000. Non-linear time series models in empirical finance.Cambridge University Press, United Kingdom.
Glasure, Y.U., 2002. Energy and national income in Korea: further evidence on the roleof emitted variables. Energy Economics 24, 355–365.
Greene, D.L., Jones, D.W., Leiby, P.N., 1998. The outlook of US oil dependence. EnergyPolicy 26, 55–69.
Hamilton, J.D., 1983. Oil and the macroeconomy since World War II. Journal of PoliticalEconomy 91, 228–248.
Hamilton, J.D., 1989. A new approach to the economic analysis of nonstationary timeseries and the business cycle. Econometrica 57 (2), 357–384.
Hansen, H., Johansen, S., 1993. Recursive estimation in cointegrated VAR-Models.Institute of Mathematical Statistics Working paper 1. University of Copenhagen,Copenhagen.
Heap, A., 2005. China_the engine of a commodities super cycle. Citigroup GlobalMarkets Paper, March 31.
Jeon, B., Chiang, T., 1991. A system of stock prices in world stock exchanges: commonstochastic trends for 1975–1990? Journal of Economics and Business 43, 329–338.
Johansen, S., 1992. Determination of cointegration rank in the presence of a lineartrend. Oxford Bulletin of Economics and Statistics 54 (3), 383–397.
Johansen, S., Juselius, K., 1992. Testing structural hypotheses in a multivariate cointegra-tion analysis of PPP and the UIP for UK. Journal of Econometrics 53 (1–3), 211–244.
Jones, C.M., Kaul, G., 1996. Oil and the stock markets. The Journal of Finance 51, 463–491.Jorion, P., 1990. The exchange-rate exposure of U.S. multinationals. The Journal of
Business 63, 331–345.Kaneko, T., Lee, B.S., 1995. Relative importance of economic factors in the U.S. and
Japanese stock markets. Journal of the Japanese and International Economies 9,290–307.
Kasa, K., 1992. Common Stochastic Trend in International Stock Markets. Journal ofMonetary Economics 29, 95–124.
Kwiatkowski, D., Phillips, P.C.B., Schmidt, P., Shin, Y., 1992. Testing the null hypothesisof stationarity against the alternative of a unit root: how sure are we that economictime series have a unit root? Journal of Econometrics 54, 159–178.
Leachman, L., Francis, B.B., 1995. Long run relations among the G-5 and G-7 equitymarkets: evidence on the Plaza and Louvre Accords. Journal of Macroeconomics 17,551–579.
Lee, K.C., Pesaran, M.H., 1993. Persistence profiles and business cycle fluctuations in adisaggregate model of U.K. output growth. Ricerche Economiche 47, 293–322.
Lee, K.C., Pesaran, M.H., Pierse, R.G., 1992. Persistence of shocks and its sources in amultisectoral model of UK output growth. The Economic Journal 102, 342–356.
Masih, A.M.M., Masih, R., 1996. Energy consumption, real income and temporalcausality: results from a multi-country study based on cointegration and error-correction modelling techniques. Energy Economics 18, 165–183.
Mork, K.A., Olsen, O., Mysen, H.T., 1994. Macroeconomic responses to oil price increasesand decreases in seven OECD countries. The Energy Journal 15, 19–35.
Papaetrou, E., 2001. Oil Price shocks, stock market, economic activity and employmentin Greece. Energy Economics 23, 511–532.
Perron, P., 1988. Trends and random walks in macroeconomic time series: furtherevidence from a new approach. Journal of Economic Dynamics and Control 12,297–332.
Pesaran, M.H., Pierse, R.G., Lee, K.C., 1993. Persistence, cointegration and aggregation: adisaggregated analysis of output fluctuations in the U.S. economy. Journal ofEconometrics 56, 57–88.
Phillips, P.C.B., 1987. Time series regression with unit roots. Econometrica 55, 277–302.Phillips, P.C.B., Perron, P., 1988. Testing for a unit root in time series regression.
Biometrika 75, 335–346.Quintos, C.E., Phillips, P.C.B., 1993. Parameter constancy in cointegrating regressions.
Empirical Economics 18, 675–706.Radetski, M., 2006. The anatomy of three commodity booms. Resources Policy 31,
56–64.Sadorsky, P., 1999. Oil price shocks and stock market activity. Energy Economics 21,
449–469.Sadorsky, P., 2001. Risk factors in stock returns of Canadian oil and gas companies.
Energy Economy 23, 17–28.Schwert, G.W., 1987. Effects of model specification on tests for unit roots in
macroeconomic data. Journal of Monetary Economics 20, 73–103.Sims, C.A., 1982. Policy analysis with econometric models. Brookings Papers on
Economic Activity 1, 107–152.Stevens, P., 2005. Oil markets. Oxford Review of Economic Policy 21, 19–42.Stock, J.H., 1991. Confidence intervals for the largest autoregressive root in U.S.
macroeconomic time series. Journal of Monetary Economics 28, 435–459.