Determining solar stock returns, a specialised solar stock model that incorporates solar module prices

(1)

Determining solar stock returns, a specialised solar

stock model that incorporates solar module prices.

Bachelor thesis for Economics and Business at the University of Amsterdam Laurens Sliepenbeek, 5984084 ABSTRACT Solar energy is growing rapidly and expectations are that it could become the largest renewable energy source by 2050. Investments are in the solar industry are reaching record highs, this thesis tries to define a model that recognizes the drivers behind the returns of solar stocks specifically, adding on earlier models for renewable stock in general. This study proposes a time series model that incorporates the direct and indirect effect solar module prices. For 13 pure- play solar stocks between 2010 and 2016 a strong influence of the MSCI world index, a non-significant effect from technology stock returns and a small but significant effect from oil prices on solar stock returns. The solar module cost is found to have a marginal but significant direct effect and has an indirect effect of investor’s interpretation of the effect of oil prices on solar stocks.

(2)

2

Statement of Originality

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

(3)

Introduction

In September 2014, the International Energy Agency (IEA) issued a press release stating that solar is poised to be the world’s largest source of electricity by 2050, ahead of fossil fuels, wind, hydro and nuclear energy1_{. A combination of solar technologies could generate up to 27% of the} world’s electricity, possibly preventing six billion tonnes of carbon dioxide, an amount equal to the direct emissions from the transport sector worldwide (https://www.iea.org). The Executive Director of the IEA, Maria van de Hoeven, argued that this development was powered by a decline in costs: “The rapid cost decrease of photovoltaic modules and systems in the last few years has opened new perspectives for using solar energy as a major source of electricity in the coming years and decades.” During the COP21, the UN Climate Change Conference in December 2015, the Global Solar Council was launched, showing the dedication of countries to support the development of solar energy on its path to become the world's largest source of electricity generation. The Global Solar Council believes that a 10% share of solar in total global power generation by 2030 is within reach, from less than 1% in 20152_{. Figures for renewable} investments showed that renewables attracted more than double the $130 billion dollars committed to fossil investment in 2015, the biggest difference in recent history3_. These developments have been catching the attention of investors and have stirred an increased interest in examining the drivers behind the returns on renewable stocks (Inchauspe, 2015). Several studies (Kumar, 2011), (Managi, 2013), (Sadorsky, 2012) and (Henriques, 2008) have focused on the influence of oil prices, equities and carbon prices on the renewable stock returns. Until now the effect of the decreasing cost of solar modules has however not yet been included in research. This thesis aims to fill this gap and contribute to an understanding of the effect of declining cost of solar stock returns. The purpose of this thesis is to review the existing literature, and build on the researched produced so far with a recent dataset up to 6 January 2016 to determine if the strong decline in the prices of solar modules has influenced the returns and the determinants of solar stock returns. Therefore the research question of this thesis is: “How has the decrease in spot module price of solar energy, influenced the returns or the determinants of listed solar companies between 2010 and 2016?” In the Literature study the determinants of renewable stock returns, as discussed in the current literature will be reviewed and expectations and the hypothesises are presented. In the Methodology chapter a description of the model is give and a description of the data is given. The outcome of the research is discussed in the Empirical results chapter and the conclusions that can be made from the results are explained in the Conclusion chapter. Finally in the Discussion the limitations, implications and recommendations for further research are presented. 1_{(https://www.iea.org)} 2_{http://www.solarpowereurope.org/newsletter-december-2015/top-stories/global-solar-council-launched-at-cop21/} 3_{https://www.iea.org/topics/renewables/subtopics/solar/}

(5)

5

Literature study

While determinants of solar stocks have not been specifically researched, there is sufficient literature on the determinants of renewable stock returns. This chapter reviews the current literature on this topic and defines the requirements for a model that elucidates solar stocks returns in a more comprehensive way.

Determinants of returns for solar stocks

Oil price For regular stocks, in the absence of complete substitution between the factors of production, rising oil prices increase the cost of providing goods and providing services. These higher production costs dampen cash flows and reduce stock prices (Henriques, 2008). Rising oil prices also increase the discount rate used to value stocks, because oil prices are often seen as inflationary indicators that can result in rising interest rates. The discount rate is used in the valuation of stocks and a rise negatively impacts stock value (Henriques, 2008). However, for clean energy stocks between 2001 and 2007, the effect of rising oil prices was found to have a significant positive but weak effect on the returns. This positive relation can be explained by the effect of rising oil prices encouraging substitution towards other non-petroleum based energy sources (Henriques, 2008). Henriques and Sadorsky concluded that these results indicate that oil price movements were not as important for determining renewable stock returns as previously thought. Kumar et al. (2011) repeated the research by Henriques and Sadorsky with extended data and added carbon prices to the model. Carbon prices represent a surcharge paid to emit a ton of CO2. Kumar hypothesised that carbon prices raise the price of conventional energy and this would make conventional energy more expensive and stimulate substitution for renewable energy, thus increasing returns on renewable stocks (Kumar, 2011). While they did not find a significant effect for carbon prices on the returns of renewables, they did however found a positive significant effect for oil prices on the returns of renewable stocks. This was supported by research by Sadorsky (2012), who also found a significant effect of the oil price on the returns of renewable energy stocks for data from 2002 to 2010. According to (Managi, 2013) the contradiction between the strong relation found by Kumar et. al (2011) and the very weak (nearly zero) relation Henriques (2008) found, could possibly be explained by the structural break of the late 2007 oil prices. Mid 2008, oil prises rose to $140 a barrel and dropped back to $40 a barrel at the beginning of January 2008. Using a Markov switching VAR model, Managi (2013) found that after this structural break oil prices have positively impacted renewable energy stocks more than before the structural break. According to Managi this was due to alterations to the economic system. The term alteration refers to the response of renewable stock returns to shocks in oil price, interest and tech stocks. These alterations suggest a movement from conventional energy to renewable energy (Managi, 2013). Managi furthermore states that this movement away from conventional energy shows an imperfect substitution effect of renewable energy. The imperfect relation arises because

(6)

6 investors do not value a rise and a drop in oil price equally, causing an asymmetric relationship, where a rise in oil prices has more effect than a decrease of the oil prices (Managi, 2013). Inchauspe (2015) used a time varying model and found beta coefficients of regression of oil price excess returns on renewable energy stock excess returns varying between 0.01 for the years up until 2007 and 0.1 betas for the peak of the oil production in June 2008. Inchauspe found that the beta stayed at the same level after this peak until 2013. In his study, Inchauspe confirmed Managi’s (2013) findings regarding the changing nature of the relationship between oil prices and renewable stocks for the years 2002 to 2013. Inchauspe hypothesised that the higher beta after 2007 is the result of investors altering their view on the oil price and renewable relation, considering it more important in energy investment decisions after 2007 (Inchauspe, 2015). Bloomberg New Energy Finance (BNEF) observed that between 2013 and 2016, returns on renewable stocks rose whilst the oil price had collapsed from $100 a barrel in 2013 to $30 a barrel in the beginning of 2016 (Bloomberg New Energy Finance). Furthermore, data shows that costs of solar modules have decreased rapidly since 2013 (Bloomberg New Energy Finance). Managi (2013) hypothesised that economic recovery, a drop of the oil price to an oil price of around $30 a barrel or strong technological improvement of alternative energy (making it relatively inexpensive) could cause the relationship between oil and renewable stock returns to shift again, but delivers no evidence for this. These observations warrant a new study with recent data, to test whether the relationship between oil price and renewable stock returns between 2010 and 2016 has altered due to structural changes in the market (i.e. whether a cost decline in solar energy has an effect). A summation of these findings can be found in the figure below4_:

Variable Study Sample Effect

Oil return Henriques (2008) 2001-2007 Positive, but weak

Managi (2013) 2002-2007 Not significant Inchauspe (2015) 2002-2007 Positive, but weak Kumar et al. (2011) 2002-2010 Positive Henriques and Sadorsky (2012) 2002-2010 Positive Managi (2013) 2007-2010 Positive Inchauspe (2015) 2007-2013 Positive, but varying coefficient Technology stocks & MSCI world index Many studies have used technology stocks to explain stock behaviour of renewable stocks. Inchauspe (2015), Henriques (2008), Henriques and Sadorsky (2012), Managi (2013) and (Kumar (2011) all found significant positive correlations between technology stock indices and renewable stock returns. Sadorsky argues that this positive correlation can be explained because renewable energy companies have more resemblance to technology companies than to 4

_{For illustrative purposes the findings are sorted in chronological order of the sample period.}

(7)

7 energy companies, as they spend comparable amounts of their balance on R&D to increase the energy efficiency of their products (for instance solar panels) and lower the costs of production (Sadorsky, 2012). According to Inchauspe (2015), another possible explanation is that renewable companies compete for the same inputs as technology companies. Both technology and solar companies require high-qualified engineers, researching facilities, semiconductors, circuits etc. If there is abundance or shortage of these resources both respectively benefit or suffer (Inchauspe, 2015). Henriques and Sadorsky (2008) argue that it is likely that alternative energy companies will be regarded as energy companies in the future, which could result in a non-relation between technology stocks and renewable stock returns at that time. Henriques and Sadorsky state this will only happen if renewable markets succeed to reach mass adoption, but that until that moment renewable stocks will be viewed as technology stocks (Henriques, 2008). In 2014, 15% of the worldwide energy capacity was renewable power, generating 10% of the worldwide energy, doubling the amount of 2007 (UNEP, 2015). These numbers shows that adoption of renewables is growing. Mature companies tend to have lower risk than technology companies. In his “innovation cycle” Wustenhagen accurately described the transformation process from technology to a mature company (Wustenhagen, 2011). In this cycle, decreasing cost and market expansion are identified as the drivers behind the transformation of a technology to a full commercial and mature business. Sadorsky (2012) supported Wustenhagen’s findings as he found that an increase in sales lowers systematic risk of renewable stocks. Technology stocks are characterised by high risk and a decrease of risk in renewable stocks could decrease the effect technology stock returns have on renewable stock returns (Sadorsky, 2012). Inchauspe (2015) found that the technology proxy for returns of renewable stocks is best implemented as an additional pricing factor for renewables. When determining the effect of technology stocks, the MSCI world index should be added to the regression (Inchauspe, 2015). The MSCI world index is a broad global equity benchmark that represents large and midcap equity performance across 23 developed market countries, weighted by GDP and does not offer exposure to emerging markets. Inchauspe found a strong relation between renewable stock returns and the MSCI, finding a beta of 1.09 after 2006 for the regression of excess returns of the MSCI world on the excess returns for renewable stock. Over the same period, he found a positive beta between 0.17 and 0.26 for the regression of technology stocks excess returns on excess returns of renewable stocks over the years 2002 to 2013 (Inchauspe, 2015). The effect of the decreasing cost of solar According to Pernick, decreasing costs is one of the most important forces driving the growth in renewable energy, as it would signal markets the economic potential of these technologies to compete with conventional energy (Pernick, 2007). However, Pernick did not test whether the decreasing cost has an effect on the dynamics of the classically used variables oil prices, MSCI World Index and technology stocks. Darmstadter found that the success of the renewable energy industry's expansion could be measured by the level of its cost reduction, as lower cost would give it the power to compete with conventional energy sources (Darmstadter, 2000). The industry expansion that is being observed by a doubling of solar capacity every year and the increasing percentage of global energy production by renewables shows there is an expansion of

(8)

8 the industry (Bloomberg New Energy Finance). The expansion of industries leads to economies of scale in production, these will decrease the costs of producing renewable energy (Ramchandra Bhandari, 2009). In turn, economies of scale lead further stimulate the usage of renewable energy and increase its market share (Pernick, 2007). According to (Buzzell, 1975) a higher market share leads to a higher return on investment (ROI). Therefore, a direct positive effect on the returns of solar company returns is expected as a result of the drop in price of solar modules. Furthermore, the indirect effect can also be expected: that the decrease in solar module price has influenced the relations between solar stock returns and oil, as expected by Managi (2013).

Conceptual model

To test the conclusions that can be drawn from the literature study and the reasoning formulated the following hypothesis are presented: H1: The oil price has significant effect on the returns of solar stocks. H2: The spot module price has a significant effect on the returns of solar stocks H3: The MSCI world index has significant effect on the returns of solar stocks H4: The interaction variable spot module price*oil price is significant.

(9)

Methodology

This chapter sets out the research methodology. Firstly, the research setup is discussed, after which the model is introduced and the set of variables used in the model are explained.

Research setup

For this study the weekly data prices of oil, the MSCI World Index, a technology stock index (PSE) and solar module prices have been downloaded from Bloomberg Terminal. The data covers a sample period with weekly data from 6 January 2010 to 6 January 2016. A choice was made for this period, as Bloomberg has not recorded the prices for solar modules earlier than January 2010 and this is the only available source for weekly pricing on solar modules. The goal of this research was to review whether the solar module cost could be added to the literature as an explanatory instrument for solar stock returns. To that end, firstly, tests were performed to determine whether the variables found to have significant influence on renewable stocks in earlier studies, also have effect for solar stocks in particular. With the extended dataset up to January 2016, it is possible to establish whether previously determined relations have their expected effect. Secondly, the influence of the solar module costs on solar stock returns was tested by means of a time series regression. Thirdly, an interaction variable for solar modules price and oil price was introduced to examine if the drop in solar model prices has had an effect on the relationship between oil prices and solar stock returns. All tests used a time series regressions (to take account for panel data) with robust standard errors to empirically investigate the relationships between the returns of (i) solar stocks; (ii) the oil price; (iii); technology stocks (PSE); (iv) the MSCI World Index and; (v) solar module prices.

The Model

The following regression was estimated using a time series regression: 𝑅𝑒𝑡𝑢𝑟𝑛 = 𝛼 + 𝛽!∗ 𝑆𝑂𝐿𝐴𝑅𝑠𝑡𝑜𝑐𝑘!"#$!%&+ 𝛽!∗ 𝑂𝐼𝐿!"#$!%+ 𝛽!∗ 𝑃𝑆𝐸!"#$!%+ 𝛽! ∗ 𝑀𝑆𝐶𝐼𝑤𝑜𝑟𝑙𝑑!"#$!%+ 𝑀𝑂𝐷𝑈𝐿𝐸𝑟𝑒𝑡𝑢𝑟𝑛 ∗ 𝑂𝐼𝐿!"#$!%+ 𝜀 The time series regression used the following variables: § SOLARstock_returns stands for the returns of the solar companies. § OILreturn stands for the weekly returns on the oil price. § PSEreturn stands for the weekly return of a proxy for technology stocks. § MSCIworldreturns stand for the weekly returns for the MSCI world index. § MODULEprice stands for the price development of solar modules. § SOLAR*OILreturn is an interaction variable used to measure the effect Solar has on Oil returns.

(10)

10 The model assumed linear behaviour and assumed that there were no extreme anomalies for the data and no extreme anomalies for the time series for the research period, as most of the data was post-financial crisis. A total of 2609 observations were used in the regression. No mutations were needed for missing values. For several variables prices have been converted into returns using the formula

Return= (Price!"#$!!!!− Price!"#$!!)/Price!"#$!!.

Variables

Return of solar stocks The return of solar stocks variable (“SOLAR”) is calculated with the weekly historical prices of 13 companies that composed the Guggenheim Solar index between January 2010 and January 2016. By selecting the companies from this index, all major solar technologies (crystalline and thin-film photovoltaic solar and solar thermal) are represented (http://www.etf.com/TAN). The companies in the index have at least 1/3 of their revenue created by solar activities and consist mostly of small cap (39%) and mid cap companies (52%), and 9% large cap (http://www.etf.com/TAN). A stock is considered a pure play by the Guggenheim Solar index if more than 2/3 of its revenue is dedicated to solar activities. To compose the variable returns on solar stocks, the pure play stocks from the Guggenheim solar index were selected and the above listed formula was used. Oil price returns Oil prices are measured with the WTI futures contract, which trades at the New York Mercantile Exchange (NYMEX). The WTI futures contract is the most widely traded oil contract in the world and is used as a benchmark (Sadorsky, 2012). As such, it represents an efficient flow of information in the market (www.nymex.com). In line with the existing body of literature by Inchauspe (2015) and Sadorsky (2012), this study uses the WTI futures contract for the oil price returns variable (“OIL”). This contributes to making a comparable study. Technology stocks index (PSE) To examine the relation of technology stocks on solar energy stock return, the New York Stock Exchange Arca Tech 100 Index is used (PSE). The PSE is a price-weighted index that is composed of 100 listed tradable pure play technology companies. Companies that are included in the index are active in software, semiconductors, telecommunications, data storage, biotechnology and other high-technology sectors. This index is a multi-industry innovative technologies index that “provides a benchmark for measuring the performance of technology-utilizing companies operating across a broad spectrum of industries” (https://www.nyse.com/quote/index/PSE). The companies that compose the return on solar stocks variable however are not listed in the PSE. Previous studies of (Kumar, 2011), (Inchauspe, 2015), (Henriques, 2008) and (Sadorsky, 2012) have all used the PSE as a proxy for technology companies. These studies concluded that the explaining power of this variable is due to investors viewing renewable stocks as the same asset class as technology stocks.

(11)

11 MSCI World index The MSCI index is “a broad global equity benchmark that represents large and mid-cap equity performance across 23 developed markets countries. It covers approximately 85% of the free float-adjusted market capitalization in each country and MSCI World benchmark does not offer exposure to emerging markets” (https://www.msci.com/world). Including this variable is in line with previous work from (Henriques, 2008), (Inchauspe, 2015), (Kumar, 2011) and (Managi, 2013). Solar module spot price index The solar model price index (MODULE) is an index maintained by the Bloomberg BNEF department, that tracks market prices of solar power modules. The MODULE Index surveys spot prices for the dominant technologies of crystalline silicon, thin film silicon, cadmium telluride and copper indium gallium selenide solar energy modules. The data is collected once a day when participants inform Bloomberg on the current delivery price of a particular product, by sharing quotations they have made. As a reward the company gets access to the average prices of the past 1, 3 and 10 days that Bloomberg has composed of all entries. All quotes are gathered on a company basis, whereas each company provides only one quote. All data is collected and presented in USD. To avoid too high or too low quotes due to typing errors or other mistakes, the module spot price index excludes all quotes more than 20% above or below the average of the period selected and calculates an average from the remaining quotes. All participants must be active in the manufacturing or buying of any of the surveyed products and BNEF must manually approve their participation before they may start contributing their quotes. (Bloomberg New Energy Finance). Figure 1 shows the development of solar module prices from 2010 to 2016. Figure 1 Interaction term The interaction term is the product of MODULE and OIL returns. This term is tested to review if the relationship between the oil price and the returns on solar stocks is influenced by the decreasing price of solar modules.

(12)

Empirical results

This section shows the descriptive statistics of the data and discusses into detail the result of the regressions that were performed. The obtained outcomes are discussed and answers to the hypothesises formulated. The table below shows the results of the univariate analysis of the variables in the model.

Variable Obs. Mean Std. Dev. Min Max

MODULE_price 2,944 0.73067 0.48043 0 1.87 MSCIworld return 2,943 0.000732 0.033679 -0.43645 0.135267 OIL_return 2,943 0.003923 0.099164 -0.40628 1.849445 PSE_return 2,943 0.001594 0.04641 -0.68125 0.209911 SOLARstock return 2,607 0.004357 0.110932 -0.4083 0.768 SOLAR*Oil 2,944 61.0338 44.1344 0 155.1906 To test the original data for normality a Shapiro Wilkinson test was performed to check if the variables were normally distributed. All values showed a 1% significance of being not normally distributed. Following the outcome of the normality test, skewness and kurtosis tests were performed to find explanations for the significant non-normality. Results show that all the variables, except the PSE and the MSCI world, are jointly significant skewed and show kurtosis. To account for these non-normalities the original data, the prices were converted to returns. All the correlations between the variables are significant. While the high correlation between the interaction and spot module cost could cause multicollinearity, no problematic disturbances from multicollinearity were found when regressing the models. Three regressions were run. These are permutations of the model presented in the literature review. The first regression was the most basic regression performed. The returns of solar companies were regressed on the oil price, the MSCI world index and Tech stocks. This regression was performed to repeat the research of (Henriques, 2008), (Managi, 2013) and (Inchauspe, 2015) conducted on renewable stocks, for solar stocks. As can be seen in the table below, this regression shows a non-significant effect for OIL returns. A significant effect for both MSCI world returns and Tech stocks returns are found, showing a positive and negative coefficient respectively. The positive coefficient of 1.07 for the MSCI world returns is in line with the results of Inchauspe (2015) for a mix of renewable energy stocks in the period after 2006. For technology stocks a significant negative coefficient is found of -0.24. The magnitude of the effect of technology stocks is in line with the earlier research on renewable stocks (Inchauspe, 2015) (Henriques, 2008) however the sign has reversed. The R-squared of this model is 0.53. The second regression included the spot module price into the equation, in order to test if there is a significant impact of spot module price on the returns of solar stocks. In this regression, a non- significant oil price and a positive significant effect for the MSCI world returns of around

(13)

13 1.07 were found. Again, the regression showed a negative effect for the PSE. The added variable, solar module cost, had a significant but very small negative effect of -0.006. The R-squared of this model had reduced to nearly zero (0.008), indicating that adding the price of solar module negatively impacts the explanatory power of the model. In the final regression, the interaction variable was added, to test if the decreasing cost of solar modules affected the relationship between the returns of solar stocks and oil returns. A significant coefficient of 0.101 was found for this term. The interaction variable showed a limited effect compared to the, in this case significant, OIL returns of 0.092. When the interaction variable was added to the model, the technology proxy PSE loses its significance, showing a p-value higher that 0.7 and the MSCI world returns, although still significant, also shows less effect. The coefficient is halved compared to the first and second regression. The price of solar modules showed a small negative effect of -0.01. The R-squared of the final regression is 0.15.

Explanatory variable Model 1 Model 2 Model 3

OIL_return 0.046 0.047 0.093*** MSCIworld_return 1.074*** 1.067*** 0.589*** PSE_return -0.248** -0.238** -0.049 MODULE_price -0.007*** -0.011*** SOLAR*OIL 0.101*** Constant 0.004*** 0.009*** 0.012*** R-squared 0.5386 0.0083 0.1498 Prob > Chi^2

0.00

0.00 *** p-value<0.01

** p-value<0.05

* p-value<0.10

The results of these regressions show that the first hypothesis (“the oil price has a significant effect on the returns of solar stocks”) is accepted, but is not robust. The second hypothesis (“the spot module price has a significant effect on the returns of solar stocks”) can also be accepted, but is not robust. The third hypothesis, (“the MSCI world index has a significant effect on the returns of solar stocks”) is accepted and is robust, as the significance is found in all three models. Hypothesis four (“the interaction variable spot module price*oil price is significant”) can also be accepted.

(14)

14

Conclusion

Contributing to a small but growing body of literature on the understanding of renewable stock returns, this thesis aimed to explore the direct effect of the price of solar modules on solar stock returns and the indirect effect it has on the oil price and solar stock return relation in the period 2010 to 2016. Direct effect of solar module price on solar stock returns To determine the direct effect of the price of solar modules on solar stock returns, the determinants for solar stock returns found to be relevant in previous research were analysed. The relevant determinants (oil price, technology stocks (PSE) and the MSCI World Index) have been tested for significant influence on solar stock returns to control for their effect. Oil Regarding the effect of oil price returns on solar stock returns, only the model incorporating solar module price and the interaction variable between oil and solar resulted in a significant and positive effect for oil price in line with earlier findings by Henriques (2008), Sadorsky (2012) and Inchauspe (2015). However, the non-significance found in the regressions without the interaction term (regressions 1 and 2) is in line with observations of BNEF and the prediction by Managi (2013), that a non-significant relation would be found for time periods with a drop in oil prices to $30 a barrel. The non-significant outcome shows that for the time period 2010 - 2016, a period with decreasing oil prices, investors value the effect of oil on solar stock returns less. This appears to confirm the asymmetric effect that Managi (2013) found, where a low oil price results in less effect of oil returns on renewable stock returns. Technology stocks (PSE) The results for technology stocks show a different picture than earlier research, as there is a non-significant effect for the technology proxy on solar stock returns in the complete model. This outcome leads to the conclusion that investors do not view the returns on technology stocks as a significant determinant for solar stock returns. This result is in line with the prediction by Sadorsky (2008) that if solar companies reach mass adoption, investors will no longer regard solar companies as tech companies. The first two regressions showed a significant negative effect of technology stocks (PSE), contrary to previous research. Possibly, the competition for the same inputs as stated in Inchauspes (2015) theory, has obtained a different dynamic. Solar and tech competitors no longer mutually benefit or suffer from developments on their inputs. The negative sign could imply a zero-sum situation in which one would benefit at the expense of the other. MSCI World Index In line with previous studies, the MSCI World Index was found to have a strong positive effect on solar stock returns in all three regressions. It can be concluded that investors value the MSCI World Index as an important measure for solar stock returns over the years 2010 – 2016.

(15)

15 Solar module price The results showed a limited but significant positive effect of the decreasing cost of solar modules on solar stock returns, but only when including the interaction effect in the regression. This limited direct effect of the solar module cost does not support the theories of Darmstadter (2000) and Pernick (2007) that lower solar module prices through market expansion and economies of scale would positively influence solar stock returns. Finding a limited effect shows that the cost decrease in solar energy does not directly cause higher returns. Indirect effect of solar module price on the oil price and solar stock return relation The interaction term is significant but has only a slightly higher coefficient than that of the oil price. This shows that a combination of oil prices and solar module prices only has marginally more effect than only the oil prices. This could imply that the substitution effect that is theorised by Henriques and Sadorsky (2008) is mainly driven by oil prices and not (or much less so) by the combination of oil price and solar module cost. Explaining power of the models A total of three models were tested. The first model replicated earlier research and was used to control if adding solar module cost and subsequently adding the interaction variable contributed to the explaining power of solar stock returns. It can be concluded that the model that only included the module cost of solar, is of little value and certainly not better that the first model, as it has virtually no explaining power with a nearly zero R-squared. This is in contrast with the first model, used in the literature by (Inchauspe, 2015), (Managi, 2013), (Henriques, 2008) and (Sadorsky, 2012), that showed an R-squared of 0.58. When the interaction variable is added in the third model, the R-squared became 0.15, which is still lower than the R-squared of the first model. It can thereby be concluded that adding these variables does not contribute to the explaining power of the model. Therefore, although significant relations were found between solar module prices and the interaction variable for solar stock returns, including these variables does not contribute to the explaining power of the model. However, the R-squared is sufficient, as it is not of such a low rate that it impacts the value of the model.

(16)

Discussion

Implications and relevance Several of the sources studied (BNEF, United Nations Energy Program, International Energy Agency) predict a bright future for solar, expecting that decreasing cost would boost returns for these stocks. This direct effect has not been found in this research and investors cannot view the decrease in prices as a sign of possible or increased returns. These sources often assume there is a substitution effect behind oil and renewables. However, scientific literature on this substitution effect for renewables is (still) scarce, and further research on the mechanics of substitution between conventional energy and solar energy is needed. Limitations and suggestion for future research Solar has been subject to rapid change, possibly driven by policy. In 2010, more than 100 countries committed to renewable energy targets (Bloomberg New Energy finance). The EU agreed on aiming for 20% of its total energy consumption to be produced by renewables by 2020 and China aims for 15% of its energy consumption to derive from renewables by 2020. Finding a measure to account for policy measures could further add to the explanatory power of the model. Another possibility is that that the effect of the several variables on solar stock returns have a lagged effect, and are in fact stronger than found. This could specifically be the case for the effect of solar module costs, which possibly need longer periods of lower costs before returns are influenced. Since this research has not accounted for lagged analysis, adding this dimension could help predict delays in responses of the market for developments in predictor variables. It is important to observe that this research used a limited (small) sample of solar stock, which could imply a large variance between stocks. The sample used shows gaps in the panels used, as not every stock has the same amount of data available, often due to recent entry on the stock market. Correcting for this would have limited the sample size. A future study with more pure-play solar companies operating in markets worldwide would reduce variance and possibly improve results. As solar companies mature, it is likely more will become listed and more data will become available, improving the possible samples.

(17)

Bibliography

Bloomberg New Energy Finance. (n.d.). Burer, W. (2009). Which renewable energy policy is a venture capatalists best friend. Energy Policy (37), 4997-5006. Buzzell, B. a. (1975). Market share - A key to profitability. Harvard Business Review . Darmstadter, J. (2000). The role of renewable resources in U.S. electricity generation— experience and prospects. . Resources for the future, Climate Change Issues Brief No. 24. Henriques, S. (2008). Oil prices and the stock prices of alternative energy companies. Energy Economics (30), 998-1010. Inchauspe, R. T. (2015, November). The dynamics of returns on renewable energy companies: A state-space approach. Energy Economics . Kumar, M. M. (2011). Stock prices of clean energy firms, oil and carbon markets: A vector autoregressive analysis. Energy Economics (34). Managi, O. (2013, April). Does the price of oil interact with clean energy prices in the stock market? Japan and the World Economy . Pernick, W. (2007). The Clean Tech Revolution: The Next Big Growth and Investment Opportunity. Ramchandra Bhandari, I. S. (2009). Grid parity analysis of solar photovoltaic systems in Germany using experience curves. Solar energy . Sadorsky. (2012). Moddeling Energy Company Risk. Energy Policy (40), 39-48. UNEP. (2015). Global Trends in Renewable Energy Investment 2015. Frankfurt School-UNEP Centre/BNEF. 2015. Wustenhagen, M. (2011). Strategic choices for renewable energy investment: Conceptual framework and opportunities for further research. Energy Policy .

(18)

Appendix

Testing for normality

All variables have a P-value of zero showing that none of the variables is normally distributed.

Correlations

Interactio~n 2,467 0.05533 1355.642 18.485 0.00000 MODULE_price 2,944 0.91534 142.574 12.791 0.00000 PSE_return 2,943 0.63894 607.885 16.530 0.00000 MSCIworld_~n 2,943 0.78213 366.806 15.228 0.00000 OIL_return 2,943 0.55100 755.939 17.092 0.00000 SOLARstock~n 2,607 0.93156 103.258 11.909 0.00000 Variable Obs W V z Prob>z Shapiro-Wilk W test for normal data

Interactio~n 2,467 0.0000 0.0000 . . MODULE_price 2,944 0.0000 0.8117 . 0.0000 PSE_return 2,943 0.0000 0.0000 . . MSCIworld_~n 2,943 0.0000 0.0000 . . OIL_return 2,943 0.0000 0.0000 . . SOLARstock~n 2,607 0.0000 0.0000 . 0.0000 Variable Obs Pr(Skewness) Pr(Kurtosis) adj chi2(2) Prob>chi2 joint Skewness/Kurtosis tests for Normality

Interactio~n 0.0075 -0.4378 0.0273 0.2718 0.0178 1.0000 MODULE_price -0.0339 0.0004 0.0411 0.0958 1.0000 PSE_return 0.1194 -0.1060 0.8425 1.0000 MSCIworld_~n 0.1613 0.1836 1.0000 OIL_return 0.0864 1.0000 SOLARstock~n 1.0000 SOLARs~n OIL_re~n MSCIwo~n PSE_re~n MODULE~e Intera~n

(19)

19

Determining solar stock returns, a specialised solar stock model that incorporates solar module prices