The impact of oil price shocks on the stock returns of Alternative and Conventional European electric utilities

(1)

Master thesis for MSc finance

The impact of oil price shocks on the stock returns of

Alternative and Conventional European electric

utilities

Author: John Dekker Supervisor: dr. G.T.J. Zwart

Student ID: 2366088 Date: 10-01-2018

Abstract:

This paper analyses the impact of oil price shocks on the stock return of European alternative and conventional electric utilities. Electric utility stock indices have been made by use of an equally and capitalization-weighted method. To study the interaction between oil prices and stock returns a vector autoregressive model has been used. The interest rate, a technology index, and industrial production have been added to capture macroeconomic influences. I find that oil price shocks have a negative impact on the stock prices of conventional utilities. Next to that, large oil price increases have a positive effect on the stock prices of alternative electric utilities. The results are robust to different weighting methods, time periods, and amount of lags used.

(2)

1. Introduction

European companies are heavily dependent on oil, but little influence can be exercised on the oil price as most of the oil is imported. Nevertheless companies are greatly impacted by sudden changes in the oil price. For most sectors an increase in the oil price leads to higher production costs and consequently to lower profits. However, the impact of an oil price shock1 can be

different among countries, sectors, and can even deviate within sectors (Lee et al., 2012; Faff and Brailsford, 1999; Driespong et al., 2003).

Oil, coal and gas prices are related, because they compete as fuel (Villar, 2006; Hartley et al., 2008). This relation connects the electric utility sector to oil prices as conventional electric utilities2 often use coal and gas-fired generators (Oberndorfer, 2009). Consequently production

costs of conventional utilities increase due to oil price shocks.

However, coal and gas-fired generators are becoming less popular as the sector is transitioning to more renewable production (Armaroli and Balzani, 2015). Efficiency of renewable production is increasing and becoming more profitable, which leads to a growing amount of investments in alternative electric utilities3 (Gielen et al., 2015). Alternative electric utilities use

nontraditional sources to generate electricity. For that reason, their production costs are less affected by fossil fuel prices. This transition changes the dependency of companies, who are switching to renewable generation, on the oil price.

Increased renewable electricity production also has disadvantages, such as a higher price variance of electricity (Wozabal et al., 2015). Next to that, it is questioned whether renewable production can completely replace conventional production as renewable producers are not able to produce a stable base load, and costs of storing electricity are high. In a nutshell, the energy transition is ongoing and the amount of investments in renewable production are increasing. Therefore, it is important for investors to have a good understanding on how the impact of the oil price on the electricity sector is changing.

1 Oil price shocks are defined as a sudden increase in the price of oil relative to the price that firms expect. 2 Conventional electricity is produced in a way that does use up natural resources or harms the environment (e.g.

gas, oil, nuclear).

(4)

Surprisingly, less research has been done on the impact of oil price shocks on the electric utility sector. Moreover, most existing literature on the impact of oil price shocks on the electricity sector does not distinguish between alternative and conventional utilities. Therefore this paper adds to existing literature, what the impact of oil price shocks on the stock returns of conventional and alternative electric utilities is.

To be able to capture the impact within the sector, portfolio indices have been made by us of an equal and capitalization-weighted method. As oil prices do not impact stock prices in isolation, the interest rate, a technology index, and industrial production have been added to the model to capture macroeconomic influences. An unrestricted vector autoregressive model has been conducted with monthly data, for the period January 2002 until June 2017, to answer the main question:

What is the impact of oil price shocks on the stock returns of conventional and alternative European electric utilities?

The oil, coal and gas price are positively correlated, therefore an increase in the oil price will increase the production costs for conventional electric utilities. Alternative utilities produce electricity with for example water, wind or solar energy. Hence, oil price changes will have less impact on these companies. Therefore it is expected to find a difference between the impact of an oil price shock on the stock returns of conventional and alternative electric utilities.

(5)

2. Literature review

This part will first discuss research done on the impact of oil price shocks on the country, sector, and utility level. Second, the relation between oil, coal gas and electricity prices will be discussed. Last, the production costs of electric utilities will be addressed.

2.1 Impact of oil price shocks

Several studies have been done on the impact of oil price shocks and stock market prices. These studies can be subdivided based on their scope: the country, industry, and utility level. The following subsection will discuss the result found by these papers, starting from the impact on country level, and ending with the electric utility level.

2.1.1 Country level

Huang et al. (1996) investigated the relationship between oil futures returns and daily U.S. stock returns during the 1980s using a vector autoregressive model. Their model examines the relationship between oil futures returns and stock returns including controls for, among others, the effect of the interest rate and seasonality. They found that oil futures returns are only correlated with stock market returns in case of some individual oil company returns.

Bjornland (2009) used a structural VAR model to investigate the relation between oil price shocks and stock market booms in Norway, an oil exporting country. Correcting for the foreign interest rate, unemployment, inflation, domestic interest rate and exchange rates, he found that a 10% increase in oil prices leads to a total increase of stock prices by 3,6% after 14 months, thereafter the effect gradually dies out. He suggests the cause to be increased demand and not restrictions on the supply side.

(6)

et al. (2009) found that the reaction of U.S. real stock returns to oil price shocks also differs greatly depending on whether the shock is driven by demand or supply.

Sadorsky (1999) used a vector autoregressive model with monthly data to show that oil prices and oil price volatility both play important roles in affecting real stock returns in the United States. By using monthly data and controlling for interest rates, he found that oil price volatility shocks have asymmetric effects on the economy. In particular, oil price drops have less impact on stock returns than oil price hikes.

Driesprong et al. (2008) used a basic regression model and found that in 12 out of 18 developed countries oil prices significantly predict stock returns. Syed et al. (2006) used a multi-factor model to analyze the impact in emerging countries, and also came with strong evidence that oil price risk impacts stock price returns.

Opposed to the previous mentioned studies, some others found that international stock market returns do not respond in a large way to oil market shocks. Aspergis (2009), who used a vector autoregressive model to research if oil-market shocks affect stock prices in the U.S., concluded that the significant effects that were found proved to be small in magnitude. Lee et al. (2012) found that oil price shocks do not significantly impact the composite index4_{in the G-7}

countries. However, for the interaction between oil price changes and sector stock prices he did find a short-run causal relationship. In particular, stock returns of the consumer staples and information technology sectors were strongly impacted, followed by the utility, financial, and transportation sectors.

2.1.2 Sector level

Several studies highlight the importance of researching the impact of oil price shocks on the sector level, as the impact of broad-based stock prices may mask industry effects (Lee et al., 2012; Faff and Brailsford, 1999; Driespong et al., 2003).

The research of Faff and Brailsford (1999) focused on the correlation between oil prices and stock returns of Australian sectors. With the help of an augmented market model they found significant positive sensitivity in four oil and oil related industries, and significant negative

(7)

sensitivity in the ‘other metals’, ‘chemicals’, ‘paper and packaging’, and the ‘transport’ industry. They found, among others, that an oil price shock of 1% increased the stock return of the oil and gas industry by 0.24%.

Daily data for the period 1998-2006 has been used by Elyasiani et al. (2011) to study the interaction between stock prices and 13 U.S. industries. They came with strong evidence in support of the view that oil price fluctuations influence asset price risks at the industry level, as they found a correlation for 9 out of 13 sectors between industry excess returns and oil future returns. An excess stock return of 0.038% and 0.007% have been found for the coal and electric-gas services industry if the daily return on a one-month crude oil future increased by 1%. Scholtens and Yurtsever (2012) estimated some dynamic vector autoregressive models to look at the correlation between oil price shocks and European industries. They used monthly data for the period 1983-2007. Their model controls for the interest rate and industrial production. As most others, they found that the impact and significance of oil price shocks, substantially differ among industries. An interesting finding is that the oil price contributes for more than 10% of the variance of the electricity sector. Additionally they found that a 1% oil price shock caused an accumulated negative response of 0.1% for the electricity industry after both 12 and 24 months. Furthermore they found the same results by using a multivariate regression model instead of a vector autoregressive model.

Narayan and Sharma (2011) used a Garch (1,1) model to investigate the impact of oil price shocks on 560 listed U.S. firms. Their first conclusion was that the impact of oil prices depends on the sectoral location of the firm. Second, they came with strong evidence of a lagged effect of the oil price on firms returns. Their third finding is a threshold effect of the oil price on firm returns for 5 out of 14 sectors. Last, they unravel that oil price returns impact stock returns differently depending on firm size.

2.1.3 Utilities

(8)

a significant relation between oil price volatility and U.S. stock prices of clean energy companies was not found.

Oberndorfer (2009), who used a Garch model to study the effect of oil prices on stock returns of European oil and gas fired utilities, found that oil price hikes negatively impact the stock returns of European oil and gas fired utilities. Another main finding was that the oil price is the main indicator for energy price developments as a whole. Furthermore he found that a 1% oil price shock has a direct negative impact of 0.03% on the stock returns of European energy stocks.

Henriques and Sadorsky (2008) used a vector autoregressive model to show that oil prices and technology stocks, are each individually useful for forecasting the U.S. stock prices of alternative energy companies in the period 2001-2007. Next to that, a technology shock has a larger impact on alternative energy stock prices than a shock to oil prices. They found that a one standard deviation shock to the natural logarithm of the technology stock, increases the natural logarithm of the alternative energy index by 3%. Besides, he found no statistically significant impact of an oil price shock on alternative energy stock prices.

2.2 The relation between oil, coal, gas and electricity prices

The main theory on the price linkage between oil, coal and gas prices is based on economic substitution (Manzoor and Seiflou, 2011). They are competitive substitutes as they can all be used for electricity generation and industrial production. (Bachemeier and Griffin, 2006; (Frydenberg et al., 2014). The correlation between oil and gas is higher than the correlation between oil and coal which can, among others, be explained by the fact that coal cannot be used for transportation, unless we return to steam trains and coal-fired steamships.

The oil, gas and electricity markets are co-integrated. (Mohamaddi, 2009; Bencivenga et al., 2010). Additionally, Bencivenga (2010) found that the oil market is a source of risk for the electricity and gas market. Figure 1 on the next page shows oil, gas and electricity price changes for the period 2002-2016 and is taken from the ‘ECB Economic Bulletin, Oil prices and Euro

area consumer energy prices, Issue 2 (2016)’. It shows a lead-lag effect5_{from oil to gas prices.}

(9)

The relation between fossil fuel and electricity prices is harder to see. The impact of an oil price change is lower for electricity than for gas prices, which can be explained by the high capital intensive nature of electricity industry, which means that the fuel costs are only a part of the total costs. Next to that, the generation of electricity is partly replaced by alternative generation, which leads to less dependence on oil and gas.

2.3 Production costs of electric utilities

The merit order ranks electricity producers based on their short-run marginal costs. Renewables have the lowest marginal costs, nuclear energy the second lowest, then coal, followed by natural gas, and oil has the highest marginal costs. Utilities generate electricity if their marginal costs of production is lower than the electricity price.6 _{Higher oil prices will increase the marginal} costs for conventional utilities, leading to lower production or suspension of the operations, and therefore to less profits. Consequently, low electricity prices will increase the relative input from alternative utilities and high electricity prices the reverse. Renewable generators do not need fossil fuels for generation, and will hence be less impacted. Alternative utilities are more dependent on the electricity price and the efficiency of generation.

6 Mulder, M. (2016). Energy and Finance – Lecture 2. Retrieved from nestor.rug.nl Figure 1 – Oil, gas and electricity prices (% change per year)

(10)

2.4 Expectations

Based on the literature review above, it is expected to find that oil price shocks have a negative impact on the stock prices of conventional electric utilities. Oil, coal and gas prices are positively correlated, consequently an increase in the oil price will increase production costs of conventional utilities. As oil and electricity prices are less correlated than oil and gas prices, production costs will increase stronger than the electricity price, which results in lower profits. Second, it is expected that oil price shocks have a less severe impact on alternative electric utilities. Renewable generators do not need fossil fuels for electricity generation. Third, it is expected that technology stock shocks have a positive impact on the stock returns of alternative companies, because it increases the efficiency of renewable electricity generators.

3. Data

In this section first the variables: Brent crude oil futures, real economic activity, technology, and interest rate will be clarified, as well as the adjustments made to the them and their relevance. Then the weighting of companies in the electric utility indices will be discussed. Last, a look will be taken at the summary statistics.

For all variables applies that monthly data for the period January 2002 until June 2016 has been used. Next to that log levels have been used for the stock indices, industrial production, technology index, and the oil price index to reduce unwanted variability. The interest rate is expressed in its levels as it is already measured as a ratio. The choice of the sample period was made purely on data availability. Monthly data has been chosen based on other papers (i.a. Sadorsky, 1999; Killian and Park, 2009; Lee et al., 2012). Table 6 in the appendix shows a short description, the sources, and the abbreviations of the data used.

3.1 Brent crude oil futures

(11)

it has the advantage that they are less noisy than spot prices (Mork, 1989; Bjornland, 2009). The 1-month oil future data has been extracted from investing.com.

To be able to check if there is a difference between an oil price increase and decrease, extra oil price variables have been made. Defining the log level of real oil prices as lop, and the first difference as Δlopt = (lopt - lopt-1). Oil price increases and decreases are respectively defined as:

Δopt+ = Δopt if > X (1)

and

Δopt- = Δopt if < X (2)

,with X as 0, 1, and 2% oil price changes.7

3.2 Real economic activity

The real economy is the part of the economy that is concerned with actually producing goods and services as opposed to the part of the economy that is concerned with buying and selling on the financial markets (Isaac, 2012). Oil price shocks typically have real effects, as higher energy prices affect output via the aggregate production function by reducing the net amount of energy used in production. Consequently, an increase in oil prices leads to a rise in production costs, inducing firms to lower output (Lee et al., 2012; Scholtens, 2011; Cunado and Perez de Gracia, 2013). The real economic activity variable is measured as the volume index of production of all 28 European Union members. The volume index is seasonally and calendar adjusted and has been derived from Eurostat.

3.3 Technology

The STOXX Europe 600 Technology is used as a proxy for technology and has been extracted from investing.com. It is an index of European stocks and consists of 600 technology companies. Among them are large companies capitalized among 18 European countries, covering approximately 90% of the free-float market capitalization of the European technology market.

(12)

Investors may view alternative utilities as similar to other high technology companies. (Henriques and Sadorsky, 2008; Sadorsky, 2011). This is what happened in the late 1990s to a number of fuel companies. The stock prices of these companies rose as quickly as the NASDAQ index, but also fell as quickly as technology companies after the bubble bursts in the early 2000s. Therefore a technology stock index is used to investigate the relationship between electric utilities’ stock prices and the broad-based technology sector.

3.4 Interest rate

The interest rate is one of the most important macroeconomic variables. Previous research showed the importance of the interest rate in explaining stock price movements (Chun et al.,1986; Sadorsky, 1999, 2001, 2008), and highlighted that stock prices are more strongly affected by very short-run interest rates (Huang et al., 1996; Sadorsky, 2001; Elyasiani et al., 2011).

Interest can be seen as cost of capital. For a borrower, investing becomes less attractive if interest rates are high, because of a higher borrowing rate. Next to that, high interest rates increase the attractiveness of lending instead of keeping stocks. This will lead to a decrease in the demand for shares (Alam and Uddin, 2009).

The three-month Euribor interest rate, deducted from the OECD website8, will be used as the

short-term interest rate. The Euro Interbank Offered rate is based on the averaged interest rates at which Eurozone banks offer to lend unsecured funds to other banks in the euro wholesale money market or interbank market.

3.5 Electric utility indices

Stock price, dividend and stock split data have been obtained from ThomsonOne9. The electric

utilities used for the index are presented in table 7 and 8 in the appendix. They have been chosen and divided in alternative10 and conventional11 electric utilities based on the industry

8 Organisation for Economic Co-operation and Development.

9 ThomsonOne is the World’s leading source of intelligent information for businesses and professionals.

10 The alternative electricity sector is defined as companies generating and distributing electricity from a renewable source. Includes companies that produce solar, water, wind and geothermal electricity.

(13)

classification (ICB). Companies have been left out if not enough data on stock prices and/ or dividends was available.

The stock prices are manually adjusted for dividends and stock splits. Historical prices are adjusted by a factor that is calculated from the moment the stock begins trading ex-dividend in the following way: First the amount of the dividend is subtracted from the prior day’s price and then divided by the prior day’s price. Second historical prices are multiplied by this factor. This adjusts historical prices proportionally so that they stay rationally aligned with current prices. In case of for example a 2-for-1 stock split, prices before the split are multiplied by 2.

The renewable and conventional index are made with respectively 16 and 18 companies. The market index therefore includes 34 electric utilities. Due to the dividend adjustments, the return of the index both tracks the capital gains of the stocks over time, and assumes that dividends are reinvested back into the index. There are several methods in which an index can be made. In this paper an equally-weighted and capitalization-weighted index are used for both the alternative and conventional index. The equally-weighted method is also used to create an index for the whole electricity industry. The industry index has been left out in the capitalization-weighted method, because more than 97% of the industry index would be determined by the conventional sector, which would lead to high correlation between the industry and conventional index.

3.5.1 Equally-weighted index

The equally weighted index only takes the price of stocks into consideration when determining the index. It gives each stock the same weight, so ignores the relative size of the company. The index for a specific month, t, is computed by first taking the sum of returns of all the companies present at that time:

Total returnt = ∑𝑛_𝑖=1Returnit, (3)

in which n is the total amount of companies (i) present in a given month. The equally-weighted index is then computed by dividing the total return by the amount of companies present:

Index(EW)t = ∑𝑛_𝑖=1Returnit / nt (4)

(14)

and the return of that company is added to the total return for the remaining months. The equally-weighted index has several advantages. First, as opposed to the capitalization-weighted method, the index does not overweight overpriced stocks and underweight underpriced stocks (pricing errors are random). Second it is easy to construct. Last, the market cap and amount of alternative utilities is still relatively small compared to conventional utilities, but both are growing over time through which the equally-weighted index integrates the alternative companies well.

3.5.2.Capitalization-weighted index

The capitalization-weighted index factors in the size of the company. The index is based on the free float market capitalization methodology, and has been created in the following way. First the market capitalization of all the companies has been calculated as ‘price of share on base date’ times ‘total numbers of shares issued’.

Market Capit = SPit x SOit (5)

In which market cap is the market value of a company i, SP is the share price of company i, and SO the total amount of shares outstanding by company i. The market capitalization calculated has been multiplied with the Free Float Factor (FFF). The FFF is the percentage of shares that is in the hands of investors and can freely be traded. The sum of all the free float market capitalization is taken as the base.

∑𝑛_𝑖=1Baseit = ∑𝑛_𝑖=1Market Capit * FFFit (6)

The weights of individual companies have been determined by dividing their free float market capitalization by the total base.

Weightit = Baseit / ∑𝑛_𝑖=1Baseit (7)

The capital weighted index is calculated as the sum of the weights of all the company present at time t, times their return in that month.

Index(CW)t = ∑𝑛_𝑖=1( Weightit * Returnit) (8)

(15)

roughly mirrors the change in the total market value of all stocks. Second, the rebalancing of this index is simple as the index automatically adjusts to changes in stock prices. A big disadvantage is that the index is heavily influenced by a few companies with large market capitalizations.

New companies have been added to the portfolio later in the time series, but as the portfolio is capitalization-weighted the influence of a new company is minor. Next to that, the market cap of the companies present at the beginning of the time series is large. A time series plot of the industry, alternative, conventional and oil index are shown in figure 8 and 9 in the appendix. To be able to compare the indices each is set equal to 100 in January 2015.

3.6 Summary statistics

The summary statistics for the indices and returns are shown in table 1. The STOXX Europe 600, which gives a good overview of the European market, is used as a benchmark. The correlation coefficient between the variables are shown in table 9 in the appendix.

Table 1 – Summary statistics of the utility indices and the STOXX market index

The summary statistics show that the cap-weighted conventional and equally weighted alternative index are doing well in comparison to the STOXX 600, with annual returns of respectively 5.4% and 4.68%.12 An interesting finding is that the conventional indices are doing

better in terms of return variance in comparison to the alternative indices and the market index. The return variances of the alternative indices are found to be relatively high.

The correlation between the capitalization and equally-weighted indices are significant and around 0.8. Furthermore, the correlation between the conventional and alternative indices is quite low. All stock index returns are significantly correlated with the oil price returns.

12 Calculated as 12 times the monthly return.

Equally-weighted Cap-weighted

Industry Conventional Alternative Conventional Alternative Stoxx 600

Mean 0.24% 0.25% 0.39% 0.45% 0.18% 0.36%

Maximum 12.92% 9.40% 22.14% 23.04% 29.77% 13.47%

Minimum -16.06% -14.13% -29.44% -11.62% -22.25% -13.27%

S.D. 3.59% 3.23% 8.29% 3.11% 6.72% 4.15%

(16)

4. Vector Autoregression

In this section the vector autoregressive model used is outlined. First the reasons for choosing this model is explained. Second, unit root tests are used to determine the order of integration of the variables. Third, the methods which are used for identifying the optimal lag length are clarified. Fourth, a Lagrange-multiplier test has been conducted to test for serial correlation. Finally, a Granger causality has been used to test if past values of a variable are helpful in predicting the behavior of the other variables.

4.1 The vector autoregressive model

A vector autoregressive model will be used to investigate the correlation between oil prices and European electric utilities as it has several advantages over a multivariate regression. A main advantage of a VAR is that it does not need to provide prior assumptions about which variables are independent and which ones are dependent, because all variables are treated as endogenous. In a VAR all variables depend on the lagged values of all the variables in the formula, which allows for a rich data structure that can capture complex dynamic properties. An unrestricted VAR will be used for the estimations, because it is especially useful for describing the dynamic behavior of economic and financial series. Next to that, it is straightforward to model dynamic relations between economic variables without making many assumptions (Scholtens and Yurtsever, 2012).

The reduced form of the VAR is: ΔXt = A0 + ∑𝑃_𝑖=1Ai ΔXt-i + εt . (9)

Where Xt =(oil price index, interest rate, technology index, industrial production index, and stock industry index), the (5x1) vector of endogenous variables. Ai is the ith (5 x 5) matrix of autoregressive coefficients for i = 1, 2,…, p, and Δ the first difference operator. A0 is a column vector of deterministic constant terms.

(17)

4.2 Unit root tests

The Dickey-Fuller generalized least squares (DF-GLS), Phillips and Perron (PP), and the Augmented Dickey-Fuller (ADF) unit root tests are used to determine the order of integration of the variables. They test the null hypothesis of an unit root against the alternative of stationarity. The null hypotheses of these tests are the presence of a unit root. The alternative hypothesis is stationarity.

The intuition behind the Augmented Dickey-Fuller test is that if the series is integrated, then the level of the series (Xt-1) will provide no relevant information in predicting the change in Xt besides the one obtained in the lagged changes. The more negative the outcome of the ADF is, the stronger the rejection of the hypothesis that a unit root is present at some level of confidence. The AD-GLS is essentially the same as the ADF. The only difference is that the AD-GLS uses generalized least squares. Generalized least squares is a technique for estimating the unknown parameters in a linear regression model and can be used when there is a certain degree of correlation between the residuals in a regression model.

Table 2 - Unit root tests

Levels First difference

(18)

The PP unit root t-test statistic is the DF t-statistic robust to serial correlation, due to the use of Newey-West estimators. This adjustment can change the significance of the statistic. The test is convenient because its results are robust to unspecified autocorrelation and heteroscedasticity.

The results in table 2 above show that the variables are stationary in first differences (I(1)), as the null hypothesis of a unit root can be rejected for all tests. Therefore the VAR will be conducted in first differences. By taking the logarithmic first differences of the variables, ‘spurious regressions’ with no economic meaning are prevented.

4.3 Lag selection

The lag length of the models have been selected on basis of the Akaike Information Criterion (AIC) and the likelihood ratio test. The AIC compares the relative quality of a set of statistical models with each other. It ranks the models with different amount of lags, from best to worst. If we let K be the number of estimated parameters in the model and L the maximum value of the likelihood function for the model. Then the AIC value of the model is: AIC = 2K -2ln(L). The preferred model is the one with the lowest AIC value. Thus, the AIC rewards the goodness of fit, but also takes under and over-fitting into account. For all the models the AIC chose a lag length of 4.

The likelihood ratio test compares two models with each other. One model with p lags to a model with p -1 lags. The hypotheses of the test are shown below.

H0: The VAR model has p = p0 lags.

H1: The VAR model has p = p1 lags where p1 > po

(19)

data, it is like that there are residual seasonal patterns in the data that need to be modeled. For all the models the LR test selected a lag length of 4.

A rule of thumb for selecting the lag length is that it should not exceed one quarter of the degrees of freedom of the model, which means that for the model the lag length should not be greater than 613. There are two reasons for this rule. First, using too many lags may make it impossible

to estimate the coefficient because the OLS estimates cannot be computed. Second, using too many degrees of freedom in the VAR will lead to relative inefficient estimates.

4.4 Lagrange multiplier test for serial correlation

The Lagrange-multiplier test has been conducted on all the models to test for serial correlation in the residuals. The null hypothesis of the Lagrange multiplier test is the absence of autocorrelation at the lag order. Monthly data has been used, therefore the 1, 6 and 12th_order of autocorrelation have been tested. Results can be found in table 3 and show that the null hypothesis of no autocorrelation cannot be rejected under the 5% significance level.

Table 3 - Lagrange-multiplier test for autocorrelation

Lags Chi2 Probability

Equal-weighted -Industry 1 20.7228 0.70796 6 27.9786 0.30884 12 28.2590 0.29607 - Conventional 1 25.6875 0.42442 6 32.9829 0.13150 12 32.2770 0.15015 - Alternative 1 20.6979 0.70933 6 26.1609 0.39905 12 26.9058 0.36061 Cap-weighted - Conventional 1 29.6473 0.23777 6 31.3380 0.17807 12 29.4359 0.24610 - Alternative 1 18.1099 0.83769 6 26.9906 0.35636 12 25.5382 0.43256

The Lagrange-multiplier test, checks for serial correlation in the residuals. The probabilities are calculated from a chi-square distribution with 25 degrees of freedom. The null hypothesis is the absence of autocorrelation at the lag order.

(20)

4.5 Granger causality

When a VAR has been conducted with many lags, it can be difficult to see which variables do have a significant effect on the dependent variables and which ones do not (Brooks, 2008). This problem can be addressed by tests that restrict all the lags of a particular variable to zero. An example of such a test is the Granger causality test, which assesses if past values of a variable are helpful in predicting the behavior of the other variables. The test shows if there is a lead-lag effect present, which is a situation in which one leading variable is cross-correlated with the values of another lagging variables at later times. The null hypothesis of this test is non-causality, which means that one variable cannot be used to determine future values of the other variables. To be clear, in a model variable X granger causes variable Y, if the past behavior of variable X better predicts the behavior of Y than Y’s past values alone.

For testing the causality between the oil price, industrial production, technology, and the interest rate, the industry index has been used. The other utility indices have sequentially been used instead of the industry index to test their causality. The outcomes can be found in Table 4 below.

Table 4 - Granger causality test

Independent

Oil Tech IP R All

D epe nde nt Oil - 5.0403 9.0101* 6.2353 33.593*** Tech 9.5905** - 11.571** 9.3283* 28.209*** IP 4.2604 16.26*** - 13.799*** 72*** R 10.033*** 20.748*** 14.723 *** - 72.047*** Equal-weight Ind. 8.4913** 10.807* 4.8916 18.833*** 33.404*** Con. 11.956** 7.0658 8.9332* 22.576*** 40.401*** Alt. 8.4489** 16.041** 10.666* 17.532*** 44.68*** Cap-weight Con. 8.9298* 4.9577 13.713*** 24.148*** 45.128*** Alt. 13.073** 15.894** 19.743*** 46.123*** 73.213***

The Granger Causality tests whether one variables is able to predict the other variables. The standard model used includes oil, technology, industrial production, and the industry index. The other stock indices have sequentially been used instead of the industry index to test their

(21)

First, the results show that the oil price future index Granger causes the electric utility indices. Second, it shows that the technology stock index is useful in predicting the alternative utility indices. Third, the electric utility indices can be explained by past movements in the interest rate and industrial production. These results are important because it shows that from a Granger causality perspective, the oil price is not the only variable impacting the stock prices of electric utilities.

The technology index is Granger caused by past values of industrial production and the interest rate. The influence of industrial production on technology can be explained by greater demand for technology if industrial production increases. The interest rate can be useful in predicting technology stock prices as lower interest rates increase investments.

Except for industrial production no variable Granger causes oil prices. Past values of industrial production may predict oil prices, because higher industrial production leads to more demand for oil. All the variables are useful in predicting the interest rate. Since interest rates are a lagging economic indicator, this result is consistent with the view that increased economic growth leads to higher interest rates.

5. Results

In this part, first the impact of oil price shocks on the utility indices are discussed. Second, the impact of bigger oil price shocks are examined. Third, the results will be checked for robustness of time and amount of lags used. Last, the responses of the utility indices to technology stock and industrial production shocks will be reviewed.

5.1 Impact of oil price shocks on the utility indices

The impulse response graphs of all shocks are shown in figure 10 in the appendix. The responses are graphed with bands representing 2 standard deviations. The x axis is in months and the y axis can be seen as the percentage change in the dependent variable due to a 1% shock in the independent variable.

(22)

in figure 2 below. In each figure, the dynamic effect of an oil price shock is reported with two standard deviations around the point estimate.

Figure 2 – Cumulative impulse responses of the electric utility indices caused by an 1% oil price shock

A) Conventional (EW) B) Conventional (CW)

C IR F C IR F

Time in months Time in months

C) Alternative (EW) D) Alternative (CW)

C IR F C IR F

Time in months Time in months

E) Industry (EW)

C

IR

F

Time in months

Cumulative impulse responses with probability bands. A and B shows the cumulative response of the conventional indices. C and D show the cumulative responses of the alternative indices. E shows the cumulative response of the industry index (EW). The probability bands denote 2 standard

(23)

Figure 2A and 2B shows that an oil price shock has a negative effect on both the equally and capitalization-weighted conventional electric utility index. The effect is largest after approximately a year, where an oil price shock of 1% has reduced stock prices of conventional electric utilities by more than 0.2%. After one year the effect gradually dies out and conventional stock prices stay permanently reduced. The industry index responds a bit less negative than the conventional index.

The impact of an oil price shock on the stock prices of alternative utilities is insigifnicant. Atlhough we do see an increase in the stock prices for the first months after which the effect becomes negative and dies out. If we compare the results between the alternative and conventional indices it is obvious that the effect is much more negative for the conventional utilities.

Figure 2E shows that an oil price shock of 1% decreases the stock returns of the equally-weighted electric utility index by almost 0.2%. Scholtens and Yurtsever (2012) who studied the impact of an oil price shock on European industries for the period 1983-2007 also found that an oil price shock negatively influences the electricity sector. They found an accumulated negative response of 0.1% after both 12 and 24 months.

To assess whether the impact on the conventional and alternative index is significantly different, a paired t-test has been conducted. A period of 12 months has been used to test for a significant difference between the cumulative impulse responses to an oil price shocks of the conventional and alternative index. The mean difference has been calculated as the mean cirf of the conventional minus the mean cirf of the alternative index. Table 5 below shows that the null hypothesis of no significant difference between the cumulative impulse responses can be rejected under the 1% level.

Table 5 – Paired T-test of the cumulative impulse responses

Mean difference T-statistic Probability

Capital-weighted -0.0893 -10.260 0.000

Equal-weighted -0.0787 -7.320 0.000

(24)

5.2 Impact of bigger oil price shocks

In other literature it was found that stock returns can respond with a different magnitude to an oil price increase and decrease (i.a. Sadorsky, 1999). Next to that, the impact of the shocks may be different if the oil price changes are higher (Narayan and Sharma ,2011). Therefore this section looks at the difference between oil price increases and decreases. The capitalization-weighted conventional (red) and alternative (green) indices have been used to make the graphs in figure 3. The specifications of the oil price changes can be found in the data section.

The graphs shows the absence of a significant difference between an oil price increase and decrease. But if we look at the difference in graphs between oil price changes of more than 0, 1 and 2% a clear difference has been found. Oil price changes between 0 and 2% do not have a significant impact on the stock returns of alternative electric utilities. However, oil price changes of more than 2% do have a significant impact.

In graphs E and F it can be seen that oil price changes higher than 2% do have a significant positive impact on the stock prices of alternative electric utilities. A big oil price increase leads to the appreciation of alternative utilities and vice versa for a big decrease. A possible explanation for this is a higher electricity price caused by the higher oil price. The impact of bigger oil price have an insignificant impact on conventional utilities. An imaginable explanation for this is that conventional utilities do not produce or produce less due to the merit-order effect when fossil fuel prices are high.

Figure 3 – The cumulative impact of positive and negative oil price shocks on the electric utility indices

A) Oil price shock higher than 0% B) Oil price shock lower than 0%

CI

RF

C

IR

F

(25)

5.3 Robustness of oil price shocks

The results found have been subject to a series of robustness tests. First, the robustness of the time serie used has been tested by looking at results for different time spans. Figure 4 on the nexst page shows the impact of an oil shock on the equally-weighted alternative index for different time spans. The solid, dotted, and dashed line represent, respectively, the time periods from 01/2002, 01/2004, and 01/2006 untill 01/2017. The graphs confirms the main result and shows that the alternative index is robust over time.

C) Oil price shock above 1% D) Oil price shock below -1%

CI

RF

C

IR

F

E) Oil price shock above 2% D) Oil price shock below – 2%

CI RF Time in months CI RF Time in months

(26)

Figure 5 shows the cumulative impulse response functions of an oil price shock on the conventional (CW) index for different amount of lags. The solid, dashed, and dotted line represent respectively 4, 5, and 6 lags used. It is clear that the amount of lags chosen is robust as no large differences in the responses hves been found.

Figure 4 – Impulse response functions of the alternative index for different time spans

Figure 5 – Cumulative impulse responses of the conventional index for different amount of lags

C

IRF _CIRF

Impulse response function of the alternative index to an oil price shock for the period from 01/2002, 01/2004, and 01/2006 untill 01/2017 denoted with respectively the solid, dotted and dashed lines.

Cumulative impulse responses of the conventional index for 4, 5 and 6 lags used denoted with respectively the solid, dashed and dotted lines.

5.4 Impact of technology and industrial production shocks

Next to oil price shocks, industrial production, technology, and interest rate shocks have been studied. Figure 10 in the appendix shows the impulse response functions of these shocks. Figure 6 on the next page shows that a technology shock has a positive effect on the alternative indices, but no significant impact on the conventional indices. The positive impact of a technology shock on alternative utilities has been found before and can be confirmed (Sadorsky, 2011; Henriques and Sadorsky, 2007).

(27)

which leads to higher profits. Alternative utilities are dependent on for example the wind, water and the sun and can therefore not suddenly increase their generation, but do profit from higher electricity prices.

Figure 6 – The cumulative impulse responses to a 1% technology shock on conventional and alternative utilities

Figure 7 – The cumulative impulse responses to a 1% industrial production shock on conventional and alternative utilities

Conventional Alternative

CI

RF

CI

RF

Cumulative impulse responses to a 1% technology shock with probability bands. The probability bands denote 2 standard deviations. The X axis is in months and the Y axis is the cumulative impulse

response. Conventional Alternative CI RF CI RF

(28)

6. Conclusion

Growing environmental concerns and costly fossil fuels have led to the booming of investments in the renewable energy sector. The amount of renewable generation is increasing, and dependency on fossil fuels is decreasing. Conventional utilities are investing more and more in renewable production, and renewable ways of generating electricity are becoming more efficient. Therefore identifying the difference in impact of oil price shocks on alternative and conventional electricity utilities is of crucial importance for investors. Next to that, a better understanding of the relationship between the oil price and financial performance of the electricity sector is critical to the development of the sector.

In this paper the impact of oil price changes on the stock returns of alternative and conventional electric utilities is researched. A vector autoregressive model with monthly data for the period January 2002 until July 2017 has been used. In order to estimate the impact, the following variables have been used in the model: 1-month oil future prices, electric utility stock prices, industrial production index, a technology index, and the short-term interest rate. The inclusion of these variables can help us understand the effect of shocks on the electric utility industry and the difference in impact on alternative and conventional companies.

The main finding of this paper is the difference found between the impact of oil price shocks on alternative compared to conventional electric utilities. A significant negative impact of oil price increases on conventional utilities and the electric utility sector has been found. The impact is largest after approximately a year, when an oil price shock of 1% has reduced stock prices of conventional electric utilities by 0.2%. No significant impact on alternative electric utilities has been found when all oil price changes were taken into account. However, for oil price shocks larger than 2%, a positive impact on the stock returns of alternative utilities of almost 0.3% has been found.

(29)

Recommendations

For the coming years, conventional electric utilities will keep their large market share, so the impact of an oil price shock on the electricity industry will stay significantly negative. But the energy transition is ongoing, therefore risks for investors will change. Oil prices will become less relevant when the energy transition continues. Although higher oil price shocks will still have a significant impact. Among other technology stock shocks will become pertinent. Therefore, to minimize risk, it is recommended to investors to hold both conventional and alternative electric utilities to hedge against oil price shocks.

Reflection

(30)

7. Appendix

Table 6 - Key variables

Name Description Abbr. Source

Oil price UK crude oil Brent futures price index (01/2015) deflated by inflation opt Investing.com

Real economic activity Industrial production index (01/2004). Seasonally and calendar adjusted ipt Eurostat

Short-term interest rate Three-month Euribor interest rate rt OECD

Inflation Change in CPI adjusted for taxes and energy prices Πt OECD

Equally-weighted index of stock returns Equally weighted index (01/2015) of the stock returns

Indt (EW)

Thomsonone

Cont (EW)

Altt (EW)

Capitalization-weighted index of stock returns

Capitalization weighted index (01/2015) of the conventional utilities’ stock returns

Cont (EW)

Thomsonone

Altt (EW)

(31)

Table 7 – Conventional electric utilities, market cap in 2017

Table 8 – Alternative electric utilities, market cap in 2017

Conventional utilities First data Market cap total Market cap sector

Energa SA 31-12-2013 0.33% 0.34%

Tauron Polska Energia SA 30-6-2010 0.50% 0.52%

PGE Polska Grupa Energetyczna SA 31-12-2009 1.35% 1.39%

ENEA SA 30-1-2009 0.41% 0.42%

Drax Group Plc 30-6-2008 0.69% 0.71%

EDF (Électricité de France SA) 30-12-2005 3.30% 3.39%

Engie 29-7-2005 13.96% 14.34%

CEZ as 31-1-2002 1.74% 1.79%

Public power corporation 31-12-2001 0.13% 0.13%

EDP - Energias de Portugal SA 31-7-2000 4.02% 4.14%

Iberdrola SA 31-5-2000 22.70% 23.32%

Red eléctrica corp. SA 31-1-2000 4.22% 4.34%

Enel SpA 31-12-1999 21.80% 22.40% RWE AG 30-7-1999 5.89% 6.05% Fortum Oyj 29-1-1999 3.72% 3.83% A2A SpA 31-8-1998 1.26% 1.30% Endesa SA 31-10-1997 3.39% 3.48% SSE Plc 31-10-1997 7.90% 8.12% Total 97.32%

Alternative utilities First data Market cap total Market cap sector

Iniziative Bresciane SpA 31-7-2014 0.04% 1.46%

2Valorise NV 31-5-2013 0.00% 0.10%

Frendy Energy SpA 29-6-2012 0.01% 0.41%

ABO Invest AG 31-5-2011 0.04% 1.60%

Arise AB 31-3-2010 0.16% 6.03%

Edisun Power Europe AG 30-9-2008 0.01% 0.23%

EDP - Renováveis SA 30-6-2008 0.59% 22.05%

Terna Energy SA 30-11-2007 0.09% 3.25%

2G Energy AG 31-7-2007 0.02% 0.90%

Voltalia 31-5-2006 0.07% 2.71%

PNE wind AG 30-11-2005 0.11% 4.01%

Alerion clean power SpA 31-1-2003 0.04% 1.33%

Albioma 29-3-2002 0.26% 9.61%

Falck renewables SpA 28-2-2002 0.09% 3.48%

Verbund AG 31-12-1999 0.68% 25.52%

Arendals Fosseokompani ASA 31-10-1997 0.50% 18.77%

(32)

Figure 8 - Time series plot of the equally weighted indexes

Figure 9 - Time series plot of the capitalization-weighted indexes

Table 9 - Correlation between indices

Equally-Weighted Cap-Weighted

Ind Conv Alt Conv Alt Oil IP Tech R

EW Ind 1 Conv 0.7951*** 1 Alt 0.8418***_0.373*** ₁ CW Conv 0.6428*** 0.8571*** 0.2762*** 1 Alt 0.7643*** 0.4114*** 0.814*** 0.2695*** 1 Oil 0.3163*** 0.2668*** 0.2251*** 0.147** 0.2738*** 1 IP 0.124* _0.1356* _0.0472* _0.1352* _0.1163 _0.1112 ₁ Tech 0.487*** 0.5591*** 0.2886*** 0.5174*** 0.327*** 0.1788** 0.122* 1 R 0.0921 0.1301* _0.0174 _0.1689**_0.0388 _0.1428*_0.4234***_{0.06 1}

The table above shows pairwise correlation coefficients of the first differences of the variables. ** and * denote repectively significance at the 1% and 5% level

0 50 100 150 200 250 300 ja n-02 ok t-02 jul -0 3 apr -0 4 ja n-05 ok t-05 jul -0 6 apr -0 7 ja n-08 ok t-08 jul -0 9 apr -1 0 ja n-11 ok t-11 jul -1 2 apr -1 3 ja n-14 ok t-14 jul -1 5 apr -1 6 ja n-17

Utilities index (cap weighted), 01-2015 = 100

Oil index P. Conventional index P. Alternative index 0 50 100 150 200 250 300 350 feb -0 2 no v-02 aug -0 3 m ei -0 4 feb -0 5 no v-05 aug -0 6 m ei -0 7 feb -0 8 no v-08 aug -0 9 m ei -1 0 feb -1 1 no v-11 aug -1 2 m ei -1 3 feb -1 4 no v-14 aug -1 5 m ei -1 6 feb -1 7

Utilities index (average), 01-2015 = 100

_Conventional index Industry index

(33)

Figure 10 – Impact of oil, technology, industrial production, and interest rate shocks on the electric utility indices R espons e of Indu ds try (E W )_

Oil Tech IP Interest

(34)

8. References

Adkins, L., Carte Hill, R., 2011. Using stata for principles of econometrics 4. John Wiley & Sons Inc, New Jersey.

Alam, M., Uddin, G.S., 2009 . Relationship between interest rate and stock price: Empirical evidence from developed and developing countries. International Journal of Business and Management 4, 43-51.

Armaroli, N., Balzani, V., 2015. Solar Electricity and Solar Fuels: Status and Perspectives in the Context of the Energy Transition. Chemistry – A European Journal 22, 32-57.

Aspergis, N., Miller, S., 2009 . Do structural oil-market shocks affect stock prices? Energy economics 31, 569-575.

Bachmeier, L., Griffin, J., 2006. Testing for Market Integration: Crude Oil, Coal, and Natural Gas. The Energy Journal 27, 55-72.

Basher, S.A., Sadorsky, P., 2006. Oil price risk and emerging stock markets. Global Finance Journal 17, 224-251.

Bencivenga, C., Sargenti, G., D’ecclesia, R., 2010. Energy markets: crucial relationship between prices. Mathematical and Statistical Methods for Actuarial Sciences and Finance, 23-32.

Bjornland, H.C., 2009. Oil price shocks and stock market booms in an oil exporting country. Scottisch Journal of Political Economy 56, 232-254.

Brooks, C., 2008. Introductory Econometrics for Finance 2. Cambridge University Press, Cambridge.

Cunado, J., Perez de Gracia, F., 2014. Oil price shocks and stock market returns: Evidence for some European countries. Energy economics 42, 365-377.

Driesprong, G., Jacobsen, B., Maat, B., 2008. Striking oil: another puzzle. Journal of Financial Economics 89, 307-327.

Elyasiani, E., Mansur, I., Odusami, B., 2011. Oil price shocks and industry stock returns. Energy economics 34, 966-974.

(35)

Faff, R.W., Brailsford, T.J., 1999. Oil price risk and the Australian stock market. Journal of Energy Finance and Development, 69-87.

Frydenberg, S., Onochie, J., Westgaard, S., Midtsund, N., Ueland, H., 2014. Long-term relationships between electricity and oil, gas and coal future prices—evidence from Nordic countries, Continental Europe and the United Kingdom. OPEC energy review 38, 216-242. Gielen, D., Saygin, D., Wagner, N., 2015. Synergies between renewable energy and energy efficiency: A working paper based on remap 2030. International Renewable Energy Agency. Hamilton, J., 2004. Oil shocks and aggregate macroeconomic behavior: the role of monetary policy. Journal of Money 36, 265–286.

Hartley, P., Medlock, K., Rosthal, J., 2008. The relationship of natural gas to oil prices. The Energy Journal 3, 47-65.

Henriques, I., Sadorsky, P., 2008. Oil prices and the stock prices of alternative energy companies. Energy economics 30, 998-1010.

Huang, R., Masulis, R., Stoll, H., 1996. Energy shocks and financial markets. The Journal of Futures Markets 16, 1-27.

Isaac, G., 2012. The benefits of the new economy: Resolving the global economic crisis through mutual guarantee. ARI publishers, Canada.

Ivanov, V., Killian, L., 2005. A practitioner’s guide to lag order selection for VAR impulse response analysis. Studies in Nonlinear Dynamics & Econometrics 9. 2-34.

Killian, L., Park, C., 2009. The impact of oil price shocks on the U.S. stock market. International economic review 50, 1267-1287.

Lee, B., Yang, C., Huang, B., 2012. Oil price movements and stock markets revisited: A case of sector stock price indexes in the G-7 countries. Energy economics 34, 1284-1300.

Manzoor, D., Seiflou, S., 2011. Are crude Oil, Gas and Coal Prices Cointegrated? Iranian economic review 16, 29-51.

Mohamaddi, H., 2009. Electricity prices and fuel costs: Long-run relations and short-run dynamics. Energy economics 31, 503-509.

(36)

Oberndorfer, U., 2009. Energy prices, volatility, and the stock market: Evidence from the Eurozone. Energy policy 37, 5787-5795.

Papapetrou, E., 2001. Oil price shocks, stock market, economic activity and employment in Greece. Energy economics 23, 511-532.

Sadorsky, P., 1999. Oil price shocks and stock market activity. Energy economics 21, 449-469. Sadorsky, P., 2012. Correlations and volatility spillovers between oil prices and the stock prices of clean energy and technology companies. Energy economics 34, 248-255.

Scholtens, B., Yurtsever C., 2012. Oil price shocks and European industries. Energy economics 34, 1187-1195.

Villar, J.A., Joutz, F., 2006. The relationship between crude oil and natural gas prices. Energy information administration, office of oil and gas, 1-43.

The impact of oil price shocks on the stock returns of Alternative and Conventional European electric utilities