Oil price risk and stock markets

Oil price risk and stock markets

APT in a global perspective

August 2007

Oil price risk and stock markets

APT in a global perspective

Johannes Jacques Schoonman

Program:

MSc in Economics (Finance and Investments)

Faculty: Faculty

of

Economics, University of Groningen

Table of contents

1. Introduction ... 4

2. Theory & Literature ... 5

2.1 Theoretical background... 5

2.2 Previous studies ... 10

3. Data and methodology ... 19

3.1 Data ... 19

3.1.1 Country index returns... 19

3.1.2 Crude oil price... 23

3.1.3 Other factors... 26

3.2 Methodology ... 30

3.2.1 Arbitrage Pricing Theory ... 30

3.2.2 Other issues ... 31

3.2.3 Robustness tests ... 32

4. Results ... 33

4.1 Determine factor sensitivities... 33

4.1.1 Relation between oil price changes and equity index returns ... 33

4.1.2 Other issues ... 33

4.1.3 Robustness test for oil price sensitivity... 35

4.1.4 Determine sensitivities for Arbitrage Pricing Theory... 37

4.2 Pricing of risk factors... 45

5. Conclusion... 47

6. References ... 49

1. Introduction

Crude oil is one of the most valuable fossil fuels, used in virtually all production processes in the world, either directly or indirectly. Therefore, it is not surprising that world media devote a lot of attention to the availability and price of crude oil. Oil prices are presented daily in the newspaper and unexpected changes in the crude oil price are often considered as the cause of shifts in stock markets. However, the influence of oil price changes on stock markets is studied less often than one would expect, based on the frequently supposed relation by the media.

Chen, Roll and Ross (1986) is the first study that employs the Arbitrage Pricing Theory (APT) to investigate whether oil price changes (among other factors) influence stock returns in the US market. Ferson and Harvey (1994) extent Chen, Roll and Ross (1986) by investigating whether oil price changes drive international equity markets by employing an international model, based on the APT. Another study, by Driesprong, Jacobsen and Maat (2005) provides evidence that oil price changes predict stock market returns worldwide. In their thirty-year sample of monthly data for developed stock markets, they find statistically significant predictability of oil price changes in 12 out of 18 countries, as well as the world market index. Driesprong, Jacobsen and Maat (2005) conclude that oil price changes lead stock market returns by one month. From the results of previous studies, it can be concluded that the relation between oil price changes and stock (market) returns is less straightforward than is often presented and still open for debate. Taking the results of previous studies into account, the research questions of this study are:

• Is there a negative relation between oil price changes and country stock indices? • Is oil price risk a priced risk in the international capital market?

Oil price risk is defined as the sensitivity of a country’s equity index to oil price changes. To answer these questions, the International Arbitrage Pricing Theory, developed by Solnik (1982), is employed. Fifty country stock indices (of both developed as well as developing countries) are examined for their sensitivity to oil price changes using time series regressions. The oil price sensitivities derived from time series regressions are used as a proxy for oil price risk. Using cross-sectional regressions it is investigated whether there exists a risk premium for oil price risk in the international capital market. Apart from oil price changes, other factors are taken into account to improve the explanatory power of the model. The possibility that oil price changes lead stock market returns is also taken into account. Taking the methodologies of previous studies into consideration, this research is a combination of the work by Ferson and Harvey (1994), Driesprong, Jacobsen and Maat (2005). To make the research questions testable, two hypotheses are formulated:

Ha,1: There is no relation between oil price changes and country equity indices

Ho,2: The oil price risk of country equity indices is priced in the international capital market

Ha,2: The oil price risk of country equity indices is not priced in the international capital market

The paper is organized as follows. In section 2 an extensive overview of the theoretical background of the risk-return relation is presented by means of the mean-variance analysis, CAPM and APT. The second part of section 2 provides the methodology and results of previous studies. Section 3 presents the data and methodology that are used in this study. Section 4 presents the empirical results. Section 5 provides the conclusions.

2. Theory & Literature

2.1 Theoretical background

Variance is the most widely accepted measure to quantify risk beared by investors. Assets that are more risky have a higher variance on return. The theory, developed by H. Markowitz (1959) that analyzes the trade-off between expected portfolio returns and variance is known as mean-variance analysis. In this section a theoretical overview is provided of the mean-variance analysis, Capital Asset Pricing Model and the development towards the Arbitrage Pricing Theory1_.

Mean-variance analysis relies on two assumptions.

The first assumption is that in making investment decisions, investors care only about the means and variances of returns of their portfolio over a particular period. The portfolios relevant for choice by investors are mean-variance efficient, which means (I) they minimize portfolio return variance, given expected return, and (II) they maximize expected return given variance. Investors prefer a higher mean return as it implies that on average, they will be wealthier. A lower variance is preferred as it implies that there will be less dispersion in the possible wealth outcomes.

The second assumption is that financial markets are frictionless. In frictionless markets, all investments are tradable at any price and in any quantity (with no short sales restrictions). In addition, there are no transaction costs, regulations, or tax consequences of asset purchases or sales.

The mean-variance analysis starts with a mean-standard deviation diagram that shows the feasible set. The feasible set is the set of expected mean and standard deviation outcomes that are achieved from all possible portfolios of risky stocks; see figure 1. The shaded area including the black boundary represents the feasible set. Portfolios lying at the upper left boundary of the feasible set, called the efficient frontier of risky stocks, provide the highest return given the level of risk. Based on the afore mentioned assumptions, investors wish to hold portfolios lying on the efficient frontier.

1 _{This theoretical overview is derived from Grinblatt and Titman (2002) and Elton, Gruber, Brown and}

Figure 1: The feasible set

If a risk free investment opportunity is added to the feasible set, it changes the shape of the feasible set and the efficient frontier. The line from the risk-free return that lays tangent to the efficient frontier in point T is called the Capital Market Line (CML). Figure 2 shows this. Point T identifies the tangency portfolio. The CML provides the best trade-off between risk and return compared to any other combination of the risk free investment and a portfolio on the efficient frontier. When investors have homogeneous beliefs; all investors reach the same conclusions about the means and standard deviations of all feasible portfolios and will all hold the tangency portfolio. By choosing to hold the tangency portfolio, investors all want to invest in the same combination of risky assets. Under the assumptions of the mean-variance analysis, and assuming the existence of a risk-free asset, all investors want to select portfolios on the Capital Market Line. They can decrease (increase) their desired level of risk by investing part of their wealth in (short selling) the risk-free investment, thereby investing in a portfolio on the CML.

Figure 2: The capital market line

•

•

•

•

•

Expected standard deviation of portfolio return

E xp ect ed p or tfo lio r et ur n

Efficient frontier of risky stocks Minimum variance portfolio

•

•

Expected standard deviation of portfolio return

)

(

r

r

r

r

=

+

β

−

)

var(

)

,

cov(

_r

r

r

=

β

The Capital Asset Pricing Model (CAPM) of Sharpe (1964), Lintner (1965) and Black (1972) is based on the mean-variance analysis. The main insight of the CAPM is that the variance of a stock by itself is not an important determinant of the stock’s return. What is important is the market beta of the stock, which measures the covariance of the stock’s return with the return on a market portfolio, scaled by the variance of that index (see equation 2). All investors together hold the total supply of risky stocks in the market. The market portfolio is a portfolio where the weight on each asset is the market value (also called market capitalization) of the asset divided by the market value of all risky assets. The market portfolio therefore consists of the same combination of risky stocks as the tangency portfolio; under the assumptions of the CAPM, the tangency portfolio is equal to the market portfolio. By identifying the tangency portfolio2 _{the relation between the relevant risk of an investment and its}

expected return can be derived:

(1)

where

r

r

market portfolio,

r

(2) While appealing because of the simplicity of only relating the risk-return trade-off to the market portfolio, empirical tests provide evidence against the validity of the CAPM. For example, test results of Reinganum (1981) demonstrate that estimated betas are not systematically related to average returns across stocks. The average returns of high beta stocks are not significantly different from the average returns of low beta stocks. That is, portfolios with widely different estimated betas based on standard market indices do not appear to measure a “risk which is priced in the market”. Lakonishok and Shapiro (1986) examine the monthly returns of all stocks traded on the New York Stock Exchange (NYSE) and find that returns on individual stocks are not specifically related to their degree of systematic risk, but is significantly related to their degree of market capitalization. They conclude that the traditional market beta is not able to explain the cross-sectional variation in return; only size in terms of market capitalization can significantly explain it. Fama and French (1992) also study the monthly returns of NYSE stocks and find an insignificant relationship between beta and average returns. They conclude that the CAPM cannot describe average stock returns during the period 1941-1990 and only market capitalization and the ratio of book value to market value have significant explanatory power for portfolio returns. A central prediction of the CAPM is that the market beta

2 _{The tangency portfolio is derived when the ratio of the risk premium of every stock and portfolio to its}

t i J j ij j i i F r _, 1

ε

β

α

∑

∑

β

λ

λ

alone is sufficient to explain differences in stock returns. The fact that several researches find other explanatory variables that explain stock returns casts doubt on the strength of the CAPM.

A new and different approach to explain the pricing of assets is the Arbitrage Pricing Theory (APT), developed by Ross (1976, 1977). In both the CAPM and the APT different firm-specific forces can influence the expected return on individual stocks. These idiosyncratic effects tend to cancel out in well-diversified portfolios. However, a well-diversified portfolio of stocks is not risk free because there are common economic forces that influence all stock returns. These forces are not firm specific and can therefore not be diversified away. According to the CAPM these common economic forces are captured in the systematic risk factor denominated as the market risk. The APT description of equilibrium is more general than that provided by a CAPM-type model since pricing can be affected by influences beyond the market portfolio. The APT requires that the return of any stock be linearly related to a set of common factors. The common factors are proxies for those events in the economy that affect a large number of different investments. If

r

(3)

for all stocks i (i = 1,…,N).αi is the intercept term, βij is the sensitivity of stock i to factor

F

1,…,J).

ε

ε

(

)

[

]

E

ε

ε

ε

E

(

ε

ε

)

=

0

So far, the model described is equal to a general multi-factor model. The contribution of APT is in demonstrating how and under what conditions one can go from a multi-factor model to a description of equilibrium. The APT is based on the law of one price: items that are the same cannot sell at different prices. Within the APT this means that e.g. two portfolios that have the same expected risk cannot sell at a different expected return; there exists “no arbitrage in expectations”. Because the no arbitrage conditions should hold for any two portfolios of stocks, it is not necessary to identify all risky assets or a “market portfolio” to test the APT.

The APT that arises from the return-generating process in equation (3) is:

k c E P= ( )

( )

[

]

( )

P

c

k

dk

c

E

c

E

d

P

c

P

dP

₊

₌

₋

₊

where λj (j = 1,…,J) is the increase in the expected return of stock i for a one-unit increase in βij. Thus,

λj is the expected excess return for bearing the risk associated with

F

the risk factor stock i is exposed to (via sensitivity measure βij) is said to be priced. A portfolio (p) that

is perfectly diversified (for which the unique risk is zero) and with no factor exposures ( βpj = 0 for all

j=1,…,J); such a portfolio has zero risk, and from equation (4) expected return is λ0. It follows that λ0

is the risk free rate of return. An important characteristic of the model is that it is very general. Although the model allows describing equilibrium in terms of any multi-factor model, it provides no evidence as to what might be the appropriate model. All the theory specifies is a structure for asset pricing; factors that should affect expected returns are not specified. This study introduces an international APT. In order to apply an APT in a global setting, an additional assumptions is needed. National equity markets are assumed to be perfectly integrated in a global economy, without barriers to ‘foreign’ equity investments.

Factor determination

The purpose of this study is to find the influence of oil price changes on global equity markets. To account for possible indirect effects of oil price changes on global equity markets via macroeconomic variables, additional macroeconomic factors are added in the factor model. Macroeconomic time-series are chosen that are proxies for common economic forces that affect stock prices. As the APT does not specify which (macroeconomic) factors affect expected returns one should use a theoretical framework that explains stock price developments to determine which factors are appropriate. By the diversification argument that is implicit in equilibrium models, only general economic state variables will influence the pricing of large stock market aggregates. Systematic variables that affect the cost of capital or (expected) cash flows that are distributed to shareholders (dividends) will influence stock market returns. Additionally, variables that have no direct influence on current cash flows but do for example describe the changing investment opportunity set are relevant. As derived from Chen, Roll and Ross (1986), stock prices can be written as expected discounted cash flows:

(5) where P is the stock price, c is the expected cash flow stream and k is the discount rate. The discount rate is an average over time. This implies that actual returns in any period are given by:

subsection macroeconomic variables are discussed that were found to be systematic risk factors in previous studies.

2.2 Previous studies

Since the introduction of the APT by Ross (1976, 1977), many studies have been conducted that model equity returns as a function of innovations in macroeconomic variables. Apart from the theoretical reasoning in section 2.1, results from previous studies are taken into account to determine which macroeconomic factors are expected to explain equity returns. This section presents results of previous studies that examine the impact of various macroeconomic variables on stock (index) returns. As this research investigates whether oil price risk affect international stock index returns, special attention is paid to studies that examine the influence of oil price risk on stock (indices). Apart from studies that employ the APT using macroeconomic variables, results from studies that use other methodologies to investigate the relation between oil prices and equity returns are presented. At the end of this section, an overview is presented of the methodological properties and conclusions of the discussed studies. The set of factors taken into consideration is derived from, among others: Chen, Roll and Ross (1986), Hamao (1988), Ferson and Harvey (1994) and Burmeister, Roll and Ross (2003). The variables considered and discussed in this subsection are:

• Oil prices • Inflation

• Industrial production • Confidence risk • Time horizon risk • Exchange rate risk

• (World) equity market index

Oil prices

Oil prices did not fluctuate much before 1973. A few large western oil companies known as the “Seven Sisters”, being well organized and able to negotiate as a cartel, stabilized the oil price through price and production controls during much of the twentieth century. The influence of the Seven Sisters declined when Arab states began to gain control over oil prices and production, through the formation of Organization of the Petroleum Exporting Countries (OPEC) in 1960. The OPEC really gained power by the 1970s. The Yom Kippur War that started on October 6th _{1973 which let to an oil}

0 10 20 30 40 50 60 70 80 1965 1970 1975 1980 1985 1990 1995 2000 2005 Year P ri ce p er b arrel in US $

Source: 2007 Federal Reserve Bank of St. Louis: http://research.stlouisfed.org The graph clearly shows the increased volatility in crude oil prices since 1973. The numbers indicate events that had an impact on the oil price. 1 indicates the start of the Yom Kippur War. 2 and 3 respectively indicate the Iran/ Iraq War and the Gulf war. 4 indicates the Asian crisis. 5 indicates the price impact of the terrorist attacks in the United States on 9/11/2001 due to decrease in demand. 6 indicates the recent oil price hike due to the combined effect of the Iraq war and the increased demand for oil by rising Asian economies. The crude oil market is the largest commodity market in the world. Nowadays, total world consumption is around 87 million barrels (about 1425 million litre) per day. According to the International Energy Agency, consumption of crude oil is expected to increase evenly with around 40% over the next 25 years (see figure 4).

Figure 4: World consumption of crude oil in million barrels per day

0 20 40 60 80 100 120 140 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030 Year M illio n b ar re ls p er d ay E E E E E

Note: ‘E’ is the estimation operator Source: 2007 International Energy Agency: http://eia.doe.gov3

3

Combination of historical consumption data from March 2007 ‘International Petroleum Monthly’ and forecasts published in ‘The International Energy Outlook 2006’, both published by the Energy Information Administration Figure 3: Development of crude oil price (West Texas Intermediate Oil Price (US$/Barrel))

As Adelman (1993, page 537) poses: “Oil is so significant in the international economy that forecasts of economic growth are routinely qualified with the remark: ‘Provided there is no oil shock’.” The International Monetary Fund estimates that a $5 price increase a barrel reduces global economic growth by 0.3% in the following year4_{. While many studies focus on the effects of oil price changes on}

the economy5_{, relatively few studies analyze the relation between oil prices and stock market prices.}

Most studies that examine the relation between oil price changes and stock prices use the Arbitrage Pricing Theory. Chen, Roll and Ross (1986) is the first study that investigates, among other macroeconomic factors, the impact of innovations in the crude oil price on US stock price returns using the APT. They find evidence that oil price risk is priced in the (sub-) period 1958-67. Hamao (1988) applies roughly the same methodology as Chen, Roll and Ross to the Japanese stock market for the period (1975-1984). Innovations in the oil price are found not to be priced in the Japanese stock market. Interpretation provided by Hamao (1988) is that the pricing influence is already taken into account by other factors.

However, Kaneko and Lee (1995) did prove oil price risk to be a significant factor in determining equity returns in Japan. Kaneko and Lee attribute the difference in findings primarily to difference in sample period and empirical methodology. They used a Vector Auto-Regression (VAR) model for the period 1975-1993. Other studies that investigate the influence of oil price changes and stock returns using VAR methodology are, among others: Jones and Kaul (1996), Huang, Masulis and Stoll (1996) and Ciner (2001). Jones and Kaul (1996) find that the reaction of Canadian and US stock prices to oil price innovations can be completely accounted for by the impact of these innovations on real cash flows. However, the results for Japan and the UK are not as strong. They find evidence that oil prices do have an effect on aggregate stock returns. Huang, Masulis and Stoll (1996) find evidence for significant Granger causality from oil futures to stocks of individual oil companies. However, they detect no impact on a broad-based index like the S&P 500. Based on this result, they conclude that the influence of oil price shocks on the aggregate economy is more of a myth than reality. Ciner (2001) also investigates the relationship between oil futures price changes and the S&P 500 index in the sense of a VAR. Ciner (2001) however suggests that the fact that Huang, Mausulis and Stoll (1996) found no evidence for a relation between oil futures’ price changes and the S&P could be due to the fact that the tests that Huang, Masulis and Stoll (1996) rely on are not powerful enough to detect nonlinear linkages6_{. Ciner (2001) relies on nonlinear causality tests of the effect of daily oil futures price}

changes on the S&P 500 index for the 1980s and 1990s. Tests of a linear relationship between oil

4

“The impact of higher oil prices on the Global Economy,” prepared by the research department of the International

Monetary Fund (2000), www.imf.org/external/pubs/ft/oil/2000/oilrep.pdf.

5 _{See Jones, Leiby and Paik (2003) for an extensive treatment on the numerous studies that examine the relation}

between oil prices and macro economic variables

6 _{Ciner (2001) defines a non-linear relation as one that is influenced asymmetrically by increases and decreases}

futures price changes and the S&P 500 index are rejected; tests of a nonlinear relationship however cannot be rejected. The nonlinear causal relationship is stronger in the 1990s than in the 1980s.

Ferson and Harvey (1994) use the APT to empirically examine the returns and expected returns of eighteen national equity markets. Factors they investigate are similar to those investigated by Chen, Roll and Ross (1986). However, Ferson and Harvey (1994) use international risk factors in stead of national. In addition, they include an exchange rate factor. They find that a four-factor model consisting the world market return, a world currency-risk factor, the change in long-term expected inflation and a world oil factor has most explanatory power, explaining an average of 36% of equity return variation. They do not provide conclusive evidence that exposure to oil price risk is priced in an international equilibrium model. An explanation provided for the lack of evidence is that in their study, they include the tumultuous times of the Yom Kippur War that started I 1973, that led to an oil price hike.

Basher and Sadorsky (2006) study the impact of oil price risk on a set of emerging stock market returns. They use an international multi-factor model in line with Pettengill, Sundaram and Mathur (1996) that allows for both conditional and unconditional risk factors to investigate the relationship between oil price risk and emerging stock market returns. For the relationship between oil price beta and returns, when using an unconditional model, the estimated coefficient for the oil price beta is positive and significant. Results differ with data frequency used. In case of daily data, only sensitivity to oil price increases is associated with a risk premium. However, when weekly data is used, only sensitivity to oil price decreases are associated with a risk premium. In case of monthly data both sensitivity to increases and decreases are associated with a risk premium.

Driesprong, Jacobsen and Maat (2005) report evidence that unconditional changes in oil prices predict stock market returns worldwide. Stock returns tend to be lower after oil price increases and higher if the oil price falls in the previous month. In their thirty-year sample of monthly data for developed stock markets, they find statistically significant predictability in 12 out of 18 developed countries as well as the world market index. For shorter time series of emerging markets, results are similar. The predictability results are not only significant but also robust with respect to sample periods, different kinds of oil prices they consider, well known calendar effects and inclusion of economic variables. Investors seem to underreact to information in the oil price: a rise in oil prices lowers future stock market returns. Driesprong, Jacobsen and Maat (2005) argue that one way to explain these findings is underreaction based on the gradual information diffusion hypothesis proposed by Hong, Lim and Stein (2000). Although Driesprong, Jacobsen and Maat (2005) provide evidence of linkages between oil price innovations and stock price returns, they do not investigate whether this oil price exposure is non-diversifiable and priced in the international capital market.

most 5 out of 13 countries show a significant relation with contemporaneous oil price changes, when oil prices are measured in local currency. Their study does not encompass the question whether oil price risk is a priced factor in the sense of an equilibrium pricing model.

Inflation

As Bilson, Brailsford and Hooper (2000) point out, common stocks are traditionally viewed as a hedge against inflation, due to the fact that equity represents a contingent claim on the realm of assets of the firm. In the presence of inflation, the value of the contingent claims will be revised upward. Therefore, proportionate increases in prices should not affect the real rates of return on equity (Day, 1984). However, the monetary assets of the firm (i.e. cash, securities, receivables and debt) will be independent of fluctuations in the price level. Hence, it is only the real component of the firm that will be hedged against changes in inflation (Hong, 1977). Empirical tests have found a negative relationship between inflation and nominal stock returns (Fama and Schwert, 1977; Gultekin, 1983). Explanation from Ferson and Harvey (1994) is that higher inflation may signal higher levels of economic uncertainty which make consumers worse off, therewith lowering expenditures, company sales and eventually cash flows. As can be seen in table 1 and 2, studies have included an inflation factor when employing a multifactor model that explains stock returns. Inflationary risk can be priced within APT if national equity markets differ in their exposure to a world consumer price index.

Industrial Production

Several studies find national industrial production as a risk factor in the sense of a multi factor model; e.g. Chen, Roll and Ross (1986) and Hamao (1988). Ferson and Harvey (1994) are the first to examine an industrial production index as a risk measure in an international multifactor model. To measure international industrial production they use a weighted average of industrial production growth rates in the G7 countries, using a measure of relative production shares as the weights. They find no risk premium for exposure to changes in the industrial production index the examined countries. As this research examines other/ more countries, a different time frame and uses a different measure for international industrial production, the factor is examined. As posed by Chen Roll and Ross (1986), industrial production affects opportunities facing investors and the real value of cash flows.

Confidence risk

exposed to changes in the yield curve. Chen Roll and Ross (1986) argue that it is natural to think of confidence risk as a direct measure of the degree of risk aversion implicit in pricing. In any month when the return on risky corporate bonds exceeds the relatively safe bond by less than the long-term average, investor confidence is said to be higher as investors are less risk averse and therefore require a lower compensation for incurred risk. An unexpected increase in investor confidence will put more investors in the market for “risky” stock(s) indices, increasing their price and producing a positive return for those who already held them. Similarly, a drop in investor confidence leads to a drop in the value of these investments. Some stocks have a negative exposure to the confidence risk factor, suggesting that investors tend to treat them as a "safe haven" when their confidence is shaken (Burmeister, Roll and Ross, 2003). Chen Roll and Ross (1986) and Hamao (1988) do find evidence that confidence risk is a priced factor in national equity markets.

Time horizon risk

Time horizon risk is the unanticipated changes in investors’ desired time to payouts. Time horizon risk is the change in difference between the 10 years US treasury rate and the 3 months Treasury bill. A positive realization of time horizon risk means that the return on long term bonds exceeds the short term bond by more than the long-term average; indicating that investors require a higher compensation for holding investments with relatively longer times to payout. Differences between the rate on bonds with a long maturity and a short maturity affect the value of payments far in the future relative to near-term payments (Elton, Gruber, Brown and Goetzmann, 2003). The difference in impact of time horizon risk on stock(index) returns is caused by the influence via discount rates. The sensitivity differs across stock(indices) depending on its growth characteristics; because of the expected timing of (expected) dividend payments. Expectation is that equity in economies that are strongly emerging, therewith generally having a strong demand for cash for further growth and a relative late distribution of dividends, will be more sensitive and is negatively affected by increases in the spread. Chen Roll and Ross (1986) provide evidence that time horizon risk is accompanied by an equity risk premium for the US market. Also several other researches have demonstrated the relation between term structure and equity risk premia e.g. Campbell (1986), Fama and French (1989) and Chen (1991).

Exchange rate risk

previous APT studies (Hamao, 1989; Ferson and Harvey, 1994; Basher and Sadorsky, 2006). Because of the large number of countries examined, this study includes a single index in the international APT to approximate exchange rate risk (in line with Ferson and Harvey (1994) and Basher and Sadorsky (2006)).

Market risk

Table 1: Literature overview: Multifactor Asset pricing Models

Author Year Study Object Area Period Frequency Model Oil variable

Other independent

variables

Empirical results: pricing outcome with (t-values)

Dependent variables

Results with regard to

oil price risk Conclusions

α: 13,641 (2,755) NYSE: -2,922 (-0,633) INDPR: 15,056 (3,054) EXINFL: -0,148 (-1,600) UNINFL: -0,950 (-2,376) RISKPR: 10,393 (2,972) TERM: -6,860 (-1,879) α: 18,787 (2,167) VWTSE: 5,889 (0,680) INDPR: 7,956 (1,581) EXINFL: 0,505 (2,233) RISKPR: 16,489 (1,806) TERM: -0,01849 (-1,914) WDRET: 0,717 (0,049)* FOREX: 0,602 (0,340)* EXINFL: 0,211 (0,252)* OIL: 0,155 (0,167)* WDRETUP: 0,732 (2,302) WDRETDOWN: -2,116 (5,658) OILUP: 3,379 (1,937) OILDOWN: 5,655 (3,343) WDRET NYSE FOREX RISKPR UNINFL TERM MOTY TED STRIR INDPR OIL NYSE

Spread between 90-day Eurodollar and 90 days US t-bill rate Weighted average of industral production growth rates Value-weighted NYSE index

Return on MSCI world index

Change in basket of exchange rates, relative to US Dollar Unexpected component of inflation measure

Month-of -the Year

Short-term real interest rates (or average of..) Changes in oil price factor as described in 8th column

Multifactor asset pricing

model

Log difference of WTI crude oil futures contracts

(NYMEX traded)

WDRET, FOREX, OIL

Find strong evidence that oil price risk impacts stock price returns in emerging markets although the relationship is not robust over data frequencies used. Results are stronger when oil price increases and decreases are seggregated.

Find several macroeconomic factors to be significantly priced. Market index NYSE has insignificant influence on pricing when combined with the macroeconomic factors. Innovations in

real per capita consumption and innovations in an index of oil price

showed no overall effect on asset pricing

NYSE Stock Portfolio

Return

Oil betas were insignificant for pricing

in overall period but found significantly priced in sub-period 1958-67. Pricing was 6,111% with t-value 1,978 Hamao 1988 Empirical investigation of the Arbitrage Pricing Theory

in the Japanese equity market using Japanese macroeconomic factors Japan Monthly APT Log difference of Producer Price Index/ Crude pertroleum series 1992 - 2005 Daily/ Weekly/ Monthly Chen, Roll and Ross 1986 Empirical investigation of the Arbitrage Pricing Theory

in the U.S. equity market using U.S. macroeconomic

factors

U.S.

Basher and Sadorsky 2006

Study the impact of oil price changes on emerging stock

market returns using an international multi-factor model that allows for both

unconditional and conditional risk 21 emerging stock markets NYSE, INDPR, EXINFL, UNINFL, RISKPR, TERM 1975-1984 WDRET, FOREX, UNINF, EXINFL, TED,STRIR, INDPR APT 1958- 1984 Monthly Multifactor asset pricing model Change in price per barrel at wellhead ('74-89') and posted WTI price ('69-'73) Japanese NSE Stock Portfolio Return Country MSCI returns in US dollars Significantly priced in one subperiod (1980's). Together with WDRET, FOREX and

EXLTINF explains over 36% of equity return variations Sensitivity to oil price

changes are not significantly priced in

the Japanese stock market.

Oil price futures affect emerging stock market index returns in a non-linear fashion. Results depend on data

frequency

*) Measures the monthly excess return; standard error in parenthesis Ferson and

Harvey 1994

Empirical investigation of multifactor asset pricing models for the returns and

expected returns on 18 national equity markets

Market index VWTSE has insignificant influence on pricing when combined with the macroeconomic factors. An exchange rate factor, and innovations in

the Arab oil price showed no influence on asset pricing

Country MSCI returns Log difference of

the Arab light spot price (in japanese Yen) VWTSE, INDPR, EXINFL, RISKPR, TERM

Multifactor Asset Pricing Models

Cannot reject the unconditional mean-variance efficiency of the MSCI world equity index. Furthermore, tests do not reject the hypothesis that the returns are consistent with a four-factor model. 16 OECD Countries + Singapore and Hong Kong 1970 - 1989 Monthly

Value-weighted NYSE index

Table 2: Literature overview: Other studies examining oil price risk

Author Year Study Object Area Period Frequency Model Oil variable

Other independent variables Dependent variables Jones and Kaul 1996

Relationship between oil price shocks and internationl stock markets via changes in current/ and expected future

real cash flows.

U.S., Canada, Japan and United Kingdom 1947 -1991 Quarterly Standard cash-flow/ dividend valuation model Producer price index of oil and related products

INDPR, UNINFL, OIL,

RISKPR, STRIR, TERM

Real stock market index returns

Huang Masulis and

Stoll 1996

Examines the correlations between daily returns of oil futures contracts and stock returns, with particular attention paid to "oil stocks"

and stock market indices

U.S. 1979 - 1990 Daily Vector Autoregressi on Heating oil futures and Crude oil futures

contracts (NYMEX traded)

OIL, 1MTB Company stock returns

Ciner 2001

Examines the dynamic linkages between oil prices

and the S&P500 stock market index U.S. 1979 - 2000 Daily Vector Autoregressi on Heating oil futures and Crude oil futures

contracts (NYMEX traded) OILFUT, DOTW, MOTY, EXLTINFL

S&P500 index returns

Driesprong, Jacobsen and

Maat 2005

Investigate whether changes in oil prices predict stock

returns

48

countries 1973 - 2003 Monthly OLS regression

WTI, Dubai and Brent spot prices and Brent and light crude oil

futures

DIVD, EXLTINFL,

MOTY

MSCI Country index returns

FOREX DIVD UNINFL RISKPR EXLTINF RREG DOTW TERM MOTY INDPR STRIR 1MTB OIL OILFUT

Change in basket of exchange rates, relative to US Dollar Unexpected component of inflation measure

Change in measure of expectation of long-term inflation

Oil futures returns are not correlated with stock market returns, even contemporaneously, except in the case of oil company returns.

Oil price futures affect stock index returns in a non-linear fashion

Dividend yield

Difference between Baa bonds and long term gov. bonds Datastream total return country

index (local currency) Log difference of

OPEC oil price

The reaction of oil price shocks on U.S. and Canadian stock prices can be completely accounted for by their impact on real cash flows

Other studies examining oil price risk

Oil prices show a significant, 1 month lagged relation with 25 out of 48 MSCI Country index returns and oil returns. The MSCI world shows a significant

(negative) relation with all five oil series

Only 4 out of 15 countries show a significant relation between stock index returns in Asia-Pasific region and oil prices (only when oil prices are measured in local currencies and conditionality is taken into account. Examines the relationship

between realized stock index return and oil and exchange rate factors

15 countries in Asia-Pacific region Weekly International factor model 1994 - 2004 OIL, FOREX, RREG Nanda and Hammoudeh 2006 Day-of-the-Week Month-of -the Year

Rolling regression on country returns

Difference between long term governemt bonds and t-bills Weighted average of industral production growth rates Short-term real interest rates (or average of..)

Changes in oil price factor

One Month T-bill Crude oil futures price

3. Data and methodology

Within this research the Arbitrage Pricing Theory is used to investigate whether oil price risk is priced in global equity markets. Monthly data is used as this is used in several comparable studies (Chen, Roll and Ross, 1986; Hamao, 1989; Ferson and Harvey, 1994. See table 1.). Moreover, monthly data allows the inclusion of monthly-publicized macroeconomic variables (inflation and industrial production). The period studied is from January 1995 until February 2006. This period is chosen as this is the longest period possible within the data constraints. Before 1995 considerably less data on country equity market indices are available. Within this section, the data and methodology are presented. Subsection 3.1 describes the data. Subsection 3.2 presents the methodology.

3.1 Data

3.1.1 Country index returns

In order to investigate the question whether oil price risk is priced in global equity markets, index returns of 50 national stock indices are selected. Total returns on Morgan Stanley Capital International (MSCI) country indices are used to proxy national stock indices. Total return indices measure the market performance of the index, including price performance and income from dividend payments. The income from dividend payments is reinvested in the index and thus makes up part of the total index performance. Of the 50 countries, 23 are developed economies, 27 are emerging economies (as classified by MSCI-Barra). All MSCI country indices are selected that run since January 1st _{1995 or}

before and are still active. This approach is used to be able to perform a meaningful cross-sectional regression analysis with enough observations. Ferson and Harvey (1994) use 18 countries in their analysis (16 OECD + 2). Basher and Sadorsky (2006) use 21 emerging stock markets in their analysis. This study uses considerably more country index returns than both comparable studies. In line with Ferson and Harvey (1994) and Basher and Sadorsky (2006), all series are in US Dollars so investment decisions are made from the perspective of an investor who has a US Dollar trading account. Country index returns are obtained by taking the log first difference of the monthly MSCI country index levels. Table 3 presents the statistical descriptives of the return on the MSCI country indices. Russia has the highest mean return on its stock market index during the period under consideration, averaging at 20.9% on an annualized basis. The average return of the MSCI world index is 6.7% per year. Several (South-) East Asian countries have negative average equity index returns over the examined period. A plausible explanation for the negative return of the Asian stock markets is the fact that the examined period includes the period of the Asian financial crisis (mid-1997).

To analyze the distribution of returns, the measures for skewness and kurtosis are presented. Skewness is a measure of asymmetry of the distribution of the series around its mean. The skewness of a symmetric distribution, such as the normal distribution, is zero. 40 out of 50 MSCI country index returns have negative skewness which means that their distribution has a long left tail. 10 out of 50 country returns have positive skewness (their distribution has a long right tail). Kurtosis measures the peakedness or flatness of the distribution of the series. The kurtosis of the normal distribution is 3. For 48 out of 50 MSCI country index returns kurtosis exceeds 3, which means that the distribution is peaked (leptokurtic) relative to the normal. For 2 out of 50 MSCI country index returns kurtosis is less than 3, which means that the distribution is flat (platykurtic) relative to the normal distribution.

Jarque-Bera test statistics are determined to test whether the series is normally distributed. The test statistic measures the difference of the skewness and kurtosis of the series with those from the normal distribution. The reported probability is the probability that a Jarque-Bera test statistic exceeds (in absolute value) the observed value under the null hypothesis. On a 5% significance level, we reject the hypothesis of normal distribution for 42 out of 50 MSCI country index returns.

Table 3: Statistical descriptives of country index returns

Mean¹ Median Maximum Minimum Std.Dev. Skewness Kurtosis Jarque-Bera Probability Series Name²

Argentina 0.005 0.010 0.377 -0.357 0.115 -0.306 4.365 12.489 0.002 MSARGT$ Australia 0.006 0.010 0.122 -0.131 0.050 -0.448 3.174 4.651 0.098 MSAUST$ Austria 0.008 0.014 0.186 -0.253 0.059 -0.712 5.349 42.129 0.000 MSASTR$ Belgium 0.007 0.009 0.163 -0.222 0.054 -0.941 6.253 78.849 0.000 MSBELG$ Brazil 0.007 0.012 0.341 -0.418 0.114 -0.529 4.084 12.826 0.002 MSBRAZ$ Canada 0.010 0.017 0.160 -0.220 0.057 -0.922 5.061 42.698 0.000 MSCNDA$ Chile 0.002 0.001 0.185 -0.303 0.068 -0.610 5.217 35.749 0.000 MSCHIL$ China -0.005 -0.006 0.382 -0.324 0.109 0.260 4.805 19.709 0.000 MSCHIN$ Colombia 0.010 0.016 0.265 -0.276 0.098 -0.249 3.541 3.017 0.221 MSCOLM$ Czech Republic 0.011 0.021 0.263 -0.327 0.085 -0.550 4.522 19.687 0.000 MSCZCH$ Denmark 0.010 0.016 0.118 -0.183 0.052 -0.744 4.134 19.538 0.000 MSDNMK$ Egypt 0.016 0.006 0.281 -0.147 0.088 0.639 3.420 10.096 0.006 MSEGYT$ Finland 0.011 0.015 0.327 -0.356 0.109 -0.296 4.154 9.399 0.009 MSFIND$ France 0.008 0.013 0.151 -0.190 0.055 -0.379 3.975 8.524 0.014 MSFRNC$ Germany 0.006 0.011 0.203 -0.228 0.062 -0.709 4.987 33.287 0.000 MSGERM$ Greece 0.009 0.002 0.268 -0.280 0.086 -0.016 3.916 4.686 0.096 MSGDEE$ Hong Kong 0.003 0.002 0.283 -0.289 0.077 -0.134 5.189 27.149 0.000 MSHGKG$ Hungary 0.016 0.023 0.351 -0.522 0.104 -0.914 7.975 156.855 0.000 MSHUNG$ India 0.005 0.006 0.199 -0.195 0.083 -0.144 2.435 2.244 0.326 MSINDI$ Indonesia -0.003 0.008 0.466 -0.550 0.149 -0.364 5.597 40.631 0.000 MSINDF$ Ireland 0.006 0.014 0.127 -0.195 0.053 -1.003 4.805 40.661 0.000 MSEIRE$ Israel 0.009 0.015 0.159 -0.201 0.065 -0.575 3.574 9.217 0.010 MSISRA$ Italy 0.007 0.005 0.179 -0.153 0.061 -0.082 3.178 0.328 0.849 MSITAL$ Japan -0.001 -0.004 0.222 -0.124 0.061 0.484 3.356 5.942 0.051 MSJPAN$ Jordan 0.008 0.001 0.165 -0.172 0.049 0.264 4.525 14.536 0.001 MSJORD$

Notes: Period: (Period: January 1995 until February 2006), Frequency: Monthly 1) Mean return: average log first difference of the MSCI country index 2), Series name: Datastream code for the unedited MSCI country index

Table 3: Statistical descriptives of country index returns, Continued

Mean Median Maximum Minimum Std.Dev. Skewness Kurtosis Jarque-Bera Probability Series Name²

Korea 0.004 0.003 0.477 -0.412 0.123 0.210 5.310 30.777 0.000 MSKORE$

Malaysia -0.003 0.003 0.400 -0.403 0.103 -0.476 7.251 105.940 0.000 MSMALF$

Mexico 0.010 0.022 0.190 -0.377 0.090 -1.031 5.311 53.555 0.000 MSMEXF$

Morocco 0.008 0.007 0.212 -0.165 0.052 0.293 4.807 20.146 0.000 MSMORC$

Netherlands 0.006 0.011 0.120 -0.254 0.056 -1.120 6.183 84.583 0.000 MSNETH$

New Zealand 0.002 0.007 0.198 -0.233 0.063 -0.554 4.688 22.772 0.000 MSNZEA$

Norway 0.008 0.011 0.197 -0.334 0.069 -0.910 6.750 97.015 0.000 MSNWAY$ Pakistan 0.001 -0.004 0.317 -0.476 0.120 -0.322 4.845 21.315 0.000 MSPKAKI$ Peru 0.008 0.011 0.304 -0.410 0.083 -0.810 7.841 145.466 0.000 MSPERU$ Philippines -0.009 -0.006 0.393 -0.374 0.094 0.166 5.880 46.943 0.000 MSPHLF$ Poland 0.007 0.015 0.339 -0.430 0.106 -0.273 4.898 21.769 0.000 MSPLND$ Portugal 0.006 0.010 0.246 -0.191 0.063 -0.008 4.295 9.367 0.009 MSPORD$ Russia 0.017 0.044 0.436 -0.904 0.185 -1.135 7.190 126.793 0.000 MSRUSS$ Singapore 0.001 0.007 0.386 -0.420 0.090 -0.369 8.580 176.895 0.000 MSVSNF$

South Africa 0.005 0.010 0.177 -0.369 0.082 -1.098 5.833 71.764 0.000 MSSARF$

Spain 0.011 0.012 0.236 -0.232 0.064 -0.271 4.926 22.355 0.000 MSSPAN$

Sri Lanka -0.001 -0.001 0.395 -0.290 0.103 0.394 5.190 30.251 0.000 MSSRIL$

Sweden 0.010 0.015 0.193 -0.249 0.076 -0.548 4.058 12.955 0.002 MSSWDN$

Switzerland 0.008 0.009 0.185 -0.160 0.050 -0.405 4.698 19.776 0.000 MSSWIT$

Taiwan -0.002 -0.014 0.248 -0.232 0.090 0.054 2.912 0.109 0.947 MSTAIW$

Thailand -0.008 -0.013 0.475 -0.414 0.131 0.102 5.429 33.177 0.000 MSTHAF$

Turkey 0.013 0.032 0.513 -0.553 0.168 -0.322 4.524 15.275 0.000 MSTURK$

United Kingdom 0.005 0.006 0.102 -0.132 0.039 -0.479 3.741 8.195 0.017 MSUTDK$

United States 0.008 0.011 0.118 -0.106 0.042 -0.403 3.051 3.646 0.162 MSUSAM$

Venezuela 0.002 -0.004 0.480 -0.638 0.143 -0.859 7.948 153.191 0.000 MSVENF$

World 0.006 0.011 0.085 -0.144 0.041 -0.788 4.030 19.806 0.000 MSWRLD$

Notes: Period: (Period: January 1995 until February 2006), Frequency: Monthly, 1) Mean return: average log first difference of the MSCI country index 2), Series name: Datastream code for the unedited MSCI country index

0 5 10 15 20 25 30 35 2001 2002 2003 2004 2005 2006 Year

ARAB OPEC WTI BRENT

3.1.2 Crude oil price

In previous research that investigates the impact of oil price risk on stock markets a variety of oil price factors is used. In this research, the widely traded Brent oil price benchmark is selected to measure the oil price. In line with Driesprong, Jacobsen and Maat (2005), the change in the U.S. dollar spot price per barrel of Brent crude oil is selected as the oil price factor. Figure 5 and table 4 present the price developments and a correlations matrix of the most frequently traded oil price benchmarks. It clearly shows the strong co-movement of the various oil prices. Moreover, Driesprong, Jacobsen and Maat (2005) found that the type of crude oil benchmark that is used to analyze the relation with stock indices does not have a significant effect on regression results.

Note: BRENT is the Crude Oil-Brent Dated FOB U$/BBL. WTI is the Crude Oil-WTI Spot Cushing U$/BBL. OPEC is the OPEC Oil Basket Price US$/Bbl. ARAB is the Arabian Gulf Arab Light Crude Oil Spot Price US$/Barrel.

Source: Datastream Table 4: Oil price correlations

BRENT WTI OPEC ARAB

BRENT 1.00

WTI 0.93 1.00

OPEC 0.97 0.96 1.00

ARAB 0.92 0.90 0.95 1.00

Note: BRENT is the Crude Oil-Brent Dated FOB U$/BBL. WTI is the Crude Oil-WTI Spot Cushing U$/BBL. OPEC is the OPEC Oil Basket Price US$/Bbl. ARAB is the Arabian Gulf Arab Light Crude Oil Spot Price US$/Barrel.

Source: Datastream Determine lag-length

The first step in determining the relation between oil price changes and country equity index returns is to find the most relevant lag length. Only one lag length can be selected for all countries, otherwise it is not possible to determine the pricing of the exposure to one oil price risk factor using a cross sectional regressions. Table 5 contains the estimation results for the time series regressions where the contemporaneous and lagged Brent oil price variables, up to six months, are included. The

Table 5: Determination of appropriate time lag for Brent oil price variable Period:

01/1995 - 02/2006 c βbrent βbrent-1 βbrent-2 βbrent-3 βbrent-4 βbrent-5 βbrent-6

Australia 0.01 0.07*** -0.03 -0.04 0.02 -0.04 0.00 0.02 Austria 0.01 0.00 0.01 -0.07 0.02 -0.04 0.00 0.03 Belgium 0.01*** -0.06 -0.07*** -0.06 -0.02 -0.01 -0.01 -0.01 Canada 0.01 0.11* 0.01 0.07 0.03 -0.01 -0.04 0.06 Denmark 0.01*** 0.06 -0.06 0.01 0.08*** 0.05 -0.03 0.01 Finland 0.01 -0.02 -0.24* 0.07 0.09 -0.05 -0.01 0.13 France 0.01 0.02 -0.09** 0.01 0.03 -0.01 -0.03 0.03 Germany 0.01 0.01 -0.12** 0.01 0.01 -0.01 -0.03 0.02 Greece 0.01 -0.01 -0.07 -0.11 -0.05 0.02 0.06 0.01 Hong Kong 0.00 0.13*** 0.06 0.07 0.00 -0.07 -0.06 0.03 Ireland 0.01 -0.03 -0.06 -0.06 0.01 0.03 -0.04 -0.01 Italy 0.01 0.04 -0.13* 0.00 0.00 -0.02 0.02 0.02 Japan 0.00 0.09*** 0.06 0.03 0.08 -0.03 -0.06 0.02 Netherlands 0.01 0.03 -0.07 -0.04 0.03 -0.04 -0.03 -0.01 New Zealand 0.00 0.04 -0.01 -0.04 0.00 -0.03 0.02 0.05 Norway 0.01 0.16* 0.03 0.01 0.01 0.01 0.00 0.01 Portugal 0.01 -0.08 -0.07 -0.08 0.02 0.01 0.03 0.05 Singapore 0.00 0.09 0.06 0.08 0.01 -0.05 -0.01 0.07 Spain 0.01*** -0.03 -0.09*** -0.02 0.00 -0.02 -0.02 -0.02 Sweden 0.01 0.11*** -0.10 0.06 0.07 -0.01 -0.03 0.05 Switzerland 0.01*** -0.04 -0.06 -0.05 0.02 -0.02 -0.04 0.03 United Kingdom 0.01*** 0.02 -0.08** -0.01 0.00 -0.01 -0.04 0.00 United States 0.01** 0.01 -0.08* 0.01 0.00 -0.04 -0.07** 0.00 Mean developed 0.01 0.03 -0.05 -0.01 0.02 -0.02 -0.02 0.03 Argentina 0.00 0.25* 0.03 0.01 -0.10 0.08 0.00 -0.03 Brazil 0.01 0.07 -0.03 0.11 0.03 -0.01 -0.10 0.11 Chile 0.00 0.1*** -0.02 0.06 0.00 0.01 -0.07 0.01 China -0.01 0.17*** 0.04 0.07 0.05 -0.1 0.06 -0.02 Colombia 0.01 -0.01 0.05 0.06 0.02 0.01 0.00 -0.07 Czech Republic 0.01 0.02 0.04 0.10 0.09 0.04 0.00 0.00 Egypt 0.01 0.16* -0.02 -0.02 0.15** 0.18* 0.09 0.06 Hungary 0.02 -0.04 0.00 0.04 0.06 0.05 0.02 -0.03 India 0.00 0.17* -0.04 -0.03 0.01 0.05 0.10 0.12*** Indonesia -0.01 0.14 0.07 0.21*** 0.17 0.03 -0.19 0.01 Israel 0.01 -0.03 -0.04 0.00 0.09 0.02 -0.04 0.1** Jordan 0.01 -0.01 -0.04 -0.01 0.03 0.05 0.02 0.03 Korea 0.01 0.18 -0.15 -0.10 -0.01 0.00 -0.12 -0.06 Malaysia -0.01 0.07 0.08 0.13 0.06 0.07 0.11 0.05 Mexico 0.01 0.18* -0.01 0.06 -0.04 -0.06 -0.04 -0.02 Morocco 0.01*** 0.01 -0.05 -0.11* -0.06 0.08*** -0.02 0.02 Pakistan 0.00 0.09 -0.04 0.13 -0.02 0.10 0.03 0.16*** Peru 0.01 0.09 0.05 0.04 -0.08 -0.03 0.00 0.01 Philippines -0.01 0.06 -0.01 -0.02 -0.06 -0.02 -0.07 0.06 Poland 0.00 0.00 0.02 -0.04 0.06 0.02 0.03 0.06 Russia 0.02 0.03 -0.02 0.31** 0.09 -0.05 -0.15 0.15 South Africa 0.00 0.15** 0.06 -0.01 0.03 -0.03 -0.03 0.12 Sri Lanka 0.00 -0.04 0.02 -0.05 -0.03 0.08 -0.05 0.03 Taiwan 0.00 0.12 -0.07 0.02 -0.07 -0.09 0.03 0.01 Thailand -0.01 0.25** -0.11 0.07 -0.03 0.04 -0.12 0.03 Turkey 0.01 0.04 -0.19 0.08 0.16 0.02 0.11 -0.04 Venezuela 0.00 0.16 0.08 0.02 0.12 -0.14 0.01 0.02 World MSCI 0.00 0.02 -0.06*** 0.01 0.01 -0.02 -0.05*** 0.01 Mean developing 0.00 0.09 -0.01 0.04 0.03 0.02 -0.01 0.03

Mean all countries 0.00 0.06 -0.03 0.02 0.02 0.00 -0.02 0.03 Notes: Estimated using heteroscedasticity consistent Ordinary Least Squared regression. Total number of time series observations after adjustment for the time lags= 128 (months). Mean is average of all countries’ coefficients.

3.1.3 Other factors

Table 6: Overview of explanatory variables

Symbol Name factor Definition Source Computations

Rf US$ 3 month _{Treasury bill} Monthly return on 3 months U.S. T-_Bill Datastream, Series Code: _{“FRTBS3M”} Natural log of (1+(T-bill series/12)) to _{convert from annual to monthly return}

Ri

Return on MSCI country equity index return

Monthly total return on MSCI

country i price index in US$ Datastream7

Natural log of (MSCI country i index

it / MSCI country i indext-1)

WORLD Total excess return of _{MSCI world index}

Monthly excess return of MSCI world price index over US$ 3 month Treasury bill

Datastream, Series Code:

“MSWRLD$(PI)” Natural log of (MSCI world indext / MSCI world indext-1) -Rft

OIL Oil price

Relative price change per month in Crude Oil-Brent Dated FOB in U$ per barrel

Datastream, Series Code:

“OILBRNP(P)” Natural log of (Price of Crude Oilt / Price of Crudet-1)

INFL World inflation

Relative monthly change in IMF seasonally adjusted world Consumer Price Index

International Financial Statistics8_: Series Code:“00164...ZF...”

1) Adjusted for seasonalities. 2) Natural log of (World CPIt / World CPIt-1)

INDPR Industrial production Relative monthly change in seasonally adjusted industrial production index in OECD countries

OECD MAIN ECONOMIC INDICATORS10, Series Code: “OCNOCIPDG”

Natural log of (industrial production indext / of industrial production indext)

EXR Exchange Rate Risk

Relative monthly change in exchange rate index of trade weighted basket of currencies versus US$

Federal Reserve Board of StLouis (http://research.stlouisfed.org/fred2/) , Source ID: “ TWEXMMTH”

Natural log of (Exchange rate indext / Exchange rate indext-1)

TIMEHOR Time Horizon Risk

Spread between 10 year US Gov. Bond and return on 3 months U.S. T-bill

Datastream, Series Code: 10Y US Gov. Bond: “FRTCM10” and 3M US T-Bill: “FRTBS3M”

(10Y US Gov. Bond – 3Months US T-Bill)-period (1995-2006) average spread

CONF Confidence Risk

Spread between return on U.S. corporate bonds with Moody’s Baa rating minus 10 year US Gov. Bond

Datastream, Series Code: 10Y US Gov. Bond: “FRTCM10” and Baa rated corp. bonds: “FRCBBAA”

(Return on Baa rated corp. bonds - 10Year US Gov. Bond) - period (1995-2006) average spread

Table 7 provides an overview of the statistical descriptives of the included independent variables. Accompanying time-series graphs of these variables are presented in appendix 3. Despite the use of log first differences, only for the oil price and industrial production variables, the hypothesis of normal distribution is accepted on a 5% significance level. Especially the distribution of the inflation displays a high degree of skewness and kurtosis. Removing the outliers in the inflation variable from early 1995 and half 1998 does reduce the degree of skewness and kurtosis to a level that the normality hypothesis is accepted. However, to avoid artificially improvement of the characteristics of the model, the data is not adjusted.

Table 7: Statistical descriptives of independent variables

BRENT WORLD INFL EXR INDPR TIMEHOR CONF

Mean 1.0% 0.2% 0.4% 0.0% 0.2% 0.0% 0.0% Median 1.3% 0.6% 0.4% 0.3% 0.2% -0.2% -0.1% Maximum 33.6% 8.2% 2.1% 3.2% 1.3% 2.1% 1.6% Minimum -34.8% -14.9% 0.0% -4.8% -1.1% -2.1% -0.8% Std. Dev. 11.2% 4.1% 0.3% 1.6% 0.4% 1.1% 0.6% Skewness -0.10 -0.79 2.41 -0.63 -0.32 0.37 0.66 Kurtosis 3.47 4.04 13.44 3.03 2.95 2.19 2.52 Jarque-Bera 1.46 19.82 738.68 8.95 2.32 6.67 11.02 Probability 0.48 0.00* 0.00* 0.01** 0.31 0.04** 0.00*

Notes: Period: January 1995 until February 2006. Frequency: Monthly. BRENT: Change in Brent oil price, WORLD: Excess return on MSCI world index, INFL: Change in IMF world Consumer Price Index, EXR: Monthly change in exchange rate index of trade weighted basket of currencies versus US$, INDPR: Change in industrial production index of OECD countries, TIMEHOR: Time Horizon risk; spread between 10 year US Government bonds and 3 months US Treasury Bill, CONF: Confidence risk; Spread between return on US ‘Baa rated’ bonds and 10 year U.S. Government bonds.

* = significant at 1% level, ** = significant at 5% level, *** = significant at 10% level.

Table 8: Bivariate correlation matrix

BRENT BRENT(-1) WORLD INFL EXR INDPR TIMEHOR CONF

BRENT 1.00 BRENT(-1) -0.14 1.00 WORLD 0.07 -0.17 1.00 INFL -0.02 0.13 0.09 1.00 EXR -0.07 -0.01 -0.08 -0.14 1.00 INDPR 0.13 0.08 0.03 0.02 0.02 1.00 TIMEHOR 0.11 0.06 0.04 -0.26 -0.26 0.06 1.00 CONF -0.02 -0.01 -0.20 -0.39 -0.19 -0.26 0.36 1.00

Notes: Period: January 1995 until February 2006. Frequency: Monthly. BRENT: Change in Brent oil price, BRENT(-1): One month lagged change in Brent oil price, WORLD: Excess return on MSCI world index, INFL: Change in IMF world Consumer Price Index, EXR: Monthly change in exchange rate index of trade weighted basket of currencies versus US$,

INDPR: Change in industrial production index of OECD countries, TIMEHOR: Time Horizon risk; spread between 10 year

US Government bonds and 3 months US Treasury Bill, CONF: Confidence risk; Spread between return on US ‘Baa rated’ bonds and 10 year U.S. Government bonds.

t i J j j ij i i F r _, 1

ε

β

α

∑

BRENT-1 WORLD INFL EXR INDPR TIMEHOR CONF 1.25 1.52 1.08 1.14 1.33

Variance Inflation Factor (VIF)

1.24 1.16

Notes: BRENT(-1): One month lagged change in Brent oil price, WORLD: Excess return on MSCI world index, INFL: Change in IMF world Consumer Price Index, EXR: Monthly change in exchange rate index of trade weighted basket of currencies versus US$, INDPR: Change in industrial production index of OECD countries, TIMEHOR: Time Horizon risk; spread between 10 year US Government bonds and 3 months US Treasury Bill, CONF: Confidence risk; Spread between return on US ‘Baa rated’ bonds and 10 year U.S. Government bonds.

3.2 Methodology

3.2.1 Arbitrage Pricing Theory

The APT follows a two-step process. In the first step, the ex post sensitivities of country indices to the relevant variables are examined using the multifactor time-series regressions as depicted in equation (7).

(7)

where ri (i = 1,…,50) denotes the MSCI country index returns. αi is the intercept term. βij is the ex post

sensitivity of index i to variable Fj (j = 1,…,J). εi is the unique risk component of stock i. A stepwise

procedure is used in selecting the independent variables of equation 7. First, only the excess return on the MSCI World index is inserted, as previous research provides evidence of its explanatory power. Moreover, cross sectional regressions on country sensitivities to the MSCI World index helps to determine the degree of capital market integration.

In the second set of time series regressions, the excess return on the World MSCI index and changes in the Brent oil price are taken into consideration. In line with Driesprong, Jacobsen and Maat (2005) oil price changes lead country equity index returns by one month. Therefore, the Brent oil price variable is one-month lagged.

In the third round of time series regressions, in line with Chen, Roll and Ross (1986) and Ferson and Harvey (1994), macroeconomic variables are included in addition to the World MSCI index and the Brent oil price. The macroeconomic variables included are investor confidence risk, time horizon risk, and changes in worldwide inflation, industrial production and exchange rates.

8 _{The determination of the variance inflation factors is independent from which dependent variable (MSCI}

2 1 2 1 2 1 2 − −

+

+

=

α

α

ε

βσ

σ

ε

β

α

∑

In the second step, cross-sectional regressions are performed on the average MSCI country index returns to investigate whether oil price risk is priced in the global equity market.

Explanatory variables are the variable sensitivities (measured by beta), that are estimated in the first step using time series regressions. In the absence of risk-less arbitrage opportunities, The APT that arises is:

(8)

where

r

(j = 1,…,J) is the ex ante extra additional required return for the sensitivity of MSCI country index i to a one-unit increase in βij.. If λj is significant, the risk factor to which country index i is exposed is

priced.

3.2.2 Other issues

One of the assumptions of an ordinary least square (OLS) regression is that the residual term has a standard normal distribution (ut ~ N(0,σ2)). By construction, the error term always has a mean of zero.

To test whether the error term is normally distributed, the Bera-Jarque test is employed. Another assumption is that the error terms are uncorrelated with one another over time, if this is not the case, it is stated that they are ‘autocorrelated’. To test for autocorrelation, the Durbin-Watson test is employed. Moreover, it is assumed that the variance of the error term is constant, this characteristic is known as homoscedasticity. If this assumption is not met (the error term does not have a constant variance), it is stated that they are ‘heteroscedastic’. Heteroscedasticity can be demonstrated using White’s general test for heteroscedasticity. In case non-normality and/ or heteroscedasticity is found in the error term of the regressions, the use of OLS regression is not to be appropriate. A way to cope with these characteristics in the data is to use the generalized autoregressive conditional heteroscedasticity (GARCH) model. Most popular is GARCH(1,1) as it is sufficient to capture the volatility clustering in the data. If OLS is found to be inappropriate in our study, we adjust the time series regressions to fit the GARCH (1,1) specification:

(9) with:

(10) where equation (9) is written as a function of exogenous variables with an error term. σit2 is known as

the conditional variance since it is a one-period ahead estimate for the variance calculated based on any past information thought relevant. The conditional variance specified in equation (10) consists of

three terms: a constant, a GARCH term which specifies the equation for the standard error and an ARCH term which specifies the equation for the conditional variance as dependent on the value of the previous period error term.

3.2.3 Robustness tests

To test the robustness of the main results, the time series are divided between developed and developing countries. This separation is also used to present separate cross sectional regressions. This separation can bring interesting results as the degree of capital market integration and oil usage differs between developed and developing countries.

As time series of MSCI country indices of developed countries are available since 1988 (and for some even longer), the relation between Brent oil price changes and 23 developed countries’ MSCI country index are also investigated for the time period: January 1988 until February 2006. Herewith, robustness of the relation over a longer time period is investigated.

Figure 6 displays the development of the Brent oil price in comparison to average MSCI country indices. Notable is the recent co-movement between the oil price and the average of the MSCI country indices. This could result in differences in outcome between the relation found for developed countries, for the period 1995-2006 and the period 1988-2006.

Figure 6: Development of Oil price in comparison to average of MSCI country indices

0 50 100 150 200 250 300 350 400 450 500 '88 '89 '90 '91 '92 '93 '94 '95 '96 '97 '98 '99 '00 '01 '02 '03 '04 '05 '06 Year Ave ra ge M S C I count ry i nd ex 0 10 20 30 40 50 60 70 80 90 B re nt O il pr ic e ( U S $/B ar re l)

Average of 23 country MSCI indices (1/1/1981=100) Average of 50 country MSCI indices (1/1/1995=100) Brent oil price

Notes: Period: January 1995 until February 2006. Frequency: Monthly. Countries included are all 50 countries (both developed and developing) and 23 countries (only developed), listed in appendix 1