The effect of oil price on the stock returns of Norwegian firms

1

The effect of oil price on the stock returns of

Norwegian firms

Abstract

In this empirical paper I study the effect of oil price on Norwegian stock returns, by way of analysing that effect on a firm level. I study this effect due to the recent claim of the Norwegian government that Norway has become too vulnerable to oil shocks. The sample

period is from January 2000-December 2016. The analysis is conducted by using a fixed effects model, in so doing controlling for firm and time fixed effects. I find that the effect of

oil price on stock returns is negative and significant. However, the significant effect is not substantial enough to endorse the claim of Norway being too vulnerable to oil shocks. This negative finding is inconsistent with prior research conducted on Norway and is also not in

line with the stated hypothesis.

Faculty of Economics &Business BSc Economics & Business

Specialization Finance & Organization Bachelor Thesis

Name: R. van Wechem Student number: 10807020 Supervisor: S.R. Changoer Date: 31-01-2018

2

Statement of originality

This document is written by Riwan van Wechem who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

3

Table of contents

4

1. Introduction

Norway has a national wealth fund that was created with the revenues of the oil and gas from the North Sea (Sheppard, 2017). Nowadays, this fund has a net worth of about 1 trillion US dollars and has stakes in different companies worldwide (Norges Bank, 2018). Recently, the Norwegian government was considering to sell their stakes in big oil companies such as Chevron, Shell, BP, ExxonMobil and Total claiming that Norway is too vulnerable to oil shocks (Sheppard, 2017).

In my thesis, I test this claim by examining the effect of the crude oil price on Norwegian stock returns in the period of January 2000 to December 2016. To examine this effect, I run seven separate regressions. These different regressions serve as a way of comparing the individual effect of the main explanatory variable under different circumstances. I also conduct two robustness tests.

I find a significant negative relation between oil price changes and stock returns, which suggests that an increase in oil price results in a decrease in Norwegian stock returns. This finding is inconsistent with prior research conducted on Norway. As, Bjørnland (2009) finds a positive relation between oil price and stock returns in Norway. The claim from the Norwegian government that Norway has become too vulnerable to oil shocks does not seem to hold. After all, the effect might be significant, but it is not substantial.

My thesis contributes to prior research by examining the effect of oil price changes on Norwegian stock returns in a period that, according to the literature I reviewed, has not been examined before. To be precise, Bjørnland (2009) and Park and Ratti (2008) study the effect of oil price changes up to the year of 2005. In order to test the claim of the Norwegian government, a more recent study is required. For instance, Figure 1 shows that the oil price changes since 2005 have been of a greater extent than before 2005. Besides that, both Bjørnland (2009) and Park and Ratti (2008) use a VAR model to capture the effect of oil price changes on stock returns. I study this effect in a different way by using the fixed effects model on a firm level instead. Bjørnland (2009) finds that a 10% increase in oil price will be followed by a 2.5% increase in Norwegian stock returns. Also, Park and Ratti (2008) find a positive relationship between oil price and stock returns in Norway

The remaining sections are structured in the following order. In section I discuss literature and formulate my hypothesis. In Section II, I discuss the method and sample. In Section III, I present the results and perform robustness checks. Last, section IV concludes this paper by reflecting the results to the stated research question.

5

Figure 1. Graph of Brent Crude oil price in Norwegian Krone from January 2000 until December 2016, monthly observations. Note: The sudden price drop starting around 2008 is noteworthy. This observation is at the start of the financial crisis.

2. Literature review

Oil is a resource used to create energy. Besides oil, Figure 2 shows that there are also other energy sources claiming a share in the world energy consumption. These other energy sources include natural gas, coal, nuclear energy, hydro energy and renewables (BP, 2017).

The world’s dependency on oil became clear during big oil shocks in 1973 and 1979. According to BP (2017) oil still is the biggest resource for energy consumption worldwide. Oil accounts for a third of the world’s energy consumption and for two years straight it is gaining market share (BP, 2017). 0 100 200 300 400 500 600 700 800 1/ 1 /2 00 0 12 /1 /2 0 00 11 /1/20 01 10 /1/20 02 9/1 /200 3 8/1 /200 4 7/1 /200 5 6/1 /200 6 5/1 /200 7 4/1 /200 8 3/1 /200 9 2/1 /201 0 1/1 /201 1 12 /1/20 11 11 /1/20 12 10 /1/20 13 9/1 /201 4 8/1 /201 5 7/1 /201 6 Oi l p ri ce Time

6

Figure 2. Chart of world energy consumption of 2016 in million tonnes of oil equivalent (mtoe). Source: BP Statistical Review of World Energy (2017).

Since 1973, the price of crude oil is mainly controlled by the Organization of the Petroleum Exporting Countries (OPEC) (Driesprong et al., 2008). The OPEC itself accounts for roughly 42 percent of total oil production worldwide (BP 2017). Yet the OPEC has the ability to influence the volatility of oil price by increasing or lowering production (BP, 2017). This results in the oil price being not that volatile. For example, if the oil price suddenly drops due to an oil shock, the OPEC will just lower its production, so the price will increase.

There are two characteristics that determine the price of oil: the density and sulphur content (Driesprong et al., 2008). When the sulphur content is low it is called sweet and when sulphur content is high it is called sour. The density is called light when it is low, and a high density is characterized by heavy. Oil with high sulphur content and high density tends to have higher production costs than oil with low sulphur content and low density. For example, West Texas Intermediate commonly has a higher price than Brent oil, since it is lighter and sweeter (Driesprong et al., 2008). Although Figure 3 shows that since 2011 the price of Brent has exceeded the price of West Texas Intermediate (BP, 2017).

-500.0 1000.0 1500.0 2000.0 2500.0 3000.0 3500.0 4000.0 4500.0 5000.0 North America S. & Centr. America Europe & Eurasia Middle East Africa Asia Pacific World mt o e

World energy consumption 2016

Oil Natural Gas Coal Nuclear Energy Hydro Electric Renewables

7

Figure 3. Graph of historical oil prices in US dollars over the period 1973-2016. Dubai, Brent and West Texas Intermediate are shown as the three main benchmarks for crude oil. Dubai is shown as the black dotted line, Brent as the black solid line and WTI as the grey solid line.

There are three main benchmarks for crude oil. The biggest benchmark accounts for about 40 to 50 million barrels per day and is the Brent oil benchmark. The West Texas Intermediate accounts for about 12 to 15 million barrels per day, whilst Dubai serves as a benchmark for about 10-15 million barrels per day (Driesprong et al., 2008).

Both Brent and West Texas Intermediate are characterized as a form of light sweet crude oil (Driesprong et al., 2008). However, Dubai is characterized as a medium and sour type of crude oil (Hammoudeh et al., 2008).

The total world production of oil in 2016 is around 92 million barrels per day (BP, 2017). This implies a yearly oil production of about 34 trillion barrels. Oil itself is used in several sectors such as power generation, buildings, petrochemicals, industry, aviation and shipping, road freight and passenger vehicles (IEA, 2017).

The predictive effect of oil price on firm stock return can be explained by their impact on expected earnings (Jones and Kaul, 1996). For oil consuming industries where oil is a key input, an increase in oil price can lead to higher production costs and thus lower expected cash flows. This in turn, can lead to lower stock prices and lower stock returns. However, for oil producing industries this is vice versa (Xu, 2015). Therefore, I expect that oil price predicts stock returns.

There are several studies finding a positive effect of oil price changes on Norwegian

0 20 40 60 80 100 120 19 73 19 75 19 77 19 79 19 81 19 83 19 85 19 87 19 89 19 91 19 93 19 95 19 97 19 99 20 01 20 03 20 05 20 07 20 09 20 11 20 13 20 15 Oi l p ri ce

Historical oil prices

8

stock returns, such as Bjørnland (2009) and Park and Ratti (2008). Driesprong et al. (2008) examine the predictive effect of oil price changes on stock market returns worldwide and find that an increase in oil price lowers stock returns. Also, Jones and Kaul (1996) find that oil price changes have a negative effect on stock returns in the US, Canada, Japan and UK in a post war period. Looking further, Narayan and Sharma (2011) find a positive relation between oil price changes and firm returns for the energy and transportation sector, while other sectors show a decline in return when the oil price rises. El-Sharif et al.(2005) prove in the UK that the relationship between oil price changes and stock returns in the oil and gas sector is always positive and mostly highly significant. Chen et al. (1986) find that the predictive effect of oil price on stock returns is insignificant for the US in the period 1953-1983. Last, Aspergis and Miller (2009) find that oil price shocks do not affect international stock returns to a large extent. In a more recent study Chiang and Keener (2017) find a negative relation between oil price changes and stock return in both an in- and out-of-sample framework.

Both Bjørnland (2009) and Park and Ratti (2008) use a VAR model to capture the effect of oil price changes on stock returns. I test the effect of oil price on stock returns in a different way by using the fixed effects model on a firm level. Based on the findings of Bjørnland (2009), who found that a 10% increase in oil price will be followed by a 2.5% increase in Norwegian stock returns, and Norway being an oil exporting country, I expect that an increase in oil price will lead to an increase in Norwegian stock returns.

3. Research Method

3.1 Method

To test my hypothesis, I estimate the following fixed effect model, in which standard errors are clustered by both firm and month:

Rit = αi + β1Oilpriceit + β2BtMit + β3Sizeit + β4Momit + β5Niborit + β6Inflationit + β7Rmit

+ β8Crisisit +Fixed effects + εit

Where:

R = The stock return of a Norwegian firm in percentages.

Oil price = The price per barrel of oil in Norwegian Krone. I use the Brent oil price from the Oseberg oilfield as a commodity benchmark for the oil price

9

Size = The natural logarithm of the market value of a firm. The market value of a firm equals the number of shares outstanding times the share price.

Mom = Momentum effect.

Nibor = The Norwegian Interbank Offered Rate, which is the three months offered interest rate in Norway.

Inflation = The rate at which purchasing power is declining, because goods and services become more expensive.

Rm = The return on the market. The market proxy is the Oslo Børs.

Crisis = A dummy variable for the financial crisis which takes the value of one in the years 2008 and 2009.

Fixed effects = Firm fixed effects and time fixed effects. ε = The error term.

I use the Brent oil price from the Oseberg oilfield as a commodity benchmark for the oil price, because the Oseberg oilfield is located on the southwestern coast of Norway. Importantly, Hammoudeh et al (2008) argue that Brent is typically used as a benchmark for oil in Europe and West Texas Intermediate for North America. Since I examine firms in Norway, a country in Europe, I choose Brent as a benchmark and not West Texas Intermediate. Another reason to choose for Brent as an oil benchmark is because it is the biggest oil benchmark worldwide (Driesprong et al., 2008).

I include BtM, because Fama and French (1992) find that book to market equity is a good predictor in describing stock returns. The correlation between book to market value and average returns is positive.

I include Size, because Fama and French (1992) find that it is a significant factor in describing stock returns. Smaller firms tend to have too high average returns, while larger firms seem to have too low average returns (Fama and French, 1992). The correlation between size and average returns is negative.

I include Mom, because Fama and French (2012) find momentum effect in the regions North America, Europe and Asia Pacific. The momentum effect is part of a behavioural theory which is elaborated by Hong and Stein (1999). They classify two sorts of traders, namely ‘newswatchers’ and ‘momentum traders’. The newswatchers possess private information and base their own forecast on it, whereas momentum traders only rely on price changes that happened in the past. First, positive news will make the newswatchers respond to it and the price will rise, although not to its full potential. This leads to an underreaction. Second, the

10

momentum traders notice the price increase and act upon it, causing the price to rise further. This price increase will attract even more momentum traders which eventually will lead to the price exceeding its long-run actual value. This mechanism results in an eventual overreaction and that will in the long run decrease the price until it reaches its equilibrium.

I include Nibor, because according to Sadorsky (1999) the effect of the interest rate on stock returns is negative and also statistically significant. Therefore, it is a variable that needs to be included in the regression model in order to prevent a bias in the estimation.

By using the formula (newCPI-oldCPI)/oldCPI*100 the monthly inflation rate between January 2000 and December 2016 has been calculated as a percentage. I include Inflation, because Feldstein (1980) finds that an increase in inflation will cause the stock price to rise until it is in proportion with the price increase of inflation. In this way the ratio of share price to real earnings is maintained.

The Oslo All Exchange (Oslo Børs) is used as a market index for Norway. The return of the market is calculated by looking at the percentual change per month of the total return index. I include Rm, because Fama and Frech (1992) find that return on the market captures a substantial amount of variation of stock returns.

I also include Crisis, because in the years 2008 and 2009 there are large outliers in stock returns. Thus, not controlling for this period could lead to a bias. Only in the years 2008 and 2009 there are large outliers in stock returns, therefore I only choose 2008 and 2009 as years for Crisis. Another reason to include Crisis as a control variable is because Iyer et al. (2014) find evidence that the banking crisis from 2007-2009 negatively impacts the ability of a firm to obtain a loan from a bank. This in turn reduces the liquidity of a firm, which reduces firm’s stock return.

The use of panel data is necessary in order to account for a different effect of oil price on Norwegian stock return due to the sort of industry. This means that multiple firms will be compared over time. Some firms in this dataset are not traded on the Norwegian exchange market, the Oslo Børs, but are traded via the Norwegian Over-The-Counter market (OTC). The firms listed on the OTC market cannot be used and will be deleted from the dataset.

According to Stock and Watson (2015), the fixed effects model controls for firm fixed effects and for time fixed effects. By making use of firm fixed effects, all the variables that vary across firms but stay the same over time are being controlled for and will not cause a bias (Stock and Watson, 2015). As for time fixed effects, the idea is the same, although this controls for values of a variable that do change over time but are the same for each firm.

11

across firms and there is heteroscedasticity. Clustered standard errors allow for autocorrelation and heteroscedasticity without violating the multiple regression assumption that each observation is independent (Stock and Watson, 2015). This is due to the fact that it allows the residuals to correlate within a cluster (firm) but assumes that they are uncorrelated across clusters. To get two dimensional clustered standard errors, meaning for both firm and for month, I generated an additional variable that grouped the firms and months into one variable. This results in having 21,093 clusters.

3.2. Sample

To test the hypothesis, I collect data from Datastream, the website of Ken French and the website of Statistics Norway. The data for the Brent oil price and Norwegian stock returns is available on Datastream. Monthly observations of individual firms are used. The firms are selected from the Oslo Børs and between the period of January 2000-December 2016.

Datastream does not provide data about a firm’s book to market value, although it does about a firm’s market to book value. One divided by the market to book value gives the book to market value. The market capitalizations of all the firms are retrieved from Datastream as well. By taking the natural logarithm of these values the variable Size is created for each firm.

The data library from Kenneth R. French (2018) his website provides data for the momentum variable and the other variables that are included in robustness test (2), shown in Table 4. This data includes monthly observations of the European momentum factor, including Norway.

Data about the inflation is obtained from the website of Statistics Norway. This website provides historical data about the Consumer Price Index (CPI) in Norway. For the data about the interest rate the Norwegian Interbank Offered Rate (Nibor) is used as a benchmark, because it is the three-month interest rate for Norway (Bjørnland, 2009). Another reason to choose the Nibor as a benchmark, is because it is calculated by the Oslo Børs. Since I choose the Oslo Børs to be the market proxy, I also choose their calculated interest rate. This data is available in a monthly frequency and is obtained from Datastream.

My dataset contains 158 different listed Norwegian firms over a time period of 16 years. These firms are obtained from the Oslo Exchange All Share (Oslo Børs), which contains 169 listed firms. However, since I make use of data from 2000 until the end of 2016 the firms of whose data starts in 2017 are not used. I choose for the period to start in 2000, because starting from 2000 the oil price changes tend to be greater than before, as can be seen in Figure 3. Those

12

greater changes can be more useful in examining the effect of oil price changes on stock returns. All the observations of the firm’s stock return give 21,093 observations. These firms are from multiple industries such as Alternative Energy, Automobiles & Parts, Banks, Beverages, Chemicals, Construction & Materials, Electricity, Electronic & Electrical Equipment, Financial Services (Sector), Food producers, Forestry & Paper, Gas Water & Multiutilities, General Industrials, General Retailers, Health Care Equipment & Services, Household Goods & Home construction, Industrial Engineereing, Industrial Metals & Mining, Industrial Transportation, Leisure Goods, Media, Mining, Mobile Telecommunications, Nonequity Investment Instruments, Nonlife Insurance, Oil & Gas Producers, Oil Equipment & Services, Personal Goods, Pharmaceuticals & Biotechnology, Real Estate Investment & Services, Software & Computer Services, Support Services, Technology Hardware & Equipment and Travel & Leisure.

Because R and BtM show large outliers in the dataset, I winsorize the data. With winsorization the values of the first percentile are replaced by the smallest value of the 2nd

percentile and the values of the last percentile are replaced by the largest value of the 99th

percentile.

Table 1 presents the summary statistics of the model. As well as, Table 1 shows that the total amount of observations of all the returns of all firms are 21,280. This amount of observations implies that although other variables might have more observations, the total amount of observations in the regression cannot be higher than 21,280. Since the stock return of a firm is the dependent variable on which the regression is being conducted, the number of observations in the regression can never exceed that of the dependent variable.

13

Table 1 Summary statistics

Variable Observations Mean Std. dev. Min Max

R 21280 0.51 13.97 -38.80 50.79 Oil price 32232 417.81 158.34 178.11 689.52 BtM 21194 1.01 1.10 -0.55 6.67 Size 21382 7.30 1.91 3.28 12.20 Mom 32232 0.86 4.10 -11.42 10.26 Nibor 32232 3.50 2.09 0.99 7.43 Inflation 32232 0.17 0.44 -0.74 1.48 Rm 32232 0.96 5.85 -17.55 12.74 Crisis 32232 0.12 0.32 0 1

Summary statistics of the dependent and independent variables from January 2000-December 2016, 204 months. The values of R, Mom, Nibor, Inflation and Rm are in percentages. Oil price is measured in Norwegian Krone, BtM is measured as a book to market ratio and Size is measured as the natural logarithm of the market value.

4. Results

4.1 Regression results

Table 2 presents the results of the regression analysis. The coefficient on Oil price is significant at the 5% significance level. The correlation between a firm’s stock return and the oil price is negative with a coefficient of -0.009. Meaning that an increase in Brent Crude oil price with one unit will lead to a 0.009 decrease in a firm’s stock return while holding all other variables constant. When only taking the individual effect of oil price on stock return as in model (3), the coefficient remains negative but changes slightly to -0.042 at the 1% significance level. This is inconsistent with prior research, since Bjørnland (2009) finds that a 10% increase in oil price will lead to a 2.5% increase in Norwegian stock returns. This somewhat contradictive result could be due to the inclusion of the variables Mom, Nibor and Inflation. Table 3 presents the correlation of those variables with Oil price and that correlation makes the coefficient slightly negative. On the contrary, when those variables are excluded as in model (7), the coefficient of Oil price is positive at the 10% significance level. This result comes more

14

near the findings of Bjørnland (2009). That oil price has a statistically significant effect on stock return is consistent with the hypothesis. However, I conclude that the effect is minimal and besides that negative. This contradicts the expected positive correlation.

The coefficient on the variable BtM is significant at the 1% significance level and shows a negative correlation between book to market value and stock return. A one unit increase leads to a 0.974 decrease in stock return. Its significance is consistent with prior research from Fama and French (1992), while the negative coefficient is not.

The coefficient on Size is also significant at the 1% significance level. This variable has a positive coefficient, where a one percent increase in Size results in a 1.876/100 unit increase in stock return. This correlation is inconsistent with the findings of Fama and French (1992), who claim that larger firms have lower expected returns.

In contrast, the coefficient on Mom is insignificant. Fama and Frech (2012) find that in the Europe region momentum does have significant explanatory power for stock return. Therefore, this insignificance is unexpected, although the negative correlation of -0.056 with stock return is consistent with the long-term mechanism of overreaction.

Furthermore, the coefficient on Nibor is insignificant and negative. This negative result is in line with Sadorsky (1999), who finds that the effect of oil price on stock return is negative and statistically significant. Sadorsky (1999) argues that a change in interest rate has an impact on the profit of a firm, because a change in interest rate means a change in the price charged for a loan. That theory supports the significant result of Sadorsky (1999). However, I find an insignificant result, and this can be due to different time frames. Sadorsky (1999) studies a period that covers the years 1947-1996 and I study the period from 2000-2016. The economic circumstances in which Sadorsky (1999) conducts his research are obviously different from the ones of this research. Different economic circumstances imply different results of the factors explaining stock returns.

The coefficient on Inflation is insignificant as well. The Inflation coefficient is positive with a coefficient of 1.588. The positive coefficient supports the theory of Feldstein (1980), who argues that an increase in inflation will cause the stock price to rise until it is in proportion with the price increase of inflation. In this way the ratio of share price to real earnings is maintained. However, the insignificant result is not in line with the significant result of Feldstein (1980). This might be due to different time frames as well. Feldstein (1980) examines a period from 1967-1976 and it could be that certain economic circumstances were different than in the period of 2000-2016. Again, different economic circumstances imply different results of the factors explaining stock returns. This could explain Inflation not being a

15 significant variable for explaining stock return.

Different from the Nibor and Inflation, the coefficient on Rm is statistically significant at the 1% significance level. The coefficient of 0.895 means that when the return on the market increases with one unit, the stock return of a Norwegian firm increases with 0.895. This positive coefficient is in line with what I would expect, since Fama and French (1992) find a strong positive result between Rm and stock returns.

Also, the coefficient on Crisis is significant at the 1% significance level and has a negative coefficient of -3.813. This means that if there is a financial crisis, this leads to a decrease in stock returns with 3.813. Iyer et al. (2014) argue that the banking crisis from 2007-2009 negatively affected the ability of firms to obtain credit from banks. This reduces the liquidity of the firm and therefore negatively impacts firms’ stock returns. This theory supports the negative coefficient of Crisis.

Table 2 presents the different regression models. Model (4) serves as the base model and the models (5) to (7) serve as alternative models to see whether excluding the insignificant variables from model (4) give different results for the coefficient on Oil price. In model (5) the Oil price coefficient changes to being insignificant and besides that the coefficient turns from negative to positive. This means that momentum and oil price are correlated with each other. For model (6), the Oil price coefficient is negative and statistically significant at the 1% level. This as well can be explained by a correlation between the macroeconomic variables and Crisis. What is noteworthy in model (7) is that Oil price shows a positive coefficient while being significant at the 10% significance level. This result of model (7) is consistent with the findings of Bjørnland (2009), who finds that a 10% increase in oil price will be followed by a 2.5% increase in stock returns. Therefore, model (7) might be a better model than model (4) in predicting Norwegian stock return, because the negative coefficient of model (4) is inconsistent with the findings of Bjørnland (2009).

16 Table 2 Regression analysis Model Regressor (1) (2) (3) (4) (5) (6) (7) Oil price -0.004*** (0.001) -0.004*** (0.001) -0.042*** (0.005) -0.009** (0.004) 0.003 (0.004) -0.022*** (0.007) 0.013* (0.007) BtM -0.974*** (0.173) -0.974*** (0.173) -0.974*** (0.173) -0.974*** (0.173) Size 1.876*** (0.182) 1.876*** (0.182) 1.876*** (0.182) 1.876*** (0.182) Mom -0.056 (0.085) -0.006 (0.119) Nibor -0.701 (0.587) 0.163 (0.369) Inflation 1.588 (0.978) 1.204 (0.993) Rm 0.895*** (0.104) 0.981*** (0.141) 0.927*** (0.057) 0.937*** (0.100) Crisis -3.813*** (1.264) -7.253*** (2.111) 0.706 (1.681) -10.000*** (3.242) Entity effects?

no yes yes yes yes yes yes

Time effects?

no no yes yes yes yes yes

Clustered standard errors?

no yes yes yes yes yes yes

Observations 21280 21280 21280 21093 21093 21093 21093

R2 _{0.002 0.012}

0.182 0.197 0.197 0.197 0.197

The regression results are estimated using panel data for 158 Norwegian firms. For each variable the coefficients are denoted, and its statistical significance is shown at the *10%, **5% or ***1% significance level. Under the coefficients the standard errors for each variable are denoted in parentheses.

17 4.2 Correlation results

Table 3 presents the correlations among regressors. This table has been created to detect problems with multicollinearity. None of the variables seem to be highly correlated with each other. Only Nibor is moderately negatively correlated with Oil price. This moderate correlation however, is not a threat to the assumption that all the variables should be independent from each other. It only has the consequence of higher standard errors.

Table 3

Correlations among regressors

Regressor Oil price BtM Size Mom Nibor Inflation Rm Crisis 1. Oil price 1 2. BtM -0.01 1 3. Size 0.11*** -0.22*** 1 4. Mom 0.07*** -0.04*** 0.02** 1 5. Nibor -0.47*** -0.11*** -0.04*** 0.06*** 1 6. Inflation 0.03*** 0.00 0.01 0.13*** 0.05*** 1 7. Rm -0.04*** -0.03*** 0.01* 0.01 -0.22*** 0.00 1 8. Crisis 0.12*** 0.01 0.01 -0.08*** 0.15*** 0.00 -0.10*** 1

The statistical significance is shown at the *10%, **5% or ***1% significance level.

4.3 Sensitivity analysis

In order to get a better understanding of the base model, conducting a sensitivity analysis is very helpful. I perform 2 robustness tests in which some variables of the base model are replaced by similar alternative variables, as shown in Table 4. For both robustness tests I winsorize the data and use clustered standard errors.

In robustness test (1) I replace certain control variables by the Fama and French (2012) variables. Those variables are similar to the variables that I use in this research and they are significant in the research of Fama and French (2012). Therefore, the new variables could strengthen the insight of the coefficient on Oil price. The coefficient on Oil price is in robustness test (1) positive at the 1% significance level. The positive coefficient of 0.019 contradicts the negative coefficient of -0.009 in Table 2.

The new variables involve BtM being replaced by HML, which stands for high minus low book to market ratio. SMB means small minus big market capitalization and this variable

18

replaces Size. The Nibor variable is replaced by Rf, which is characterized by the risk-free rate. The variable Rm is replaced by Rm-Rf, which can be seen as the market risk premium. All the new variables are retrieved from Kenneth R. French’s data library (2018) and represent the Europe region. Given the lower R2 of 0.182 compared to the R2 of 0.197 of the base model from Table 2, I conclude that the base model is a better model in predicting Norwegian stock returns.

For robustness test (2) I change the benchmark for the oil price from Brent to WTI (West Texas Intermediate). The other variables remain the same as in the base model. I perform this robustness test, because WTI is after Brent the biggest oil benchmark in the world (Driesprong

et al., 2008). Besides that, both benchmarks are a form of sweet light crude oil, although WTI

is slightly sweeter and lighter (Driesprong et al., 2008). Because Brent and WTI have similarities, but are also slightly different, it is interesting to see whether using WTI as an oil benchmark gives different results. Table 4 shows that this modification results in the coefficient of oil price being negative at the 1 % significance level. The coefficient of -0.020 is slightly more negative than in the base model. Robustness test (2) provides more significant results than the base model and has the same R2_{. In that sense the model with WTI as an oil benchmark}

might be better than the base model. However, the negative coefficient of WTI is inconsistent with Bjørnland (2009).

19

Table 4 Sensitivity analysis

Robustness (1) Robustness (2)

Regressor Coefficient Regressor Coefficient

Oil price 0.019*** (0.004) WTI -0.020*** (0.005) HML 0.252 (0.183) BtM -0.974*** (0.173) SMB 1.776*** (0.196) Size 1.876*** (0.182) Mom 0.010 (0.117) Mom 0.303** (0.118) Rf 51.513*** (6.849) Nibor -1.171*** (0.424) Inflation -1.176 (1.024) Inflation 2.014** (1.009) Rm-Rf 0.374*** (0.135) Rm 0.698*** (0.100) Crisis 7.458*** (1.494) Crisis 1.534 (1.255) Observations 21280 Observations 21093 R2 0.182 R2 0.197

In robustness test (1) BtM is replaced by HML, Size by SMB, Nibor by Rf and Rm by Rm-Rf. In robustness test (2) the original oil benchmark Brent Crude is replaced by another oil benchmark WTI (West Texas Intermediate). The coefficients of each variable are denoted with the standard errors underneath in parentheses. The statistical significance is shown at the *10%, **5% or ***1% significance level.

5. Conclusion

In this paper the effect of the oil price on Norwegian stock returns has been studied. By using the fixed effects model I have used a different approach than previous studies conducted on the effect of oil price on stock returns in Norway. Also, the findings of the previous studies conducted on Norway covered a period until 2005, while I examined a period of 2000-2016.

20

Therefore, my different approach could contribute to prior research conducted on the effect of oil price on stock returns in Norway.

I find that when oil price increases with one unit, the stock returns decrease with 0.009. This negative result contradicts the stated hypothesis and the findings of prior research on Norway. However, in an alternative model with the same R2 the coefficient of oil price shows to be positive and also statistically significant, as shown in Table 2.

The claim from the Norwegian government that Norway has become too vulnerable to oil shocks does not seem to hold. After all, the effect might be significant, but it is not substantial.

Robustness test (2) from Table 4 presents more significant variables than the base model, but the coefficient of oil price is again negative. Moreover, the WTI makes less sense as an oil price benchmark for Norway, since Brent Crude oil from Oseberg is produced next to the coast of Norway.

The limitation in this research is the absence of certain macroeconomic variables such as unemployment rate or the Gross-Domestic Product. These variables could increase the explanatory power of stock returns and thereby produce better insights. In this paper I control for the effect of the financial crisis and do this for the years 2008 and 2009. Yet the impact of the financial crisis can last longer than 2009 and maybe controlling for a longer period would have given different results. Future research should focus more on the volatility of the oil price, since the volatility can give a good insight whether the Norwegian government is vulnerable to oil shocks in the short run.

21

6. References

Aspergis, N., Miller, S.M. (2009). Do structural oil-market shocks affect stock prices?. Energy

Economics, 31(4), 569-575.

Bjørnland, H.C. (2009). Oil price shocks and stock market booms in an oil exporting country.

Scottish Journal of Political Economy, 56(2), 232-254.

BP (2017). BP Statistical Review of World Energy June 2017. Retrieved January 5, 2018, retrieved from https://www.bp.com/content/dam/bp/en/corporate/pdf/energy- economics/statistical-review-2017/bp-statistical-review-of-world-energy-2017-full-report.pdf

Chen, N., Roll, R., & Ross, A. (1986). Economic Forces and the Stock Market. The Journal of

Business, 59(3), 383-403.

Chiang, I.E., Hughen, W.K. (2017). Do oil futures prices predict stock returns? Journal

of Banking and Finance, 79, 129-141.

Consumer price index, (2017, December 11). Retrieved December 14, 2017, retrieved from https://www.ssb.no/en/priser-og-prisindekser/statistikker/kpi

Driesprong, G., Jacobsen, B., Maat, B. (2008). Striking oil: Another puzzle? Journal of

Financial Economics, 89(2), 307-327.

El-Sharif, I., Brown, D., Burton, B., Nixon, B., & Russel, A. (2005). Evidence on the nature and extent of the relationship between oil prices and equity values in the UK, Energy

Economics, 27(6), 819-830.

Fama, E.F., French, K.R. (1992). The cross-section of expected stock returns, Journal of

Finance, 47. 427-465.

Fama, E.F., French, K.R. (2012). Size, value, and momentum in international stock returns,

Journal of Financial Economics, 105(3), 457-472.

Hammoudeh, S.M., Ewing, B.T., & Thompson, M.A. (2008). Threshold Cointegration Analysis of Crude Oil Benchmarks, The Energy Journal, 29(4), 79-95.

Hong, H., Stein, J.C. (1999). A unified theory of underreaction, momentum trading, and overreaction in asset markets, Journal of Finance, 54(6), 2143-2184.

22

IEA (2017). World Energy Outlook 2017. Retrieved January 23, 2018, retrieved from

https://www.iea.org/weo2017/

Iyer, R., Peydró, J., da-Rocha-Lopes, S., & Shoar, A. (2014). Interbank Liquidity Crunch and the Firm Credit Crunch: Evidence from the 2007-2009, The Review of Financial Studies, 27(1), 347-372.

Jones, C.M., Kaul, G. (1996). Oil and the Stock Markets, The Journal of Finance, 51(2), 463-491.

Narayan, P.K., Sharma, S.S. (2011). New evidence on oil price and firm returns. Journal of

Banking and Finance, 35(5), 3253-3262.

Norges Bank Investment Management, (2018). Retrieved January 29, 2018, retrieved from https://www.nbim.no/en/the-fund/

Kenneth R. French – Data Library, (2018). Retrieved January 18, 2018, retrieved from

http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html

Park, J., Ratti, R. A. (2008). Oil price shocks and stock markets in the U.S. 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.

Sheppard, D. (2017, November 16). Norway fund to sell off oil shares. Financial Times, p. 21.

Stock, J.H., Watson, M.W. (2015). Introduction to Econometrics. Harlow, Essex: Pearson Education Limited.

Xu, B. (2015). Oil prices and UK industry-level stock returns. Applied Economics, 47(25), 2608-2627.

23

7. Appendix

Proof of fixed effects model and OLS fixed effects estimator: Yit = αi + β1X1,it + βkXk,it + εit

This is the fixed effects model with firm i at time period t. Yit-Y̅it = (αi-α̅i) + β1(X1,it - 𝑋̅1,it) + βk(Xk,it - 𝑋̅k,it) + (εit-ε̅it)

With Y̅it = (1/T)∑𝑇_𝑡=1 Y̅it and 𝑋̅1,it and ε̅it are defined equally

This is the first step to obtain the OLS fixed effects estimator. The average is subtracted from each variable.

𝑌̃it = β1𝑋̃it + … + βk𝑋̃k,it + 𝜀̃it

With 𝑌̃it = Yit - 𝑌̅it , 𝑋̃it = Xit - 𝑋̅it and 𝜀̃it = εit - 𝜀̅it

This is the second step to obtain the OLS fixed effects estimator. Now β1 can be