The link between crude oil spot prices and stock indices of developed and emerging markets

(1)

The link between crude oil spot prices and stock indices of

developed and emerging markets

By Jan Willem Veeningen Student number: 2073455

First supervisor: dr. J.H. (Henk) von Eije Faculty of Economics and Business

University of Groningen Study program: MSc Finance Date: January 11th, 2018

Abstract

(2)

1 Introduction

When all information would be included in prices, markets would be efficient. Over the years, information technology has developed with the introduction of internet. With internet being introduced, a huge set of information has become available to the public. One could argue that more information should be included in prices because of faster and more voluminous information diffusion. To this end, this paper looks at changes over the years in terms of informational links between oil prices and stock indices.

A body of literature is devoted to establishing the link between oil price and stock indices. This link could provide a possibility to predict future returns, which could lead to profitable investment opportunities. However, within this body of literature, there is no consensus on whether the sign of the link between the two is positive of negative. A paper by Driesprong et al. (2008) pursued to see how well the average monthly oil price can predict the stock index of the next month. In these tests they varied with the set of days that was used to calculate the monthly oil price to create a lag between oil price and stock index, where they found that adding a lag to the average monthly oil price provided the strongest link. Keeping this in mind and looking at developing information technologies, one could wonder whether with improving information diffusion the characteristics of the link between oil prices and stock indices have changed. At the same time a financial theory related question could be answered: have markets become more efficient over time?

The goal of this paper is to study the link between crude oil price returns and stock index return, another goal is to find whether the sign of the link between oil prices and stock indices is positive or negative and whether it has changed over time. Moreover, tests are done to see whether weak form efficiency is present in terms of information diffusion of oil prices into stock indices. It contrasts to the paper of Driesprong et al. in that daily prices are used as data input instead of average monthly prices.

The tests of Driesprong et al. are first reproduced and comparable results were found for the period that was also used in that paper (oil prices until 2003). Here, tests were conducted over a sample including more recent years and results show that this link indeed has decreased in more recent years. This result hints towards increased information diffusion and to study this phenomenon further, subsequent tests were done with daily prices of crude oil and stock indices. In this paper weak form efficiency is assumed to exist if daily changes in crude oil prices are influencing market indices at the same day. However, when a lag is discovered in the information diffusion of this link, the conclusion can be drawn that the weak form efficiency is not established. When markets are not efficient, investors can use information to predict prices and make profits by using predictive investment strategies. Nevertheless, when efficiency increases over time, predictability of prices decreases, and investors are less able to use information to predict the market.

As it turns out, a serious number of significant results was found with lagging information of oil prices into stock indices. Because the information of oil lags into the stock indices, in this study the term ’lag’ will be used to describe how long it takes for the information of crude oil prices to be incorporated into stock indices.

Looking from a more meta level at the results, one can see that all tests show that the link has gone from negative in the period of 1990-1992 to positive from 2005 onwards. However, when comparing developed markets to emerging markets, the first period of significant negative results in not present in emerging markets, this may be caused by financial market mechanisms not being as developed yet.

(3)

1 Introduction

indices is observed over time. In general, many studies take a longer sample period of 20, 30 years or longer to analyze the general relationship of prices of financial products and stock indices. Regressions are done with sample periods of 1 year, so that one can accurately see when changes occur.

A common belief in financial markets in the years before 2000, was that the oil price and the stock index had a negative relation. This means that if the oil price would go up, the stock index would go down and vice-versa. Nowadays this belief has flipped, and it is believed that oil prices have a positive relation with the stock index. To this end, tests were done to look at when the link between oil prices and stock indices changed from negative to positive. The results of this test show that in the period 2002-2005 the link became significantly positive, whereas the link in the period 1990-1993 was significantly negative. The period between 1993 and 2002 shows in general no effects and can be considered as a transition period for this change.

(4)

2 Background

The theme which this paper studies is the speed of incorporation of oil prices in stock indices. Next to that, the sign of the link is studied. When oil prices rise, the logistics behind products also rise in costs. That is why these logistic firms may pursue a strategy that includes shorting stocks of corporate transportation companies when oil prices are high and going long when oil prices are low. Also consumers are affected by higher oil prices: it makes going to work or stores more expensive. This can lead to companies having less revenues and increased expenses. These beliefs can have effect on how investors allocate their money. On the other hand, oil companies are major factors that influence stock indices. That is why when oil prices increase, oil companies’ stock price increase, which in turn increase stock indices. For these reasons it is interesting to look into the sign of the relationship between crude oil prices and stock indices and whether it has changed over time. In this section literature on the link between commodities and stock indices will be elaborated.

2.1 Price informativeness and commodity markets

The paper by Sockin and Xiong (2013) studies the role of futures market speculation. Specula-tors believe that they can predict future market movements and for a market to be predictable. Questions asked in the paper of Sockin and Xiong (2013) are: How do commodity markets aggre-gate information about the global economy? How do informational frictions affect commodity prices and demand? One example is a world where suppliers are subject to supply shocks but there are also informational frictions. What follows is that goods producers partially attribute a lower commodity price to weaker global economy, which leads them to reduce their commodity demand. Thus, the negative price impact of a supply shock is amplified. In literature it is common to ignore informational frictions and assume that agents directly observe both demand and supply shocks. But these informational frictions might be present after all and cause the weak-form efficiency to be invalidated. In the end, their analysis systematically shows that both spot and futures prices of key industrial commodities can serve as price signals for the strength of the global economy. This signal may also have predicting power to indicate the future move-ment of the economy of a country, a link between commodity prices and stock indices may thus be present in financial markets.

2.2 Predictability in oil markets

In literature the links between stock indices and commodity prices have been studied bountifully. One of the reasons is that investors may use this information to predict the market (on a long term). Predicting prices can be valuable if one is correct in predicting, but can also incur great losses; particularly if the established historical relations change.

An example of a study establishing the predictive accuracy of crude oil futures prices is the study of Abosedra and Baghestani (2004), where month ahead crude oil futures prices are evaluated. In their paper a benchmark is used to compare the future oil prices, the benchmark is constructed to be naive, such that it is not a good tool to forecast oil prices. They find that using the futures price cannot outperform the naive benchmark. In the paper of Sørensen (2009), the under-reaction to information is studied in oil prices and stock returns. Sørensen finds that oil price changes that are caused by exogenous events can predict stock returns. But he also questions whether the source of predictability can be used in a investment strategy.

(5)

2 Background

is authored by Driesprong et al. (2008). This study takes the average monthly crude oil price return of month t-1 and uses it in a regression for the average monthly return on stock indices of month t with data from 1973 to 2003. During these tests, a lag is added in the set of days that is used to calculate the monthly return on crude oil prices. One day is added of month t-2 and the last day of month t-1 is removed. The results show that adding a lag of 6 days provides the strongest predictability. This test is reproduced in this paper and extended to check whether this lag has become shorter over the years.

Ferrer Lape˜na et al. (2017) study the time-varying causality between crude oil and stock markets for six developed markets in countries that are oil importing. The results show that there are time-varying causal relations between oil price changes and stock returns. Moreover, Ferrer et al. find that in times of financial turmoil, the causal interactions seem to intensify in period of turbulence, such as terrorist attacks, the global financial crisis of 2008 and the European debt crisis of 2009-2011.

Similarly, Boldanov et al. (2016) study the correlation between oil and stock market volatilities and find that the correlation between the oil and stock market volatilities changes over time to both positive and negative values. These papers prove to be an indication that the informational link between oil prices and stock markets has changed over time.

The hypothesis to be tested that is concerned with predictability: The predictability of stock indices using crude oil spot prices on a monthly basis has been reduced over time.

2.3 Efficiency in markets

Weak-form efficiency is the theory that public information is directly reflected in the price. The most seminal paper that coins the theoretical and empirical literature on the efficient markets model is the paper by Fama (1970). Markets are called efficient where prices always fully reflect available information. Under this efficiency there are three distinctions: weak form, where the information set is only historical prices. Semi-strong form where prices adjust efficiently to other information that obviously is publicly available. The strongest form of efficiency is where prices always fully reflect all available information.

(6)

2 Background

2.4 Emerging versus developed markets

Studies done by Aastveit et al. (2015) and Long et al. (2014) show that research done in financial markets often make the distinction between emerging markets and developing markets. The results in these studies show that a difference is present between the two. Aastveit et al. (2015) show that demand shocks in oil are far more important in emerging markets than in developed markets. Long et al. (2014) find that emerging markets show much higher volatility in currency betas when comparing it to currency betas in developed markets and that especially in emerging markets, the market betas tend to be lower than currency betas. A paper by Hull and McGroarty (2014) studies the relationship between price efficiency and market development in emerging markets. Hull finds that more advanced markets exhibit the least memory persistence and appear the most efficient. Because these studies show that there is a difference between emerging and developed markets and the before mentioned link of efficiency and degree of development, the difference between emerging and developed markets is studied. The (sub)hypothesis to be tested is: Emerging markets show lagging development financial mechanisms compared to developed markets.

2.5 Positive of negative sign of the relation

In literature the link between stock indices and oil prices has been discussed extensively, but there is no real consensus about the sign of the link. In order to be able to predict, the sign of the link must be at least universal. This is, however, not fully the case. In table 1, an overview is given of authors that establish a negative, positive or no link at all in their study.

Table 1: Classification of authors on their conclusion on the relationship of oil prices and stock indices

Negative link No link Positive link

Author Sample Author Sample Author Sample

Sadorsky (1999) 1947-1996 Wei (2003) 1974-1976 Basher (2006) 1992-2005 Papapetrou (2001) 1989-1999 Cong (2008) 1996-2007

Miller (2009) 1971-2008 Huang (1996) 1983-1990 Nandha (2008) 1983-2005 Chen (1986) 1953-1983

Kling (1985) n/a Park (2008) 1986-2005

Jones (1996) 1950-1991

Negative: A set of researchers find a negative link between returns on crude oil prices and stock indices. Results in the study of Sadorsky (1999) show that positive shocks to oil prices reduce real stock returns. Papapetrou (2001) finds that for Greece positive oil price shock depress real stock returns. Similarly, in the study of Miller and Ratti (2009) findings lead to the conclusion that stock market indices respond negatively to increase in the oil price in the long run. In the study of Nandha and Faff (2008) the conclusion is that oil price rises have a negative impact on equity returns for all sectors except mining, and oil and gas industries. Kling (1985) argues that increases in crude oil prices may have been followed months later, first by declines in stock prices [...]. Lastly, Jones and Kaul (1996) report that there is a stable negative relationship between oil prices changes and aggregate stock returns.

(7)

2 Background

After constructing a general-equilibrium model, he finds a link that is negligible. Cong et al. (2008) study interactive relationships between oil price shocks and Chinese stock market. Also, Cong et al. find no statistically significant impact on the real stock returns of most Chinese stock market indices. Another approach is taken by Huang et al. (1996), where a vector autoregressive approach is used to examine the relation between oil futures and stock returns. Huang finds that oil futures returns are not correlated with stock market returns. Also, Chen et al. (1986) examine the impact of an index of oil price changes on asset prices and found no overall effect. Park and Ratti (2008) find mixed results where eight out of thirteen European countries show a significant increase in short-term interest rate when there are increases in oil price, but no convincing evidence for the U.S. and for Norway.

Positive: Basher and Sadorsky (2006) study the impact of oil price changes on a large set of emerging stock market returns. For daily and monthly data, findings point towards oil price increases having a positive impact on excess stock market returns in emerging markets. For weekly and monthly data, oil price decreases seem to have a positive and significant impact on emerging market returns.

As can be seen in this section, the consensus does not exist. The papers discussed in this section are from varying time periods but overall there are advocates for each stand. From the table it can be seen that in general the negative link is established by earlier papers and that a positive link or no link is described in later papers. That is why the following hypothesis is founded:

The sign of the link between crude oil spot prices and stock indices has changed from negative to positive.

2.6 Predictability versus descriptive analysis

In this section papers study the predictability of stock indices. Investors are interested in any information that can help predict future stock prices, such that they can make a profit. Other papers study the efficiency of stock indices and commodity markets. In this study, the main goal is to establish whether a weak form efficiency can be found between oil markets and stock indices.

2.7 Hypotheses

To conclude the background section, an overview of the hypotheses is given.

• The predictability of stock indices using crude oil spot prices on a monthly basis has been reduced over time.

(8)

in-3

Data

3.1 Data collection

For this research data are required about the daily prices of crude oil and stock indices of developed markets. Data for stock indices and crude oil prices were obtained from Datastream on October 13th, 2017.

The developed markets that are considered for this study are labeled as ’developed’ by Morgan Stanley Capital International, which was founded in 1969 and has provided equity, fixed income, hedge fund stock market indexes, and equity portfolio analysis tools. The smallest time slot in which these data were available was daily, thus daily is the time slot that is taken in the regression tests. To reproduce the paper of Driesprong, these data were converted to monthly data by taking averages of each month. The countries with developed markets that are included in the regressions are displayed in table 2. 23 stock index returns of developed market are considered: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Hong Kong, Ireland, Israel, Italy, Japan, New Zealand, Netherlands, Norway, Portugal, Singapore, Spain, Sweden, Switzerland, the United Kingdom and the United States of America. Next to the developed markets, emerging markets were included in the regressions. Data were less available for emerging markets, but still daily date could be retrieved from 1992 onwards.

Because the period of 1992 seemed relevant for this research, the emerging makets that were not available for 1992 onwards from Datastream were removed from the analysis. The 24 emerging market countries considered are: Brazil, Chile, China, Columbia, Czech Republic, Egypt, Greece, Hungary, India, Indonesia, Korea, Malaysia, Mexico, Pakistan, Peru, Philippines, Poland, Qatar, Russia, South Africa, Taiwan, Thailand, Turkey and United Arab Emirates and can be found in table 3.

3.2 Descriptive statistics

(9)

3 Data

Table 2: Descriptive statistics for data of stock index returns of the developed markets used in this study.

Country Starting date Mean Std. dev. Kurtosis Skewness # obs.

Australia 1-1-1988 1.000311 0.01386 21.5062 -1.1513 8045 Austria 1-1-1988 1.000252 0.01545 11.8227 -0.0229 8045 Belgium 1-1-1988 1.000288 0.01308 11.5881 -0.1761 8045 Canada 1-1-1988 1.000312 0.01219 14.5448 -0.5918 8045 Denmark 1-1-1988 1.000497 0.01296 9.9623 -0.2262 8045 Finland 1-1-1988 1.000451 0.01198 9.8903 -0.1032 8045 France 1-1-1988 1.000311 0.01409 9.8989 0.0125 8045 Germany 1-1-1988 1.000303 0.01480 9.0799 -0.1175 8045 Hong Kong 1-1-1988 1.000405 0.01539 41.2320 -1.4950 8045 Ireland 1-1-1988 1.002150 0.01557 12.5922 -0.3784 7761 Israel 1-1-1993 1.000192 0.01385 7.6328 -0.2150 6456 Italy 1-1-1988 1.000141 0.01566 8.8874 -0.0609 8045 Japan 1-1-1988 1.000161 0.01433 10.1555 0.07685 8045 Netherlands 1-1-1988 1.000339 0.01320 11.1256 -0.0701 8045 New Zealand 1-1-1992 1.000142 0.01038 14.5145 -0.3020 6739 Norway 1-1-1988 1.000361 0.01697 14.1348 -0.4301 8045 Portugal 1-1-1992 1.000123 0.01141 9.8390 -0.1120 6739 Singapore 1-1-1988 1.000302 0.01350 23.3277 -0.5964 8045 Spain 1-1-1988 1.000323 0.01538 11.3885 0.0654 8045 Sweden 1-1-1988 1.000462 0.01675 8.7581 0.1291 8045 Switzerland 1-1-1988 1.000357 0.01159 9.3758 -0.1853 8045 UK 1-1-1988 1.000253 0.01241 13.7197 -0.2167 8045 USA 1-1-1988 1.000352 0.01122 25.2839 -0.8430 8045

Notes: Descriptive statistics on MSCI stock indices for 23 developed markets. Values in the table are based on daily stock index return data. Starting on varying dates and ending 9-30-2017.

In table 2 the descriptive statistics are shown for the 23 developed countries that are used in this study. The average mean of the returns lies around 1.0003 with an average standard deviation of 0.014. A perfectly normally distributed data set would exhibit a kurtosis value of 3. The data set used in this study exhibits a positive kurtosis value, which means that the distribution has heavier tails and a sharper peak than a normal distribution would have. Skewness is a measure how the tails of a distribution are distributed. The table shows that skewness of these series are not extraordinary.

(10)

3 Data

Table 3: Descriptive statistics for data of stock index returns of the emerging markets used in this study.

Country Starting date Mean Std. dev. Kurtosis Skewness # obs.

Brazil 1-1-1992 1.000591 0.02432 8.8229 0.1742 6523 Chile 1-1-1992 1.000265 0.01285 14.1410 0.1468 6523 China 1-1-1993 1.000092 0.01885 9.1004 0.2715 6261 Columbia 1-1-1993 1.000395 0.01547 13.0740 0.1306 6261 Czech Republic 1-1-1995 1.000292 0.01674 14.1588 0.1473 5740 Egypt 1-1-1995 1.000448 0.01702 29.7485 -1.3512 5740 Greece 1-1-1992 0.999895 0.02259 10.8859 -0.0702 6523 Hungary 1-1-1995 1.000542 0.02173 11.9455 0.1369 5740 India 1-1-1993 1.000377 0.01658 11.0466 0.1522 6261 Indonesia 1-1-1992 1.000584 0.01810 13.7517 0.1755 6523 Korea 1-1-1992 1.000418 0.01813 8.3714 0.2456 6523 Malaysia 1-1-1992 1.000221 0.01611 67.8126 0.8843 6523 Mexico 1-1-1992 1.000617 0.01427 8.9683 0.2478 6523 Pakistan 1-1-1993 1.000212 0.01721 10.1097 -0.2569 6261 Peru 1-1-1993 1.000563 0.01785 9.6436 0.0020 6261 Philippines 1-1-1992 1.000327 0.01456 14.2703 0.4803 6523 Poland 1-1-1993 1.000502 0.02175 7.4036 0.0211 6261 Qatar 6-1-2005 1.000126 0.01477 14.9245 -0.3008 3023 Russia 1-1-1995 1.000733 0.02896 14.2622 0.1678 5740 South Africa 1-1-1993 1.000392 0.01733 8.1259 -0.1441 6261 Taiwan 1-1-1992 1.000214 0.01511 6.0104 0.1173 6523 Thailand 1-1-1992 1.000275 0.01787 17.4472 0.9634 6523 Turkey 1-1-1992 1.000532 0.02887 9.7088 0.1521 6523 UAE 6-1-2005 1.000002 0.01888 15.4275 -0.2174 3023

Notes: Descriptive statistics on MSCI stock indices for 24 emerging markets. Values in the table are based on daily stock index return data. Starting on various dates and ending 9-30-2017.

Table 3 shows the descriptive statistics for the emerging markets. As can be seen the mean and kurtosis are similar to the ones of the developed countries. Similar to the series of the developed markets, the series of the emerging markets show no extraordinary values for skewness.

To describe the crude oils that are studied in this paper, table 4 contains the characteristics of series with Dubai crude oil price returns. In this study, crude oil prices are assumed to be exogenous, thus they are assumed not be predictable.

Table 4: Characteristics of oil price changes Dubai Mean 1.005 Maximum 1.218 Minimum 0.676 Std. dev. 0.024 # obs. 8045

Notes: Characteristics of the daily crude oil price return. The time series is extracted with a starting date of 1-12-1987

(11)

3 Data

limited data availability, only daily data could be retrieved and not hourly or shorter. The choice was made to take Dubai oil, because this crude oil benchmark is used in literature frequently (Sørensen (2009); Driesprong et al. (2008); Agusman and Deriantino (2008); Le and Chang (2015)). In these articles WTI (Western Texas Intermediate) is also used frequently. That is why WTI crude oil is also used to check whether this crude oil price delivers comparable results. To make the data of the oil prices and stock indices comparable, returns are taken instead of prices according to the following equation. Next to that, the natural logarithm is taken, to reduce skewness in the return series and to reduce outliers. In equation 3.1 the equation is shown that is used to calculate the returns of both oil prices and stock indices.

log( Pt Pt−1

(12)

4 Methodology

The regressions that are done here are Ordinary Least Squares regressions and the tests in this paper are done with samples of 1 year, to keep the tests uniform.

4.1 Using monthly data to perform a reproduction

In this test a part of the paper of Driesprong et al. (2008) is reproduced to see whether similar results are retrieved. The regression equation is given in equation 4.1. Monthly averages are calculated for both the returns on crude oil spot prices and the stock index. The independent variable for the regression is the return on the stock index of country i of month t. However, the set of days that is being used to calculate the average monthly return for the oil price varies among tests. In the set of days for the monthly oil price return one day is added of the month before (month t-2) and the last day of month t-1 is removed from the set. The tests are being done are to test for which lag the link between the independent and dependent variable is the strongest. The lag (k) is varied from 0 to 10 days. The regressions are done separately for 18 developed countries and 13 emerging countries. Also for every lag a separate regression was performed.

Sit= α + βiOit−k + it (4.1)

where

Sit represents monthly average return on the stock index of country i of month t.

βi represents the to be estimated coefficient.

Oit−krepresents monthly average crude oil price i returns of month t-1 with a lag k between

0 and 10.

it represents the error term

4.2 Using daily data to increase informativeness of tests

To be able to say more about what happens on a daily basis, daily data are used to see what the results state on smaller time intervals. Equation 4.2 defines the Ordinary Least Squares regression that is done to establish the link between daily crude oil spot price returns and daily stock index returns. In this study there are some variations applied to the equation which will be described in subsequent sections.

Sit= α + βitOit+ it (4.2)

where

Sit represents daily return on the stock index of country i on day t.

βit represents the coefficient to be estimated.

Oit represents daily return on the crude oil spot price on day t.

(13)

4 Methodology

4.2.1 Testing the weak form efficiency

For efficiency to exist in markets, information must be directly incorporated in market prices once it is publicly available. For this to be tested, one would need all the information that should be incorporated in the price and test whether this information is incorporated from the moment it is available. In this study, the tests done are to see whether there is significance on t=0, but also around t=0. For efficiency to exist, there must be no significant results other than on t=0. To this end, regressions are done for both developed and emerging markets with Dubai and WTI oil prices. A lag of t=-2 until t=0 is tested to see whether significant results exist outside t=0. In the test only historical prices are used, thus the weak form of efficiency is tested.

4.2.2 Testing the sign change of the link between returns on the crude oil spot price and the stock indices

In this test a yearly sample is taken for each year to see from year to year, what kind of sign is present in the link. For each year and for each country, separate regressions are done to establish the p-values and T-statistics. In the results sections the number of significant results will be analyzed and whether these significant results have positive or negative signs.

4.2.3 Testing the difference between emerging markets and developed markets

(14)

5 Results

Tables in this section show only t-statistics. The magnitude of the coefficients are left out, because the t-statistics contain sufficient information for this study. Coefficients can be provided on request.

5.1 Reproducing Driesprong with monthly data

In this section the results of Driesprong et al. (2008) will be reproduced and extended in terms of sample size. Namely, the sample period that was used in the mentioned paper was 1973 until 2003, but here the sample period is 1988 until 2017. which makes the results more topical. The test was conducted over three sample periods of 10 years, namely: 1988-1997, 1998-2007 and 2008-2017. The set of 18 countries are categorized as ”developed” by the MSCI according to Driesprong et al. (2008), thus these countries will also be used to reproduce the study. The OLS regression establishes coefficients between the monthly average stock index returns and the monthly average oil price return with varying lags. As described in the methodology, ten lags are tested and compared. The sets of returns that are compared are the average monthly stock index returns of month t=0 and the crude oil price returns of month t=-1. This is done for periods of 10 years, so 120 months of stock index returns are compared to the prior months of crude oil price returns.

For the crude oil price returns, a lag is added from month t-2 and the most recent day of the set is deleted. This means that when a lag is added, for the set of days that is used to calculate the monthly average crude oil price is altered. To provide an example, assume that the stock index return of the month December is trying to be predicted. Then the crude oil price return of November is used. When the value of the lag is 1, this means that the most recent day of the month November is deleted and the most recent day of October is added. Similarly, when the lag is 2, the two most recent days are deleted from the set of days of November and the two most recent days of October are added.

Table 5: Results for the MSCI stock index of Germany for three sample periods of ten years and ten different lags testing for significance in regressions with Dubai crude oil price returns

Sample: 1988 to 1997 Sample: 1998 to 2007 Sample: 2008 to 2017

t t-stat t-stat t-stat

1 -2,184 -2,591 1,368 2 -2,365 -2,843* 1,332 3 -2,717* -3,055* 2,013 4 -2,509 -2,525 1,796 5 -3,715* -2,183 1,277 6 -3,558* -2,612 1,362 7 -3,327* -2,118 2,131 8 -2,779* -1,972 2,028 9 -2,615* -1,690 1,827 10 -2,423 -1,786 1,862

Taking Germany as an example with the regression equation: Sit= α + βiOit−k+ it. Results with bold font and an * are significant on a 1% level.

(15)

5 Results

for Germany. As can be seen, the table shows in the period 1988 to 1997 significant results on a 1% level for -3,-5,-6,-7,-8 and -9. For the following period, 1998 to 2007, the table shows significant (1% significance level) results at a lag of -2 and -3. In the last sample period of 2008 to 2017, there are no significant (1% significance level) results present. The results in Germany show the strongest lag of 5 days in the period of 1988-1997 and in 1998 to 2007 the strongest lag for a lag of 3 days. What can be observed is that the lag becomes shorter from the first to second sample period, and then disappears. This means that the predictability of stock indices, using crude oil price returns has decreased. This is a phenomenon that is observer in 6 of the 18 countries tested in reproducing Driesprong et al. This test that is displayed in table 5 is repeated for 18 countries on significance levels of 1%, which is the same significance level that was used in the reproduced paper. The summary of the test results for every country are displayed in table 6.

Table 6: Test results for 18 developed countries as reproduced from the study of Driesprong et al. (2008)

Sample period 1988-1997 1998-2007 2008-2017

Australia n/a n/a n/a

Austria 5 n/a n/a

Belgium n/a n/a n/a

Canada n/a n/a 10

Denmark 6 3 n/a

France n/a 3 n/a

Germany 5 3 n/a

Hong Kong n/a n/a n/a

Italy 5 2

Japan 10 n/a n/a

Netherlands 6 n/a n/a

Norway n/a n/a 10

Singapore n/a n/a 10

Spain 5 3 n/a

Sweden 5 3 n/a

Switzerland 5 n/a n/a

UK 5 3 n/a

USA n/a 6 7

# of countries with shorter lag 6 # of countries with longer lag 1

(16)

5 Results

were no significant results. This implies that the link between crude oil returns and stock index returns can be used to a lesser degree for predictability purposes. When this lag is becoming shorter, one can think in terms of weak-form efficiency, that it is decreasing. That is the reason why in the following section, daily data are taken to further study the matter.

5.2 Using daily data

In this section the results will be discussed that were retrieved from the regression with the daily crude oil price returns and daily stock index returns as input. The sample period for every regression in this section is 1 year. For developed countries the total period that is tested per year is 30 years. Table 7 shows the results for the regressions for each of the 30 years and for each of the 23 developed markets. Stock indices returns of 23 developed markets are taken as dependent variable in the regression and crude oil price returns are taken as independent variable. The two variables are taken from the same day. When the link of the same day is present, the relationship is considered to be efficient. The hypothesis that is tested here is the one where efficiency has increased over time. Contrary to the reproduction Driesprong et al. (2008), the significance level is 5%, which is used throughout the remaining part of the results section. All results are significant on a 5% significance level, unless stated otherwise. Also the list of developed and emerging markets will be updated according to MSCI categorization of 2017. This categorization includes 23 developed markets and 24 emerging markets, which will be used in the following analyses.

5.2.1 Time difference

The commodity exchanges where financial products on the crude oils in this study are traded are the Chicago Mercantile Exchange & Chicago Board of Trade (CME) and Dubai Mercantile Exchange Limited (DME). Financial products on Dubai crude oil are traded on the DME and financial products on WTI are traded on the CME. The CME and DME trade financial products on commodities almost continuously (the CME closes from 16:00 to 17:00 each day), so trading is possible at all times of the day, except on holidays and during weekends.

On the DME, trading hours are from 16:00 North American Central Standard Time (CST or GMT(-5)) until 16:00 CST the next day. Dubai is 10 hours ahead and the CME exchange operates from 17:00 to 16:00 CST. As can be seen the commodity exchange operate at almost all hours of the day. The crude oil price that is provided by Datastream is the average price per day.

(17)

5

Results

(18)

5

Results

Table 7: T-statistics of regressions of developed markets and Dubai oil prices

1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

Country t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat

Australiaa -0.13 -1.69 -1.74 -0.92 0.80 2.33* 3.32* 1.80 0.54 1.00 1.88 0.54 -1.97* -1.62 -0.15 Austria -0.22 0.18 -3.11* -6.78* -7.73* 1.83 1.32 -1.53 -0.61 0.05 -0.77 -0.23 -1.80 -1.70 -0.00 Belgium -0.20 -0.37 -3.19* -7.58* -8.21* 0.10 0.66 -0.79 -0.24 -0.70 -1.90 -1.81 -1.90 -2.38* -0.17 Canada 1.13 2.14* -1.20 -1.39 0.28 1.32 1.14 1.21 1.80 0.52 0.22 0.23 -1.04 -0.83 0.85 Denmark 0.04 -0.07 -2.52* -6.79* -6.00* 1.75 1.00 -0.46 1.12 0.50 -0.22 0.68 0.04 -0.72 0.10 Finland 0.43 1.30 -0.66 -2.48* -2.18* 1.21 1.23 -1.49 1.05 1.79 0.32 0.34 0.49 -0.43 0.02 France 0.05 0.83 -3.10* -7.93* -8.51* -0.8 -0.46 -0.83 0.81 0.22 -0.72 -0.52 -0.58 -0.98 0.67 Germany 0.16 0.93 -3.20* -6.86* -6.7* 1.37 1.09 -0.29 0.87 0.57 -0.30 -0.02 -0.16 -1.51 -0.37 Hong Konga -2.01* -3.86* -3.60* -1.05 -1.59 -1.05 0.12 0.64 -0.48 0.51 1.01 0.94 -0.23 -0.81 0.63 Ireland 0.10 0.78 -0.58 -4.81* -6.25* 1.10 1.41 -0.08 -0.50 -0.31 -0.65 -0.10 -0.70 -0.22 2.03*

Israel n/a n/a n/a n/a 1.27 1.17 1.41 2.20* 0.85 -0.04 0.30 -0.89 -1.40 -0.77 -0.77

Italy 0.12 0.29 -3.54* -7.36* -5.92* 0.85 0.30 -0.60 0.98 0.50 -0.38 -0.65 -0.25 -0.57 0.16

Japana -1.42 -4.80* -4.27* -1.68 0.90 2.26* 0.40 0.92 -0.32 -0.91 0.58 1.51 -0.01 0.84 0.40

New Zlnda n/a n/a n/a n/a n/a 0.10 1.90 2.54* -0.02 -0.37 0.90 1.98* 0.83 -2.10* -2.76*

NL 0.69 0.98 -1.05 -4.73* -6.1* 1.14 0.61 -1.2 2.34* 1.25 0.24 0.42 0.30 -0.27 1.13

Norway 0.87 1.61 -0.98 -3.87* -3.09* 1.16 0.10 -0.83 1.09 2.72* 1.85 3.14* 2.78* 1.90 2.78*

Portugal n/a n/a n/a n/a 0.71 -0.01 -1.20 -1.78 -0.91 -0.40 -0.65 -1.06 0.46 -0.40 -0.70

Singaporea -0.93 -7.06* -5.43* -0.58 1.24 0.86 0.55 0.10 -0.69 -0.43 0.29 1.05 0.84 0.05 1.08 Spain 0.38 0.28 -4.81* -9.11* -8.86* 0.09 1.80 0.46 1.36 0.60 -0.75 -0.97 -1.10 -1.12 0.20 Sweden -0.39 0.53 -4.59* -7.50* -5.52* 0.70 0.68 -0.09 3.11* 1.09 0.45 1.04 -0.05 -1.22 -0.12 Switzerland 0.59 0.44 -3.88* -7.79* -7.55* 0.76 -0.85 -2.07* 1.26 1.07 -0.20 -0.98 -0.25 -1.42 -0.81 UK -0.27 -0.4 -1.73 -4.64* -4.81* 0.58 0.74 -1.37 0.09 -0.03 -0.20 -0.70 -0.44 -0.71 0.62 USA 0.17 -0.28 -6.50* -7.16* -3.55* -0.29 -0.13 0.24 1.28 -0.4 -0.60 -0.44 -0.68 -0.58 0.16 Average -0.04 -0.41 -2.99 -5.06 -3.97 0.81 0.77 -0.14 0.64 0.38 0.03 0.15 -0.30 -0.73 0.22 # pos. 0 1 0 0 0 2 1 2 2 1 0 2 1 0 2 # neg. 1 3 13 15 15 0 0 1 0 0 0 0 1 2 1

Notes: Results for 23 developed markets with Dubai crude oil prices with regression equation: Sit = α + βitOit+ it. All reported t-values are based on heteroscedasticity consistent standard errors.

* Significant on a 5% significance level.

(19)

5

Results

(20)

5

Results

Table 7: continued

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

Country t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat

Australiaa 0.79 2.36* 3.34* 1.70 3.85* 5.25* 6.42* 6.09* 5.22* 4.14* 3.51* 3.50* 4.41* 4.92* 2.84* Austria -1.17 0.93 2.39* 3.34* 4.03* 8.12* 9.48* 11.40* 12.10* 9.32* 7.34* 3.00* 4.97* 8.51* 8.17* Belgium -0.40 -1.38 1.05 2.20* 3.08* 8.28* 9.83* 11.00* 12.40* 10.50* 7.79* 2.78* 3.18* 5.53* 4.99* Canada 1.21 1.56 5.01* 9.00* 8.80* 14.10* 16.60* 18.60* 15.00* 12.20* 10.40* 8.48* 11.30* 15.20* 15.50* Denmark -0.82 -0.32 1.97* 3.06* 3.99* 10.50* 11.40* 10.00* 11.2* 9.01* 6.48* 1.82 1.87 5.23* 6.1* Finland 0.71 -1.39 -0.01 1.52 2.77* 8.76* 9.92* 10.6* 11.9* 10.9* 8.35* 2.97* 3.21* 5.64* 5.22* France -0.01 -2.09* 0.67 3.04* 4.11* 11.10* 12.70* 13.10* 12.70* 11.00* 9.16* 3.53* 4.02* 7.31* 7.23* Germany -0.31 -1.76 -0.17 1.83 3.24* 10.30* 12.00* 13.10* 12.40* 10.90* 9.16* 2.96* 3.21* 6.41* 6.78* Hong Konga -0.69 -1.00 2.32* 0.99 1.54 3.55* 6.45* 5.83* 4.80* 4.78* 3.17* 2.31* 3.71* 3.83* -0.48 Ireland 1.21 0.40 1.40 2.81* 3.11* 6.96* 8.30* 9.40* 9.80* 9.20* 8.23* 2.33* 3.48* 6.65* 5.92* Israel -0.86 -0.10 -0.89 1.56 6.87* 7.58* 6.61* 7.32* 7.04* 3.59* 0.13 2.74* 3.70* 2.07* -0.44 Italy 0.45 -1.13 0.50 2.64* 3.71* 11.2* 12.7* 12.7* 11.4* 9.80* 8.35* 3.13* 3.96* 7.91* 8.02* Japana -2.26* -0.87 2.92* 1.91 5.27* 6.83* 7.33* 4.93* 4.54* 4.92* 1.79 2.79* 7.46* 8.33* 1.66 New Zlnda -0.25 -1.37 -0.66 0.89 1.42 4.00* 5.40* 6.22* 5.18* 4.33* 3.14* 0.78 1.88 2.9* 1.81 NL 0.23 -2.18* 0.33 2.0* 2.65* 10.4* 11.80* 12.10* 13.00* 11.00* 8.93* 2.55* 2.76* 6.25* 6.30* Norway 1.17 1.70 3.88* 5.24* 6.21* 13.10* 14.30* 14.00* 14.90* 13.20* 8.76* 6.25* 8.82* 10.40* 9.52* Portugal -0.21 -1.69 -0.57 0.08 1.29 8.42* 9.75* 9.74* 9.40* 7.53* 6.19* 2.75* 5.10* 9.33* 10.00* Singaporea -0.64 -1.87 1.91 1.71 1.53 1.88 2.43* 3.68* 3.09* 2.73* 2.88* 3.02* 4.34* 4.69* 1.94 Spain -0.41 -2.12* 0.17 1.89 3.67* 10.30* 12.10* 11.00* 10.40* 8.34* 6.77* 2.74* 4.35* 7.90* 7.13* Sweden 0.81 -1.07 0.68 2.48* 3.04* 9.17* 10.60* 11.70* 13.50* 11.20* 7.97* 2.89* 4.28* 7.68* 7.51* Switzerland -0.36 -0.93 1.51 2.30* 2.74* 9.89* 11.00* 11.10* 11.80* 9.79* 7.75* 2.06* 3.39* 6.76* 6.56* UK 0.21 -1.82 1.79 4.02* 4.43* 12.30* 14.00* 13.80* 13.00* 11.30* 10.10* 4.48* 5.96* 9.24* 8.94* USA -0.81 -3.27* -1.77 -0.1 0.63 7.23* 10.60* 15.90* 12.70* 10.40* 10.60* 5.65* 5.95* 9.07* 9.15* Average -0.11 -0.85 1.21 2.44 3.57 8.69 10.11 10.72 10.38 8.73 6.83 3.29 4.59 7.05 6.11 # pos. 0 1 7 12 18 22 23 23 23 23 21 21 21 23 18 # neg. 1 4 0 0 0 0 0 0 0 0 0 0 0 0 0

Notes: Results for 23 developed markets with Dubai crude oil prices with regression equation: Sit = α + βitOit+ it. All reported t-values are based on heteroscedasticity consistent standard errors.

(21)

5 Results

5.2.2 Increased efficiency

In table 7 two periods of high numbers of significant results can distinguished. The first period is from 1990 till 1992 and the second period is from 2006 till 2017. In both periods the majority of the countries have significant results for t=0. The results in the section performing a repro-duction suggest that crude oil price returns of month t=-1 can be used to predict the stock index returns of month t=0. To study this effect, the independent variable is taken as t=-2, t=-1 and t=0 to see whether oil price take time to diffuse into stock indices. When significant results exist on t=-2 or t=-1, weak-form efficiency is not present because crude oil price information is not directly reflected in the stock index, but with a lag. What this means is that when there are a high number of significant results for t=-1, one could use the crude oil price of t=-1 to predict stock indices in the countries where t=-1 is significant. To be able to say something about the degree of efficiency, a ratio has been established that compares how much information is lagging from one and two days before.

Table 8: Number of significant results around taking the independent variable for t=-2, t=-1 and t=0 for 23 developed countries and Dubai crude oil.

1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

Dubai(-2) 0 2 6 0 1 3 1 3 0 1 2 1 0 0 6

Dubai(-1) 0 1 12 9 2 1 3 3 0 0 0 3 0 2 2

Dubai(0) 0 1 14 20 18 3 2 2 2 1 0 2 1 1 1

Ratio 0.00 3.00 1.29 0.45 0.17 1.33 2.00 3.00 0.00 1.00 0.00 2.00 0.00 2.00 8.00

Table with the number of significant results on a 5% significance level, varying in the lag. The lags included in this table are for t=-2, t=-1 and t=0. Also the ratio of information contained in t=-1 and t=-2 compared to t=0 is given. Table 8: continued 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Dubai(-2) 0 11 1 0 2 4 0 0 0 0 0 8 0 2 2 Dubai(-1) 0 3 2 5 3 6 6 6 7 5 6 13 9 7 6 Dubai(0) 0 5 8 13 17 21 21 20 21 20 19 16 19 19 19 Ratio 0.00 2.80 0.38 0.38 0.29 0.48 0.29 0.30 0.33 0.25 0.32 1.31 0.47 0.47 0.42

Table with the number of significant results on a 5% significance level, varying in the lag. The lags included in this table are for t=-2, t=-1 and t=0. Also the ratio of information contained in t=-1 and t=-2 compared to t=0 is given.

(22)

5 Results

5.2.3 Changing sign

As mentioned before, consensus about whether stock index returns and crude oil price returns are positively or negatively connected does not exist. To study this, sample period of one year are taken with daily data of stock index returns as a dependent variable and crude oil price returns as the independent variable in an OLS regression. For the stock indices, 23 developed (as categorized by MSCI) markets are taken and for the crude oil spot price benchmark, Dubai oil is taken. In table 7, the T-statistics are separated by their sign to see the change in sign over the years. The table shows that the period of 1990 to 1992 contains results where the majority is significant on a 5% significance level and negative. Furthermore, nearly all T-statistics are negative for this period. This implies that for all the developed markets, the relationship between crude oil spot prices and stock index returns was negative.

Table 7 contains the results for using Dubai crude oil price returns, but the same regressions are done with the same developed markets, but with WTI crude oil returns as independent variable. To show this visually, figure 1 shows the behavior of the relationship of both crude oil price returns with stock indices.

1985 1988 1990 1993 1996 1998 2001 2004 2007 2009 2012 2015 2018 −5 0 5 10 Dubai(t=0) WTI(t=0)

Figure 1: Average t-statistics for Dubai and WTI crude oil price returns as independent variable and stock indices as dependent variable for 23 developed markets.

In figure 1 the results are displayed visually to compare the results between different crude oil price returns. The graph show very similar results for Dubai and WTI, which makes the results of this study more generalizable. The figure shows, first, that for both Dubai and WTI crude oil prices that in the period of 1990 to 1992, crude oil price returns have a negative relation with stock indices. Secondly, both crude oil price returns show a positive relation with stock indices in the period from 2006 to 2017.

5.2.4 Emerging markets

(23)

5 Results 1992 1994 1997 2000 2002 2005 2008 2011 2013 2016 2019 0 2 4 6 8 Dubai(t=0) WTI(t=0)

Figure 2: Average t-statistics Dubai and WTI for 24 emerging markets of significant results on a 5% significance level

(24)

6 Conclusion

In this paper the link between crude oil prices and stock indices has been studied. Oil prices proved to have a significant impact on stock indices when looking at both developed and emerging markets. This impact has been studied to see whether it has changed over the years and whether this impact on stock indices of crude oil price changes has a positive or a negative sign.

The relevance for the academic world is to see how the speed of information diffusion changes over the years when looking at the incorporation of crude oil price returns in stock indices. This study adds to the literature about efficiency in markets and what kind information is in the set of information that determines stock prices. This impact is relevant for investors because they can base investment strategies, especially when a lag is present. When a lag is present in the information inclusion in stock indices, then the weak form of efficiency is violated. In case of a lag present in the link between oil prices and stock indices, one can use historical prices for investment strategies to make better decisions and to enhance returns.

The results in this study show that predicting stock returns becomes more difficult and may become impossible if the relations change (the sign can change). What can be seen in the results section is that the sign has changed in the period between 1993 and 2005, moreover the link is decreasing again after 2015, which could be a precursor of another sign change.

Initially, the paper by Driesprong et al. (2008) was reproduced, and additional tests were done to see whether the results of that paper have changed over time. Monthly averages were taken to see whether oil price changes could be used to predict stock indices. The initial paper found that a lag of 6 days would provide the strongest link to base predictions on. The tests have been reproduced for three periods of ten years. Comparable results were found for the earlier periods. Thereafter, additional tests were done to study the changes of the lag over the years. The lag could decrease because information technology has improved, which would increase the speed of information diffusion. In the results a shorter lag was indeed found.

To test efficiency more thoroughly, daily prices were taken for oil and stock indices. The daily prices were used for tests in different ways, resulting in differing key variables, namely: number of significant results changing over the years, significance of regressions with lags and signs of the coefficients.

The number of significant results showed differences over the years for Dubai crude oil prices, but also between developed and emerging markets. For developed markets the majority of the countries analyzed had a significant negative link between the oil prices and stock indices of the same day in two periods, namely: 1990-1992. Between 1993 and 2005, this significance was mostly not present and from 2006 onwards the majority of the countries is significant once more, but now positive relations. This information shows that oil prices were relevant in determining stock index returns, but not in the period between 1992 and 2005.

Further tests show that significance around t=0 is also common in the results, which means that a lag is present in inclusion of information in the stock indices. For efficiency to exist, prices must be incorporated in the stock index of the same day. However, in the results section it can be seen that in the significant periods (1990-1992 and 2006-2017), there are also large amounts of significant links for oil prices of t=-1. This is partly because of time differences, where markets where oil prices are announced close later than stock indices of markets included in this paper.

The information that would be interesting for investment strategies is whether the significant link between crude oil price and stock indices is positive or negative.

(25)

6 Conclusion

prices were not relevant for stock indices and the link was around 0. And lastly, in the second significant period (2006-2017) there was a clear positive link between crude oil price changes and stock indices. This is contradicting evidence to what is the majority in literature that found a negative or no link between crude oil prices and stock indices.

6.1 Discussion and future work

Limited data availability did not allow to analyze the efficiency matter on an hourly (or even shorter) basis, data were only available on a daily basis. Investors of today have access to information technologies, such as the internet, to provide them with information on often a real time basis. Therefore, one would expect the investors to react immediately on extra information that could influence the stock index. Thus, markets will become more efficient over time, with the arrival of improved information technologies. For future research it would be interesting to study the efficiency of markets on an hourly (or shorter) basis.

The tests done do not have the statistical and informational power to say something about whether markets are efficient. But when a lag is discovered, the conclusion can be drawn that efficiency is not fully present. Next to that, the set of information that influences the stock indices is much larger than only oil prices. In this study, only the link between crude oil prices and stock indices is studied.

The ideal study case would be were there is unlimited availability of small time intervals (hours, quarters of an hour) of all relevant information that influence the stock prices. This set of information can also change over the years as was seen in this study, where oil seemed to be not relevant in stock indices in the period from 1993 to 2005.

(26)

Bibliography

Aastveit, K. A., Bjørnland, H. C., and Thorsrud, L. A. (2015). What drives oil prices? emerging versus developed economies. Journal of Applied Econometrics, 30(7):1013–1028.

Abosedra, S. and Baghestani, H. (2004). On the predictive accuracy of crude oil futures prices. Energy policy, 32(12):1389–1393.

Agusman, A. and Deriantino, E. (2008). Oil price and industry stock returns: Evidence from indonesia. 21st Australasian Finance and Banking Conference 2008 Paper.

Basher, S. A. and Sadorsky, P. (2006). Oil price risk and emerging stock markets. Global finance journal, 17(2):224–251.

Boldanov, R., Degiannakis, S., and Filis, G. (2016). Time-varying correlation between oil and stock market volatilities: Evidence from oil-importing and oil-exporting countries. Interna-tional Review of Financial Analysis, 48:209–220.

Chen, N.-F., Roll, R., and Ross, S. A. (1986). Economic forces and the stock market. Journal of business, pages 383–403.

Cong, R.-G., Wei, Y.-M., Jiao, J.-L., and Fan, Y. (2008). Relationships between oil price shocks and stock market: An empirical analysis from china. Energy Policy, 36(9):3544–3553. Driesprong, G., Jacobsen, B., and Maat, B. (2008). Striking oil: Another puzzle? Journal of

Financial Economics, 89(2):307–327.

Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical work. The journal of Finance, 25(2):383–417.

Ferrer Lapeña, R., Jareño Cebrián, F., Shahzad, S. J. H., and Jammazi, R. (2017). Time-varying causality between crude oil and stock markets: What can we learn from a multiscale perspective?

Huang, R. D., Masulis, R. W., and Stoll, H. R. (1996). Energy shocks and financial markets. Journal of Futures Markets, 16:1–27.

Hull, M. and McGroarty, F. (2014). Do emerging markets become more efficient as they develop? long memory persistence in equity indices. Emerging Markets Review, 18:45–61.

Jiang, Z.-Q., Xie, W.-J., and Zhou, W.-X. (2014). Testing the weak-form efficiency of the wti crude oil futures market. Physica A: Statistical Mechanics and its Applications, 405:235–244. Jones, C. M. and Kaul, G. (1996). Oil and the stock markets. The Journal of Finance, 51(2):463–

491.

Kling, J. L. (1985). Oil price shocks and stock market behavior. The Journal of Portfolio Management, 12(1):34–39.

Le, T.-H. and Chang, Y. (2015). Effects of oil price shocks on the stock market performance: Do nature of shocks and economies matter? Energy Economics, 51:261–274.

(27)

Bibliography

Miller, J. I. and Ratti, R. A. (2009). Crude oil and stock markets: Stability, instability, and bubbles. Energy Economics, 31(4):559–568.

Nandha, M. and Faff, R. (2008). Does oil move equity prices? a global view. Energy Economics, 30(3):986–997.

Papapetrou, E. (2001). Oil price shocks, stock market, economic activity and employment in greece. Energy economics, 23(5):511–532.

Park, J. and Ratti, R. A. (2008). Oil price shocks and stock markets in the us and 13 european countries. Energy economics, 30(5):2587–2608.

Sadorsky, P. (1999). Oil price shocks and stock market activity. Energy economics, 21(5):449– 469.

Sockin, M. and Xiong, W. (2013). Feedback effects of commodity futures prices. National Bureau of Economic Research.

Sørensen, L. Q. (2009). Oil price shocks and stock return predictability. EFA 2009 Bergen Meetings Paper.

Tabak, B. M. and Cajueiro, D. O. (2007). Are the crude oil markets becoming weakly efficient over time? a test for time-varying long-range dependence in prices and volatility. Energy Economics, 29(1):28–36.

Wang, Y. and Liu, L. (2010). Is wti crude oil market becoming weakly efficient over time?: New evidence from multiscale analysis based on detrended fluctuation analysis. Energy Economics, 32(5):987–992.

(28)

7 Appendix

(29)

7

App

endix

Table A.1: T-statistics of regressions of 23 developed markets and WTI oil prices

1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats

Australiaa 0.45 1.07 -0.82 -2.62* -3.07* -1.15 0.87 1.68 0.64 0.60 -0.04 -0.0 -0.07 0.25 1.49 Austria -0.40 -0.32 -2.54* -5.26* -6.12* -0.11 1.40 0.7 -0.71 0.51 0.12 -0.20 -1.97* -1.38 -0.07 Belgium 0.92 0.61 -2.07* -6.41* -7.11* -0.0 0.76 0.3 -0.6 -0.76 -0.06 -0.36 -1.73 -1.92 -0.18 Canada 1.14 1.1 -0.96 -1.27 -0.54 -0.32 -1.67 -0.85 0.44 -0.06 1.29 1.78 0.68 0.45 1.92 Denmark -0.21 0.90 -0.8 -5.12* -4.72* 0.5 0.63 0.85 -0.62 -0.07 0.38 0.44 0.61 0.40 1.24 Finland 0.01 0.43 -1.32 -3.5* -2.75* 0.78 0.82 -0.84 0.1 1.71 1.03 0.41 1.1 0.3 0.55 France 0.06 0.78 -2.33* -5.77* -6.27* 0.2 0.40 -0.58 -0.65 -0.17 0.27 0.39 0.63 0.05 1.05 Germany 0.89 0.88 -1.98* -5.43* -5.77* 0.69 -0.5 -0.48 0.38 0.32 0.02 0.35 0.08 -0.81 1.10 Hong Konga 0.74 0.53 -2.20* -5.02* -2.38* -0.04 -0.3 0.54 0.71 0.93 0.8 0.87 0.19 -0.78 -0.15 Ireland 0.60 0.90 -0.00 -3.26* -4.65* 0.88 0.48 -0.49 0 0.34 0.12 0.44 0.76 0.6 1.39

Israel n/a n/a n/a n/a n/a -0.67 1.60 3.08* 1.64 1.26 0.26 0.40 0.21 -0.04 0.99

Italy 0.5 1.17 -2.37* -6.16* -4.74* -0.02 -1.55 -0.9 1.12 0.54 -0.38 -0.08 0.86 0.0 0.23

Japana 0.81 1.40 -2.70* -4.16* -2.98* 0.95 1.49 1.27 -0.18 -1.05 -0.21 0.31 1.73 2.00* 1.02

New Zlnda n/a n/a n/a n/a -0.50 -0.7 -0.91 -1.0 -1.23 -0.75 -0.18 -0.21 1.24 0.45 -1.01

NL 1.55 0.74 -0.62 -3.52* -4.71* 0.78 0.29 -0.73 1.2 0.84 0.03 0.07 -0.25 -0.13 1.33

Norway 0.84 1.20 -1.00 -3.63* -3.13* 0.03 -1.0 -1.51 0.1 1.86 1.49 2.06* 2.69* 1.82 1.5

Portugal n/a n/a n/a n/a -0.30 -0.78 -0.75 -0.14 0.21 -0.70 0.58 1.74 1.18 -1.30 -2.32*

Singaporea 1.39 1.53 -4.75* -7.38* -5.* 0.78 0.78 -0.02 -0.89 -0.10 0.10 0.24 0.3 -0.40 -0.12

(30)

7

App

endix

Table A.1: Continued

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats

Australiaa 1.43 1.92 2.85* 3.98* 3.75* 7.88* 7.60* 7.24* 11.4* 9.78* 6.13* 1.14 1.95 4.27* 4.69* Austria -0.11 0.38 0.74 3.58* 4.09* 8.43* 9.39* 10.* 13.5* 11.7* 8.58* 2.01* 4.57* 8.19* 7.54* Belgium 0.22 -0.15 0.80 1.91 3.19* 7.60* 8.76* 10.7* 13.3* 12.3* 9.24* 2.51* 3.02* 5.58* 4.79* Canada 2.19* 2.* 5.64* 8.61* 8.34* 12.2* 13.6* 15.2* 17.9* 16.8* 13.7* 8.82* 11.8* 16.4* 16.2* Denmark 1.66 1.16 1.12 2.9* 3.93* 9.60* 10.1* 9.80* 11.6* 10.* 6.3* 0.78 1.36 4.83* 5.83* Finland 1.70 0.2 1.05 1.87 3.44* 7.78* 8.3* 9.49* 13.3* 13.* 9.52* 2.2* 3.39* 6.20* 5.36* France 1.1 -0.37 0.67 3.* 4.50* 9.65* 10.8* 12.3* 14.1* 13.3* 10.5* 2.85* 4.00* 7.51* 6.9* Germany 0.5 -1.9 -0.41 1.80 3.66* 9.61* 10.9* 12.4* 14.0* 13.4* 10.5* 2.38* 3.13* 6.43* 6.45* Hong Konga -0.27 -0.02 1.01 1.6 2.78* 5.50* 5.86* 5.95* 6.70* 5.26* 2.96* 0.8 0.88 2.45* 2.44* Ireland 1.14 0.82 0.98 2.32* 3.09* 6.47* 7.11* 8.17* 10.9* 10.5* 8.39* 1.51 3.26* 6.57* 5.39* Israel -0.53 -1.85 0.09 -0.20 1.66 5.54* 6.0* 5.28* 6.07* 6.36* 3.80* 0.7 2.5* 2.78* 1.05 Italy 1.03 0.16 0.25 2.25* 3.78* 9.84* 10.7* 11.5* 13.0* 12.* 9.66* 2.61* 4.33* 8.5* 8.03* Japana 1.06 2.14* 2.2* 2.32* 1.58 2.94* 1.65 -0.39 3.4* 2.99* -0.04 -0.25 -0.01 0.26 0.65 New Zlnda _-1.02 _-1.32 _-0.3 _0.58 _1.3 _7.06* _7.68* _8.53* _10.7* _8.89* _6.73* _2.58* _4.74* _8.86* _9.54* NL 0.85 -0.88 -0.10 1.57 2.94* 8.71* 9.81* 11.2* 13.9* 12.7* 10.2* 2.22* 3.06* 6.65* 6.10* Norway 1.33 2.41* 3.48* 5.49* 7.04* 12.2* 13.3* 13.* 16.4* 15.7* 10.6* 5.35* 8.68* 11.6* 10.4* Portugal -1.31 -2.44* -0.50 1.98* 2.34* 3.17* 3.75* 4.40* 5.21* 4.53* 2.68* -0.0 1.30 3.26* 2.44* Singaporea -0.64 -0.08 1.92 2.62* 3.29* 5.31* 6.67* 8.5* 10.4* 8.93* 5.28* 0.56 2.87* 4.83* 4.35* Spain 0.64 -0.15 0.75 1.93 3.66* 8.82* 10.0* 10.5* 11.2* 10.2* 8.27* 2.29* 4.68* 8.54* 7.22* Sweden 1.10 -0.31 0.42 2.5* 3.51* 8.* 9.58* 11.3* 15.1* 13.7* 9.76* 2.5* 3.6* 7.13* 6.90* Swtzrlnd 1.00 0.40 0.59 2.17* 3.07* 9.14* 9.98* 10.5* 13.2* 12.* 8.68* 1.31 2.85* 6.44* 6.20* UK 0.30 -1.3 1.53 3.86* 4.38* 10.7* 11.4* 11.* 14.* 13.3* 10.8* 4.28* 6.0* 9.29* 8.75* USA -0.33 -4.14* -1.86 -0.64 0.48 6.2* 8.60* 12.3* 13.2* 12.3* 12.5* 5.55* 5.86* 8.27* 7.78*

Notes: Results for 23 developed markets with WTI crude oil prices with regression equation: Sit= α + βitOit+ it. All reported t-values are based on heteroscedasticity consistent standard errors.

a _{In this regression the independent variable (crude oil price return) is taken for t=-1, to account for the time difference.}

(31)

7

App

endix

Table A.2: T-statistics of regressions of 24 emerging markets and Dubai crude oil prices

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats t-stats

Brazil 1.02 0.72 0.59 1.38 0.99 -1.22 -0.71 0.06 0.01 0.66 0.64 -1.59 -0.35

Chile 0.15 1.12 2.09* 1.76 1.76 0.80 -0.61 0.53 0.49 -1.45 -0.27 -0.39 -0.44

Chinaa _n/a _0.81 _1.04 _0.9 _0.35 _0.33 _0.62 _0.63 _0.36 _0.33 _0.60 _1.29 _0.16

Columbia n/a 0.2 0.29 0.08 0.0 0.04 1.24 1.29 0.0 -0.48 -1.1 -1.5 0.1

Czech Republic n/a n/a n/a -0.15 -1.17 -1.17 -0.31 -0.13 -1.08 -0.82 0.8 0.71 1.5

Egypt n/a n/a n/a 0.67 -1.99* -1.52 0.12 0.00 0.09 1.26 1.85 -0.9 -1.27

Greece 0.35 0.56 0.23 0.05 -0.58 -0.15 0.58 -0.67 -2.41* -1.88 -0.27 0.63 1.08

Hungary n/a n/a n/a 0.64 1.25 0.91 0.26 -0.17 -0.09 -0.29 0.31 0.89 1.61

Indiaa n/a 0.9 0.32 -0.96 0.22 0.63 0.33 -0.29 -0.27 -0.21 -0.73 0.55 -0.04 Indonesiaa 0.35 1.15 -0.62 -1.41 -0.13 0.34 0.02 -0.36 -1.31 -0.84 0.14 -1.0 -1.3 Koreaa -0.37 -1.14 -1.3 -0.10 -0.07 -0.47 0.40 0.46 -0.26 0.04 1.22 1.21 -1.31 Malaysiaa -0.06 -0.00 0.78 1.39 0.8 0.41 0.42 0.67 0.55 -0.47 -1.08 -0.90 -0.47 Mexico -0.27 -0.75 1.03 1.20 0.24 -0.03 0.76 1.03 0.74 0.49 0.37 -0.53 -0.48 Pakistana n/a -0.71 -0.74 -0.6 -1.33 -0.70 0.49 0.71 0.8 0.61 -0.37 -2.11* -1.36 Peru n/a -0.69 -1.8 -0.87 1.51 -0.18 -1.29 0.10 0.99 -0.64 -1.66 -0.39 1.99* Philippinesa 3.06* 2.10* -0.3 -0.02 -0.16 -1.67 0.04 0.85 -0.57 -0.3 0.01 -0.74 -0.22 Poland n/a -0.37 -1.30 -0.71 0.38 -1.84 -1.37 -0.71 -1.87 -1.04 1.15 1.42 1.43

Qatar n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a

(32)

7

App

endix

2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

Brazil 2.00* 3.0* 4.88* 11.8* 14.8* 17.5* 13.9* 11.* 8.08* 3.97* 6.35* 9.30* 8.15* Chile 0.72 1.12 1.55 8.4* 10.* 10.5* 10.2* 8.90* 5.* 2.38* 7.43* 10.4* 8.25* Chinaa _0.47 _4.30* _2.21* _1.99* _3.61* _6.40* _5.37* _4.44* _4.85* _3.90* _3.09* _4.55* _5.19* Columbia 1.15 2.46* 3.20* 8.44* 10.2* 10.2* 8.38* 7.47* 4.99* 3.80* 10.3* 13.* 11.7* Czech Republic 3.37* 3.9* 3.66* 10.* 11.4* 1* 9.27* 7.93* 6.19* 1.37 3.56* 5.34* 4.08* Egypt 0.0 0.09 0.35 3.91* 3.15* 2.89* 3.43* 0.68 -0.51 -1.55 1.08 2.07* 1.47 Greece 1.42 2.03* 3.16* 9.25* 8.86* 7.27* 7.33* 4.79* 2.64* 1.72 3.02* 4.51* 4.36* Hungary 2.6* 3.20* 3.89* 8.72* 10.1* 10.4* 10.4* 8.98* 6.18* 1.22 2.64* 5.15* 4.43* Indiaa -0.41 1.90 1.40 -1.25 -0.1 2.54* 2.60* 0.66 0.41 1.60 0.80 1.24 1.81 Indonesiaa 0.17 2.31* 1.72 2.80* 3.61* 3.52* 3.58* 3.81* 3.17* 2.19* 2.31* 2.7* 2.26* Koreaa -1.30 2.96* 2.13* 0.79 1.92 4.59* 5.63* 5.57* 5.55* 3.56* 2.41* 3.53* 3.24* Malaysiaa -0.06 1.00 1.66 3.45* 5.40* 7.94* 6.13* 4.80* 3.70* 5.07* 7.19* 7.14* 6.04* Mexico 2.26* 2.22* 1.88 6.74* 10.2* 14.6* 12.2* 8.28* 3.41* 2.25* 5.24* 7.69* 6.14* Pakistana 0.63 0.83 1.1 -0.08 -0.04 0.37 1.80 1. -0.30 -0.11 2.61* 3.69* 1.61 Peru 3.54* 5.86* 6.1* 11.6* 14.7* 15.* 10.3* 7.1* 5.90* 4.62* 8.04* 10.2* 8.45* Philippinesa 0.8 1.45 1.59 4.04* 6.49* 8.80* 6.72* 5.18* 3.52* 2.28* 3.84* 5.19* 4.10* Poland 2.55* 3.26* 3.55* 10.0* 11.4* 11.4* 11.6* 10.4* 7.64* 1.76 4.03* 6.20* 4.38* Qatar 0.32 0.54 0.48 2.29* 1.63 0.54 2.80* 2.18* 0.32 -0.77 2.30* 4.73* 3.98* Russia 1.6 3.67* 4.32* 8.06* 9.98* 12.7* 12.7* 11.8* 10.3* 3.92* 6.96* 13.4* 13.6* South Africa 2.59* 3.48* 4.67* 10.0* 10.7* 10.2* 11.6* 9.7* 7.21* 3.00* 5.58* 8.6* 7.61* Taiwana -1.23 1.18 0.80 3.06* 4.72* 6.4* 6.56* 6.* 6.41* 1.97* 1.42 3.37* 3.68* Thailanda -1.15 2.65* 2.16* 1.63 2.9* 4.29* 3.69* 3.8* 3.79* 3.12* 2.45* 1.62 0.95 Turkey 2.07* 2.79* 2.83* 7.59* 9.7* 10.0* 7.87* 6.50* 3.20* 1.00 2.86* 4.52* 4.44* UAE 0.48 -0.2 -1.15 4.28* 4.19* 3.0* 3.80* 1.81 -1.40 -1.37 2.89* 5.2* 2.60*

Notes: Results for 24 emerging markets with Dubai crude oil prices with regression equation: Sit= α + βitOit+ it. All reported t-values are based on heteroscedasticity consistent standard errors.

(33)

7

App

endix

Table A.3: T-statistics of regressions of 24 emerging markets and WTI crude oil prices

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Brazil -0.54 0. -1.73 -1.69 0.98 -0.77 -0.40 0.18 -0.44 0.9 1.44 -0.53 -0.22

Chile 1.30 0.90 -1.74 -0.91 0.49 0.14 -0.46 0.59 0.54 -0.35 1.27 0.37 -0.25

Chinaa _n/a _1.09 _0.29 _0.15 _-0.24 _-0.18 _0.48 _0.64 _-0.51 _-1.36 _-0.18 _0.75 _0.18

Columbia n/a 0.56 1.07 0.69 -1.49 -1.54 -0.38 0.61 0.90 -0.41 -0.5 0.66 0.91

Czech Republic n/a n/a n/a -0.30 -1.19 -1.17 1.26 1.1 -0.36 0.1 0.21 -0.35 0.76

Egypt n/a n/a n/a 0.93 -0.22 -1.09 -0.12 0.50 -0.02 1.20 2.01* -1.07 -1.49

Greece -0.50 0.39 1.79 1.29 -1.55 -0.82 0.70 -0.18 -0.85 -0.28 0.11 0.68 0.84

Hungary n/a n/a n/a 0.44 -0. 0.22 0.51 -0.37 0.00 0.52 1.14 1.40 1.08

Indiaa n/a -0.32 0.24 0.66 -1.80 -0.70 2.31* 1.81 -0.80 -2.30* -1.48 0.90 -0.65 Indonesiaa 1.3 1.37 1.16 0.45 0.83 1.58 0.10 0.47 1.1 0.18 0.81 1.26 1.60 Koreaa 0.18 -0.11 0.12 0.20 -0.0 -0.31 -0.74 -0.42 -0.50 -0.15 1.46 0.45 0.9 Malaysiaa -1.11 0.47 0.65 0.17 -0.2 1.22 1.34 0.64 0.15 0.57 -0.14 1.00 1.84 Mexico -2.08* -1.62 -0.86 -0.50 0.31 0.26 1.82 1.3 0.2 0.81 1.13 -0.19 -0.63 Pakistana n/a -0.77 -1.01 -0.58 -0.41 -0.12 0.1 0.29 -0.35 0.59 1.57 -0.92 -1.32 Peru n/a -1.60 -3.0* -1.0 0.92 -0.27 0.47 1.12 0.66 0.18 0.51 1.38 1.58 Philippinesa 0.38 1.0 -0.65 -0.51 1.02 0.67 0.83 0.15 -0. 0.75 0.53 0.00 -0.54 Poland n/a -1.00 -0.48 0.68 0.02 -1.20 -0.01 1.09 0.68 1.17 2.04* 0.77 0.39

Qatar n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a

(34)

7

App

endix

2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

Brazil 1.80 3.12* 4.54* 10.50* 12.40* 15.10* 16.60* 14.40* 9.16* 4.31* 6.59* 9.75* 8.65* Chile 0.82 1.93 1.64 7.57* 8.47* 8.73* 11.8* 10.80* 6.30* 2.40* 6.49* 9.57* 7.84* Chinaa _1.09 _1.59 _2.78* _5.32* _6.03* _7.31* _7.46* _5.80* _3.11* _0.0 _0.46 _2.72* _3.37* Columbia 0.85 2.36* 2.99* 7.83* 9.53* 9.9* 9.68* 9.41* 6.21* 4.04* 9.65* 13.* 11.4* Czech Republic 1.9 3.33* 3.80* 9.69* 10.3* 9.94* 9.50* 8.42* 6.34* 0.33 3.16* 5.47* 4.07* Egypt -0.51 -0.31 0.01 3.90* 2.31* 1.39 3.70* 0.99 -0.49 -1.85 0.58 2.27* 1.88 Greece 0.73 1.98* 3.43* 8.43* 8.19* 7.34* 6.94* 4.80* 4.25* 1.41 1.77 3.68* 3.86* Hungary 2.09* 3.87* 4.54* 7.15* 8.75* 10.0* 10.80* 10.30* 6.74* 0.78 2.85* 4.82* 3.64* Indiaa 0.19 3.80* 4.2* 5.87* 6.90* 8.1* 8.70* 7.19* 3.98* 0.85 1.61 3.28* 3.40* Indonesiaa 1.75 3.05* 3.47* 3.75* 3.94* 5.16* 6.45* 4.69* 1.45 -0.18 1.89 2.4* 1.43 Koreaa 2.96* 2.98* 3.74* 4.9* 4.48* 4.22* 6.23* 4.90* 2.11* 0.99 2.30* 3.28* 2.37* Malaysiaa 0.91 0.45 2.52* 2.95* 3.9* 6.4* 6.86* 5.15* 1.4 -0.17 1.45 3.48* 4.50* Mexico 1.44 1.26 1.11 5.7* 8.39* 11.9* 12.9* 9.37* 4.55* 2.3* 4.85* 7.11* 5.54* Pakistana -0.48 -0.26 0.93 -0.02 -0.30 -0.0 1.27 0.92 -0.18 -0.60 1.00 1.96 0.92 Peru 2.11* 5.31* 5.93* 9.98* 12.0* 13.1* 11.6* 9.39* 6.5* 4.70* 8.03* 9.52* 7.77* Philippinesa 0.66 2.18* 1.49 3.00* 2.17* 0.48 2.07* 0.75 -0.58 -1.36 1.64 2.86* 1.98* Poland 0.81 2.00* 3.97* 9.2* 10.9* 11.9* 13.3* 12.0* 7.57* 0.9 4.6* 6.45* 3.90* Qatar 0.23 0.37 0.56 2.79* 1.51 -0.12 2.20* 0.98 -0.08 -1.7 1.63 3.65* 2.49* Russia 1.22 3.48* 4.40* 7.56* 9.56* 13.0* 14.6* 13.6* 10.6* 3.10* 6.37* 13.40* 13.9* South Africa 2.27* 3.60* 4.77* 9.15* 9.45* 9.89* 14.4* 12.* 7.18* 2.41* 5.14* 8.30* 7.66* Taiwana 1.27 0.84 2.74* 3.68* 4.09* 4.87* 6.53* 4.79* 1.44 0.02 2.71* 3.19* 2.03* Thailanda 1.54 1.83 3.16* 4.71* 5.59* 6.01* 6.46* 6.18* 1.60 0.15 2.71* 4.36* 3.66* Turkey 0.26 2.05* 3.20* 6.87* 8.80* 10.00* 8.87* 7.05* 2.23* -0.15 2.11* 4.22* 4.54* UAE 1.30* 0.34 -0.73 3.92* 3.27* 2.03* 2.76* 1.00 -0.44 -1.29 2.21* 4.30* 1.49

Notes: Results for 24 emerging markets with WTI crude oil prices with regression equation: Sit = α + βitOit+ it. All reported t-values are based on heteroscedasticity consistent standard errors.