-4 (2), 2011

European Journal of Economic and Political Studies

Effects of Crude Oil Price Changes on Sector Indices of Istanbul Stock Exchange
Cengiz Toraman1 - aatay Baarr2 - M. Fatih Bayramolu3

Abstract Oil price is one of the most important economic factors directing the world economy. A small change in oil prices has positive or negative affects on all the economic factors. Aim of this study is to investigate the affects of oil price changes on Istanbul Stock Exchange (ISE) 100 composite index, services index, industrial index and technology index of ISE. Long-run relationship is tested by Cointegration tests and short-run relationship is tested by Vector Error Correction Model (VECM). According to the results of the study, 32.71 % of the forecasting error variance of industrial index and 16.40% is explained by crude oil prices of the forecasting error variance of ISE 100 index. About the other indices; 12.60 % of the forecasting error variance of services index, 11.82 % of the forecasting error variance of financial index and 5.38 % of the forecasting error variance are explained by crude oil prices. Consequently, investors purposing to invest in ISE Market should consider the oil prices and especially investors of ISE industrial index and ISE 100 index should follow the developments in crude oil prices.

Keywords: Oil Price, ISE, Market Indices, Cointegration, VECM

(Corresponding author) Gaziantep University, cengiztoraman2004@[Link] Balkesir University Dogus University

109

Cengiz Toraman-aatay Baarr-M. Fatih Bayramolu

Introduction Oil price is one of the most important indicators of the economy. When we look at the literature, we can easily see that a small change in oil prices has either a negative or a positive effect on economy. Oil, first and foremost in energy sector, is used as an input in many sectors. In addition, it is a security tool. It reveals that oil prices must be considered when financial decisions are made (Bayramolu and Hamzaebi 2010: 303). Effects of oil prices on economy differ whether the country is an oil-importer or an-oil exporter country (Akgn 2006: 83). Oil production in Turkey can be seen at Table 1. Turkey is an oil-importer country as it can be seen in the table.

Table 1: Oil Import, Production and Consumption in Turkey (Million Tons) Production 2001 2002 2003 2004 2005 2006 2007 2008 2009 2,55 2,44 2,37 2,27 2,28 2,17 2,13 2,16 2,40 Import 23,14 23,70 24,02 23,91 23,39 23,78 23,44 21,83 14,21 Consumption 29,90 30,60 31,00 31,00 30,20 29,50 30,50 30,90 28,80

Resource: Collected from TSI and GDPA

Changes in oil prices affect oil-importer countries more. High oil prices cause a decline in real national income of oil-importer countries. Furthermore, oil is an input used for real sector production and this raises inflation via input prices. Objective of this study is investigating the affects of crude oil prices on services index, financial index, industrial index, technology index and ISE 100 composite index of ISE following the studies widely made for USA and European Stock Exchanges.

110

European Journal of Economic and Political Studies

Many studies were made about the relationship between the crude oil prices and stock exchange indices for different economies. In the first part of the study, a literature review is made in order to introduce the variables and the models applied for different countries. In the second part, data set and variables of the model are explained. In the following part, a VAR (Vector Autoregression) model is used to explain the relationship between crude oil prices and ISE sector indices. Last section evaluates the results of the model and concludes the paper.

Literature Review Sadorsky (1999) investigated oil prices and effects of oil price volatility on stock exchange returns. Sadorsky used a dataset including USA industrial index, interest rates and oil prices between 1947:11996:4 periods. According to the evidence, oil prices and oil price volatility affect economic activity. On the contrary, changes in economic activity have no effect on oil prices. It is also concluded that the stated effects expand after 1986. Papapetrou (2001) used a VAR model using monthly data of oil prices, real prices of stock Exchange, interest rates and employment rate between 1989:1 and 1996:6 period. It is concluded that changes in oil prices have effects on real economic activity and employment. Nonetheless, oil prices explain the movements of stock exchange to a large extend, but index returns have no effect on real activity and unemployment. Cunado and Garcia (2005) discuss the subject on a macroeconomic perspective, investigating the effects of oil price shocks on both economic activity and consumer price index. A co integration analysis, to find a long-run relationship, is made using the variables of six Asian Countries (Japan, Singapore, South Korea, Malaysia, Thailand and The Philippines) between 1975Q:1-2002Q:2 considering the data of real oil price levels, exchange rates, inflation rates, and industrial index and GDP; in addition they look for causality relationship to detect dynamic relationships. Unlike from the previous studies, different oil-output and oil- CPI definitions are used in order to measure 2000 oil price shock. As conclusion, no long-run relationship was found between oil prices and economic activity. In the short-run, a causality relationship was found between the variables. In all the analyzed countries, significant relationships between inflation and oil prices were detected, but these effects did not continue in the long-run.

111

Cengiz Toraman-aatay Baarr-M. Fatih Bayramolu

Cong and others (2008) made a VAR model discussing the relationship between oil price shocks and Chinese stock exchange market for different indices. Monthly data of Brent crude oil prices, exchange rate, interest rates, Shanghai Stock Exchange and Shenzhen Stock Exchange between 1996:1 and 2007:12 were taken while building different VAR models. As a conclusion no significant relationship was found between Chinese economic activity and stock exchange market. Furthermore, different indices of the stock exchange market have no significant response to oil price shocks. Effects of oil price shocks on real return of the stock exchange in 13 European countries between 1986:1 and 2005:12 periods was investigated by Park and Ratti (2008). With alternative VAR models, it becomes obvious that in many countries oil price shocks and economic activity, in addition oil price shocks and stock exchange market have a significant relationship. Another important result of the study is the varying relationships due to the countries. Odusami (2008) analyzed whether the non-linear oil price effects on US stock exchange can be explained by the unexpected shocks in crude oil market. Even it is concluded in the study that unexpected crude oil price shocks have non-linear effects on the return of US stock exchange market, lagged values of crude oil futures are found to have negative effects on the return of US stock exchange. Also, the relationship between OPEC meetings and US stock exchange returns was analyzed but no significant relationship was detected. A VECM was build by Miller and Ratti (2009) using monthly data of OECD countries between 1971 January and 2008 March. A long-run relationship was found between crude oil prices and world stock exchange indices for the stated countries. Results of the study reveal that there is a negative relationship between crude oil prices and world stock indices but the relationship weakens after the period 1999:9. Apergis and Miller (2009) tested the volatility of oil prices using a VECM considering 3 different factors: oil-supply shocks, global aggregate-demand shocks, and global oil-demand shocks. Eight countries were chosen (Australia, France, Canada, Germany, Italy, Japan, United Kingdom and USA). It was concluded that structural shocks of oil prices have a significant effect on explaining the return ratios of the stock exchange. Eryiit (2009) analyzed the effect of oil price changes on the sector indices of ISE between 2000 January and 2008 November using daily data and OLS method. Oil price changes have significant effects on Electricity, Whole sale and retail trade,

112

European Journal of Economic and Political Studies

Insurance, Holding, Investment, Wood, Paper, Printing, Basic metal, Metal products, Machinery and nonmetal and mineral product Indices at the 5 % significance level. In conclusion, oil price changes have important positive effects in Wood, Paper and Printing, Insurance and Electric sub-sector indices. In their study, Narayan and Narayan (2010) investigated the effects of oil prices on Vietnam Stock Exchange Prices during the period between 2000-2008. In addition to oil prices and stock exchange prices, nominal interest rates were also included in the study. It was found that stock exchange prices, oil prices and nominal interest rates are co integrated and oil prices have a significant effect on stock exchange prices in a positive direction. Gogineni (2010) tested the effect of daily oil price changes on stock exchange returns for different industries. It was concluded that return of industries highly depended on oil and also industries slightly depended on oil have sensitivity to oil prices. Furthermore, sectors are both cost-sensitive and demand-sensitive and these factors differ due to the industries. Lee and Chiou (2011) examined the asymmetric effects of oil prices on stock exchange indices using daily oil trade data between 1992 and 2008, and besides they investigated the effects of the regulations. They found a negative relationship for the variables, but they also stated that this relationship is valid in the periods when regulations for high price volatilities. Thus, importance of oil price shocks in the periods of economic fluctuations.

Methodology In this part of the study, the effects of oil prices on sector indices of ISE Market with a VECM using time series methods of co-integration and Granger Causality. For that purpose, data set will be defined first, and then time series properties of the series will be tested. Data Set and Variables Daily data between the period 02.01.200915.02.2011 is used in the study. The variables are crude oil prices (Brent Oil price per barrel for crude oil in US Dollars), ISE 100 index, industrial index, technology index, financial index and services index. Data set and definitions can be seen in Table 2.

113

Cengiz Toraman-aatay Baarr-M. Fatih Bayramolu

Table 2: Data Set Variable dbrent dISE100 dsnai dmali dhizmet dtekno Definition Crude Oil Prices ISE100 Compound Index ISE Industrial index ISE Financial Index ISE Services Index Index Brent Oil price per barrel for crude oil in US Dollars ISE Composite Index (Differentiated) ISE Industrial index (Differentiated) ISE Financial Index (Differentiated) ISE Services Index (Differentiated) (Differentiated)

The Model In this part of the study, a VAR model is used to explain the relationship between crude oil prices (Brent Oil price per barrel for crude oil in US Dollars), ISE 100 index, industrial index, technology index, financial index and services index. In this method, each variable is determined by the lagged values of both its own and the other variables. The purpose is to present the dynamic relationships for models in which more than one time series is considered (Bozkurt 2007, 76). In econometrical models, while some variables are explained by other variables in the model, a group of variables only used for explanatoriness. Variables that are explained in the system are called endogenous variables; and variables that help to explain the system are called exogenous variables (Kutlar 2005: 333). Economic variables interact each other continuously. For this reason, it is not always sufficient having one variable as endogenous variables and the other variables as exogenous variables in a model. A VAR model will help to test the causality between the variables without defining them as endogenous and exogenous (Bozkurt 2007:75).

Empirical Results First of all, the stationary of each variable were tested by unit root tests to determine their level of stationary. Stationary variables should be used in the

114

European Journal of Economic and Political Studies

model. Then, a VAR model is estimated, granger causality, impulse, response functions and variance decomposition were tested. With this method, the purpose is to present a dynamic system. A series with a constant arithmetic mean and variance, showing no systematic change or periodic fluctuations is called stationary time series (Ik et al. 2004, 330-331). Stationary of the series is a prior condition for forecasting. Spurious regressions, showing unreal relationships between the variables, should be obtained when non-stationary time series are used (zata and Esen 2010, 60). Stationary of the variables are tested the unit root tests of Augmented DickeyFuller (ADF) and Philips-Peron (PP). None of the series is found to be stationary as result of ADF and PP test results. All the series become stationary when their first differences are taken. For this purpose, differenced series are used in the analysis. Results of the unit root tests are presented in Table 3.
Table 3: ADF and PP Unit Root Test Results TEST ADF(c) ADF(t) PP (c) PP(t) dbrent -23.50 -23.48 -23.51 -23.49 dISE100 -22.43 -22.42 -22.42 -22.42 dsinai -21.77 -21.75 -21.75 -21.73 dtekno -21.13 -21.13 -21.06 -21.05 dmali -22.76 -22.76 -22.76 -22.76 dhizmet -21.85 -21.83 -21.85 -21.82

Cointegration Tests Long-run relationship between two or more non-stationary variables can be tested with cointegration tests. Cointegration method provides separating long-run effects and short-run effects and also estimating the adjustment rate to the long-run values (Badurlar 2008: 230). Results of ADF and PP tests show us that all the variables are stationary at the first difference. Fort his purpose, Johansen cointegration test developed by Johansen (1988) is used to present the long-run relationship among the variables. Lagged ratios of the variables are specified by VAR model before the cointegration test. Lagged ratio is taken as 2 according to the Akaike information

115

Cengiz Toraman-aatay Baarr-M. Fatih Bayramolu

criteria. Lagged ratio determination criteria, Johansen cointegration test, trace test and maximum Eigenvalue test results are presented in table 4 and Table 5.

Table 4: Trace Test Results

0.325333

988.3555

95.75366

0.0001

At most 1 (r1)

0.306577

779.7815

69.81889

0.0001

Table 5: Maximum Eigenvalue Test Results

Hypothesized No. of CE(s)

Eigenvalue

Probability

None *

0.325333

40.07757

0.0001

At most 1 (r1)

0.306577

194.0408

33.87687

0.0001

Trace test and Maximum Eigenvalue test results shows that we can reject the null hypothesis: Ho: There is no cointegration, And we can state that there is a long-run relationship between the variables. When there is a long-run relationship between the variables, Engle and Granger (1987) suggest a VECM to determine short run causality relationship between the variables.

Vector Error Correction Model (VECM) VECM suggest that a bias in the long-run relationship can be corrected. The main advantage of the error correction model, it enables to benefit the sub information of

116

European Journal of Economic and Political Studies

the series in the short-run and long-run. It also provides to eliminate the spurious regression (Sevktekin and Nargeleekenler 2010, 483-484). Therefore, for long-run relationship, a dynamic specification of VECM can be defined as: (1) (1) (2) (2) ECTt-1 is the lagged value of error correction model. Coefficients 1 and 2 show the 2and from the equation 1 equilibrium ratio. When cointegration is considered, from the equation are tested whether they are significant in group by F-test and also coefficients of the error correction model 1 and 2 are tested whether 2 or 2significant not (zer and Trkylmaz 2010, 101).

2 2

Granger Causality Test When there is a cointegration relationship between the variables, there must be at least one way causality relationship between the variables. Results of the causality test using vector error correction model is presented in Table 6.

Table 6: Granger Causality Test Results

0.235945 9.104088 0.736108 0.787781

0.8887 0.0105 0.6921 0.6744

117

Cengiz Toraman-aatay Baarr-M. Fatih Bayramolu

H0: Crude Oil Prices does not Granger cause Services Index H0: ISE 100 does not Granger cause Crude Oil Prices. H0: Industrial index does not Granger cause Crude Oil Prices H0: Technology Index does not Granger cause Crude Oil Prices H0: Financial Index does not Granger cause Crude Oil Prices H0: Services Index does not Granger cause Crude Oil Prices

1.640007 0.641355 11.75543 4.091373 1.036068 4.850526

0.4404 0.7257 0.0028 0.1293 0.5957 0.0885

According to the Granger Causality test results, a mutual causality relationship is seen from crude oil prices to industrial index, from industrial index to crude oil prices. Accordingly, crude oil prices granger cause industrial index and industrial index granger cause the crude oil prices.

Impulse Response Functions After testing causality, responses of the indices to an impulse in crude oil prices are analyzed. Impulse response functions reveal the effects of an unexpected shock given to a variable on the future values of its own and also other variables. As a result of impulse response functions, dynamic relationships can be observed among the variables and adjustment process can be detected.

Graph 1: Graphs of Impulse-Response Functions

R e s po n s e of D IMK B 1 0 0 t o D B R E N T

10

11

12

10

11

12

R e s po ns e of D MAL I t o D B R E N T
600 500 400

R e s p o ns e o f D S IN AI to D B R E N T

118

240 220 200

1 .6 1 .2 0 .8 0 .4 0 .0

300 250 200

European1 5 Journal of Economic and Political Studies 0

R e s po ns e of D MAL I t o D B R E N T
600 500 400 300 200 100 240 220 200 180 160 140 120

R e s p o ns e o f D S IN AI to D B R E N T

10

11

12

10

11

12

R e s p o ns e o f D H IZM E T to D B R E N T
180 160 140 120 100 80 60 70 60 50 40 30 20 10

R e s po ns e of D T E K N O to D B R E N T

10

11

12

10

11

12

Graph 1 shows the impulse response functions of the variables. According to the impulse response function, an unexpected change of crude oil prices has impact on ISE 100 indexes, financial index and technology during 2 periods and afterwards the effects disappear. An unexpected change of crude oil prices has impact on industrial index and services index during 4 periods and afterwards the effects disappear. Variance Decomposition Variance decomposition shows the percentage of changes of each variable explained by the changes in the other variables and changes of that variable. Results are defined as percentages and 100 in total (ztrk 2008: 120).Variance decomposition explains the change in one variable as shocks affecting all the endogenous variables. Variance decomposition gives information about the dynamic properties of the system. Purpose of the variance decomposition is to reveal the effect of each random shock to forecasting error term.

119

Cengiz Toraman-aatay Baarr-M. Fatih Bayramolu

Table 7: Variance Decompositions of dIse 100 Index Period 1 6 12 dbrent 15,04 16,06 16,40 dISE100 84,96 82,76 82,67 dmali 0 0,15 0,14 dsnai 0 0,55 0,35 dhizmet 0 0,29 0,27 dtekno 0 0,19 0,17

Table 8: Variance Decompositions of dsinai Index Period 1 6 12 dbrent 17,80 29,36 32,71 dISE100 62,81 56,92 55,62 dmali 8,62 6,78 6,19 dsnai 10,77 6,33 4,81 dhizmet 0 0,06 0,05 dtekno 0 0,55 0,62

Table 9: Variance Decompositions of dservices Index Period 1 6 12 dbrent 11,80 12,49 12,60 dISE100 55,68 55,62 55,78 dmali 19,20 20,88 21,30 dsnai 7,35 6,68 6,46 dhizmet 5,94 4,21 3,75 dtekno 0 0,12 0,09

Table 10: Variance Decompositions of dmali Index

120

European Journal of Economic and Political Studies

1 6 12

12,82 12,02 11,82

84,19 84,78 85,59

2,99 2,007 1,70

0 0,61 0,36

0 0,38 0,38

0 0,18 0,14

Table 11: Variance Decompositions of dtekno Index Period 1 6 12 dbrent 5,61 5,42 5,38 dISE100 35,23 37,29 37,89 dmali 3,62 4,47 4,63 dsnai 3,66 3,53 3,33 dhizmet 0,59 0,49 0,29 dtekno 51,27 48,78 48,47

Variance decompositions of dISE 100 index, dmali index, dsinai index, dhizmet and dtekno are shown in tables 7, 8, 9, 10, 11 respectively. Results are evaluated in accordance with the importance of the variables. As a result of the variance decomposition tests, 32.71 % of the error variance of industrial index and 16.40 % of the ISE 100 index is explained by crude oil prices; for the other variables, 12.60 of the forecasting error variance of services index and 11.82 % of the forecasting error variance of financial index and 5.38 % of the forecasting error variance of technology index is explained by crude oil prices. As a conclusion, among the different sectors, industrial index is affected by the changes of crude oil prices at the most and technology index is affected by the changes of crude oil prices at least.

Conclusion Consequently, a small change in crude oil prices affects all the economic factors positively or negatively. 16.40 % of the ISE 100 index is explained by crude oil prices. When we look at the other indices, 32.71 % of the error variance of industrial index, 11.82 % of the forecasting error variance of financial index, 12.60 of the forecasting error variance of services index and 5.38 % of the forecasting error variance of technology index is explained by crude oil prices. The industrial index is the most

121

Cengiz Toraman-aatay Baarr-M. Fatih Bayramolu

sensitive index and the technology index is the least sensitive index among all the variables. Similar to the previous studies, indices are linked to the crude oil prices in some extent. Crude oil is one of the important raw materials that are used by the manufacturers composing industrial index. For this reason, industrial index is heavily affected crude oil prices. Under this conclusion, investors purposing to invest in ISE Market should consider the oil prices and especially investors of ISE industrial index and ISE 100 index should follow the developments in crude oil prices. Changes in crude oil prices would shed some light on the rise and fall of ISE indices. Further researches can be made for other indices such as sub sector indices.

References Akgn, Ali. 2006. Petrol Fiyatlarndaki Deiimlerin ISE-100 Endeksine [Link] Lisans Tezi.,Seluk niversitesi. Apergis, N. and Miller, S.M. 2009. Do Structural Oil Market Shocks Affect Stock Prices. Energy Economics, 31, 569575. Badurlar, . ner. 2008. Trkiyede Konut Fiyatlar ile Makro Ekonomik Deikenler Arasndaki likinin Aratrlmas. Anadolu niversitesi Sosyal Bilimler Dergisi 8: 223-238. Bayramolu, F. Mehmet. and Hamzaebi Cokun. 2010. Ham Petrol Fiyat Hareketlerinin Snflandrlmasnda Yapay Zeka Teknolojilerinin Kullanm., leri Teknolojiler altay 2010 Bildiriler Kitab, Kocaeli, Nisan 303317 Bozkurt, Hilal. 2007. Zaman Serileri Analizi. Bursa: Ekin Yaynevi. Cong, R.G, Wei, Y.M, Jiao, J.L and Fan, Y. 2008. Relationships Between Oil Price Shocks and Stock Market: An Empirical Analysis From China. Energy Policy 36: 3544 3553. Cunado, J. and Gracia, F.P. 2005. Oil Prices, Economic Activity and Inflation: Evidence For Some Asian Countries. The Quarterly Review Of Economic and Finance 45: 6583. Engle, R.F. and Granger, Cwj. 1987. Cointegration and Error Correction: Representation, Estimation and Testing. Econometrica 55: 251276.

122

European Journal of Economic and Political Studies

Eryiit, M. 2009. Effects Of Oil Price Changes On The Sector Indices Of Istanbul Stock Exchange. International Research Journal of Finance and Economics. 25: 209216. Gogineni, S. 2010. Oil and The Stock Market: An ndusry Level Analysis. The Financial Review, Forthcoming, [Link] Ik, Nihat, Acar, Mustafa and Ik, H. Bayram. 2004. Enflasyon ve Dviz Kuru likisi: Bir Ebtnleme Analizi. Sleyman Demirel niversitesi ktisadi ve dari Bilimler Fakltesi Dergisi 9: 325340. Johansen, S.1988. Statistical Analysis of Cointegration Vectors. Journal of Economic Dynamics and Control 12: 231-254 Kutlar, Aziz. 2005. Uygulamal Ekonometri. Ankara: Nobel Yayn Datm. Lee, Y.H. and J.S. Chiou. 2011. Oil Sensitivity and its Asymmetric Impact on the Stock Market. Energy 168174 Miller, J.I. and Ratti, R.A. 2009. Crude Oil and Stock Markets: Stability, nstability and [Link] Economics 31: 559568. Narayan, P.K. and Narayan, S. 2010. Modelling The mpact Of Oil Prices On Vietnams Stock Prices. Applied Energy 87: 356361. Odusami, B.O. 2008. Crude Oil Shocks and Stock Market Returns. Applied Financial Economics, Forthcoming. [Link] zata, Erkan and Esen, Ethem. 2010. Reel cretler le stihdam Arasndaki likinin Ekonometrik Analizi. Anadolu niversitesi Sosyal Bilimler Dergisi 10: 5570. zer, M. and Trkylmaz, S. 2005. Trkiyede Enflasyon ile Enflasyon Belirsizlii Arasndaki likinin Zaman Serisi Analizi. ktisat letme ve Finans Dergisi 229: 93104. ztrk, Beyamil. 2008. Makroekonomik Faktrlerin Istanbul Menkul Kymetler Borsas Ulusal100 Endeksi ve Volatilitesi zerindeki Etkilerinin ncelenmesi (19972006). Yksek Lisans Tezi., Istanbul Teknik niversitesi. Papapetrou. E., 2001. Oil Price Shocks, Stock Market, Economic Activity and Employment in Greece. Energy Economics 23: 511-532. Park, J. and Ratti, R.A. 2008. Oil Price Shocks and Stock Markets in the US and 13 European [Link] Economics 30: 25872608. General Directorate Of Petroleum Affairs. 2011. World Oil and Gas Consumption.

123

European Journal of Economic and Political Studies

Accessed March 04, 2011. [Link] dogal_gaz_tuketimi.php Sadorsky, P., 1999. Oil Price Shocks and Stock Market Activity. Energy Economics 21: 449469. Sar, Aydn. 2008. Parasal Gre Gre Trkiyede demeler Bilanosu Dengesinin Salanmasnda Otomatik Denkleme Mekanizmalarnn Etkinlii. Cumhuriyet niversitesi, ktisadi ve dari Bilimler Dergisi 9: 112. Sevktekin, Mustafa and Nargeleekenler, Mehmet. 2010. Ekonometrik Zaman Serileri Analizi, Eviews Uygulamal. Ankara: Nobel Yayn Datm. Turkish Statistical Institute. 2011. Foreign Trade Statistics, Crude Oil Import. Accessed March 04, 2011. [Link]

124