Determinants of returns on renewable energy stocks : an empirical study using a multifactor linear regression model

(1)

1

Determinants of returns on renewable energy stocks,

an empirical study using a multifactor linear regression model

Victor Obermüller, 11087617

University of Amsterdam, Spui 21, 1012 WX Amsterdam, The Netherlands

Presented to the faculty of Economics & Business on the 31.01.2018 in partial fulfilment of the requirement for the degree

Bachelor of Science in Economics and Business (track: Finance & Organization)

Supervised by Dr. Philippe Versijp

Abstract

This study investigates the effect of oil prices, gas prices, technology stock prices, and interest rates on the performance of clean energy stocks. While it is widely accepted that oil prices influence returns on clean energy stocks, there has been little research on other energy-related variables, such as gas prices. The novelty of this research is adding gas prices to a multifactor linear regression model that uses time series data in four different time regimes (over the period of 2001-2017). While the data fails to demonstrate a significant effect of gas prices, significance on the effect of technology stocks and oil prices can be found. However, in the time regime from 2012 to 2017 no significance of any explanatory variable can be found which suggests a cyclical change in the business conditions of the renewable energy sector.

(2)

2

Statement of Originality

This is to certify that to the best of my knowledge, the content of this thesis is my own work. This thesis has not been submitted for any degree or other purposes. I certify that the

intellectual content of this thesis is the product of my own work and that all the assistance received in preparing this thesis and sources have been acknowledged.

(3)

3

This quote represents the forum of discussion observed in recent years. The terms ‘renewable1’, ‘clean2’, and ‘alternative’ have become increasingly relevant when we think of where we draw our energy from. This is primarily due to concerns about climate change and a general environmental consciousness among consumers and investors. Hence, the clean energy sector has accomplished a remarkable growth over the last decade. In terms of numbers, since 2004, investments in renewable energy increased on average by 15.56% annually, reaching a peak of $349 billion in 2015 (Bloomberg New Energy Finance, 2017). Furthermore, the figure is expected to increase to around $400 billion per year by 2040. According to the International Energy Outlook, this would make renewable energy globally the fastest growing energy source between 2015 and 2040 (U.S. Energy Information Administration, 2017).

Given these significant developments, investors have been attracted and gained a strong interest in what exactly drives the price of renewable energy stocks. With accurate estimates of correlation coefficients, investors are able to optimize their portfolio, hedge against fluctuations, and derive prices. Such analytical techniques lie at the heart of modern finance. Thus, it becomes academically relevant to examine the main drivers of the financial performance of renewable energy companies. However, the factors influencing the clean energy sector are often complex and counterintuitive. To date, only very little is known about the interaction between the performance of clean energy stocks and individual macroeconomic factors. This study aims to fill this gap.

Besides the fact that consumers become environmentally more aware, energy security issues further motivate a substitutive movement towards renewable energy sources. Rapidly increasing demand for fossil energy sources in emerging markets, combined with supply

1,2 _{Most renewable energy sources are considered as clean energy sources. In this paper, the term will be} used interchangeably.

(5)

5 shortages that arise from geopolitical tensions, will accentuate concerns about dependency on fossil sources. Thus, fossil energy prices are predicted to increase in upcoming years, which should stimulate the demand and supply of renewable energy (U.S. Energy Information Administration, 2018). Therefore, this study considers the prices of fossil sources, such as crude oil and natural gas, as potential drivers of clean energy stocks’ performance.

Research by Henriques and Sadorsky (2008), which is the closest identified study to this present paper, uses a vector autoregressive model (VAR) to analyse the interaction between alternative energy stocks, oil prices, interest rates, and a proxy for technology companies. However, the authors did not incorporate gas prices. This paper builds on this approach by also taking into account natural gas prices as a variable. Moreover, the time frame considered in previous studies will be extended significantly - up to 2017 - and divided into four different time regimes. A multifactor linear regression model is used to ultimately answer the question: to what extent can oil and natural gas prices determine returns from clean energy stocks?

The purpose of this paper is to provide an empirical analysis of the relationship between returns of clean energy stock prices, oil prices, natural gas prices, short term interest rates, and stock prices of technology companies. Additionally, it the model controls for market fluctuations. All variables are expressed in logarithmic returns to reduce adverse variability.

The remainder of this study is organized as follows. The following section consists of a literature review that addresses earlier research on the same topic and derives hypotheses that will be tested. This is followed by a ‘Methodology’ section, which discusses the data used and provides a description of the model. In the ‘Empirical Results and Analysis’ section, results will be presented, followed by a discussion that embeds this research’s findings in the current literature. Concluding remarks in section five will eventually answer the aforementioned research question and provide recommendations for further research.

(6)

6

2. Literature Study

To date, there is little research investigating the link between clean energy stock prices and various macroeconomic variables. Recognizing the importance of this dynamic energy sector, the following chapter reviews the current literature on the topic and defines the requirement for a model that determines returns of clean energy stocks in a more comprehensive way.

2.1 Oil

There has been extensive literature about the effect of changing oil prices on stock prices. Previous literature by Hamilton (2003) indicates that rising oil prices negatively affect the economy and financial markets. Data between 1949 and 2001 shows that most of the U.S recessions are primarily results of inflationary oil price upsurges (oil shocks). Rising oil prices increase production expenses of goods, lower cash flow, and decrease stock prices (Darby, 1982). On the other hand, more recent literature concluded a positive relationship between rising oil prices and stock prices, especially when it comes to clean energy stocks (Cunado and Perez, 2005; Henriques and Sadorsky, 2008). These contradictory findings suggest that there are some industry sectors that benefit from higher oil prices, whereas others do not. One example of a positive relationship could be the renewable energy industry.

In order to determine how the renewable energy industry possibly benefits from changes in oil prices, Henriques and Sadorsky (2008) employed a VAR approach in their research paper. Using data from 2001 to 2007, the model shows that rising oil prices have a significantly positive effect on clean energy stock returns. However, the effect is rather weak with a correlation coefficient of 0.11. More explanatory significance was given to the relationship between technology stock prices and clean energy stock prices. As a possible explanation for the oil price effect, the authors suggest that inflationary oil prices motivate a substitutive movement towards non-fossil based energy sources (Henriques and Sadorsky, 2008). Nevertheless, concluding that shocks to oil prices have only a small effect on clean energy stock prices encourages further research.

Kumar et al. (2012) repeated the analysis of Henriques and Sadorsky (2008) and expanded its focus. By using the same methodological approach, the author extended the dataset up to 2008 and added carbon prices as an explanatory variable. Kumar et. al (2012) hypothesize that the price per ton of CO2 emitted (carbon price) drives conventional energy

(7)

7 prices and therefore has an impact on the demand of renewables. While the authors did not find a significant effect of carbon prices, they did, however, find a significant positive effect of oil prices on the returns of clean energy stocks. The correlation coefficient for three different clean energy indices was found to range from 0.31 to 0.59. It indicates that extending the sample period from 2007 up to 2008 gave oil prices a stronger explanatory power for determining the returns of clean energy stocks.

These results are in line with the findings of Sadorsky (2012), who confirms a significant relationship between oil prices and clean energy stock prices with a coefficient of 0.28, between the years 2002 and 2010. However, this leads to the question of why the relationship has been considerably weaker before 2007 and stronger afterwards (Henriques and Sadorsky 2008, Kumar et. al, 2012).

Managi and Okimoto (2013) address this contradiction of strength among the variables oil and clean energy stocks. By using a Markov-Switching (MS) VAR model, the paper aims to explain interactions between oil prices and clean energy stock prices and considers the possibility of certain asymmetric effects between the variables. Managi and Okimoto (2013) hypothesize that changes in the relation between oil prices and clean energy stock prices are due to a structural break at the end of 2007. This was the period when the US economy entered a recession with a simultaneous surge in the price of oil. In contrast to previous studies, Managi and Okimoto (2013) found a positive relationship between oil prices and clean energy prices after a structural break, in the period from 2007 to 2010.

The result adds to the existing literature and indicates that the contradiction between the strong relation found by Kumar et. al (2011) and the weak relation found by Henriques and Sadorsky (2008) is due to a structural break in the price of oil. Furthermore, it suggests a substitutional movement from conventional energy to clean energy, where structural changes in oil prices are possibly a major cause of the shift. However, this more of a permanent process than a transitory phenomenon (Managi and Okimoto, 2013).

Perhaps the most recent paper about this topic is that of Inchaupse, Ripple, and Trück (2015). The authors further extend the data set up to 2014, which contains twelve years of observations, including the time of the Great Recession (2007-2012) and the two following years. By using a multi-factor asset pricing model with time-varying coefficients, Inchaupse et al. (2015) confirm Managi and Okimotos’ (2013) result regarding the changing nature of the link between the oil price and clean energy stocks. The beta-factor was found to range

(8)

8 between 0.01 and 0.1, having a peak in June 2008, when the oil price had a strong surge up to more than $140/barrel. Consequently, Inchaupse et al. (2015) hypothesize that in times of high oil prices, oil has more explanatory significance as a determinant of clean energy stocks. Moreover, a substitutional movement from oil towards renewable energy is supported when oil prices rise.

To illustrate the findings from the previous line of research:

*chronically sorted by years **significant results

This study builds on the current line of research by significantly extending the time frame considered in previous studies up to 2017 (Henriques and Sadorsky, 2008; Kumar et al., 2012, Managi and Okimoto 2013; Sadorsky, 2012; Inchaupse, 2015). Furthermore, a changing nature of the relationship between clean energy stocks and oil prices is expected and therefore will be distinguished between four different time regimes: (A) the total period of 01/2001 - 12/2017; (B) the period of substantial surge in the oil price, 01/2001 - 06/2007; (C) the period of the Great Recession, 06/2007- 12/2012; (D) the long-term recovery of the financial markets until the present point in time, 01/2013 - 12/2017. Given the results of discussed literature, a significant positive relationship between oil prices and clean energy stocks is hypothesized in all three periods until the end of 2012. For the last period, a non-significant relationship is hypothesized and motivated in the following paragraph. The following null-hypothesis is derived:

H1: Oil price returns do not affect clean energy stock returns, in period (A), (B), and (C).

Author Time frame* Relationship between oil prices

and clean energy stocks** Henriques & Sadorsky (2008) 2001-2007 positive

Kumar et al. (2012) 2001-2008 positive

Sadorsky (2012) 2002-2010 positive

Managi & Okimoto (2013) 2007-2010 positive (after structural break) Inchaupse et al. (2015) 2002-2014 positive (time-varying beta

(9)

9

2.2 Substitution Effect & Natural Gas

Comparing Fig. 1 and Fig. 2 gives useful insights in the dynamics between oil prices and the clean energy sector (see Appendix 7.1, p.28). Fig 1. shows the quarterly amount of global financial investments in renewable energy by sector. It can be observed that the average annual rate peaked in 2008 and 2011 with a consecutive downwards trend. As shown in Fig 2., both points in time were periods with downward trends in the oil price. The given commonality indicates that high oil prices seem to boost investment in renewables, while cheap oil decelerates the demand for renewable energy. This mechanism, however, appears to be disrupted from 2014 onwards, when low oil prices did not slow down investment in renewable energy.

Henriques and Sadorsky (2008), Managi and Okimoto (2013), Sadorsky (2012), and Inchaupse (2015) do have similar findings in their research and explain such a causality with a substitution effect. As Sadorsky (2012) stresses in his paper, it is common media wisdom that oil prices are an important driver of clean energy stock prices. Henriques and Sadorsky (2008) describe in more detail that in times of rising oil prices, consumers and investors perceive a strong stimulus for substituting fossil based energy sources with renewables. However, it is a rather imperfect substitution away from conventional energy sources. The imperfect relation arises because investors do not value a surge and a fall in the oil price equally. As Managi and Okimoto (2013) state, an increase in oil prices has a larger effect than a decrease in oil prices. But why does investors’ attitude change towards clean energy sources when oil prices rise?

It is widely accepted that oil prices represent the price of energy in general. Investors associate high oil prices with high electricity prices from conventional sources, which consequently improves the competitive position of renewable energy sources from an economic point of view (Rojas and Stinson, 2015). Though this logic may seem valid at first, it is actually a misconception.

Oil and clean energy operate in different markets (Nyquist, 2015). Whereas oil is primarily used for various kinds of transportation fuel, only very little is used to generate power. In fact, oil accounts for less than 1 percent of power generation in the US and Canada. Globally, the figure is around 5%. In contrast, natural gas is a major player in power generation (27 percent in the United States and 18.6 percent in Europe) and in effect it is about to become the floor price for power (International Energy Agency, 2016). Assuming the real threat to integrating clean energy into our power networks is coming from cheap

(10)

10 electricity rather than from cheap oil, natural gas could be a stronger challenger. The historic correlation between the price of oil and the demand for renewable energy appears to have weakened in today’s global markets. Therefore, this study assumes that the link between the financial performance of clean energy stocks and oil prices has become weaker in period (D), i.e. 2013-2017. Instead, this paper hypothesizes a significant correlation coefficient between renewable energy stock returns and natural gas prices. The following null-hypotheses are derived:

H2: Oil price returns affect clean energy stock returns in period (D).

H3: Natural Gas price returns do not affect clean energy stock returns in period (D).

2.3 Technology Stocks

Considering the discussed studies, in all past cases technology shares were included in the regression model. This motivates the decision to include an index of technology stocks in the proposed model of this present study, suggesting that some investors might see clean energy firms and high-technology companies as entities of the same asset class (Inchaupse et al., 2015). Furthermore, most of the previous studies indicate a significant positive effect of technology stock prices on clean energy stock prices. The correlation coefficients range from 0.61 up to 0.88 respectively, which suggests stronger explanatory power of technology stock prices compared to oil prices (Henriques & Sadorsky, 2008; Kumar et al., 2012; Inchaupse et al., 2015; Sadorsky, 2012; Managi & Okimoto, 2013). According to Sadorsky (2012), a high correlation coefficient could possibly be explained by the fact that technology stocks and clean energy stocks share certain commonalities. Both sectors often rely on the success or failure of a specific technology. This is further reflected in the comparable high amounts that both sectors spend on Research & Development (R&D). Moreover, Inchaupse et al. (2015) explain that both technology and clean energy companies depend on similar resources, such as high-skilled labor force, researching facilities, semiconductors (solar). If a fundamental technology fails or a shortage of these inputs arises, both sectors will face similar risks, so that stock price movements are correlated. Hence, this paper hypothesizes a relationship between clean energy stock returns and technology stock returns for the periods (A), (B), and (C) (2001-2017, 2001-2007, 2007-2012). This results in the following null-hypothesis:

H4: Technology stock returns do not affect clean energy stock returns in the periods (A), (B),

(11)

11 However, Henriques and Sadorsky (2008) recognize the rapidly changing nature of the renewable energy sector and expect a transition movement from technology companies towards mature companies. In other words, when a clean energy company will reach mass adoption, it will be seen as an energy company in a mature stage. Since mature companies tend to face lower risks than companies in an earlier stage of the business cycle, this implies a weaker relationship between clean energy stock prices and technology stock prices (Wüstenhagen & Menichetti, 2011).

The renewable energy market is rapidly evolving through its strong expansions and growth. Global grid installments of renewables increased from 2011 to 2014 by 40 percent, which significantly improves the economics of renewables. In production, economies of scale are increasingly present and continue to drive down costs (Nyquist, 2015).

These are developments towards a mature stage of a company and support the hypothesis of a weakening relationship between clean energy stock returns and technology stock returns in the most recent fourth period (2013-2017). Consequently, the following null-hypothesis is derived:

H5: Technology stock returns affect clean energy stock returns in period (D).

2.4 Interest Rate

Previous business cycle research has shown the importance of short-term interest rates in indicating recent and future economic and financial growth. More precisely, a significant relationship has been shown between interest rates and stock price movement of energy companies (Chen, 1991; Sadorsky, 2001, Kumar et al., 2015). Therefore, interest rate will be included as a variable in this research’s model. The following null-hypothesis is derived:

(12)

12

2.5 Conceptual Model

This paper attempts to test conclusions drawn from the literature study, which are summarized below as null-hypotheses:

H1: Oil price returns do not affect clean energy stock returns, in period (A), (B), and (C).

H2: Oil price returns affect clean energy stock returns in period (D).

H3: Natural Gas price returns do not affect clean energy stock returns in period (D).

H4: Technology stock returns do not affect clean energy stock returns in the periods (A), (B),

and (C).

H5: Technology stock returns affect clean energy stock returns in period (D).

(13)

13

3. Research Methodology

In this section, a detailed description of the empirical model is explained, including the data collection and sampling procedure.

3.1 Data

In this research, 4424 daily observations from 01/01/2001 until 31/12/2017 will be used, retrieved from DataStream®. Weekend days are not included, since data is only available for trading days from Monday until Friday. The variables considered in the model are the following: the stock index of clean energy, expressed as NEX, the technology stock price index PSE, Oil prices, natural GAS prices, and short-term interest rates (rf). Furthermore, a variable to control for the global market fluctuations will be added, expressed by MSCI. The following describes each variable in more detail.

3.1.1 NEX

The WilderHill New Energy Global Innovation Index (NEX) is an accurate global benchmark of the financial performance of the clean energy sector. The NEX was chosen in line with previous research (Inchaupse et al., 2015; Kumar et al., 2012) and can be viewed as the first and leading global benchmark for renewable, clean, and alternative energy stocks. The index is a modified dollar-weighted benchmark, composed of 94 publicly traded companies in 22 countries. Furthermore, it is well-diversified across individual sub-sectors of the renewable energy industry. Its main components are ‘Energy Efficiency’ companies with 30.67%. This sector includes firms focusing on improvements in efficiency of the existing power generation and distribution systems. The sector is closely followed by ‘Solar’ and ‘Wind’, with 23.37% and 25.66% respectively (retrieved from http://www.nexindex.com/). Hence, the focus of the NEX lies on clean power generation and efficiency enhancement of these sources. The exact composition in Q1 of 2018 is illustrated in Fig. 3.

(14)

14

(Source: http://www.nexindex.com/Constituents_And_Weightings.php)

3.1.2 PSE

The Arca Tech 100 Index (PSE) is a pure play technology index, compromising companies across a broad spectrum of industries, which are listed on leading stock exchanges and OTC facilities. The dollar-weighted index provides an accurate benchmark for the financial performance of technology stocks across 15 different industries. Given the commonalities between technology companies and renewable energy firms (Inchaupse et al., 2015), it is worth commenting on the composition of both indices. Despite their broad diversification, the PSE and NEX do not list similar companies (retrieved from

https://www.nyse.com/quote/index/PSE). This enables a consideration of both indices as explanatory variables.

3.1.3 Oil

Oil is associated with conventional fossil fuel energy. Therefore, it is suitable for this study in order to examine the general impact of oil on clean energy stocks and a substitutional movement from conventional energy sources towards renewable energy generation. Moreover, it is the most widely traded physical commodity in the world (Henriques and Sadorsky, 2012). In this paper, the oil price is measured using the spot price of the West Texas Intermediate (WTI) crude oil, determined by the major trading hub for WTI in Cushing, Oklahoma. The WTI is a grade of crude oil and commonly used as an accurate benchmark in oil pricing.

3.1.4 Natural Gas

In this study, the Henry Hub Spot Price of natural gas will be used as a benchmark for gas pricing. It can be considered as adequate proxy for the global gas market, because it is the

Fig.3 - Sector Weights

Key Sectors

Index Sector Weights - To Start Q1 2018*

ECV Energy Conversion 1.00%

EEF Energy Efficiency 30.67%

ENS Energy Storage 7.64%

RBB Renewables - Biofuels & Biomass 7.74%

ROH Renewables - Other 3.92%

RSR Renewable - Solar 23.37%

(15)

15 pricing point for natural gas futures on the New York Mercantile Exchange (NYMEX). These futures are the third-largest traded physical commodity contracts in the world.

3.1.4 Interest Rate

Since previous research found a significant relationship between stock price movements and short term interest rates (Chen, 1991; Sadorsky, 2001), this paper will use the yield on a 3-month US Treasury bill as a variable to investigate the link between clean energy stocks and interest rates.

3.1.5 Market

The explanatory power of the global market movements is not of primary interest to this paper. However, in order to test the relative relationship between the other independent variables and the clean energy stock returns, it will be controlled for market fluctuations. As a proxy for global market movements, the MSCI World Equity Index will be used. The Index measures the market performance of 4,500 large and mid-cap companies that have a global presence. Often, it is described by financial media as a benchmark for the world’s stock markets. However, it does not include stocks from emerging market countries (https://www.msci.com/world).

3.1.6 Sample Period

In this study, the data will be analyzed by distinguishing between four different time regimes. This is motivated by the assumption of a structural break in 2008 which was found by previous studies (Managi & Okimoto, 2013; Inchaupse et al., 2015). Additionally, running individual regressions of different time periods is expected to provide more insights into the future development of the link between clean energy stock returns and macroeconomic variables. The first period (A) contains the total sample from 01/01/2001 to 29/01/2017. The second period (B) includes 1684 observations from 01/01/2001 to 15/06/2007. The third period (C) includes 1444 observations over the time of the Great Recession, from 16/06/2007 to 31/12/2012. The fourth sample period (D) contains 1296 data points until the present point in time, from 01/01/2013 to 29/12/2017.

(16)

16

3.2 Empirical Model

A multiple linear regression model will be used to investigate the link among clean energy stock returns and the following explanatory variables. The sample set consists of time series data. Regression estimates are predicted for each time regime, (A), (B), (C), and (D). The regression model looks as follows:

ln_NEX = 𝛼0 + 𝛽1 ln_MSCIt + 𝛽2 ln_rft + 𝛽3 ln_PSEt + 𝛽4 ln_Oilt + 𝛽5 ln_GASt + 𝜀

where:

ln_NEXt = logarithmic rate of return of The WilderHill New Energy Global Innovation Index.

ln_MSCIt = logarithmic rate of return of the MSCI World Equity Index.

ln_rft = logarithmic rate of return of the three-months yield on a US Treasury bill.

ln_PSEt = logarithmic rate of return of the Arca Tech 100 Index (PSE).

ln_Oilt = logarithmic rate of return of the West Texas Intermediate (WTI) spot price.

ln_GASt = logarithmic rate of return of the Henry Hub Natural Gas spot price. 𝜀 = residual.

3.2.1 Explanatory and Control variables

The given model can be viewed as a log-log model, both depended and independent variables are transformed to natural logarithmic returns. The explanatory variables - rft, PSEt, Oilt, GASt - are calculated by taking the log of all daily returns in order to reduce unwanted variability (heteroscedasticity). Moreover, the use of log returns creates a more comparable metric for measurements. Also, the dependent variable, NEX, is expressed in logarithmic returns. Logarithmic returns for various prices at time t are calculated by:

(1) ln 𝑃𝑡+1

𝑃𝑡 = 𝑟/

Besides using logarithmic rates of return, this study runs the same model with arithmetic returns. This is primarily done in order to reduce for uncertainty and check for robustness of the empirical results. Additionally, the model controls for global financial market movements by considering daily returns of the MSCIt as control variable.

(17)

17 3.2.2 Testing for Multicollinearity

Given similar characteristics of the two broad indices, MSCIt and PSEt, this paper tests for possible multicollinearity among the explanatory variables. Both benchmarks include a broad range of international companies in developed markets. In a few instances, both indices list the same companies (e.g. Apple, Microsoft), and thus, could be too closely related.

Consequently, a test for multicollinearity is conducted, using variance inflation factors

(VIFi). The test quantifies how much variances is inflated, which happens when

multicollinearity exists. To calculate the VIFi for 𝛽i the following formula is used:

(2) 𝑉𝐼𝐹₃ = 1 1−𝑅_𝑖2

3.2.3 Robust Standard Errors

To avoid that standard errors of correlation coefficients are biased and inconsistent, this paper will use Huber-White’s Robust Standard Errors approach for all conducted regression analyses. This technique obtains standard errors which can be considered as heteroscedasticity-consistent. This is done in order to allow the model for heteroscedastic residuals.

(18)

18

4. Empirical Results and Analysis

In this section, empirical results are analysed, presented, and interpreted. Moreover, comparisons with previous research are discussed. In total, four OLS regression analyses are conducted.

4.1 Descriptive Statistics

Table 1 provides descriptive statistics for each discussed variable. The standard deviation represents the volatility of the log returns which, as seen in Table 1, are proven to be relatively high for the variables ln_GAS and ln_rf, with 0.416 and 0.231 respectively. This implies that the yield on a 3-months U.S. treasury bill and the returns on natural gas are, on average, riskier than the other variables, including the global market. This is further illustrated by the values for minimum and maximum observations. Both variables have considerably larger or smaller observations, for both, minimum and maximum values (ranging from -0.569 to 0.623 and -2.732 to 1.897 respectively). These anomalies might be due to severe fluctuations in the years of the financial crisis, in particular 2008. This is visible in the scatter plots of both variables’ logarithmic returns (see Appendix 7.2.2, p.31).

Table 1. Summary statistics of daily returns1

N Mean Std. Dev. Min Max

ln_NEX 4434 4.73e-05 0.0141 -0.105 0.121 ln_MSCI ln_PSE 4434 4434 0.000123 0.000280 0.0100 0.0141 -0.0732 -0.0812 0.0910 0.108 ln_OIL 4434 0.000184 0.0238 -0.171 0.164 ln_GAS ln_rf 4443 4434 -0.000285 -0.000322 0.416 0.231 -0.569 -2.732 0.623 1.897 1_{Figures are presented as continuously compounded daily returns (01/01/2001 to 31/12/20017).}

The correlation matrix in Table 2 shows that the MSCI and PSE are strongly correlated with returns on clean energy stocks (NEX), with coefficients of 0.827 and 0.734 respectively. This indicates that technology stocks and clean energy stocks tend to move closely together which may be due to commonalities of both sectors. A strong correlation with the MSCI, however, was expected in order to control for global market movements. Considering the energy related variables Oil and GAS, while oil price returns are moderately correlated with clean energy

(19)

19 stock returns, gas price returns just find a weak correlation. This is not in line with earlier expectations of this paper, where a stronger correlation with gas prices compared to oil prices was hypothesized. However, this is calling for further analyses to better understand both relationships.

Table 2. Correlation coefficients1 between the independent variables

ln_NEX ln_MSCI ln_PSE ln_rf ln_Oil ln_GAS

ln_NEX 1 ln_MSCI 0.8768 1 ln_PSE 0.7338 0.8512 1 ln_rf 0.0536 0.0919 0.0980 1 ln_Oil 0.4107 0.4436 0.3290 0.0322 1 ln_GAS 0.0659 0.0610 0.0157 0.0283 0.0577 1

1_{Correlation coefficients are measured on the base of the (1) time regime (01/01/2001 to 31/12/20017).}

The strong correlation between MSCI and PSE supports the claim that due to similar characteristics of both indices collinearity might arise. Therefore, a test for multicollinearity is conducted, as illustrated in Table 3. VIF stand for variance inflation factor, whereas 1/VIF defines the tolerance. A tolerance value lower than 0.1 could be interpreted as one variable is considered a linear combination of another independent variable. However, the results are higher than 0.3425 and indicate there is no multicollinearity present in this model.

Table 3. Test for Multicollinearity

Variable VIF 1/VIF1

ln_MSCI 2,92 0,342451 ln_PSE 2,73 0,366804 ln_Oil 1,12 0,892774 ln_rf 1,01 0,882999 ln_GAS 1,01 0,994054 Mean VIF 1,76

(20)

20

4.2 Regression Results and Discussion

For each of the four time regimes an OLS regression is conducted. The results are summarized in Table 4. Since this paper is using a log-log regression model, the coefficients are to interpret as follows. For example, a 1% increase in the variable MSCI increases the

returns on clean energy stocks (NEX) by 1,073% (𝑒9.;<=∗?@A 9.;9 _{= 1,0107). The results are}

robust across all explanatory variables. The R-squared of the different time regimes ranges from 59,2% up to 79,5%, indicating that most of the variation in clean energy stock returns is explained by the model. Furthermore, the control variable MSCI is found to be significant in all four regimes.

4.2.1 Oil

The first two hypotheses, regarding the effect of oil on clean energy stocks, can be rejected. A significant positive relationship between returns on clean energy stocks and oil price returns is

Table 4. Linear regression analysis, logarithmic daily returns

Time Regime1 (A)

VARIABLES ln_NEX ln_NEX ln_NEX ln_NEX

ln_MSCI 1.073*** 0.719*** 1.324*** 1.058*** (0.0306) (0.0438) (0.0441) (0.0552) ln_rf -0.00131** -0.0184 -0.00266*** 9.24e-05 (0.000595) (0.0120) (0.000908) (0.000630) ln_PSE 0.0776*** 0.231*** -0.0996*** 0.0253 (0.0192) (0.0214) (0.0356) (0.0421) ln_Oil 0.0280*** 0.0219*** 0.0245** 0.0141 (0.00590) (0.00832) (0.0115) (0.0103) ln_GAS 0.000873 -0.00267 0.00572 -0.000645 (0.00266) (0.00332) (0.00673) (0.00444) Constant -0.000114 0.000324* -0.000600*** 1.89e-05 (0.000116) (0.000184) (0.000226) (0.000166) Observations 4,424 1,684 1,444 1,296 R-squared 0.702 0.605 0.795 0.592

Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1 (Time regimes1_{described exactly in section 3.1.6)}

(21)

21

found in the first three time regimes, (A), (B), and (C). However, in period (D) there is no significant relationship. Thus, H1 and H2 are rejected. This is in line with previous research by Henriques and Sadorsky (2008) who found a weak positive relationship between 2001 and 2007. Therefore, this paper confirms a positive relationship in the period (B) from 2001 until 06/2007, with a weak coefficient of 0.0219. This period describes the time before the Great Recession coinciding with an upsurge of the oil price. This supports the claim of Inchaupse et al. (2015) that rising oil prices have an impact on clean energy stock performance. Furthermore, a significant positive impact of oil on clean energy stocks is found in period (C), with a coefficient of 0.0219. This is in line with Managi and Okimoto (2013) and Inchaupse et al. (2015) who found a positive relationship after a structural break in the beginning of 2008. However, as contribution to exciting literature, this paper hypothesizes that the relationship between oil price returns and clean energy stock returns weakens as observations getting closer to the present point in time. This is represented in time regime (D). As

expected, no significant relationship can be found. Thus, H2 is rejected. This may be due to a

change in investors perspectives about the operating locus of oil and renewables. As indicated earlier, both variables operate in different markets which consequently weakens the relationship. While only very little oil is used to generate power, the primary focus of renewables is producing power. Furthermore, the economics of renewables are improving (i.e. production efficiencies and cost reductions) which increases grid installments and enhances

the independence of renewables towards conventional energy sources (Nyquist, 2015).

4.2.2 Natural Gas

Considering the previous line of research, this paper was the first study that attempts to examine an effect of natural gas on clean energy stocks. However, in contrast to the expected positive impact, the regression shows no significant relationship in all four time regimes. Consequently, H3 cannot be rejected.

Although, natural gas supplies 22% of the energy used globally it does not seem that cheap gas prices of the recent years influence the financial performance of clean energy stocks (see Appendix 7.1.2, p.29), (International Energy Agency, 2017). Furthermore, this finding contests the assumption of renewables being influenced by electricity prices, and thus linked to gas prices. This result could be possibly explained by aggressive governmental policies which consistently subsidize investments in renewable energy and promote grid

(22)

22

installments. Consequently, renewable energy stocks appear to become less sensitive towards energy related variables in general.

4.2.3 Technology Stocks

The regression shows a significant effect of technology stocks on clean energy stocks in the periods (A), (B), and (C). Thus, H4 is rejected. This is in line with previous research by

Henriques & Sadorsky (2008), Kumar et al. (2012), Inchaupse et al. (2015), Sadorsky (2012), and Managi & Okimoto (2013). While the coefficient during the Great Recession (C) was negative and weak with -0.0996, the coefficient for the prior years between 2001-2007 was strong with 0.231. This change in sign may be due to severe fluctuations during the financial crisis which influenced the relationship between both variables. A relationship is generally supported by the common characteristics of both sectors, such as similar inputs and dependence on success or failure of individual technologies.

However, during the last period (D) from 2012 until 2017, a slightly positive, but not significant coefficient was found. Hence, the regression failed to establish a link between his technology stocks and clean energy stock performance in the period (D). Consequently, H5 can be rejected. This supports the earlier expectations of Henriques and Sadorsky (2008) and Wüstenhagen & Menichetti (2011). The authors recognized the fast-changing nature of the renewable energy sector and forecast a transformation process from technological stage towards a mature phase of the sector. This happens through enhancements in technologies, production and cost efficiencies. According to Wüstenhagen & Menichetti (2011), a mature company tend to face lower risk than a firm in a growth stage, such as technology companies. This implies that the link between both variables weakens which is finally supported by the findings of this model.

4.2.4 Interest Rate

In this study, no consistency can be found of the link between clean energy stock returns and interest rate changes. While over the total period (A) a significant negative effect was found, the model failed to proof a relationship between both variables in the periods (B) and (D). Furthermore, the significant effects in period (A) and (C) are negative and weak with coefficients of -0.00131 and -0.00266 respectively. This is inconsistent with the previous findings of Kumar et al. (2012) who found a positive significant relationship from 2001 until 2008. This may be due to anomalies in the yield during the financial crisis and its aftermath

(23)

23 which are illustrated in the scatter plot of the variable ln_rf (see Appendix 7.2.2, p.31). Consequently, H6 cannot be rejected.

4.2.5 Arithmetic returns

The descriptive statistics of the variables, using simple returns, are presented in Table 6 (see Appendix 7.2.1, p. 30). In contrast to the summary statistics of log returns, there are slightly higher standard deviations for variables with greater variance. This is consistent with the expectation of this study. As returns get further away from zero, arithmetic and logarithmic returns will produce increasingly different numbers. However, as shown in Table 5, the regression analysis with arithmetic returns does not show different results compared to the regression with log returns (see Appendix 7.2.1, p. 30). This contributes to the robustness of the empirical results of this present study.

(24)

24

5. Conclusion

5.1 Summary

This thesis analyses the relationship between the stock performance of clean energy companies and energy related factors, such as oil and gas prices. Furthermore, the effects of interest rates and technology stocks are investigated. By using a multifactor linear regression model, this research shows that the variation of clean energy stocks between 2001 and 2012 is significantly explained by past movements in oil prices and technology firms’ stock prices. Furthermore, this study anticipates a change in the historic correlation between the price of oil and the performance of renewable energy firms. For the fourth period (D) between 2012 and 2017, there was no significant relationship found which questions the logic of a substitutional movement from oil towards renewables in times of high oil prices. Additionally, there is no effect of technology stock prices on clean energy stock returns, in the fourth period (D). This supports the claim of a cyclical change in the nature of the renewable energy sector. The commonalities between the technology and the renewable energy sector seem to weaken which points to a mature stage of the business cycle. Furthermore, this paper fails to establish a link between gas prices and clean energy stocks. This result might be because of constant governmental subsidies which desensitizes renewable energies from external influences (such as cheap conventional electricity prices). Finally, in contrast to the previous line of research, this paper fails to prove a significant effect of interest rates on clean energy stocks. This might be explained by strong fluctuations in the yield on interest rates during the time of the Great Recession. In order to answer the aforementioned research question, this study concludes that oil prices can be seen as determinants of clean energy stock returns, on the contrary, gas prices did not show a significant effect.

5.2 Concluding Remarks 5.2.1 Implications for practice

The findings of this research show a tendency of the renewable energy sector to become less sensitive towards energy-related variables. Continuous subsidies seem to keep the demand for renewable energies constant despite fluctuations in oil prices. However, a cut in subsidies could abruptly change the situation. Investors should be aware of this when evaluating the

(25)

25 riskiness of investments in renewable energy companies and determining their hedging strategies.

5.2.2 Limitations & recommendations for further research

Diversity of the NEX – On the one hand, it is beneficial that the NEX represents the renewable energy sector accurately – however – on the other hand, there are many sub-sectors with distinct characteristics. For instance, solar and wind may be influenced by electricity prices whereas biofuel companies have other influential factors. Thus, further research should distinguish between individual sectors of the renewable energy industry when establishing the determinants of clean energy stocks.

Limitations of the model – this study adds gas prices to the analyses of previous studies. Nevertheless, there may be other possible drivers of clean energy stock prices. For example, further research could investigate the effect of domestic electricity prices and policy decision (about governmental subsidies).

(26)

26

6. Bibliography

Bloomberg New Energy Finance. (2017). New Energy Outlook 2017. Retrieved from https://about.bnef.com/new-energy-outlook/.

Chen, N. F. (1991). Financial investment opportunities and the macroeconomy. The Journal

of Finance, 46(2), 529-554.

Cunado, J., & De Gracia, F. P. (2005). Oil prices, economic activity and inflation: evidence for some Asian countries. The Quarterly Review of Economics and Finance, 45(1), 65-83.

Darby, M. R. (1982). The price of oil and world inflation and recession. The American

Economic Review, 72(4), 738-751.

Hamilton, J. D. (2003). What is an oil shock? Journal of econometrics, 113(2), 363-398.

Henriques, I., & Sadorsky, P. (2008). Oil prices and the stock prices of alternative energy companies. Energy Economics, 30(3), 998-1010.

International Energy Agency (IEA). (2016). Electricity Information 2017. Retrieved from https://www.iea.org/newsroom/energysnapshots/oecd-electricity-production-by-source-1974-2016.html.

Inchaupse, J., Ripple, R. D., & Trück, S. (2015). The dynamics of returns on renewable energy companies: A state-space approach. Energy Economics, 48, 325-335.

Kumar, S., Managi, S., & Matsuda, A. (2012). Stock prices of clean energy firms, oil and carbon markets: A vector autoregressive analysis. Energy Economics, 34(1), 215-226.

Managi, S., & Okimoto, T. (2013). Does the price of oil interact with clean energy prices in the stock market? Japan and the World Economy, 27, 1-9.

(27)

27 Nyquist, S. (2015, June 1). Lower oil prices but more renewables: What’s going on?.

Retrieved from https://www.mckinsey.com/industries/oil-and-gas/our-insights/lower-oil-prices-but-more-renewables-whats-going-on.

Rojas, V.A., & Stinson, P. (2015, January 5). Why Falling Oil Prices Don't Hurt Demand For Renewable Energy. Retrieved from https://www.forbes.com/sites/2015/01/05/why-falling-oil-prices-dont-hurt-demand-for-reneable-energy/#37c03fc02c12.

Sadorsky, P. (2001). Risk factors in stock returns of Canadian oil and gas companies. Energy

economics, 23(1), 17-28.

Sadorsky, P. (2012). Correlations and volatility spillovers between oil prices and the stock prices of clean energy and technology companies. Energy Economics, 34(1), 248-255.

U.S. Energy Information Administration (EIA). (2017). Annual Energy Outlook 2017. Retrieved from https://www.eia.gov/outlooks/aeo/.

U.S. Energy Information Administration (EIA). (2018). Short Term Energy Outlook 2018. Retrieved from https://www.eia.gov/outlooks/steo/.

Wüstenhagen, R., & Menichetti, E. (2012). Strategic choices for renewable energy

investment: Conceptual framework and opportunities for further research. Energy

(28)

28

7. Appendix

7.1 Charts

Fig. 1 (Investment in Clean Energy by sector; quarterly 2004-2017; Source: Bloomberg New Energy Finance)

Fig. 2 (Spot price of West Texas Intermediate (WTI) Crude Oil; adjusted for inflation 2006/2018; Source: http://www.macrotrends.net/1369/crude-oil-price-history-chart)

(29)

29

7.1.2 NEX, GAS, & OIL – (2001 to 2017)

0 5 10 15 20 G AS $/ Mcf 1/1/2000 1/1/2002 1/1/2004 1/1/2006 1/1/2008 1/1/2010 1/1/2012 1/1/2014 1/1/2016 1/1/2018 Date 0 50 100 150 O IL $ /b arre l 1/1/2000 1/1/2002 1/1/2004 1/1/2006 1/1/2008 1/1/2010 1/1/2012 1/1/2014 1/1/2016 1/1/2018 Date 100 200 300 400 500 N EX 1/1/2000 1/1/2002 1/1/2004 1/1/2006 1/1/2008 1/1/2010 1/1/2012 1/1/2014 1/1/2016 1/1/2018 Date

(30)

30

7.2 Regression Results: 7.2.1 Arithmetic returns

Table 5. Linear regression analysis, arithmetic returns

Time Regime1 (A)

VARIABLES NEX NEX NEX NEX

MSCI 1.070*** 0.718*** 1.322*** 1.060*** (0.0304) (0.0436) (0.0440) (0.0546) INTR_RATE -0.000565 -0.0182 -0.00156** 0.000317 (0.000455) (0.0122) (0.000734) (0.000317) PSE 0.0778*** 0.231*** -0.103*** 0.0249 (0.0193) (0.0214) (0.0362) (0.0419) OIL 0.0276*** 0.0221*** 0.0247** 0.0126 (0.00590) (0.00826) (0.0116) (0.0101) GAS 0.00123 -0.00221 0.00593 -0.000614 (0.00252) (0.00314) (0.00667) (0.00418)

Constant -6.82e-05 0.000339* -0.000467** 3.93e-05

(0.000116) (0.000184) (0.000229) (0.000166)

Observations 4,434 1,684 1,446 1,304

R-squared 0.702 0.606 0.794 0.592

Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1 (Time regimes1 described exactly in section 3.1.6)

Table 6. Summary statistics of daily returns1

N Mean Std. Dev. Min Max

NEX 4434 0.000147 0.01409 -0.0996 0.1283 MSCI PSE 4434 4434 0.000173 0.000379 0.01002 0.01410 -0.0706 -0.0779 0.0952 0.1145 OIL 4434 0.000467 0.02379 -0.1571 0.1783 GAS INTR__RATE 4443 4434 0.000597 0.023062 0.04306 0.35608 -0.4340 -6 0.8649 5.6667 1_{Figures are presented as arithmetic daily returns (01/01/2001 to 31/12/20017).}

(31)

31 7.2.2 Scatter plots: -. 2 -. 1 0 .1 .2 ln_Oil

01jan2000 01jan2002 01jan2004 01jan2006 01jan2008 01jan2010 01jan2012 01jan2014 01jan2016 01jan2018

Time -. 1 -. 0 5 0 .05 .1 ln _ N EX

Time -3 -2 -1 0 1 2 ln _ rf

Time -. 5 0 .5 ln _ G AS

Time -. 1 -. 0 5 0 .05 .1 ln _ PSE

Time -. 1 -. 0 5 0 .05 .1 ln _ MSC I

Determinants of returns on renewable energy stocks : an empirical study using a multifactor linear regression model