UNIVERSITY OF GRONINGEN Faculty of Economics and Business
OIL PRICE SHOCKS AND STOCK MARKETS EMPIRICAL EVIDENCE FOR THE EURO AREA
MSc Economics
Student Name: Cenk Yurtsever Student Number: S1667491
Supervisor: Prof. Dr. Bert Scholtens
Co-assessor: Prof. Dr. Gerard Kupers
OIL PRICE SHOCKS AND STOCK MARKETS EMPIRICAL EVIDENCE FOR THE EURO AREA
Cenk Yurtsever 2008 June
Abstract
In this thesis, I study the sensitivity of the euro area industry stock market returns to the oil price shocks using with monthly data from 1983.8-2007.11 for 38 industries.
Using a multivariate Vector Autoregressive (VAR) model with 5 variables (interest rate,
real oil price changes, industrial production, total stock market return and industrial
stock market return) as well as impulse response function and variance decomposition,
I conclude that oil price changes have a negative effect on stock returns for almost all
industries except oil & gas producers, oil equipments, services & distribution, industrial
metals & mines and mining. The significance of the results differ when I use different oil
price specification but robust to a different VAR ordering. When I check for the effect of
asymmetric oil price changes, I conclude that negative oil price shocks have more effect
than positive oil price shocks on stock market returns. I also conclude that oil price
shocks contribute more than interest rate and industrial production into the variance
error of real industrial stock market returns. This result emphasizes the impact of oil
price shocks on stock markets. However, the effect of industrial production and interest
rate on the stock markets increase after 1999. It shows that independent from oil shocks,
introduction of a common monetary policy within the euro area and the integration of
the euro area economies caused a higher exposure of stock markets to interest rate and
industrial production changes.
Contents
1 INTRODUCTION 1
2 LITERATURE REVIEW 5
2.1 Oil Prices and Economic Activity . . . . 5
2.2 Oil Prices and Stock Markets . . . . 8
3 OIL AND THE EURO AREA ECONOMY 12 4 THEORETICAL FRAMEWORK 21 4.1 Vector Autoregression (VAR) . . . . 21
4.2 Impulse response functions . . . . 22
4.3 Variance decomposition . . . . 22
4.4 Unitroot . . . . 23
4.5 Cointegration . . . . 23
5 DATA DESCRIPTION AND MODEL 25 5.1 Data description . . . . 25
5.2 Model . . . . 32
6 DATA ANALYSIS 34 6.1 Unit Root . . . . 34
6.2 Cointegration Test . . . . 38
6.3 Impact of oil price shocks on stock market . . . . 38
6.3.1 Impulse response functions and accumulated response . . . . 39
6.3.2 Variance Decomposition . . . . 56
6.3.3 Robustness Tests . . . . 65
6.3.4 Asymmetric effect of oil price shocks . . . . 69
6.3.5 Historical decomposition of the macroeconomic variables’ effect on stock returns . . . . 76
7 CONCLUSION 84
List of Figures
1 Market Capitalization of EMU stock markets . . . . 2
2 Oil intensity of EMU countries . . . . 12
3 EMU stock market index and crude oil price (monthly data, 1983.8 - 2007.11 in e) . . . . 27
4 Oil Prices in US dollars and Euro . . . . 28
5 Alternative oil price specifications . . . . 29
6 Orthogonalized impulse response functions. (Linear Specification) . . . . . 42
7 Orthogonalized impulse response functions. (SOP Specification) . . . . 45
8 Orthogonalized impulse response functions. (NOPI Specification) . . . . . 48
9 Orthogonalized impulse response functions. (NOPD Specification) . . . . 51
B.1 Orthogonalized impulse response functions. (Linear Specification) . . . . . 91
B.2 Orthogonalized impulse response functions. (SOP Specification) . . . . 94
B.3 Orthogonalized impulse response functions. (NOPI Specification) . . . . . 97
B.4 Orthogonalized impulse response functions. (NOPD Specification) . . . . 100
List of Tables
1 DRC Coefficient across ESA95 Industry Classification . . . . 15
2 List of variables . . . . 27
3 Descriptive statistics: montly return data, 1983.8 to 2007.11 . . . . 30
4 Correlation coefficients among different oil price specifications. . . . . 32
5 Results of PP Unit Root test for real stock returns . . . . 35
6 Results of PP Unit Root test for macroeconomic variables an oil price . . 36
7 Results of KPSS Unit Root test for real stock returns . . . . 36
8 Results of KPSS Unit Root test for macroeconomic and oil price variables 37 9 Cointegration test using the Johansen procedure . . . . 38
10 Orthogonalized impulse response functions of real industrial stock returns market to oil price shocks with different specifications. (Ordering: ∆lr, oil, ∆lip, emt, isr) . . . . 40
11 Accumulated response of industrial stock returns to oil price shocks. . . . 54
12 Variance decomposition of forecast error variance after 24 months. (linear specification) . . . . 57
13 Variance decomposition of forecast error variance after 24 months. (SOP specification) . . . . 59
14 Variance decomposition of forecast error variance after 24 months. (NOPI specification) . . . . 61
15 Variance decomposition of forecast error variance after 24 months. (NOPD specification) . . . . 63
16 Multivariate Regression Market Model Results. . . . 67
17 Variance decomposition of forecast error variance after 24 months. (Asym- metric effect of oil price shocks on stock return) . . . . 73
18 Variance decomposition of forecast error variance after 24 months. (linear oil specification) . . . . 78
19 Variance decomposition of forecast error variance after 24 months. (SOP oil specification) . . . . 81
B1 Variance decomposition of forecast error variance after 24 months (linear specification) . . . 103
B2 Variance decomposition of forecast error variance after 24 months. (SOP specification) . . . 105
B3 Variance decomposition of forecast error variance after 24 months. (NOPI
specification) . . . 107
B4 Variance decomposition of forecast error variance after 24 months. (NOPD
specification) . . . 109
1 INTRODUCTION
Oil is considered to be one of the most influential commodities which has a great impact on an economy. Especially after 1970s, many of recessions were preceded by a great increase in oil prices. Hence this situation caused economists to pay a special attention particularly to this commodity.
Several channels have been offered as explanations to account for the negative cor- relation between oil prices and economic activity. The most fundamental explanation is classic supply shock in which rising oil prices are indicative of the reduced availability of an important input to production. The other fundamental explanation is demand shock which means an income transfer from oil-importing countries to oil-exporting countries that reduces aggregate demand and slows economic activity in oil-importing countries.
The other explanations can be considered as the effect of oil price shocks to developments in the financial markets such as contradictionary monetary policy, adjustment cost, co- ordination problems, uncertainty and financial stress. 1
Stock markets in the euro area play a far less prominent role than in the United States which traditionally relies more heavily on market based financing in stead of bank intermediation. However, several studies showed that stock markets have assumed a somewhat more prominent role in the euro area over recent years. The overall importance of the stock market is quite often gauged by the ratio between the market value of domestic shares traded on a country’s stock exchange (market capitalization) and nominal gross domestic product (GDP). As we can see from Figure 1 , market capitalization / gdp ratio has become almost four times higher in recent years compared to the 1990s which clearly indicates the increase of stock markets’ importance in the euro area.
Although the bulk of the literature examines the relationship between economic ac- tivity and oil price change, surprisingly little research have been conducted for financial market and oil price change relation. Some of exceptions are Hamao (1988), Jones and Kaul (1996) , Sadorsky (1999) ,Papapetrou (2001), Park (2007).
Therefore in this thesis, it is my interest to study the relationship between oil price shocks and stock market in the euro area countries. 2 Main research questions of this study are; What is the response of Euro area stock markets to the oil price shocks? How does it differ among the industries? If it exists, is this relationship symmetric? 3 Do the
1
See Brown, Y¨ ucel, and Thompson (2003) for detailed discussion.
2
The euro area corresponds to the countries which use Euro ( e) as their currency. They are currently;
Germany, France, Belgium, Italy, Ireland, Spain, Portugal, Netherlands, Luxembourg, Greece, Finland, Austria, Slovenia, Cyprus and Malta.
3
I also check for asymmetry effect of oil price changes since there has been a hot debate about this issue in
both literature and press. e.g. Mork (1989), Hamilton (1996), Sadorsky (1999),Park (2007)
Figure 1: Market Capitalization of EMU stock markets
Source: World Federation of Exchanges statistics, Eurostat and author’s own calculations.
List of EMU stock markets: Athens Stock Exchange, Berlin Stock Exchange, Euronext Brussels, Cyprus Stock Exchange, Irish Stock Exchange, Dusseldorf Stock Exchange, Euronext Amsterdam, Frankfurt Stock Exchange, Hamburg Stock Exchange, Helsinki Stock Exchange, Euronext Lisbon, Ljubljana Stock Exchange, Luxembourg Stock Exchange, Madrid Stock Exchange, Milan Stock Exchange, Munich Stock Exchange, Euronext Paris, Stuttgart Stock Exchange, Vienna Stock Exchange
introduction of a common monetary policy and integration of Euro area economies cause a difference?
In order to provide answers to these question I examine Datastream stock market indexes of Euro area for 38 different industries classified according to FTSE industry classification system.
There are several studies which test for the European stock market integration. In one of these studies, Yang, Min, and Li (2003) investigated the market integration among European stock markets examining the possible impact of the European Monetary Union (EMU) on stock market linkages. They concluded that the EMU has significantly strengthened stock market integration among its member countries, but lessened linkages with a non-member country (UK) in the same region. As EMU stock markets become more integrated and move together, a common approach for asset selection for investors moved from country orientation to industry orientation in order to enjoy benefits of di- versification. (Ferreira and Ferreira (2006))
This paper analyses the relationship between oil price shocks and stock markets using
a Vector Autoregression (VAR) model with 5 variables in following ordering; interest
rate, real oil price change, industrial production, real total stock market return and
real industrial stock market return. I use monthly data from 1983.08 - 2007.11 with a
few exceptions. For oil price specification I test 4 different specifications: linear (log first
difference), SOP (scaled oil price by Lee and Ratti (1995) ), NOPI (Net Oil Price Increase by Hamilton (1996)) and NOPD (Net Oil Price Decrease).
Among the study I examine the following issues:
1. I check how the euro area stock markets response to oil price shocks using impulse response functions and variance decomposition. Following the literature 4 I define oil price shocks with different oil price specifications such as linear, NOP (Net Oil Price) and SOP (Scaled Oil Price). I check robustness with different lag ordering (real oil price change, interest rate, industrial production, real total stock market return and real industrial stock market return) and also with a classical multivariate linear regression model.
2. I analyze the asymmetry pattern of stock markets in response to oil price shocks with variance decomposition and Wald coefficient test for pre-1999 and post-1999 periods.
3. I compare the effect of oil price, interest rate and industrial production (hereby economic activity) changes on stock market returns for pre-1999 and post-1999 periods to see if monetary and economic integration among the euro area countries have changed the effect of these variables on stock markets.
The main results of the study are summarized as follows:
1. In general, my findings indicate that oil price increases have a negative effect on stock returns for almost all industries except oil & gas producers, oil equipments, services
& distribution, industrial metals & mines and mining industries. The significance of the results differ when I use different oil price specification.
2. According to the literature 5 oil price changes have an asymmetric effect on the economy and financial markets such that oil price increases have a more significant effect than oil price decreases. When I check for the effect of asymmetric oil price changes, I conclude a different result than common perception. In linear oil price specification, since overall magnitude of the effect of negative oil price changes on stock markets are more than positive oil price changes, stock markets in the euro area are more influenced by an oil price decrease than an oil price increase. A historical decomposition for pre&post-1999 periods prove a change in this pattern such that positive oil price changes influence stock markets more than negative oil price changes in pre-1999 period but this case becomes opposite in post-1999 period.
4
Hamilton (1996), Lee and Ratti (1995)
5
e.g. Mork (1989), Hamilton (1996), Sadorsky (1999)
When I use SOP specification as a proxy of oil price volatility I observe that SOPD (scaled oil price decrease) has more explanatory power on stock market returns than SOPI (scaled oil price increase) on average for both pre-1999 and post-1999 periods.
But a formal asymmetry test shows that oil price shock’s effect on stock markets is not asymmetric for most of industries which is consistent with a recent study of Park (2007).
3. In addition to these results, I also conclude that effect of oil price changes on the stock markets decrease after 1999 but on contrary, effect of industrial production and interest rate on the stock markets increase after 1999. This result is consistent with the perception that introduction of a common monetary policy within the euro area and the integration of the euro area economies caused a higher exposure of stock markets to interest rate and industrial production changes.
This paper contributes to the literature of oil price and stock market in the following ways: First, I analyze the effect of oil price changes on stock market return of the euro area countries and to my knowledge, it is the first paper which examines this relationship for the euro area economies on industrial level. Second, I address the change of asymmetry pattern of oil price changes in pre&post-1999 periods. Third, I compare the exposure of stock market returns to oil price, interest rate and industrial productions variables for pre&post-1999 periods.
A common approach in the literature is to use a multivariate regression model (MLRM) to study the relationship between industrial stock returns and oil price changes. 6 In my paper, a robustness test with MLRM shows that results differ significantly with this model and many of the industries’ exposure become insignificant when I use MLRM. Hereby, fourth, I show that model selection matters for the results of a quantitative research for the oil price and stock market relationship.
The reminder of the paper is organized as follows: In Section 2, I review the existing literature about the effect oil price shocks on economic and financial activities. Section 3 provides information about the importance of oil particularly for the euro area economies.
In Section 4, I discuss the theoretical framework of my study. Section 5 states data description and model. Section 6 is the part in which I discuss the empirical results from data analysis. I conclude in Section 7.
6
e.g. Faff and Brailsford (1999), Sadorsky (2001), Nandha and Faff (2007)
2 LITERATURE REVIEW
Since oil price changes hit the economies towards both supply and demand channels, a special attention to oil price shocks have been paid both in the literature of economics and finance. These numerous studies which I cover below mainly differ from each other with respect to questions they ask and the way they approach to the problem. In this section I will mention theoretical and empirical studies as follows: (1) oil prices and economic activity and (2) oil prices and stock market.
2.1 Oil Prices and Economic Activity
In his seminal paper Hamilton (1983) has examined the impact of oil price shocks and the US economy in 1948-72. Using Granger-causality test and a seven variable VAR system 7 he has found that changes in oil prices Granger caused changes in GNP and unemployment.
Burbidge and Harrison (1984) have used a similar VAR model for US, Germany, Canada, Japan and the UK. Using impulse response functions they have concluded that the impact of oil price shocks on industrial production in the US is sizable while in Japan, Germany and Canada it is relatively small. Price level impacts on the US and Canadian economies are substantial, while they are smaller but still significant in Japan, Germany and the UK.
Gisser and Goodwin (1986) have found that oil price shocks affect a set of macro variables. Using Hamilton’s data they detect a relationship between the crude oil price and employment. Furthermore, they examine whether oil price shocks have a different impact on the macroeconomy before 1973 than after. However, they could not provide support for that hypothesis.
Extending the results of Hamilton (1983), Mork (1989) has investigated the asym- metric response to oil price changes by specifying real price increases and decreases as separate variables. Following theoretical work with Gilbert (Gilbert and Mork 1986) this study established an aggregate basis for both positive and negative GDP responses to oil price shocks, offered the first specification of an oil price shock as separate variables for in- creases and decreases. This study has triggered the subsequent asymmetry examinations in the literature.
A controversial response to Hamilton’s study came from Hooker (1996). Applying the Chow stability test and Granger causality test, he found that there is a breakdown
7
Two output variables (real GDP, unemployment), three price variables (implicit price deflator for nonfarm
business income, hourly compensation per worker, import price), M1 an.d oil price.
of the negative oil price-macroeconomy relation after 1973.Q3. A quick response to this argument came from Hamilton (1996) and he showed that using NOPI (Net Oil Price Increase) specification for data after 1973.Q3 results are still consistent with the historical correlation between oil shocks and recessions. This brilliant idea of using a non-linear specification shows that if one wants a measure of how unsettling an increase in the price of oil likely to be for the spending decisions of consumers and firms, it seems more appropriate to compare the current price of oil with where it has been over the previous years rather than during the previous month alone.
Another popular non-linear oil price specification, scaled oil price (SOP) was first proposed by Lee and Ratti (1995). The objective SOP specification is to account for volatility of oil prices by using GARCH. According to Lee and Ratti (1995), oil price changes are likely to have a greater impact on GDP in an environment where the oil price has been stable than where the oil price changes frequently. Using the SOP oil price specification, Ferderer (1996) provided support of oil price shocks’ adverse impact on the economy by showing that oil price volatility helps to forecast aggregate output movements in the U.S. economy.
Not surprisingly, the bulk of the oil price - economic activity relationship literature has examined the case for US. But in recent studies, authors have studied the relationship for other countries as well. Cunado and Perez de Gracia (2003) have analyzed the relationship between oil price and economic activity for many European countries 8 . They mainly have used Granger causality test to check whether oil price changes have an effect on economy.
They have calculated world oil price as the ratio between producer price index (PPI) of US for crude oil divided by PPI of US for all commodities. National oil prices are calculated using exchange rates. They have used three different oil price specification:
real oil price change, net oil price increase (NOPI) and scaled oil price increase (SOPI).
Using a granger causality test, when world oil price is used oil price changes cause the industrial production for 7 out of 14 countries. However, if the national oil price changes or positive oil price change in the world or NOPI in the world oil price is used it causes industrial production growth in more countries. With regard to asymmetric effect of oil price changes on the economic activity they have concluded that oil price increases have a negative and significant effect on industrial production in many of countries while oil price decreases have an insignificant effect.
Jim´ enez-Rodr´ıguez and S´ anchez (2004) have studied the effects of oil price shocks on the real economic activity of the main industrialized countries (individual G-7 coun-
8
UK, Germany, France, Italy, Ireland, Belgium, Austria, Spain, Finland, Netherlands, Denmark, Greece,
Sweden
tries, Norway and the euro area as a whole) They have used a multivariate VAR analysis with both linear and non-linear(scaled oil price increase (SOPI), scaled oil price decrease (SOPD) and NOPI) oil price specifications. They have concluded that the interaction be- tween oil prices and macroeconomic variables is significant, with the direction of causality going in at least one direction in all countries, and in both directions in most countries.
The effects of an increase in oil prices on real GDP growth were found to differ substan- tially from those of an oil price decrease, providing evidence against the linear approach that assumes that oil prices have symmetric effects on the real economy. They also have found that an increase in oil prices have a significant negative impact on the GDP growth in all oil-importing countries but Japan. For Japan the model fails to identify any nega- tive real effect of oil prices, possibly due to the special circumstances undergone by the Japanese economy. With regard to the two net oil-exporting countries under considera- tion, they have concluded that oil prices affect Norwegian GDP growth positively ,while having a negative impact on oil exporter UKs economic activity (relating to the stan- dard Dutch disease effect). In addition, the effects of an oil price hike on GDP growth are overall strongest for the US, although the euro area countries (Germany, France and Italy) exhibit similarly strong real effects when they use non-linear modelling.
Robalo and Salvado (2008) have investigated the impact of oil price shocks on the Portuguese economy. Using a multivariate VAR methodology and NOPI oil specification they conclude that, as for most of industrialized countries, the nature of oil price change - economic activity relationship changed in the mid-1980s. They also showed that the main Portuguese macroeconomic variables have become progressively less responsive to oil shocks and the adjustment towards equilibrium has become increasingly faster.
There is also a discussion of monetary policy and oil price changes relationship. Bohi (1991) has argued that the recessions that followed the big oil shocks were caused not by oil shocks themselves, but rather by Federal Reserve’s contractionary response to inflationary concerns attributable in part to the oil shocks.
Taking this argument into account, Bernanke, Gertler, and Waston (1997) (BGW)
have analyzed how much of an economic recession is attributed to oil price increases and
contractionary monetary policy by considering the argument that economic declines are
caused by oil price shocks and monetary policy shocks. They use 4 different measures of
oil price shocks : the log of the nominal producer price index for crude oil and products,
Hoover & Perezs oil dates, the log-difference of the relative price of oil when that change
is positive and otherwise is zero, and Hamilton’s Net Oil Price Increase (NOPI). They
have found that Hamilton’s NOPI is the most appropriate indicator for the investigation
of the macroeconomic effect of oil prices, in that oil price shocks are followed by an
output decline and price increase. They also have checked how systematic monetary policy changes affect the economy and then determine what portion of the last five US recessions are attributed to oil price shocks and the Feds monetary policy shocks. They have concluded that the majority of the impact of an oil price shock on the economy is explained by contractionary monetary policy in response to inflationary pressures caused by oil price shocks.
However, Hamilton and Herrera (2004) argue that potential of monetary policy to avert the contradictionary consequences of an oil price shock is not as great as suggested by the analysis of BGW. Using a different lag level in their VAR analysis, they concluded that oil price shocks have a bigger effect on the economy than suggested by BGW’s VAR analysis. Therefore taking monetary policy implications into account, this result strengthens the argument of adverse impact of oil price shocks on economy.
2.2 Oil Prices and Stock Markets
Although a vast literature examines the relationship between economic activity and oil price change, surprisingly little research has been conducted for financial market and oil price change relation. These studies mainly differ from each other with the methodology, data period and stock market level they examine. But generally, a very common approach in the asset pricing literature is to use a standard market model 9 augmented by oil price and some other factors to gauge the effect of oil price or volatility shocks on stock markets.
Chen, Roll, and Ross (1986) have tested whether innovations in macroeconomic vari- ables are risks that are rewarded in the stock market. They addressed the difficulty of using macroeconomic time series since their smoothing and averaging characteristics, in short holding periods, such as a single month, these series cannot be expected to cap- ture all the information available to the market in the same period. Stock prices, on the other hand, respond very quickly to public information. The effect of this is to guarantee that market returns will be, at best, weakly related and very noisy relative to innova- tions in macroeconomic factors. Taking this argument into account, they have concluded that the spread between long and short interest rates, expected and unexpected inflation, industrial production and the spread between high- and low- grade bonds explain the innovation in stock prices but however, oil price is not prices and therefore does not have a significant effect on stock market.
Hamao (1988) applied this method to Japanese stock market also including interna- tional factors because of their importance for international trade in Japanese economy.
9
The market model says that the return on a security depends on the return on the market portfolio and
the extent of the security’s responsiveness as measured by beta.
He concluded that oil price changes and unexpected change in foreign exchange rates are not priced in the Japanese stock market for the sample period of 1975-1984. An opposite conclusion for Japanese stock market came from Kaneko and Lee (1995). It differs from Hamao (1988) primarily due to the difference in sample period and empirical methodol- ogy. In his paper they employed a VAR analysis for period 1975.1 - 1993-12 and found that changes in oil prices are the most significant in Japanese stock market return.
Jones and Kaul (1996) have investigated the reaction of the US, Canadian, Japanese and UK stock prices to oil price shocks using quarterly data. Utilizing a standard cash- flow dividend valuation model, they found that for the US and Canada this reaction can be accounted for entirely by the impact of the oil shocks on real cash flows. The results for Japan and the UK were not as strong.
In addition to examination of the extend that stock market and oil market are contem- poraneously correlated, Huang, Masulis, and Stoll (1996) have examined the efficiency of oil oil futures and stock markets, namely, the extend to which price changes or returns in one market lead returns in the other. They have concluded that in the period of the 1980s, there is virtually no correlation between oil futures returns and the returns of various stock indices. But this paper was criticized by Ciner (2001) and retested relying on a nonlinear causality test. His study has provided evidence that oil shocks affect stock index returns. Moreover he found that the linkage between oil prices and the stock market was stronger in the 1990s.
Sadorsky (1999) has investigated the dynamic interaction between oil price and other
economic variables including stock returns using an unrestricted VAR with US data. He
presented variance decompositions and impulse response functions to analyze the dynamic
effect of oil price shocks. Variables include industrial production, interest rate of a 3-
month T-bill, oil price (measured using the producer price index for fuels), real stock
returns (calculated using the difference between the continuously compounded returns on
the S&P 500, and the inflation measured using the consumer price index).He has found
that oil price changes and oil price volatility have a significantly negative impact on real
stock returns. He also has found that industrial production and interest rates responded
positively to real stock returns shocks. In particular, he splited the full sample period
into two sub-periods, pre-1986 and post-1986, because in 1986 the oil price declined
significantly and the oil price has been more volatile since 1986. In post-1986 period
oil price changes and oil price volatility have a larger impact on the economy than in
the pre-1986 period. However when he used asymmetric oil price shocks (positive oil
price changes and negative oil price changes), positive shocks explain more forecast error
of variance in real stock returns, industrial production and interest rates than negative
shocks during the full sample period. For the post-1986 period, positive and negative oil price shocks explain almost the same fraction of forecast error variance of real stock returns, while in the pre- 1986 period positive oil price shocks contribute more to the forecast error variance in real stock returns than negative oil price shocks. In the case of oil price volatility over the full period and two sub-periods, positive oil price volatility shocks (SOPI) had a greater influence on stock returns and industrial production than negative oil price volatility shocks (SOPD). In the post-1986 period, oil price movements explain more forecast error variance in stock returns than interest rates.
Papapetrou (2001) has examined the dynamic relationship among oil prices, real stock returns, interest rate, economic activity and employment for Greece. Using a multivariate VAR approach with generalized impulse response and generalized variance decomposition analysis he concluded that oil price changes affect real economic activity and employment.
He has also found that oil prices are important in explaining stock price movements.
Park (2007) has used a similar methodology with Sadorsky (1999) for examining relationship between oil price shocks and stock market relation. In his paper he analyzed this relationship for the US and 13 European countries 10 with monthly data from 1986.1 - 2005.12. Using an unrestricted VAR model with 4 variables (interest rate, real oil price changes, industrial production and real stock returns) he has concluded that in most oil importing countries oil price shocks have significantly negative effect on the stock market, while among oil exporting countries only Norway shows a significant positive response of real stock returns to oil price shocks. In addition to that, he also investigated the impact of interest rate(monetary) shocks on stock market and taking into account the response of monetary policy to oil price shocks, he concluded that oil prices play a crucial role in the stock market of oil importing countries. On the contrary, in oil exporting countries oil price shocks have a smaller impact on the stock market than interest rate shocks and monetary policy does not respond to oil price shocks.
Faff and Brailsford (2000) have employed a GMM-based approach to the restrictions imposed by a two-factor(market and oil price) pricing model and they have found evidence that Australian industries are heterogeneous with respect to the sensitivity of their equity returns to oil prices.
Nandha and Faff (2007) have studied the effect of oil price shocks on 35 Datastream global industry indices for the period from 1983.4 to 2005.9. Using a standard market model augmented by oil price factor, they found that oil price rises have a negative impact on equity returns for all sectors expect mining, and oil and gas industries. They
10
Austria, Belgium, Denmark, Finland, France, Germany, Greece, Italy, Netherlands, Norway, Spain, Sweden
and UK.
also proved the evidence of asymmetric response to oil price increases and decreases.
A firm-specific study by Al-Mudhaf and Goodwin (1993) has examined the returns from 29 oil companies listed on the New York Stock Exchange. Their findings suggested a positive impact of oil price shocks on ex post returns for firms with significant assets in domestic oil production. Sadorsky (2001) have shown that stock returns of Canadian oil and gas companies are positively sensitive to oil price increases.
Another study for Canadian oil and gas stock return have been conducted by Boyer and Filion (2007). They included natural gas prices and industry specific factors to explain the stock return of oil and gas firms. They also examined how the factors affect differently producers and integrated firms, and how differently they affect crude oil intensive versus natural gas intensive firms. They have found that the stock return of Canadian energy stock is positively associated with returns on the Canadian stock market, appreciations of crude oil and natural gas prices, growth in internal cash flows and proven reserves, and negatively with interest rate.
Scholtens and Wang (2008) have studied the oil price sensitivities of 96 NYSE listed oil & gas firms’ returns by using a two-step regression under two different arbitrage pricing models and they concluded that return of most of the oil and gas firms’ stocks are positively associated with the market return and increase of the spot crude oil price and insensitive to the default premium and term premium. They also showed that inclusion of firm specific factor returns improves the model’s fit.
In general, literature suggests an adverse and asymmetric impact of oil price shocks on economy and financial markets. The effect of oil price shocks on stock markets differs whether oil is used as an input or output for an industry. But some factors such as the degree of competition and price elasticity effect the ability of passing on their higher fuel costs to customers and thus minimizing the negative impact of oil price shock. (Nandha and Faff (2007))
Considering all these arguments, this paper examines oil price and stock market re-
lation, asymmetric pattern of oil price shocks and change of this asymmetry pattern in
pre&post 1999 periods in the euro area stock markets on industrial level. By doing these,
I reveal if this relationship is relevant for the euro area stock markets as well and introduc-
tion of a common monetary policy in 1999 conduction by European Central Bank (ECB)
and integration of economic and financial activities change the way of this relationship.
3 OIL AND THE EURO AREA ECONOMY
As one of the biggest oil importing areas, European countries are highly depended on oil as an energy resource. From gasoline to chemicals, shampoos, synthetic textile products, many industries use crude oil in different forms in their production process.
According to the results of a quantitative exercise carried out by the IEA in collabora- tion with the OECD Economics Department and with the asistance of the International Monetary Fund Research Department, a sustained $10 per barrel increase in oil prices from $25 to $35 would result in the euro zone countries which are highly dependent on oil imports as a whole losing 0.5% of GDP and rising inflation by 0.5% in 2004. IEA (2004) Oil intensity measures how much energy is required in a country or region to generate one unit of GDP. In Figure 2 we observe a decreasing oil intensity between 1995 and 2005 which is a result of the combined effect of higher energy prices , energy conservation programmes and more recently CO2 abatement policies, and other economic factors.
(EAE (2006))
Figure 2: Oil intensity of EMU countries
Source: Eurostat and author’s own calculations
Since I study oil price effect on the industrial level, it is wise to check for the importance of oil in industries’ production processes. Oil can be used as an input commodity for some industries (e.g. industrial transportation), but it can also be output of other industries.
( e.g.oil & gas producers) In this situation, an oil price change has a different effect on different industries according to whether this change effects their revenue or cost side.
For example, a priori, industrial transportation (oil & gas producer) industry devotes a
relatively high proportion of their cost (revenue) to oil-based products such as gasoline
or diesel. Since a stock price reflects the present discounted value of future earning expectations of a company, ceteris paribus, an oil price increase is expected to cause decrease (increase) of stock value of this company. But as Faff and Brailsford (1999) note, the impact of oil price change on equity prices will depend on the ability of firms to pass on the effect to customers through changing good prices.
In order to gauge the oil price sensitivity of the industries, following the methodology in Faff and Brailsford (1999), I construct Table 1 from the extracted data of European System of Account (ESA95) input-output tables of 2004. Unfortunately, classification systems of Financial Times Stock Exchange (hereafter FTSE) and ESA95 are not consis- tent. To make them comparable, I attempt to link the ESA95 industries to FTSE. But some of linkages are not perfect or at least mixed since some of different ESA95 industries are mentioned with the same corresponding FTSE industry. (e.g. Mining of metal ore and Other mining and quarrying industries are labeled as ”mining”)
In Table 1 third column gives the ”Coke, refined petroleum and nuclear fuels” in- put prices for various industries and fourth column gives the output of the corresponding industry at basis prices. Direct Requirement Coefficients (DRC) are obtained by calculat- ing ”Coke, refined petroleum and nuclear fuels” input prices as a percentage of the total output price of a particular industry. To make comparison between industries more easily interpretable, I calculate Relative Direct Requirement Coefficients (RDRC) dividing the DRC of a particular industry to the DRC average of the whole industries. The RDRC can be seen as a relative oil intensity of a particular industry and a value bigger than 1 means that this industry has a comparative oil intensity than average of the industries.
Some of the industries such as Industrial Transportation, Food Producers, Mining, Oil & Gas Producers, Chemicals, Industrial Metals, Electricity, Oil Equipment, Services
& Distribution have higher RDRC than other industries. Not surprisingly, ESA95 Air Transportation industry has highest RDRC, at almost 7 times the economy-wide average.
Another industry with high RDRC is Oil & Gas Producers which has RDRC value of 6.8.
For these industries, ceteris paribus, I may predict a significant sensitivity to oil price shocks. Of course, the sign of response depends on whether oil is used as an input or output for these industries.
It is also useful to consider the industries that have extremely low RDRC values. One
notable case is FTSE Real Estate industry which seems to be consistently low (0.055)
as it relates to ESA95 Real Estate Activities industry. Another industries with low
RDRC values are Banking, General Financial, Life Insurance, Nonlife Insurance, Health
Care Equipments & Services, Support Services, General Retailers, Food & Drug Retailers,
Leisure Goods and Travel & Leisure industries. Accordingly, ceteris paribus, I may predict
a negligible role for the oil price shocks for these industries.
It is worth noting that, because the comparison between ESA95 and FTSE classifi-
cation systems is not perfect, there are some industries which show mixed signals with
respect to their RDRCs. For example ”Manufacture of coke, refined petroleum products
and nuclear fuel” and ”Manufacture of rubber and plastic products” industries are both
linked to FTSE ”Chemicals” industry. But although the former has a RDRC which ex-
ceeds unity (2.205), the latter has a RDRC which is lower than unity (0.541). Therefore
when it comes to interpret oil intensity of Chemical industry as a whole, one should
consider these mixed results.
T able 1: Direct Requiremen t Co efficien ts across ESA95 Industry Classification no ESA95 Industry Sp ecification C ok e, ref. p etr. pro. & n uc. fuels (mil. e ) Output at ba- sis prices (mil. e )
DR C a Relativ e DR C b Corresp onding FTSE Global Classification 1 Agriculture, h un ting and related service activities 8497.936 268684.592 0.032 1.608 F o o d Pro ducers 2 F orestry , logging and related ser- vice activities 332.218 16541.820 0.020 1.021 F orestry & P ap er 3 Fishing, op erating of fish hatc heries and fish farms; service activities inciden tal to fishing
796.240 10103.012 0.079 4.006 F o o d Pro ducers 4 Mining of coal and lignite; extrac- tion of p eat 156.783 6106.776 0.026 1.305 Mining 5 Extraction of crude p etroleum and natur al gas; service activiti e s inciden tal to oi l and gas extrac- tion excluding surv eying
63.028 21879.747 0.003 0.146 Oil & Gas Pro ducers 6 Mining of metal ores 42.700 707.053 0.060 3.070 Mining 7 Other mining and qu arrying 1193.828 24657.016 0.048 2.461 Mining 8 Man ufacture of fo o d pro ducts and b ev erages 3773.587 564497.211 0.007 0.340 Be v erages ; F o o d Pro ducers 9 Man ufacture of tobacco pro d ucts 59.816 15019.043 0.004 0.202 T obacco 10 Man ufacture of textiles 537.204 93499.683 0.006 0.292 P ersonal Go o ds 11 Man ufacture of w earing apparel; dressing and d y eing of fur 357.792 72697.024 0.005 0.250 P ersonal Go o d s 12 T anning and dressing of leather; man ufacture of luggage, hand- bags, saddlery , harness and fo ot w ear
156.991 46455.195 0.003 0.172 P ersonal Go o ds ; House h old Go o d s Con tin ued on Next P age .. .
T able 1 – Con tin ued no ESA95 Industry Sp ecification C ok e, ref. p etr. pro. & n uc. fuels (mil. e ) Output at ba- sis prices (mil. e )
DR C a Relativ e DR C b Corresp onding FTSE Global Classification 13 Man ufacture of w o o d and of pro d- ucts of w o o d and cork, except fur- niture; man ufacture of articles of stra w an d plaiting materials
840.113 84358.683 0.010 0.506 F orestry & P ap er ; Construc- tion & Mat e ri als 14 Man ufacture of pulp, pap er and pap er pro ducts 719.508 111916.086 0.006 0.327 F orestry & P ap er ; Household Go o d s 15 Publishing, prin ting and repro- duction of rec or ded media 793.542 167387.092 0.005 0.241 Media ; Electronic & Electri- cal Equipmen t 16 Man ufacture of cok e, refined p etroleum pro ducts and n uclear fuels
26364.150 198336.916 0.133 6.757 Oil & Gas Pro ducers 17 Man ufacture of chemicals and chemical pro ducts 20391.523 470069.779 0.043 2.205 Chemic als 18 Man ufacture of rubb er and plastic pro ducts 1807.305 169942.764 0.011 0.541 Chem ic als 19 Man ufacture of other non- metallic mineral pro ducts 3290.999 156167.550 0.021 1.071 Mining 20 Man ufacture of basic metals 5452.836 214054.239 0.025 1.295 Indu st rial Metals 21 Man ufacture of fabricated metal pro ducts, except mac h inery and equipmen t
1612.305 311605.625 0.005 0.263 Industrial Metals 22 Man ufacture of mac hinery and equipmen t n.e.c. 1541.052 406857.219 0.004 0.193 Industrial Engineering ; Con- struction & Materials ; Elec- tronic & Electrical Equip- men t ; T ec hnology Hardw are & Equipmen t 23 Man ufacture of office mac hinery and computers 115.033 45655.482 0.003 0.128 T ec hnology Hardw are & Equipmen t Con tin ued on Next P age .. .
T able 1 – Con tin ued no ESA95 Industry Sp ecification C ok e, ref. p etr. pro. & n uc. fuels (mil. e ) Output at ba- sis prices (mil. e )
DR C a Relativ e DR C b Corresp onding FTSE Global Classification 24 Man ufacture of electrical mac hin- ery and ap paratus n.e.c. 912.105 184442.244 0.005 0.251 Electronic & Electrical Equipmen t ; 25 Man ufacture of radio, television and com m unication equipmen t and apparatus
322.235 126151.070 0.003 0.130 Electronic & Electrical Equipmen t ; Media ; T ec hnology Hardw are & Equipmen t 26 Man ufacture of medical, precision and optical instru m en ts, w atc hes and clo cks
350.762 92759.808 0.004 0.192 P ersonal Go o ds ; Health Care Equipmen t & S e rvice s ; T ec h- nology Hardw are & Equip- men t 27 Man ufacture of motor v ehicles, trailers and semi-trailers 1690.590 490912.151 0.003 0.175 Industrial Engineering ; Au- tomobiles & P arts ; 28 Man ufacture of oth e r transp ort equipmen t 436.344 119006.547 0.004 0.186 Industrial T ransp ortation ; Industrial Engineering ; Au- tomobiles & P arts ; Leisure Go o ds 29 Man ufacture of furnitu re; man u- facturing n.e.c. 702.008 121991.921 0.006 0.293 Household Go o ds ; T ra v el & Leisure 30 Recycling 538.609 20110.178 0.027 1.362 Supp ort Services 31 Electricit y , gas, steam and hot w ater supp ly 12698.289 291549.679 0.044 2.214 Electricit y; Oi l & Gas Pro- ducers 32 Collection, purification and distri- bution of w ater 382.473 35787.856 0.011 0.543 Gas , W ate r & Mul.Uti. 33 Construction 11729.775 919409.191 0.013 0.649 Construction & Materials 34 Sale, main tenance and repair of motor v ehicles and motorcycles; retail sale services of automotiv e fuel
2851.065 232058.574 0.012 0.625 Ge n e ral Retailers ; Indu str ial T ransp ortation Con tin ued on Next P age .. .
T able 1 – Con tin ued no ESA95 Industry Sp ecification C ok e, ref. p etr. pro. & n uc. fuels (mil. e ) Output at ba- sis prices (mil. e )
DR C a Relativ e DR C b Corresp onding FTSE Global Classification 35 Wholesale trade and commis sion trade, exc ept of motor v ehicles and motorcycles
12288.050 678874.332 0.018 0.920 Supp ort Services ; F o o d & Drug Retailers ; General Re- tailers 36 Retail trade, except of motor v e- hicles and motorcycles; repair of p ersonal and household go o ds
7075.513 534145.760 0.013 0.673 F o o d & Drug Retailers ; Gen- eral Retailers 37 Hotels and restauran ts 3406.553 392436.070 0.009 0.441 T ra v el & Leisure 38 Land transp ort ; tr ansp ort via pip elines 30752.476 321953.090 0.096 4.856 Oil Equipmen t, Services & Distribution 39 W ater transp ort 3447.024 45480.664 0.076 3.853 Industrial T ransp ortation 40 Air transp ort 10557.740 75795.031 0.139 7.081 Industrial T ransp ortation ; T ra v el & Leisure ; Aerospace & Defense 41 Supp orting an d auxiliar y trans- p ort activities; activities of tra v el agencies
5465.445 259846.465 0.021 1.069 T ra v el & Leisure ; Industrial T ransp ortation 42 P ost and telecomm unications 2254.275 312237.604 0.007 0.367 Fixed Line T elecomm unica- tions ; Mobi le T elecomm u ni- cations ; T ec hnology Hard- w are & Equipmen t 43 Financial in termediation, except insurance and p ension funding 819.917 432754.496 0.002 0.096 Banks ; Real Estate ; Eq- uit y In v estmen t Instrumen ts ; General Financial 44 Insurance and p ension funding, except compulsory so c ial securit y 298.931 184102.047 0.002 0.083 Nonlife Insurance ; Life Insur - ance 45 Activities auxiliary to financial in- termediation 467.099 101542.845 0.005 0.234 Ge n e ral Financial ; 46 Real estate activities 1031.007 960689.128 0.001 0.055 Real Estate Con tin ued on Next P age .. .
T able 1 – Con tin ued no ESA95 Industry Sp ecification C ok e, ref. p etr. pro. & n uc. fuels (mil. e ) Output at ba- sis prices (mil. e )
DR C a Relativ e DR C b Corresp onding FTSE Global Classification 47 Ren ting of mac hinery and equip- men t without op erator and of p er- sonal and h ouse h old go o ds
1594.394 104591.296 0.015 0.775 T ra v el & Leisure; General Re- tailers 48 Computer and related activities 1127.057 186955.052 0.006 0.306 Soft w are & Computer Ser- vices ; T ec hnology Hardw are & Equipmen t ; Leisure Go o ds ; 49 Researc h and dev elopmen t 544.423 53208.228 0.010 0.520 Pharmaceuticals & Biotec h- nology ; Real Estate ; 50 Other business activities 6003.975 881638.833 0.007 0.346 F o o d & Drug Retailers ; Soft- w are & Computer Services ; Supp ort Services ; General Industrials 51 Public administration and de- fence; compulsory so cial securit y 6349.196 649316.161 0.010 0.497 Aerospace & Defense 52 Education 3173.427 406181.602 0.008 0.397 Supp ort Services ; General Retailers 53 Health and so cial w ork 4152.129 676086.174 0.006 0.312 Health Care Equi pme n t & Services ; 54 Sew age and refu se disp osal, sani- tation and sim ilar activities 1068.047 80205.030 0.013 0.677 Supp ort Services 55 Activities of mem b ership organi- sation n.e.c. 546.467 57069.310 0.010 0.487 Supp ort Services 56 Recreational, cultural and sp ort- ing activities 2258.227 232546.047 0.010 0.494 Leis u re Go o ds ; T ra v el & Leisure ; General Retailers Con tin ued on Next P age .. .
T able 1 – Con tin ued no ESA95 Industry Sp ecification C ok e, ref. p etr. pro. & n uc. fuels (mil. e ) Output at ba- sis prices (mil. e )
DR C a Relativ e DR C b Corresp onding FTSE Global Classification 57 Other service ac ti vities 755.564 87447.782 0.009 0.439 Soft w are & Computer Ser- vices ; T ec hnology Hardw are & Equipmen t ; Leisure Go o ds ; Oil Equipmen t, Services & Distribution ; General Finan- cial ; Construction & Ma- terials ; Industrial T rans- p ortation ; Supp ort Services; General Retailers ; Media ; Fixed Line T elecomm unica- tions ; Mob ile T elecomm u ni- cations ; Banks 58 Priv ate households with emplo y e d p ersons 0.000 37106.279 0.000 0.000 Supp ort Services 59 Financial in termediation se rv ic es indirectly measured (FIS IM ) 0.000 0.000 0.000 0.000 Ge n e ral Financial a Direct Requiremen t Co efficien t. b Relativ e DR C is c alcul ate d as dividing the DR C of a particular industry to the DR C a v erage of th e whole industries. Source: Eurostat, Europ ean System of Accoun ts ESA95 Input-Ou tput T ables.
4 THEORETICAL FRAMEWORK
In this section of the paper, I briefly introduce the econometrical methods which I use for data analyzing in latter chapter. These methods are; Vector Autoregression (VAR), Impulse Response Functions, Variance Decomposition, Unitroot and Cointegration.
4.1 Vector Autoregression (VAR)
VAR model is first proposed by Sims (1980). It is a simple way to model the dynamic relationships among several time series variables without making many assumptions. It is indeed a multivariate analog of an autoregressive model for a single time series. 11 The main advantage of using VAR is that it does not stand on any economic theory on which the model is built and its practical ability to capture dynamic relationships among the economic variables of interest.
Let the 1 × g vector Y t denote the t th observation on a set of g variables. Then a vector autoregressive model of order p, sometimes referred as VAR(p) model, can be written as:
Y t = α +
p
X
j=1
Y t−j φ j + U t U t , ∼ IDD(0, σ) (1)
where U t is a 1 × g vector of error terms. α is a 1 × g vector of constant terms, and the φ j , for j = 1, ..., p, are g × g matrices of coefficients, all of which are to be estimated. If y it denotes the i th element of Y t and φ j,ki denotes the ki th element of φ j then i i th column of Equation 1 can be written as:
y it = α i +
p
X
j=1 m
X
k=1
y t−j,k φ j,ki + u ti (2)
This is just a linear regression, in which y it depends on a constant term and lags 1 through p of all of the g variables in the system. Thus we see that the VAR in Equation 1 has the form of multivariate linear regression model.
We can see that more linearly by this definition:
X t ≡ [1 Y t−1 ...Y t−p ] and Y
≡
α φ 1
.. . φp
Since φ j vector of coefficients tend to be mostly insignificant, Sims advices using impulse
11
For detailed discussion See Davidson and MacKinnon (1999)
response functions and variance decomposition in order to examine the contribution of independent variables on dependent variable Y t .
4.2 Impulse response functions
The rationale behind using impulse response functions is to trace out the system’s re- sponse to typical random shocks that represent positive residuals of one standard devia- tion unit in each equation in the system.
Suppose that a 2-variables VAR(1) is specified as;
y 1t
y 2t
=
φ 11 φ 12
φ 21 φ 22
y 1t−1
y 2t−1
+
u 1t
u 2t
A perturbation in u it has an one-for one effect on y it . In period t+1, that perturbation in y 1t affects y 1t+1 through the first equation and also affects y 2t+1 through the second equation. These effects work through to period t + 2, and so on. Thus, a random shock in one innovation in the VAR sets up a chain reaction over time in all variables in the VAR. Impulse response functions calculate these chain reactions.
One weakness of the analysis from impulse response function is that a perturbation in one innovation is not contemporaneously independent of the other innovations in the system, although it eventually leads to a chain reaction over time in all variables in the system. A widely used solution to this problem is to transform the innovations to produce a new set of orthogonal innovations. These innovations are pairwise uncorrelated and have a unit variance. One problem of the transformation is that the order in which the residual variables are orthogonalized can have dramatic effects on the numerical results.
4.3 Variance decomposition
While impulse response functions trace the effects of a shock to one endogenous variable
on to the other variables in the VAR, variance decomposition separates the variation in an
endogenous variable into the component shocks to the VAR. Thus, the variance decom-
position provides information about the relative importance of each random innovation
in affecting the variables in the VAR.
4.4 Unitroot
A nonstationary time series is said to be integrated to order one, or I(1), if the series of its first differences, ∆y t ≡ y t − y t−1 , is I(0). More generally, a series is integrated to order d, or I(d), if it must be differenced d times before an I(0) series results. A series is I(1) if it contains what is called a unit root, a concept that we will elucidate in the next section. Using standard regression methods with variables that are I(1) can yield highly misleading results. It is therefore important to be able to test the hypothesis that a time series has a unit root.
Many of financial and economic variables tend to have a unit root. Let’s consider such an ARMA model:
y t = 1y t−1 + u t
This is a typical I(1) series and a conventional approach for handling unit root non- stationary is to use differencing.
4.5 Cointegration
If two or more series are themselves non-stationary, but a linear combination of them is stationary, then the series are said to be cointegrated. Economically speaking, two variables will be cointegrated if they have a long term, or equilibrium, relationship between them.
Consider a VAR of order p:
y t = A 1 t t−1 + · · · + A p y t−p + Bx t + u t (3) where y t is a k-vector of non stationary I(1) variables, x t is a d-vector of deterministic variables, and u t is a vector of innovations. We may write this VAR as,
∆y t = Πy t−1 +
p−1
X
i=1
φ i ∆y t−i + Bx t + u t (4)
where:
Π =
p
X
i=1
A i − I, φ i = −
p
X
j=i+1
A j (5)
Granger’s representation theorem asserts that if the coefficient matrix Π has reduced
rank r < k, then there exist k × r matrices α and β each with rank r such that Π = αβ
0and β
0y t is I(0). r is the number of cointegrating relations (the cointegrating rank) and
each column of β is the cointegrating vector. As explained below, the elements of α
are known as the adjustment parameters in the VEC model. Johansen’s method is to
estimate the matrix Π from an unrestricted VAR and to test whether we can reject the
restrictions implied by the reduced rank of Π.
5 DATA DESCRIPTION AND MODEL
5.1 Data description
In this paper, I study the impact of oil price shocks on EMU stock markets on industrial level between 1983.8 - 2007.11. (a few exceptions are; mining industry: 1985.03 - 2007.11, tobacco industry: 1987.05 - 2007.11, software & computer srv.: 1985.07 - 2007.11). The data series and notation used are as follows: 12 Nominal oil price is the price index in US dollars of UK Brent oil from IMF database. European real oil price is obtained using the euro/dollar exchange rate and deflated using the CPI of Eurozone from IMF. Actually when I use oil price in Euro, a change in the value of index could be result of whether an oil price change or exchange rate fluctuation. A common approach in the literature is to check for both world oil price (in US dollar) and national oil price specifications. 13 But for the sake of concreteness, I only study oil prices in Euro ( e) currency. Since a decomposition of oil price change as exchange rate and pure oil price change does not make so much sense in reality, I believe this oil price specification with euro is more plausible.
Following the existing literature in measuring the value of the oil price 14 , I use oil price excluding taxes. The reason is simply that there is not a tax-including end-use prices of oil prices database for EMU countries. During the study following the existing literature I use different oil price specifications as proxies for oil price shocks as follows:
1. dlroil t (linear specification) : Monthly changes of real oil prices, the conventional first log difference transformation of real oil price variables.
dlroil t = lnroil t − lnroil t−1 = lroil t − lroil t−1
roil t : real oil price in period t in Euro ( e).
dlroilp t : real oil price increase, max(0, dlroil t ) dlroiln t : real oil price decrease, min(0, dlroil t )
2. SOP t : Scaled oil price (SOP) is first proposed by Lee et al (1995). The intuition behind this approach is to define a proxy for oil price volatilities. According to this definition, it is expected to more likely observe impact of oil prices on stock market
12
Check Appendix A for the data details.
13
Park (2007), Cunado and Perez de Gracia (2003), Jim´ enez-Rodr´ıguez and S´ anchez (2004)
14
