6.4 Methods
6.5.3 Robustness Checks
6.5.3.3 Tracking Error and Downside Tracking Error
In Table 16, I report how strong tracking error and downside tracking error estimates deviate from each other. I compute the deviation as the difference between each traditional
and downside tracking error estimate and test whether the difference is statistically significant, using Wilcoxon nonparametric rank-sum test. Differences in regression estimates confirm that about 80 percent (11 out of 14) of renewable energy indexes have higher downside volatilities, i.e. downside tracking error volatilities are larger than traditional tracking error volatilities. This means that the majority of renewable energy indexes have had periods of relatively poor risk management. Furthermore, the difference between traditional tracking errors and downside tracking errors can be large in magnitude. I find monthly differences between the two measures of up to 1.03 percent. Although, the difference can be large in individual cases, tests on the statistical significance of the deviation show that none of the differences is significantly different. Meaning that both relative risk measures, tracking error and downside tracking error, should deliver identical results. One explanation for why the two measures perform very similar is the fact that I investigate passive index investments. In other words, I would expect to find larger differences in the two risk measures when observing actively managed mutual funds as they change their investment styles more dynamically and may specifically try to reduce downside variation compared to a passive index investment with a less dynamic tracking strategy.
Rather than only examine deviations between tracking error and downside tracking error estimates overall, I can rank energy indexes according to their risk performance relative to MSCI's all-industry equity index. I rank each energy index risk performance in two steps. First, I identify the energy index with the lowest tracking error within each geographic region and type of energy index (renewable or conventional). Second, I compute the difference between the lowest index relative to its peers. For example, I separately look at the magnitude of the difference between the lowest global renewable energy index with the highest global renewable energy index. Subsequently, I compute the statistical significance of the difference, using Wilcoxon nonparametric rank-sum test. In Panel A of Table 16, I identify S&P Global Alternative Energy to have the lowest tracking error among renewable energy indexes with global investment objectives. The monthly differences for my sample of global renewable and conventional energy indexes can be large (ranging from 1.22 to 6.88 percent and 0.8 to 5.72 percent, respectively). These findings suggest that my samples of global energy indexes vary substantially, which I formally confirm with Wilcoxon's nonparametric significance test.
Panel B of Table 16 shows that both, renewable and conventional energy indexes' relative risks are identical, as I cannot reject Wilcoxon's null hypothesis that the two series
American energy firms. In line with previous observations, I find renewable Nasdaq Clean Edge index to have the lowest tracking error across the North American sample of energy indexes and except for one globally oriented index to almost have the lowest relative risks across the entire sample of renewable and conventional energy indexes relative to MSCI world. Panel D shows that across my Asian sample of energy indexes, conventional DJGL Asia Oil & Gas has the best risk protection, very closely followed by another conventional energy index, Daxglobal Asia Oil & Gas.
Overall, the robustness checks suggest that my conclusions from previous analyses hold irrespective of whether a) I use absolute downside risk measures such as semi-standard deviation, lower partial moments and worst losses, or use total absolute risk measures such as standard deviation of returns b) I use an all industry benchmark such as MSCI world, or directly compare relative investment risks between renewable and conventional energy indexes, and c) I use traditional tracking errors to compute relative investment risks or use my modified downside tracking error to compute these risks.
Table 16: Tracking Error and Downside Tracking Error Regressions relative to MSCI World
Absolute Risk Relative Risk
ID Energy Indexes Ann. Semi-Std. Dev. Ann. LPM Min. TE TE Down TE - TE Down Wilcoxon TE - TE lowest Wilcoxon TE Down - TE Down lowest Wilcoxon Panel A: Global
C1 AGAE Composite 0.441 0.374 -0.490 0.068 0.061 0.007 0.195 0.030 4.04*** 0.023 3.04***
C2 AGAE Extra Liquid 0.452 0.386 -0.512 0.066 0.063 0.003 0.328 0.028 3.90*** 0.025 2.74***
C5 AGAE Solar 0.651 0.530 -0.548 0.106 0.102 0.004 0.090 0.069 7.26*** 0.065 5.47***
C6 Daxglobal Alternative 0.329 0.265 -0.313 0.050 0.047 0.003 0.484 0.012 2.94*** 0.009 2.51**
C7 World Renewable Energy 0.455 0.373 -0.477 0.083 0.083 0.000 0.369 0.045 7.11*** 0.045 4.91***
C8 S&P Global Alternative Energy 0.336 0.296 -0.360 0.037 0.038 0.000 0.093 Lowest Lowest
C10 HFRX Alternative Energy 0.469 0.410 -0.510 0.056 0.059 -0.002 0.431 0.019 2.55** 0.021 1.40
C12 S&P Global Clean Energy 0.445 0.364 -0.392 0.051 0.048 0.002 0.464 0.013 2.42** 0.011 1.25
D1 MSCI World Oil & Gas 0.215 0.171 -0.158 0.039 0.041 -0.002 0.424 0.010 3.75*** 0.011 2.85***
D2 FTSE World Oil & Gas 0.217 0.173 -0.175 0.039 0.041 -0.002 0.682 0.010 3.80*** 0.011 3.29***
D3 TR Global Oil & Gas 0.214 0.170 -0.176 0.036 0.038 -0.001 0.652 0.008 3.16*** 0.008 2.72***
D4 DJ Titans Oil & Gas 30 0.227 0.181 -0.191 0.039 0.041 -0.002 0.403 0.010 3.98*** 0.011 3.13***
D5 MSCI World Metals & Mining 0.339 0.299 -0.387 0.052 0.058 -0.006 1.019 0.023 5.67*** 0.028 4.54***
D12 Daxglobal Coal 0.410 0.350 -0.440 0.086 0.087 -0.001 0.236 0.057 7.79*** 0.057 5.27***
D13 Daxglobal Nuclear 0.282 0.221 -0.226 0.064 0.058 0.006 0.694 0.035 5.41*** 0.028 3.54***
D17 HFRX EH: Energy 0.156 0.138 -0.177 0.029 0.030 -0.001 0.038 Lowest Lowest
Panel B: Europe
C4 AGAE Europe 0.494 0.417 -0.519 0.068 0.062 0.006 0.609 Lowest Lowest
C13 European Renewable Energy 0.496 0.430 -0.525 0.075 0.077 -0.002 0.650 0.007 0.67 0.015 1.56
D10 EURO STOXX Oil & Gas 0.201 0.159 -0.162 0.041 0.041 0.000 0.424 0.000 0.24 -0.001 0.02
D11 DJ Europe Oil & Gas 0.240 0.192 -0.212 0.041 0.042 -0.001 0.169 Lowest Lowest
Panel C: North America
C3 AGAE North America 0.436 0.363 -0.396 0.080 0.070 0.010 0.194 0.048 5.72*** 0.040 4.15***
C11 NASDAQ Clean Edge US Liq. 0.218 0.191 -0.219 0.032 0.030 0.001 0.141 Lowest Lowest
C14 Wilderhill New Energy Global Inn. 0.363 0.313 -0.431 0.041 0.041 0.000 0.191 0.009 3.15*** 0.011 1.96*
D6 S&P 500 Oil & Gas 0.210 0.168 -0.159 0.042 0.043 -0.001 0.146 Lowest Lowest
D7 DJ US Int. Oil & Gas 0.188 0.149 -0.135 0.042 0.043 -0.001 0.246 0.001 0.11 0.000 0.23
D8 NYSE Arca Oil 0.233 0.184 -0.189 0.044 0.046 -0.002 0.266 0.003 0.65 0.003 0.57
D9 DJ US Coal 0.533 0.433 -0.466 0.113 0.116 -0.003 0.399 0.071 8.66*** 0.073 6.15***
D18 NASDAQ/SIG Oil 0.368 0.297 -0.296 0.070 0.074 -0.005 0.277 0.028 3.35*** 0.031 2.42**
Panel D: Asia
C9 S&P Asia Alternative Energy 0.407 0.329 -0.295 0.082 0.083 -0.001 0.698 NA NA
D14 Daxglobal Asia Oil & Gas 0.252 0.212 -0.251 0.067 0.054 0.013 0.715 0.006 0.29 Lowest
D15 DJGL Asia Int. Oil & Gas 0.335 0.261 -0.287 0.087 0.080 0.007 0.693 0.026 3.41*** 0.026 2.68***
D16 DJGL Asia Oil & Gas 0.315 0.270 -0.367 0.061 0.061 0.000 0.865 Lowest 0.007 0.06
Notes: In this table, I report absolute and relative risk measures between my combined sample of 32 energy indexes relative to MSCI World all-industry equity index. In particular, I compute three absolute risk measures, annualised semi-standard deviations, lower partial moments and minimum returns. Relative risk measures include tracking error and downside tracking error volatilities. Columns 3 to 5 list absolute risk measures. Columns 6 to 8 report relative risk measures including the exact difference between tracking error and downside tracking error estimates relative to MSCI world index. I compute tracking errors and downside tracking errors according to equations 10 to 12. Using Wilcoxon nonparametric rank-sum test, I compute statistical significances for the difference in medians between tracking error and downside tracking error estimates in column 9. Column 10 reports the difference between the lowest tracking error estimate and energy indexes that are from the same region and energy type. For example, Panel A lists global energy indexes of which 8 are renewable energy indexes. After identifying the global renewable energy index with the lowest tracking error in that group, I compute the difference between the lowest tracking error with the 7 others belonging to the same group. Column 9 reports Wilcoxon nonparametric rank-sum test for statistical differences
6.6 Conclusion
This study investigates absolute and relative risk relations between renewable and non- renewable energy equity indexes. Using return data on fourteen international renewable energy indexes from 2000 to 2013 and eighteen conventional energy indexes over the same time period, I find strong positive correlations between their returns, indicating similarities in their underlying return generating processes. Despite strong associations between the returns of renewable and traditional energy equity indexes such as fossil-fuel generated companies, I find return volatilities to be substantially higher for 70 percent of my sampled renewable energy indexes relative to my benchmarked conventional energy indexes.
By introducing a novel approach to assess relative downside return volatilities that account for investors' asymmetric risk appetites (i.e. risk-averse investors), namely the downside tracking error, I capture the real risk exposure that matters to risk-averse investors. Using the modified tracking error for downside residual return volatilities, my empirical analysis suggests that the majority of renewable energy indexes experience higher downside tracking errors relative to their conventional energy benchmarks.
My empirical analysis leads to the conclusion that major international renewable energy indexes carry higher absolute, downside and relative investment risks compared to a large sample of conventional fossil-fuelled energy indexes. This conclusion turns out to be robust, since it holds across different risk specifications and alternative non-energy industry benchmarks. My conclusion is generally consistent with findings from the renewable energy literature that return volatilities of renewable energy indexes tend to be high (Henriques and Sadorsky, 2008; Kumar et al., 2012; Sadorsky, 2012a, b). The high return volatilities are a result of high uncertainties regarding the nature of renewable technology businesses and future prospects of the renewable energy industry in several respects. First, renewable tech companies tend to be small and technology-oriented businesses with a selected focus on few projects. Private sector investors perceive companies with only a few projects, whose outcome is unknown, as risky. Second, the renewable energy sector is very capital intensive. Due to the capital intensity of renewable energy projects, the government has substantially supported the sector by policy investments in the past, which was especially the case in the solar industry. One of the major risks of policy investing is the duration of financial support, which if discontinued (for example due to a recession) can produce devastating effects to a sector, as seen by the recent wave of bankruptcies in the German solar industry (Bohl et al., 2013). Furthermore, the weak carbon price does not contribute to the growth of the renewable
energy sector. My empirical findings also show that several more specialised or "concentrated" renewable energy indexes have substantially lower tracking error volatilities. These indexes are specialised in specific sub-segments of renewable energy technologies such as biofuels or advanced batteries. Further, direct comparisons between more traditional measures of relative risk and my alternative approach indicate that the former are well suited to explain the risk behaviour of renewable energy equity indexes as the results obtained from tracking error are generally in line with results from the downside tracking error. A formal non-parametric significance test between these two risk measures shows that the difference is tiny and not statistically significant. As the difference between the traditional tracking error and the downside tracking error is not significantly significant, the two risk proxies lead to the same result. Meaning that the traditional tracking errors sufficiently explain the risk relationship between renewable and conventional energy equity indexes.
My results may be useful to policymakers and investors as they seek to understand differences in the assessment and perception of risk in international renewable energy equity indexes. This is increasingly relevant as renewables make up larger and larger parts of total energy and power capacity. For policymakers, an investigation of these risks can help to shape future financial support in RD&D activities. In order to increase the participation of private sector investors, understanding perceived risks in renewable energy investments is of great importance. Particularly, to reduce investment risks in renewable energy to cater for more risk-averse private sector investors such as large institutional funds. Innovative products such as fixed-income renewable energy funds or green bonds have the potential to achieve this goal. I hope my findings will be useful for future research to increase the understanding of risk in renewable energy equity indexes.
My findings are, however, limited due to the following restrictions. First, my risk findings apply to equity investments in renewable energy and in particular to equity indexes only. More recently, the renewable energy sector has seen growing demand for green bonds and fixed income products, which could substantially reduce uncertainties and risks in the sector. Second, my results largely apply to renewable energy equity indexes in developed countries only. Renewable energy equity and bond markets in emerging economies are growing at enormous speeds. So far, data limitations on individual firms or indexes have made comparisons rather difficult. Further research should look at the drivers of risk in renewable energy equity indexes in relation to traditional energy equity indexes. Also, to advance the development of more appropriate risk measures to evaluate renewable energy
7
The Effects of Environmental, Social, and Governance (ESG)
Disclosure Quality on the Cost of Capital
Abstract
This chapter empirically investigates the effects of Environmental, Social, and Governance (ESG) disclosure quality on the expected cost of equity as well as the cost of debt. To investigate these effects, my analysis is based on a large sample of US S&P 500 companies over the sample period from 2004 to 2014. Using several alternative approaches to compute the expected cost of equity and debt (based on Graham and Harvey's expected market premium and inferred from several asset pricing models), my results show a negative and statistically significant association between ESG disclosure quality and my expected cost of equity and debt variables, while also controlling for company- and debt-specific characteristics. My results suggest that companies with high ESG disclosure quality have lower expected cost of equities and debt, everything else equal. Accordingly, my findings of the relation between ESG disclosure quality and the expected cost of equity and debt imply that the market prices a company's ESG disclosure quality along with other factors. My results are robust over time and alternative regression specifications.
The Effects of Environmental, Social, and Governance (ESG)
Disclosure Quality on the Cost of Capital
7.1 Introduction
This chapter empirically investigates the effects of Environmental, Social, and Governance (ESG) disclosure quality on the cost of equity and debt capital on a large sample of US S&P 500 companies from 2004 to 2014. Over the last decade, corporate disclosures of environmental, social, and governance activities have substantially increased. According to KPMG's annual survey on corporate disclosures of ESG activities of the world's largest companies, the rate of disclosure has risen from 64 percent in 2005 to 93 percent in 2013 (KPMG, 2005; 2013)92. Since then many North American companies have committed to
make ESG disclosures, and have now overtaken leading European companies in this type of disclosure (KPMG, 2013).
The rapid increase of corporate ESG disclosures raises the question: What are the motives behind companies' ESG disclosure?
The reasons for companies to disclose on their ESG activities vary, but the following three motives could potentially answer the question. First, corporations are under constant scrutiny now more than ever due to a series of recent corporate scandals that have adversely impacted society and the economy's stability. Companies who are trying to re-build that trust and to increase their reputation could have the incentive to increase disclosure on environmental, social, and governance issues. Some scholars (see e.g. Brammer and Pavelin, 2006; Lii and Lee, 2012) have argued that poorly managed ESG risks have shown to impact a company's reputation and sales. For instance, British Petroleum (BP) has been struggling for several years and have invested more than 90 billion US dollars in environmental liabilities to strengthen its reputation after the 2010 Deepwater Horizon oil spill in the Gulf of Mexico (Chazan and Crooks, 2013; Gordon, 2013). Second, the rise in ESG disclosure could also be driven by large institutional investors such as pension and investment funds who have called on companies to provide more transparency with regard to their environmental, social, and governance activities to be able to incorporate such information into their investment decision-making processes (PRI, 2014). Finally, the rapid growth of socially responsible investments in the US and globally could have triggered increased corporate disclosure on ESG activities. According to the US Social Investment Forum (USSIF, 2014), from 1995 to 2014, assets under management using socially responsible investment strategies in the US have grown from $639 billion to $6,570 billion.
92 The rate of disclosure is defined as the quality of CR (Corporate Responsibility) Reporting measured against seven key criteria, which are based on current reporting guidelines including: 1) Strategy, risk and opportunity,
Thus, in this chapter, I empirically investigate whether a reduction in companies' cost of capital explains the rise in ESG disclosure quality. I focus on the cost of equity and cost of debt because they represent the two main sources of a company's financing as well as play a pivotal role in a company's financing decision making. In addition, corporate executives feel that voluntarily increasing information to investors can reduce their companie's cost of capital (Armitage and Marston, 2008). Also, there is a long-established interest in the academic community in the relation between ESG disclosure and the cost of capital (Clarkson et al., 2013; Dhaliwal et al., 2011; Plumlee, 2008; 2010).
Based on Heinkel et al.'s (2001) and Merton's (1987) theoretical framework, and consistent with the Efficient Market Theory (which I discuss in Chapter 2 'Theory: The Efficient Market Theory'), I hypothesise that companies with high Environmental, Social, and Governance (ESG) disclosure quality have lower expected cost of equity and debt, everything else being equal. The theoretical mechanisms through which ESG disclosure quality could affect the expected cost of equity and debt are the depth's of a companies investor base, reductions in companies' beta or systematic risk, and future litigation and reputational risks (Lambert et al., 2007; Merton, 1987). I will explain the three mechanisms in more detail in Chapter 7.2.1. 'ESG Disclosure and the cost of capital', as well as why ESG disclosure quality could be "priced" in the cost of capital (See Chapter 7.2.1.4. 'ESG Disclosure and Diversification').
To empirically test whether ESG disclosure quality is related to a company's expected cost of equity and debt capital, I use a large sample of US companies based on the historical constituents of the S&P 500 index. My empirical analysis shows that ESG disclosure quality is negatively associated with all of my expected cost of equity and cost of debt variables, controlling for company-and bond-specific characteristics. My results are generally statistically significant at the 1 and 5 percent significance level and consistent across alternative proxies for the expected cost of equity (based on Graham and Harvey's, 2015 expected market premium and inferred from three different asset pricing models including CAPM, Fama/French, and Carhart using daily, weekly, and monthly data frequencies) and different regression specifications. My results are consistent with Hypothesis 1 and 2, which predict that companies with high ESG disclosure quality have lower expected cost of equities and cost of debt, everything else equal. Additional robustness tests for temporal consistency and stepwise regressions support my prior findings. On the cost of equity side, my findings are in line with Dhaliwal et al. (2011), El Ghoul et al. (2011) and Sharfman and Fernando
ESG disclosures, environmental management, or human rights reduce a company's cost of equity. On the cost of debt side, my findings tend to be generally in line with Bauer and Hann (2010) and Oikonomou et al. (2014), who find several ESG dimensions such as corporate environmental management, community strengths, and product safety and quality strengths to reduce the cost of debt.
To the best of my knowledge, my chapter is the first to investigate both, the effects of the cost of equity and debt capital on a novel ESG disclosure quality variable. I contribute to