You are on page 1of 16

Received: 26 December 2017 Revised: 20 August 2018 Accepted: 4 September 2018

DOI: 10.1002/bse.2242

RESEARCH ARTICLE

Corporate carbon risk, voluntary disclosure, and cost of capital:


South African evidence
Tesfaye T. Lemma1 | Martin Feedman1 | Mthokozisi Mlilo2 | Jin Dong Park1

1
Department of Accounting, College of
Business and Economics, Towson University, Abstract
MD, USA The study examines the interplay among corporate carbon risk, voluntary disclosure,
2
School of Economic and Business Sciences,
and cost of capital within the context of South Africa, a “rising power” in the climate
Faculty of Commerce, Law and Management,
University of the Witwatersrand, policy debate. We develop a system of simultaneous equations models and analyze
Johannesburg, South Africa
data drawn from firms traded on the Johannesburg Securities Exchange (JSE), for
Correspondence
Tesfaye T. Lemma, Department of Accounting,
the period 2010 to 2015, using the three‐stage least squares procedure. We find that
College of Business and Economics, Towson voluntary carbon disclosure is associated with lower overall (and equity) cost of cap-
University, MD, USA.
Email: tlemma@towson.edu
ital, after controlling for corporate carbon risk. We also find that firms with higher car-
bon risk tend to provide better quality carbon disclosure and signal the possibility of
high carbon risk to avoid negative market reactions resulting from concealing carbon
information. Although the capital market does not appear to incorporate individual
firm's carbon risk exposure into the required cost of capital, we find that it generally
requires higher returns for companies operating in carbon‐intensive sectors. These
findings suggest that firms could exploit the virtues of voluntary carbon disclosure
to reduce their overall (and equity) cost of capital. Our findings also imply that regu-
lators and policymakers could point to the cost of capital reducing role of voluntary
disclosure to lure firms into voluntarily providing superior quality carbon disclosures.

KEY W ORDS

corporate carbon risk, cost of capital, South Africa, voluntary disclosure

1 | I N T RO D U CT I O N research decomposes firm value into expected cash flow and cost of
capital (Barth, Cahan, Chen, & Venter, 2017). Numerous studies show
Climate change has received heightened research and policy attention that corporate capital providers incorporate the firm's carbon risk into
due to its potential to cause not only disturbances to complex ecological their capital allocation decisions (Cogan, 2006; Jung et al., 2018; Kim,
systems but also to inflict unprecedented damages to economies and An, & Kim, 2015; Kolk, Levy, & Pinkse, 2008; Li, Eddie, & Liu, 2014; Saka
the health of human population (Jung, Herbohn, & Clarkson, 2018; & Oshika, 2014; Subramaniam, Wahyuni, Cooper, Leung, & Wines,
Labatt & White, 2007). Although there has been a push towards a low 2015; Weber, 2012). The present study focuses on the interrelationship
carbon economy, the response from the corporate world varies consid- among cost of capital, corporate carbon risk, and voluntary disclosure.
erably, ranging from those that implement proactive approaches to mit- Despite the increased research interest on how the capital market
igating corporate carbon risk to others that implement reactive rewards (or punishes) corporate carbon risk and the mediating role of
solutions (Hart, 1995; Liao, Luo, & Tang, 2015). An emerging strand of carbon disclosure, there is virtually no empirical evidence based on the
research establishes that corporate carbon risk impacts firm value context of countries that have come to be known as the “rising pow-
through several mechanisms (Chapple, Clarkson, & Gold, 2013; ers”—China, India, Brazil, and South Africa—in the climate policy
Clarkson, Li, Richardson, & Vasvari, 2008; Garber & Hammitt, 1998; debate. This is a significant omission because these economies are
Hughes, 2000; Labatt & White, 2007; Matsumura, Prakash, & Vera‐ “not only gaining weight in the international political system but also
Muñoz, 2014; Peters & Romi, 2014). The hitherto value relevance have a growing impact on the ecosystem as they struggle to sustain

Bus Strat Env. 2018;1–16. wileyonlinelibrary.com/journal/bse © 2018 John Wiley & Sons, Ltd and ERP Environment 1
2 LEMMA ET AL.

economic growth and development” (Never, 2012). To an increasing Recent literature suggests that “carbon emissions have become an
extent, the future of human life on our planet would be influenced essential element in analyzing a company's risk profile” (Matsumura
by what goes on in these countries (Schmitz, 2017). Furthermore, vir- et al., 2014). We note two contrasting views on the association
tually all the available empirical evidence has focused on reporting between corporate carbon risk and voluntarily carbon disclosure. The
regimes where greenhouse gas (GHG) emissions reporting is somehow first view submits that low carbon risk firms would have higher incen-
regulated. Insights grounded on empirical examination of the eco- tives to voluntarily disclose their carbon risk information to differenti-
nomic benefits of voluntary carbon disclosure could provide informa- ate themselves from companies with higher carbon risk (Clarkson
tion that will help corporate executives in making important cost– et al., 2008; Dye, 1985; Verrecchia, 1983). Contrarily, the second view
benefit trade‐offs in allocating resources to measuring and disclosing contends that high carbon risk firms are more likely to voluntarily dis-
carbon emissions related information (Matsumura et al., 2014). close their carbon information as they would face more pressure from
South Africa provides a unique setting for four reasons. First, the stakeholders (Hahn, Reimsbach, & Schiemann, 2015; Roberts, 1992)
country's cheap and plentiful coal has created heavy dependency on and would want to convey a more positive image of the firm (Deegan,
energy‐ and emissions‐intensive industries. Also, the country's very 2007; Hahn et al., 2015). We use the inverse of the ratio of Scope 1
slow transition to low carbon, renewable, energy sources has made GHG emissions to annual sales turnover, a transformed measure of
it to be a very high GHG emissions country in the world, and by far carbon emissions intensity, as a proxy for corporate carbon risk in
the highest source of GHG emissions in Africa (Andreasson, 2017; our main analysis.
Baker, Newell, & Phillips, 2014; Hallding et al., 2011; Never, 2012; Given the potential endogeneity among corporate carbon risk,
Pegels, 2014; Schmitz, 2017). Second, despite being a very high voluntary disclosure, and cost of capital (Hassan & Romilly, 2018),
GHG emissions country, South Africa has had a voluntary carbon dis- we develop a system of simultaneous equations to examine the rela-
closure regime until very recently, which provides an opportunity to tionships using data drawn from the Johannesburg Securities
investigate the interplay among our variables of interest in a voluntary Exchange (JSE) for the period 2010 to 2015. We find that voluntary
reporting setting.1 Third, the presence of entrenched stakeholders, carbon disclosure is associated with lower overall (and equity) cost
with clashing interests, vying for influence in shaping the country's cli- of capital, after controlling for the impact of the firm's exposure to car-
mate policy and carbon risk management behavior of corporates bon risk. Further, this association is economically meaningful, with one
(Baker et al., 2014; Schmitz, 2017) promises to provide an exciting standard deviation increase in carbon disclosure score leading to a
research context for exploring corporate carbon risk exposures and 1.5% decrease in the overall cost of capital. We also find that firms
the related disclosure practices. Fourth, response rates to Carbon Dis- with higher carbon risk tend to provide better quality carbon disclo-
closure Project's (CDP)2 annual survey questionnaire suggest that sure and signal the possibility of high carbon risk to avoid negative
South African companies tend to be more enthusiastic about voluntar- market reactions resulting from concealing carbon information.
ily disclosing their carbon emission information compared with their Although the capital market does not appear to incorporate an individ-
peers in other countries in the “rising powers.”3 We investigate if this ual firm's carbon risk into the required cost of capital, we find that it
“commitment to transparency” is rewarded by the capital market. generally requires higher returns on companies operating in carbon‐
The literature on the association between voluntary carbon dis- intensive sectors. The findings suggest that firms could exploit the vir-
closure and a firm's cost of capital presents two contrasting views. tues of voluntary carbon disclosure to overcome the accentuating role
The first view contends that voluntary carbon disclosure, through its of corporate carbon risk on the overall (and equity) cost of capital.
impact on information asymmetry, estimation risk, and investor prefer- That is, high carbon risk firms could utilize voluntary carbon disclosure
ence, would have a decreasing effect on the firm's cost of capital to obtain a more favorable overall (and equity) cost of capital leading
(Beyer, Cohen, Lys, & Walther, 2010; Diamond & Verrecchia, 1991; to a suboptimal allocation of resources and yet continue to engage
Dutta & Nezlobin, 2017; Easley & O'Hara, 2004; Healy & Palepu, in activities that adversely impact the environment. It also suggests
2001; Lambert, Leuz, & Verrecchia, 2007; Merton, 1987; Verrecchia, that policy makers and regulators could point to the cost of capital
2001). The second view argues that voluntary disclosure of carbon reducing role of carbon disclosure in luring firms into voluntarily dis-
information comes with some proprietary costs and could potentially closing carbon information.
result in revelation of information that would have negative repercus- The present study's contribution is twofold. First, it illuminates
sions to the firm (Guidry & Patten, 2012; Peters & Romi, 2014; our understanding of the interaction among corporate carbon risk, vol-
Verrecchia, 1983). In our analysis, we use CDP's carbon disclosure untary disclosure, and cost of capital in a voluntary reporting setting.
score as a proxy for the quality of a firm's carbon disclosure. The findings could be used to establish the key benefits of voluntary
carbon disclosure and help us predict which firms may take avoidance
1 actions or lobby against carbon regulation (Leuz & Wysocki, 2016).
On the third of April 2017, the South African government gazetted the
National Greenhouse Gas Reporting Regulations which mandated certain types Second, it provides insights into corporate carbon risk and voluntary
of corporates to report their GHG emissions activity data into the National corporate carbon disclosure policies within the context of one of the
Atmospheric Emissions Inventory System (NAEIS).
“rising powers,” which would inform the climate policy debate. The
2
The CDP, an independent non‐for‐profit organization acting on behalf of hun-
rest of the paper is organized as follows. Section 2 presents a brief
dreds of institutional investors, holds the largest repository of carbon emission
information (Matsumura et al., 2014). review of background literature and hypotheses development. Section
3
The response rate to the CDP annual survey of South African companies is the 3 outlines the research design including issues of sample selection,
second highest in the world with 83% response rate (CDP, 2013). variable identification and definition, model specification, and
LEMMA ET AL. 3

econometric procedures. Section 4 presents the results and discus- Hypothesis 1. : Voluntary carbon disclosure is nega-
sions, and Section 5 concludes. tively associated with cost of capital.

2.2 | Voluntary carbon disclosure and corporate


2 | B A CKG R O U N D L I T ER A T U R E A N D carbon risk
H Y P O T H E S E S D E V EL O P M EN T
Corporate carbon risk refers to “threats and opportunities that are
associated with a firm's management of carbon emissions” (Jung
2.1 | Voluntary carbon disclosure and cost of capital et al., 2018; Subramaniam et al., 2015). It poses regulatory, physical,
The wider literature on information asymmetry and corporate disclo- and business risks (Labatt & White, 2007) each of which are likely to
sure contends that voluntary disclosure reduces the firm's cost of impact firms to varying degrees “depending on the actions that firms
capital (Healy & Palepu, 2001) through the following mechanisms. undertake to confront or even pre‐empt the risks and challenges
First, voluntary disclosure reduces information asymmetries between posed by climate change” caused by carbon emissions (Jung et al.,
the firm and its capital providers, which in turn reduces monitoring 2018). Voluntary disclosure theory suggests that companies with
costs. Second, firms could use voluntary disclosure of relevant infor- lower carbon risk have the incentive to voluntarily disclose their car-
mation to increase investor demand for their securities, which in turn bon risk information as they would want to differentiate themselves
increases market liquidity of their securities (Diamond & Verrecchia, from companies which have higher carbon risk (Clarkson et al., 2008;
1991; Richardson, Welker, & Hutchinson, 1999; Verrecchia, 2001) Dye, 1985; Verrecchia, 1983). Likewise, signaling theory contends that
and risk‐sharing (Merton, 1987). Third, voluntary disclosures help companies with low carbon risk are more likely to voluntarily disclose
improve companies' information environment by enhancing analysts' carbon information as they cannot be easily imitated by companies
understanding of their prospects (Beyer et al., 2010) which reduces with high carbon risk, and thus, have more incentives to voluntarily
estimation of information risk (Diamond & Verrecchia, 1991; Lambert engage in carbon disclosure (Clarkson et al., 2008; Li, Richardson, &
et al., 2007; Verrecchia, 2001). Fourth, voluntary disclosure could Thornton, 1997). Thus, both voluntary disclosure and signaling theo-
affect cost of capital of a firm through its effect on the firm's real ries imply an inverse association between voluntary carbon disclosure
decisions including production, investment, and capital allocation and corporate carbon risk (Clarkson et al., 2008).
choices (Dutta & Nezlobin, 2017; Easley & O'hara, 2004; Lambert Sociopolitical theories of which stakeholder theory and legitimacy
et al., 2007). theory are the most prominent explain disclosure based on companies'
On the other hand, Peters and Romi (2014) submit that providing response to perceived outside pressure (Hahn et al., 2015). These the-
carbon disclosures could create adverse consequences by providing ories contend that a firm could use disclosure of information about its
information that might encourage government agencies to investigate carbon risk to convey a positive image about itself and to satisfy the
firms, invoke costly litigation, providing competitors with information demands for transparency. That is, these theories submit that firms
about firm‐specific sustainability strategies, and inciting potentially with high carbon risk with threatened legitimacy would voluntarily
negative attention from environmental advocacy groups. In the same engage in disclosure of carbon‐related information to deflect or nullify
vein, both Verrecchia (1983) and Guidry and Patten (2012) contend suspicion or doubt about their environmental activities (Adams, 2004;
that disclosure of (carbon) information comes with proprietary costs Gray, Kouhy, & Lavers, 1995; Hughes, Anderson, & Golden, 2001).
in the form of actual costs of disclosure as well as eventual conse- Stakeholder theory contends that firms are accountable to all their
quences that accompany such decisions. Thus, the nature of influence stakeholders, not just their investors. Because these stakeholders have
that voluntary carbon disclosure has on the cost of capital of a firm multiple demands, and the demands may in fact conflict, the firm
remains an empirical question. needs to provide information to meet these demands (Roberts, 1992).
The empirical literature is seemingly dominated by studies that On the other hand, legitimacy theory is concerned with whether a firm
present evidence consistent with the first view that the cost of capital conforms to society's expectations of the firm (Chen & Roberts, 2010;
of a firm is likely to decrease with voluntary carbon disclosure. For Deegan, 2007). By disclosing carbon risk information, the firm can be
instance, Clarkson, Fang, Li, and Richardson (2013) document evi- seen as fulfilling its social contract (Hahn et al., 2015). Both stake-
dence that strongly corroborates the conjecture that environmental holder and legitimacy theories suggest a positive association between
disclosures enhance firm value. In a similar vein, in a study that voluntary carbon disclosure and corporate carbon risk (Clarkson et al.,
focused on S&P 500 firms, He, Tang, and Wang (2013) demonstrate 2008; Patten, 2002).
that cost of equity capital is negatively associated with carbon disclo- Again, available empirical literature on the association between
sure and the association is weaker for companies with good carbon voluntary corporate disclosure and corporate environmental risk pre-
performance. Both Freedman and Patten (2004) and Plumlee, Brown, sents mixed results. Consistent with voluntary disclosure and signaling
Hayes, and Marshall (2015) reach similar conclusions. On the other theories, Luo and Tang (2014) demonstrate that carbon disclosure and
hand, empirical investigations by Murray, Sinclaire, Power, and Gray corporate carbon risk are negatively associated. Both Al‐Tuwaijri,
(2006) document that there is no association between environmental Christensen, and Hughes (2004) and Clarkson et al. (2008) find evi-
disclosures and firm performance. Consistent with the dominant dence that supports voluntary disclosure and signaling theories. On
empirical evidence, we propose our first hypothesis (in alternative the other hand, corroborating sociopolitical theories, several studies
form) as follows: show that firms that were poorer environmental performers made
4 LEMMA ET AL.

more extensive environmental disclosures (Bewely & Li, 2000; Hypothesis 3. : There is positive association between
Campbell, 2003; Cho, Freedman, & Patten, 2012; Cho & Patten, corporate carbon risk and cost of capital.
2007; Clarkson, Li, Richardson, & Vasvari, 2011; Hassan & Romilly,
Freedman and Patten (2004) examined stock market reaction to the
2018; Hughes et al., 2001). Likewise, He et al. (2013) provide evidence
announcement of a firm's environmental performance and found a
corroborating a positive association between carbon disclosure
positive association between environmental performance and stock
and corporate carbon risk. Still, some studies found no statistically
market reaction; that is, the market reacted negatively to firm's pollu-
significant association between the two variables (see, for instance,
tion performance. However, the authors also found that firms could
Wiseman, 1982; Freedman & Wasley, 1990). In line with the dominant
mitigate the impact of poor performance on market reaction through
empirical evidence, we propose our second hypothesis (in alternative
extensive environmental disclosures. In a similar vein, He et al.
form) regarding the association between voluntary carbon disclosure
(2013) argue that firms with good carbon performance face a smaller
and corporate carbon risk as follows:
legitimacy problem, and investors are probably less sensitive to the
Hypothesis 2. : Voluntary carbon disclosure is positively carbon disclosure of such firms; thus, the association between volun-
associated with corporate carbon risk. tary carbon disclosure and a firm's cost of capital would be weaker
for firms with low carbon risk. Thus, we state the fourth hypothesis
(in the alternative form) as follows:
2.3 | Corporate carbon risk and cost of capital
Hypothesis 4. : The association between voluntary car-
Climate change poses risks to a firm's value through several channels
bon disclosure and cost of capital is weaker for firms with
including the prospect of increased costs of compliance to carbon pol-
lower carbon risk.
icies; the prospect of increased costs of operation and reduced reve-
nue due to bad weather, uncontrollable accidents, and so forth; and
reputational and market value penalties from increasing GHG emis- 3 | R E S E A RC H D E S I G N
sions (Chapple et al., 2013; Clarkson et al., 2008; Garber & Hammitt,
1998; Hughes, 2000; Labatt & White, 2007; Peters & Romi, 2014). In sync with prior studies (e.g., Sharfman & Fernando, 2008), we use
These impacts of climate change have direct bearing on firm's future the weighted average cost of capital (WACCi,t), described in the sem-
cash flows and, by extension, its ability to repay debts and maintain inal work of Modigliani and Miller (1958), as a proxy for a firm's cost
regular dividend payments. Furthermore, capital providers may face of capital, based on estimates obtained directly from INET BFA data-
reputational risks if their financial dealings with high carbon risk firms base. Both Al‐Tuwaijri et al. (2004) and Ullmann (1985) contend that
are viewed in a negative light by stakeholders which, in turn, may a firm's overall strategy is likely to codetermine its economic perfor-
affect a capital provider's ability to attract future customers, and thus, mance, environmental performance, and voluntary disclosure implying
its's ability to generate future revenue streams. Thus, several scholars potential endogeneity problems among the variables. And failure to
contend that capital providers incorporate the firm's carbon risk into account for endogeneity among the variables could result in biased
their capital allocation decisions (Cogan, 2006; Jung et al., 2018; Kolk and inconsistent results if we use ordinary or generalized‐least
et al., 2008; Subramaniam et al., 2015; Weber, 2012). squares single‐equation estimation procedures (see also Hassan &
The available empirical evidence corroborates the argument that a Romilly, 2018). In sync with this argument, He et al. (2013) imple-
firm's carbon risk exacerbates its cost of capital. For instance, Li et al. mented joint estimates of the interaction among carbon disclosure,
(2014) show that a firm's cost of equity increases with its emissions carbon performance, and cost of capital. Thus, we develop a system
intensity, although they report that they find little evidence to corrob- of simultaneous equations for examining joint interactions among
orate such an association with respect to its cost of debt. Similarly, corporate carbon risk, voluntary disclosure, and cost of capital as
Chen and Gao (2012) demonstrate that both costs of equity and debt follows:
increase with increase in a firm's carbon emissions. Still, Jung et al.
(2018) demonstrate a positive association between a firm's cost of
WACCi;t ¼ DSCRi;t þ BETAi;t þ ICOV i;t þ LEV i;t þ SIZEi;t þ GRW i;t
debt and its carbon emissions and that the association is weaker for
þ PRFT i;t þ TANGi;t þ VLT i;t þ INTENSIVEi;t ; (1a)
firms that voluntarily disclose their carbon emissions. In a related vein,
several empirical studies demonstrated that lending institutions, faced
DSCRi;t ¼ CRi;t þ LEV i;t þ SIZEi;t þ GRW i;t þ PRFT i;t þ INTENSIVEi;t (2)
with the challenges of potential damages to their reputation and abil-
ity to generate future customers, incorporate borrowers' environmen-
tal/carbon risk exposures into their capital allocation decisions CRi;t ¼ DSCRi;t−1 þ LEV i;t þ SIZEi;t þ GRW i;t þ PRFT i;t
(Coulson & Monks, 1999; Subramaniam et al., 2015; Thompson, þ INTENSIVEi;t (3)

1998; Thompson & Cowton, 2004). Nonetheless, several studies


find no association between environmental performance and a firm's where WACCi,t denotes the weighted average cost of capital of firm i
economic performance (e.g., Freedman & Jaggi, 1992; Rockness, E D
in year t estimated using WACCi;t ¼ KEi;t þ KDi:t ð1 − T Þ;
Schlachter, & Rockness, 1986). In sum, the results of empirical studies DþE DþE
are inconclusive. Thus, in line with the dominant empirical evidence, KEi,t refers to the cost of equity capital of firm i in year t estimated
we propose our third hypothesis (in alternative form) as follows: using the capital asset pricing model (CAPM); KDi.t denotes the cost
LEMMA ET AL. 5

of debt of firm i in year t; E is market value of the firm's equity; D is former is designed to capture a firm's potential carbon liabilities, the
market value of the firm's debt; and T is the firm's rate of corporate latter is designed to capture corporate carbon risk exposure (Clarkson,
taxation (Jung et al., 2018; Sharfman & Fernando, 2008). Li, Pinnuck, & Richardson, 2015; Jung et al., 2018; Wang, Li, &
Unlike prior studies that employ self‐constructed indices (Plumlee Gao, 2013). In addition, relative GHG emissions measures facilitate
et al., 2015), or dummy variables based on whether or not carbon comparisons between companies and reduction potentials become
information was publicly disclosed (Liao et al., 2015), we use the car- more transparent (Hoffmann & Busch, 2008). Thus, the corporate car-
bon disclosure score that is publicly available on annual CDP reports, bon risk (CRi,t) variable in Equation 2, is based on the GHG emissions
to proxy voluntary carbon disclosure, for two reasons. First, the CDP's data available on the annual CDP reports and computed as the inverse
carbon disclosure score is more comprehensive and covers many of the ratio of Scope 1 GHG emissions to sales revenue.4 Higher
dimensions of a firm's activity regarding climate change (Luo, Lan, & (lower) values of CRi,t suggest poorer (better) carbon efficiency and
Tang, 2012; Tang & Luo, 2014). Second, CDP has been identified as higher (lower) corporate carbon risk. We control for leverage (LEVi,t),
a leading source of quality carbon emissions information disclosure as more geared companies would have stricter debt covenants and
(Griffin, Lont, & Sun, 2017; Rankin, Windsor, & Wahyuni, 2011). The creditors would require more information to monitor the behavior of
carbon disclosure variable (DSCRi,t) in our models refers to the carbon such firms (Leftwich, Watts, & Zimmerman, 1981). We also control
disclosure score of firm i in year t. It measures the quality and compre- for firm size (SIZEi,t), as it captures several factors including financial
hensiveness of the information provided in a company's response to resources, political costs, and information asymmetry (Lemma et al.,
CDP's annual climate change survey questionnaire. The score also 2018) that motivate carbon disclosure (Shan & Taylor, 2014; Stanny
serves as a metric of good internal management, an understanding & Ely, 2008).
of climate change issues and company transparency on climate There are two contrasting views regarding the association
change; higher values indicate that the firm's carbon‐related strategies between firm growth (GRWi,t) and voluntary carbon disclosure. On
and actions are more transparent and visible than those with lower the one hand, it is argued that growing firms tend to be more finan-
values (Tang & Luo, 2014). cially constrained, and hence, may lack the resources needed to volun-
Following prior research (Gray, Koh, & Tong, 2009; Jung et al., tarily produce and disclose information on their carbon risk (Stanny &
2018), our cost of capital model (Equation 1a) includes firm size (SIZEi,t), Ely, 2008). On the other hand, it could be argued that growing firms
a proxy for collateral and inverse information asymmetry, as a control tend to have higher information asymmetry, which may induce them
variable. Both Jung et al. (2018) and Chen and Gao (2012) submit that into voluntarily disclosing their carbon information to attract potential
firm's default risk, and hence its cost of capital, increases with leverage investors (Dhaliwal, Li, Tsang, & Yang, 2011; Kim & Shi, 2011; Lemma
(LEVi,t). Companies with high growth prospects (GRWi,t) tend to have & Negash, 2012, 2013). Companies with higher profits (PRFT i,t) may
higher levels of information asymmetry (Lemma, Negash, Mlilo & want to signal that they have high quality earnings and that they could
Lulseged, 2018; Lemma, 2015; Lemma & Negash, 2011, 2012, easily afford the expenditures needed to produce and disclose carbon
2013), which may increase the risk premium that lenders would information to investors (Stanny & Ely, 2008). Finally, firms operating
require to provide capital to a firm. in carbon‐intensive sectors (INTENSIVEi,t) are exposed to higher regu-
More profitable (PRFT i,t) firms tend to have lower default risk and latory risks, and thus, may have higher incentives to voluntarily engage
benefit from lower cost of capital (Dhaliwal, Gleason, Heitzman, & in carbon disclosure (Labatt & White, 2007).
Melendrez, 2008; Ge & Kim, 2010; Lopes & Alencar, 2010). Bank- We extrapolate the argument in Al‐Tuwaijri et al. (2004) and
ruptcy recovery rates tend to be higher in firms with higher holdings control for prior year's disclosure (DSCRi,t−1) as a proxy for an upper
of tangible assets (TANGi,t), reducing a creditor's loss on repayment bound for current year's carbon risk (CRi,t), in Equation 3. Al‐Tuwaijri
and lowering the cost of debt (Bharath, Sunder, & Sunder, 2008; Jung et al. (2004) contend that investors' expectations of corporate carbon
et al., 2018). The literature suggests that lenders charge risk premia to risk are conditioned on information provided by prior carbon disclo-
those borrowers which exhibit volatile and less predictable earnings sures. A firm may feel that it is necessary or prudent to signal the
(VLT i,t) increasing their cost of debt (Francis, Khurana, & Pereira, possibility of high carbon risk to investors, thereby reducing the pos-
2005; García‐Teruel, Martínez‐Solano, & Sánchez‐Ballesta, 2014; sibility of litigation if firm value fails (Al‐Tuwaijri et al., 2004; Hassan
Trueman & Titman, 1988). Firms with higher interest coverage & Romilly, 2018). We expect a negative association between prior
(ICOVi,t) are likely to have greater operating income to cover interest year's carbon disclosure score (DSCRi,t−1) and the current year's proxy
payments, and hence, may have lower default risk. Due to for carbon risk (CRi,t). We also control for variables that account for
intersectoral differences in operational technologies and processes,
firms in carbon‐intensive sectors (INTENSIVEi,t) such as materials,
energy, and utilities face higher degrees of carbon‐risks, and hence, 4
The annual CDP reports present three “scopes” of GHG emissions: scopes 1, 2,
are likely to incur higher carbon‐related costs and liabilities, compared and 3. Scope 1 consists of direct emissions that come from sources owned or
controlled by the firm. Scope 2 emissions refer to indirect emissions caused
with those in less carbon‐intensive sectors (Clarkson et al., 2008; Luo
by the firm's consumption of electricity, heat, cooling, or steam brought into
et al., 2012; Stanny & Ely, 2008). its reporting boundary. Scope 3 emissions are those that are generated by
Prior studies use either absolute or relative GHG emissions mea- employee business travel, suppliers, customers, and contractors. We use scope
1 GHG emissions in our main analysis because they account for most of the
sures to proxy corporate carbon risk exposure (Jung et al., 2018; Salo
reported GHG emissions and a firm can be held directly accountable for such
& Ast, 2009). Nonetheless, these two proxies are designed to capture emissions (Jung et al., 2018; Matsumura et al., 2014). We consider scope 2
different dimensions of a firm's carbon emissions profile. Although the emissions later in our sensitivity analyses.
6 LEMMA ET AL.

intercompany differences in financial resources, carbon risks, and currency exchange rate differences into account.5 The mean carbon
management capabilities that help explain intercompany variations disclosure score of 83.956, compared with an average of 64.16 for
in corporate carbon risk (see Equation 3 above) by incorporating the S&P 500 firms (He et al., 2013), arguably suggests that the JSE
leverage (LEVi,t), firm size (SIZEi,t), growth prospect (GRWi,t), profitabil- 100 firms are more transparent when it comes to carbon disclosure.
ity (PRFT i,t), and intensive industry (INTENSIVEi,t), into our models Consistent with prior studies that conduct univariate analyses
(Clarkson et al., 2011; de Villiers, Naiker, & van Staden, 2011; between high and low carbon risk firms (Jung et al., 2018), we com-
McKendall, Sánchez, & Sicilian, 1999; Walls, Berrone, & Phan, 2012). pare differences in the means (and medians) of all variables. Panel A
Next, we model the attenuating (or accentuating) role of a firm's presents statistics for the sample partitioned into high (128 firm‐year
carbon risk (CRi,t) on the association between voluntary carbon disclo- observations) and low (145 firm‐year observations) carbon risk firms
sure (DSCRi,t) and cost of capital (WACCi,t). To this end, we bifurcate based on the median value of carbon risk (CRi,t). It indicates that the
the sample into high and low carbon risk companies by creating a median cost of capital of high carbon risk firms (8.8%) is lower, albeit
dummy variable (HCRi,t) that takes a value of 1 if a firm's CRi,t is insignificantly so (p‐value = 0.578), than that of low‐risk firms (9.6%).
greater than the sample median (i.e., low carbon risk subsample) and Panel A also shows that the median carbon disclosure score (DSCRi,t)
0 otherwise (i.e., high carbon risk subsample). Then, we develop an of high carbon risk firms (86) is greater, although insignificantly so
augmented model of Equation 1 by including the dummy variable (p value = 0.120), than that of low carbon firms (85). In terms of the
(HCRi,t) and its interaction with the carbon disclosure (DSCRi,t*HCRi,t) control variables, high carbon risk firms tend to have lower interest
variable. The purpose of the interaction term is to determine whether coverage, more tangible assets, lower profitability, higher earnings
a firm's carbon risk attenuates (or accentuates) the association volatility, and lower growth compared with low carbon risk firms.
between voluntary carbon disclosure and a firm's cost of capital. Thus, These results are consistent with those reported in Jung et al.,
the revised system of equation would be the following: (2018) for Australia.
Panel B presents statistics for the sample partitioned into high

WACCi;t ¼ DSCRi;t þ HCRi;t þ HCRi;t *DSCRi;t þ BETAi;t þ ICOV i;t (128 firm‐year observations) and low (145 firm‐year observations) dis-
þ LEV i;t þ SIZEi;t þ GRW i;t þ PRFT i;t þ TANGi;t closure firms based on the median carbon disclosure score. Consistent
þ VLT i;t þ INTENSIVEi;t (1b) with our expectations, Table 1 (Panel B) reveals that the median cost
of capital of high carbon disclosure firms (8.2%) is significantly lower

DSCRi;t ¼ CRi;t þ LEV i;t þ SIZEi;t þ GRW i;t þ PRFT i;t þ INTENSIVEi;t (2) (p value = 0.000), than that of low carbon disclosure firms (10.0%).
Panel B also shows that the level of corporate carbon risk for high dis-
closure firms (130.583) is slightly lower, albeit insignificantly (p
CRi;t ¼ DSCRi;t−1 þ LEV i;t þ SIZEi;t þ GRW i;t þ PRFT i;t
value = 0.704), than that of low disclosure firms (137.493). In terms
þ INTENSIVEi;t (3)
of the control variables, high carbon disclosure firms are likely to have
bigger firm size, more tangible assets, and higher profitability.
Table 2 presents pairwise correlation coefficients (both the para-
4 | SAM P LE AND R ESULTS metric and nonparametric) for the variables used in our tests. As is
to be expected, there are significant correlations among many of the
4.1 | Sample and data variables. Of note, consistent with our expectations and in line with
our observations in the bivariate tests reported in Table 1, Table 2
Financial and other firm‐level data for this study were from INET BFA
shows a negative correlation between the cost of (debt, equity, and
database, whereas data on carbon emissions and disclosure were
overall) capital and a firm's carbon disclosure score.
hand‐collected from publicly available CDP reports. Because the
annual CDP survey questionnaire covers only the top 100 firms
(JSE100), our initial sample is composed of JSE100 firms for the 6‐year
period 2010 to 2015. We excluded firms with missing carbon disclo- 4.3 | Regression results
sure scores. This process resulted in a preliminary sample of 417 Our primary goal is to investigate the interplay among corporate car-
firm‐year observations. We further eliminated companies with missing bon risk, carbon disclosure and cost of capital. Table 3 reports the
financial information resulting in a final sample of 98 firm‐year obser- results based on three‐stage least squares (3SLS) regressions for our
vations. As this is a relatively small sample, we also examine the data simultaneous equations models. As discussed elsewhere, the results
using nonparametric procedures that are robust to small samples. on the left side of Table 3 pertain to a system of equations which
includes a reduced form of the cost of capital model (that includes only
4.2 | Preliminary results the carbon disclosure score) and models for carbon disclosure and cor-
porate carbon risk. On the other hand, the results on the right side of
Table 1 presents the descriptive statistics and shows that the median
Table 3 pertain to the full model which augments the reduced model
cost of capital for all sample firms is 9.3%. The mean value of 4,185.63
with a new variable, HCRi,t, and an interaction term, HCRi,t * DSCRi,t.
for our corporate carbon risk (CRi,t) indicates that JSE 100 firms,
Each model includes all control variables as described in Section 5.
on average, emit 0.2389t of Scope 1 GHG emissions per ZAR1,000
of sales revenue, which is very high compared with 0.4838t of 5
ZAR denotes the South African official currency, the rand whereas the A$
A$1,000 for Australian firms (Jung et al., 2018), after taking the denotes the Australian dollar.
LEMMA ET AL. 7

TABLE 1 Summary statistics and results of T‐tests (Wilcoxon‐tests) for mean (and median) comparisons

Panel A: CRi,t (bottom 50% = 0, top 50% = 1) Panel B: DSCRi,t (bottom 50% = 0, top 50% = 1)
Full Sample CRi,t = 0 CRi,t = 1 DSCRi,t = 0 DSCRi,t = 1
t‐stat Wilcoxon t‐stat Wilcoxon
Mean Median Mean Median Mean Median (p‐value) (P‐value) Mean Median Mean Median (p‐value) (P‐value)
LEVi,t 0.538 0.531 0.528 0.502 0.547 0.587 0.475 0.126 0.522 0.525 0.552 0.532 0.268 0.387
SIZEi,t 16.579 16.641 16.525 16.569 16.632 16.935 0.484 0.143 16.530 16.518 16.624 16.758 0.536 0.099
ICOVi,t 16.400 6.221 11.648 5.166 21.654 7.979 0.001 0.009 17.805 6.709 15.187 6.073 0.395 0.245
TANGi,t 0.628 0.655 0.691 0.717 0.565 0.488 0.000 0.000 0.585 0.622 0.667 0.689 0.004 0.004
PRFT i,t 0.129 0.097 0.114 0.091 0.144 0.110 0.027 0.035 0.143 0.108 0.116 0.084 0.046 0.014
VLT i,t 0.530 0.329 0.558 0.382 0.502 0.302 0.374 0.024 0.502 0.325 0.557 0.331 0.380 0.498
GRWi,t 0.075 0.069 0.054 0.059 0.095 0.083 0.049 0.063 0.067 0.068 0.082 0.076 0.493 0.670
BETAi,t 0.547 0.560 0.529 0.502 0.565 0.659 0.654 0.452 0.566 0.596 0.521 0.442 0.615 0.493
DSCRi,t 83.956 85.000 84.657 86.000 83.250 85.000 0.334 0.120 73.805 77.000 92.917 94.000 0.000 0.000
CRi,t 4185.63 131.352 38.293 26.990 8363.5 627.55 0.001 0.000 3982.78 137.943 4364.69 130.583 0.883 0.704
KEi,t 10.2 10.1 10.3 10.2 10.1 10.0 0.723 0.963 10.7 10.4 9.8 9.6 0.010 0.006
KDi,t 9.5 6.1 8.8 6.2 10.3 6.0 0.397 0.625 10.2 6.7 8.9 5.8 0.429 0.024
WACCi,t 9.5 9.3 9.4 8.8 9.5 9.6 0.805 0.578 10.2 10.0 8.8 8.2 0.001 0.000

All continuous variables are winsorized at the 1st and 99th percentiles. LEVi,t is the sum of current liabilities and non‐current liabilities of firm i in year t
scaled by its total assets. SIZEi,t is the natural logarithm of annual sales turnover of firm i in year t. ICOVi,t is earnings before interest and tax of firm i in year
t scaled by its interest expenses. TANGi,t is total fixed assets of firm i in year t scaled by its total assets. PRFT i,t is earnings before interest and taxes of firm i
in year t scaled by its total assets. VLT i,t is a five‐year rolling standard deviation of earnings before interest and taxes of firm i in year t. GRWi,t is the first
difference of the natural logarithm of annual sales turnover of firm i in year t. BETAi,t is the unleveraged beta of firm i in year t, obtained from INET BFA
database. DSCRi,t is carbon disclosure score of firm i in year t, obtained from the publicly available annual CDP reports. CRi,t is the inverse of Scope 1 GHG
emissions of firm i in year t, as reported on the annual CDP reports, scaled by its total annual sales turnover. KEi,t is the cost of equity capital of firm i in year
t estimated using the CAPM and is obtained from INET BFA database. KDi,t is annual interest paid of firm i in year t scaled by its total interesting bearing
debt and is obtained from the INET BFA database. WACCi,t is the weighted average cost of capital of firm i in year t as described in Modigliani and Miller
(1958) and is obtained from the INET BFA database.

TABLE 2 Spearman/Pearson correlation coefficients

LEVi,t SIZEi,t ICOVi,t TANGi,t PRFT i,t VLT i,t GRWi,t BETAi,t DSCRi,t CRi,t KEi,t KDi,t WACCi,t
LEVi,t 1 0.216* −0.235* −0.359* 0.054 −0.239* −0.100 −0.177* −0.013 0.217* −0.114 0.032 −0.224*
SIZEi,t 0.242* 1 0.259* −0.330* 0.212* −0.136* 0.059 0.243* 0.191* 0.202* 0.187* 0.070 0.155*
ICOVi,t −0.324* −0.018 1 −0.285* 0.646* −0.261* 0.023 0.298* 0.092 0.132* 0.069 −0.079 0.125*
TANGi,t −0.075 −0.338* −0.284* 1 −0.085 0.350* −0.005 −0.027 0.090 −0.272* −0.078 −0.035 −0.136*
PRFT i,t −0.107* 0.215* 0.474* −0.226* 1 −0.379* 0.006 0.087 0.005 0.068 −0.131* 0.020 −0.109
VLT i,t −0.176* −0.206* −0.079 0.351* −0.219* 1 0.045 0.116 0.094 −0.063 0.136 * 0.098 0.177*
GRWi,t 0.032 0.011 0.031 −0.010 0.077 0.080 1 0.030 0.050 0.069 0.075 −0.093 0.007
BETAi,t −0.184* 0.223* 0.150* −0.025 0.138* 0.036 −0.075 1 −0.020 0.005 0.344* 0.010 0.353*
DSCRi,t 0.066 0.074 −0.077 0.186* −0.069 0.071 0.010 −0.015 1 −0.016 −0.192* −0.169* −0.232*
CRi,t 0.124* −0.175* −0.080 0.117* −0.062 0.211* 0.054 −0.076 0.017 1 −0.107* 0.010 0.067
KEi,t −0.211* 0.133* 0.051 −0.047 −0.050 0.183* 0.032 0.227* −0.185* −0.089 1 0.100 0.769*
KDi,t −0.032 0.057 −0.041 −0.103* −0.088 0.019 −0.047 0.003 −0.108* −0.053 0.057 1 0.483*
WACCi,t −0.233* 0.181* 0.089 −0.115* −0.031 0.087 −0.065 0.205* −0.223* −0.100 0.702* 0.404* 1

Note. All continuous variables are winsorized at the first and 99th percentiles. The Spearman (Pearson) correlations are above (below) the diagonal. LEVi,t is
the sum of current liabilities and noncurrent liabilities of firm i in year t scaled by its total assets. SIZEi,t is the natural logarithm of annual sales of firm i in
year t. ICOVi,t is earnings before interest and tax of firm i in year t scaled by its interest expenses. TANGi,t is total fixed assets of firm i in year t scaled by its
total assets. PRFT i,t is earnings before interest and taxes of firm i in year t scaled by its total assets. VLT i,t is a five‐year rolling standard deviation of earnings
before interest and taxes of firm i in year t. GRWi,t is the first difference of the natural logarithm of annual sales turnover of firm i in year t. BETAi,t is the
unleveraged beta of firm i in year t, obtained from INET BFA database. DSCRi,t is carbon disclosure score of firm i in year t, obtained from the publicly avail-
able annual CDP reports. CRi,t is the inverse of Scope 1 GHG emissions of firm i in year t, as reported on the annual CDP reports, scaled by total annual sales
turnover. KEi,t is the cost of equity capital of firm i in year t estimated using the CAPM and is obtained from INET BFA database. KDi,t is annual interest paid
of firm i in year t scaled by to total interesting bearing debt and is obtained from the INET BFA database. WACCi,t is the weighted average cost of capital of
firm i in year t as described in Modigliani and Miller (1958) and is obtained from the INET BFA database.
*The estimated coefficient is statistically significant at 10% level;
**The estimated coefficient is statistically significant at 5% level;
***The estimated coefficient is statistically significant at 1% level.
8 LEMMA ET AL.

TABLE 3 Three‐stage least squares regression results of weighted average cost of capital, carbon disclosure score, and carbon risk proxied by
the inverse of Scope 1 greenhouse gas emissions scaled by annual sales turnover

Dependent variables Dependent variables


Independent Var. WACCi,t DSCRi,t CRi,t WACCi,t DSCRi,t CRi,t
DSCRi,t −0.124*** −0.125***
(−3.37) (−2.88)
DSCRi,t−1 −428.491*** −428.426***
(−3.28) (−3.28)
CRi,t −0.001*** −0.001***
(−2.60) (−2.60)
HCRi,t −1.454
(−0.20)
DSCRi,t* HCRi,t 0.013
(0.14)
BETAi,t 0.030 0.050
(0.05) (0.08)
ICOVi,t 0.001 0.001
(0.50) (0.50)
LEVi,t −2.130 10.775* 43530.12*** −2.094 10.775* 43530.52***
(−0.99) (1.81) (4.39) (−0.95) (1.81) (4.39)
SIZEi,t 1.118*** 3.108*** 2633.143 1.114*** 3.108*** 2632.993
(2.72) (3.00) (1.38) (2.71) (3.00) (1.38)
GRWi,t −0.368 −8.777* −6497.537 −0.262 −8.777* −6497.480
(−0.21) (−1.84) (−0.75) (−0.14) (−1.84) (−0.75)
PRFT i,t −7.718** 4.798 −2086.807 −7.706** 4.798 −2086.579
(−2.51) (0.60) (−0.14) (−2.50) (0.60) (−0.14)
TANGi,t −5.002* −5.127*
(−1.75) (−1.76)
VLT i,t 0.111 0.124
(0.16) (0.17)
INTENSIVEi,t 5.498*** 9.058* 12252.01 5.527** 9.058* 12251.08
(2.61) (1.70) (1.25) (2.56) (1.70) (1.25)
CONSTANT 8.481 43.009*** 7342.256 8.881 43.008*** 7339.001
(1.14) (2.81) (0.26) (1.10) (2.81) (0.26)
Industry Indicators Yes Yes Yes Yes Yes Yes
χ2 41.37*** 35.45*** 58.30*** 41.63*** 35.45*** 58.30***

Note. All continuous variables are winsorized at the first and 99th percentiles. WACCi,t is the weighted average cost of capital of firm i in year t as described
in Modigliani and Miller (1958) and is obtained from the INET BFA database. DSCRi,t is a measure of voluntary carbon disclosure of firm i in year t, obtained
from the publicly available annual CDP reports. DSCRi,t − 1 is a one year lagged carbon disclosure score of firm i in year t. CRi,t is the inverse of Scope 1 GHG
emissions of firm i in year t, as reported on the annual CDP reports, scaled by its total annual sales turnover. HCRi,t is a dummy variable that equals 1 if the
firm's CRi,t in year t is higher than the sample median, and 0 otherwise. BETAi,t denotes the unleveraged beta of firm i in year t obtained from the INET BFA
database. ICOVi,t is earnings before interest and tax of firm i in year t scaled by its interest expenses. LEVi,t is the sum of current liabilities and noncurrent
liabilities of firm i in year t scaled by its total assets. SIZEi,t is the natural logarithm of annual sales of firm i in year t. GRWi,t is the first difference of the
natural logarithm of annual sales turnover of firm i in year t. PRFT i,t is earnings before interest and taxes of firm i in year t scaled by its total assets.
TANGi,t is total fixed assets of firm i in year t scaled by its total assets. VLT i,t is a five‐rolling standard deviation of earnings before interest and taxes of firm
i in year t. INTENSIVEi,t is a dummy variable that equals 1 if the firm operates in a carbon‐intensive industry, and 0 otherwise.
*The estimated coefficient is statistically significant at 10% level;
**The estimated coefficient is statistically significant at 15% level;
***The estimated coefficient is statistically significant at 1% level.

Table 3 shows, in both the left (and the right) systems of equations, those reported in Clarkson et al. (2013), Dhaliwal et al. (2011), and He
the coefficient of DSCRi,t on WACCi,t is −0.124 (and −0.125) is signifi- et al. (2013). In terms of economic significance, the coefficient estimate
cant at 1% level (Z = −3.37 and −2.88, respectively), suggesting a nega- of −0.125 implies that a one standard deviation in carbon disclosure
tive relationship between carbon disclosure (DSCRi,t) and the overall score (DSCRi,t) leads to a decrease in the firm's overall cost of capital
cost of capital (WACCi,t), which supports the hypothesis (H1) that by 1.5% (i.e., −0.125 * 12.156).6 As expected, although larger compa-
carbon disclosure reduces the firm's cost of capital, through its effect nies (p value = 0.006) and companies operating in carbon‐intensive
on investor preferences, information asymmetry, and estimation of sectors (p value = 0.009) tend to incur higher overall cost of capital,
information risk (Richardson et al., 1999). In addition, it is also in sync firms with higher profitability (p value = 0.012) and those with more
with the argument that carbon disclosure may be seen as a firm's cred- tangible assets (p value = 0.081) tend to incur lower overall cost of capital.
ible commitment to environmental issues in its long‐term strategic and
6
production systems (Plumlee et al., 2015). This finding is similar with 12.156 refers to the standard deviation of carbon disclosure score for our sample.
LEMMA ET AL. 9

TABLE 4 3SLS regression results of weighted average cost of capital, carbon disclosure score, and carbon risk proxied by the inverse of Scope 1
& Scope 2 GHG emissions scaled by sales turnover

Dependent Variables Dependent Variables


Independent Var. WACCi,t DSCRi,t CRi,t WACCi,t DSCRi,t CRi,t
DSCRi,t −0.122*** −0.155***
(−3.320) (−3.570)
DSCRi,t−1 −13.072 −13.016
(−3.520)*** (−3.510)***
CRi,t −0.005 −0.005
(−2.450)** (−2.450)**
HCRi,t −9.940
(−1.530)
DSCRi,t* HCRi,t 0.116
(1.470)
BETAi,t −0.030 −0.091
(−0.050) (−0.150)
ICOVi,t 0.007 0.009
(0.310) (0.430)
LEVi,t −2.025 10.844 1337.300 −2.292 10.842 1337.637
(−0.850) (1.790)* (4.740)*** (−0.910) (1.790)* (4.740)***
SIZEi,t 1.110*** 3.115 81.695 1.132*** 3.115*** 81.566
(2.720) (2.990)*** (1.510) (2.700) (2.990) (1.510)
GRWi,t −0.326 −8.465* −128.419 −0.504 −8.464* −128.371
(−0.180) (−1.770) (−0.520) (−0.290) (−1.770) (−0.520)
PRFT i,t −8.015** 4.974 −24.758 −8.724** 4.974 −24.563
(−2.080) (0.620) (−0.060) (−2.260) (0.620) (−0.060)
TANGi,t −4.404 −5.326*
(−1.410) (−1.660)
VLT i,t 0.161 −0.041
(0.230) (−0.060)
INTENSIVEi,t 5.357** 8.987* 357.510 6.003*** 8.987* 356.719
(2.530) (1.690) (1.280) (2.810) (1.690) (1.280)
CONSTANT 7.758 42.394*** 90.268 11.489 42.395*** 87.480
(1.010) (2.760) (0.110) (1.400) (2.760) (0.110)
Industry Indicators Yes Yes Yes Yes Yes Yes
X2 40.870*** 34.510*** 58.310*** 43.970*** 34.510*** 58.230***

Note. All continuous variables are winsorized at the first and 99th percentiles. WACCi,t is the weighted average cost of capital of firm i in year t as described
in Modigliani and Miller (1958) and is obtained from the INET BFA database. DSCRi,t is a measure of voluntary carbon disclosure of firm i in year t, obtained
from the publicly available annual CDP reports. DSCRi,t−1 is a one year lagged carbon disclosure score of firm i in year t. CRi,t is the inverse of Scope 1 plus
Scope 2 GHG emissions of firm i in year t, as reported on the annual CDP reports, scaled by its total annual sales turnover. HCRi,t is a dummy variable that
equals 1 if the firm's CRi,t in year t is higher than the sample median, and 0 otherwise. BETAi,t denotes the unleveraged beta of firm i in year t obtained from
the INET BFA database. ICOVi,t is earnings before interest and tax of firm i in year t scaled by its interest expenses. LEVi,t is the sum of current liabilities and
noncurrent liabilities of firm i in year t scaled by its total assets. SIZEi,t is the natural logarithm of annual sales of firm i in year t. GRWi,t is the first difference
of the natural logarithm of annual sales turnover of firm i in year t. PRFT i,t is earnings before interest and taxes of firm i in year t scaled by its total assets.
TANGi,t is total fixed assets of firm i in year t scaled by its total assets. VLT i,t is a five‐rolling standard deviation of earnings before interest and taxes of firm i
in year t. INTENSIVEi,t is a dummy variable that equals 1 if the firm operates in a carbon‐intensive industry, and 0 otherwise.
*The estimated coefficient is statistically significant at 10% level;
**The estimated coefficient is statistically significant at 5% level;
***The estimated coefficient is statistically significant at 1% level.

Furthermore, Table 3 shows that the coefficient of CRi,t on DSCRi,t −0.001 implies that a one standard deviation increase in the measure
is −0.001 at 1% level of significance (Z = −2.60), suggesting a positive of corporate carbon risk (CRi,t) would map into a 21.37 (i.e.,
relationship between the corporate carbon risk (CRi,t) and voluntary −0.001 * 21,365.01) increase in the carbon disclosure score (DSCRi,t).7
carbon disclosure (DSCRi,t), which supports the conjecture based on Also, the results indicate that more leveraged (p value = 0.071), larger
sociopolitical theories (H2). That is, it corroborates the hypothesis that firms (p value = 0.003) and firms operating in carbon‐intensive sectors
firms with high carbon risk with threatened legitimacy would voluntar- (p value = 0.088) tend to voluntarily provide more carbon information.
ily engage in disclosure of carbon‐related information to deflect or Contrarily, the extent of voluntary carbon disclosure decreases in firm
nullify suspicion or doubt with regard to their environmental activities level growth opportunities (p value = 0.065).
(Adams, 2004; Gray et al., 1995; Hughes et al., 2001). Similar results
were reported in Hassan and Romilly (2018) and He et al. (2013). This 7
21,365.01 refers to the standard deviation of the corporate carbon risk
association is economically meaningful too. The coefficient estimate variable.
10 LEMMA ET AL.

The results in Table 3 also indicate a negative regression Nonetheless, we do not find support for the hypothesis that firms
coefficient (−428.426) between CRi,t and one‐year‐lagged carbon with high carbon risk would incur higher cost of capital (H3) nor for
disclosure score of a firm (DSCRi,t−1), significant at the 1% level the argument that corporate carbon risk would affect the relationship
(Z = −3.28), suggesting a positive association between previous year's between voluntary carbon disclosure and a firm's cost of capital (H4).
carbon disclosure score and the current years corporate carbon risk. This finding is contrary to the findings documented in Jung et al.
This finding is consistent with the argument that firms with high car- (2018) and He et al. (2013). However, it is consistent with a number
bon risk may find it necessary or prudent to provide signals regarding of other studies that found no association between environmental
the possibility of poor carbon performance to avoid negative sur- and economic performance of firms (e.g., Freedman & Jaggi, 1992;
prises and future market punishments caused by withholding carbon Rockness et al., 1986). Finally, the positive coefficient between LEVi,t
information (Al‐Tuwaijri et al., 2004). He et al. (2013) reports similar and CRi,t suggests that companies with higher gearing tend to be com-
results. panies with a lower carbon risk exposure.

TABLE 5 Three‐stage least squares regression results of weighted average cost of capital, carbon disclosure score, and carbon risk proxied by
the inverse of Scope 1 GHG emissions scaled by total fixed assets
Dependent Variables Dependent Variables
Independent Var. WACC DSCR CR WACC DSCR CR

DSCRi,t ‐0.123*** ‐0.135***


(‐3.330) (‐2.980)
DSCRi,t−1 ‐4323.111*** ‐4317.193***
(‐3.080) (‐3.070)
CRi,t 0.000** 0.000**
(‐2.540) (‐2.540)
HCRi,t ‐2.724
(‐0.450)
DSCRi,t* HCRi,t 0.033
(0.460)
BETAi,t ‐0.039 ‐0.085
(‐0.060) (‐0.140)
ICOVi,t 0.010 0.009
(0.440) (0.390)
LEVi,t ‐1.911 10.350* 444761.700*** ‐2.010 10.351* 444797.400***
(‐0.810) (1.750) (4.180) (‐0.840) (1.750) (4.180)
SIZEi,t 1.110 3.116*** 28617.230 1.099*** 3.116*** 28603.560
(2.720) (3.000) (1.400) (2.680) (3.000) (1.400)
GRWi,t ‐0.346 ‐8.797* ‐72355.880 ‐0.503 ‐8.797* ‐72350.680
(‐0.190) (‐1.850) (‐0.780) (‐0.270) (‐1.850) (‐0.780)
PRFT i,t ‐8.321** 4.734 ‐26606.390 ‐8.254** 4.734 ‐26585.630
(‐2.150) (0.590) (‐0.170) (‐2.130) (0.590) (‐0.170)
TANGi,t ‐4.415 ‐4.375
(‐1.410) (‐1.370)
VLT i,t 0.205 0.165
(0.290) (0.230)
INTENSIVEi,t 5.378** 8.981* 122760.500 5.557** 8.981* 122676.400
(2.540) (1.690) (1.170) (2.550) (1.690) (1.160)
CONSTANT 7.709 43.012*** 62325.940 8.797 43.012*** 62029.380
(1.000) (2.810) (0.200) (1.090) (2.810) (0.200)
Industry Indicators Yes Yes Yes Yes Yes Yes
2
x 41.090*** 35.120*** 56.240*** 41.350** 35.120*** 56.220***

Note. All continuous variables are winsorized at the first and 99th percentiles. WACCi,t is the weighted average cost of capital of firm i in year t as described
in Modigliani and Miller (1958) and is obtained from the INET BFA database. DSCRi,t is a measure of voluntary carbon disclosure of firm i in year t, obtained
from the publicly available annual CDP reports. DSCRi,t−1 is a one year lagged carbon disclosure score of firm i in year t. CRi,t is the inverse of Scope 1 GHG
emissions of firm i in year t, as reported on the annual CDP reports, scaled by total fixed assets. HCRi,t is a dummy variable that equals 1 if the firm's CRi,t in
year t is higher than the sample median, and 0 otherwise. BETAi,t denotes the unleveraged beta of firm i in year t obtained from the INET BFA database.
ICOVi,t is earnings before interest and tax of firm i in year t scaled by its interest expenses. LEVi,t is the sum of current liabilities and noncurrent liabilities of
firm i in year t scaled by its total assets. SIZEi,t is the natural logarithm of annual sales of firm i in year t. GRWi,t is the first difference of the natural logarithm
of annual sales turnover of firm i in year t. PRFT i,t is earnings before interest and taxes of firm i in year t scaled by its total assets. TANGi,t is total fixed assets
of firm i in year t scaled by its total assets. VLT i,t is a five‐rolling standard deviation of earnings before interest and taxes of firm i in year t. INTENSIVEi,t is a
dummy variable that equals 1 if the firm operates in a carbon‐intensive industry, and 0 otherwise.
*The estimated coefficient is statistically significant at 10% level;
**The estimated coefficient is statistically significant at 5% level;
***The estimated coefficient is statistically significant at 1% level.
LEMMA ET AL. 11

4.4 | Sensitivity analyses To model the attenuating (or accentuating) role of a firm's carbon
risk (CRi,t) on the association between voluntary carbon disclosure
In our main analyses, we used only Scope 1 GHG emissions in the
(DSCRi,t) and equity cost of capital (KEi,t), we replace Equation 1b
computation of the corporate carbon risk variable. However, such a
with the Equation 1d and the revised system of equations is as
proxy may fail to adequately measure the extent of carbon risk to
follows:
which a firm is exposed. To examine the sensitivity of our results to
the choice of GHG emissions' categories in the measurement of cor- KEi;t ¼ DSCRi;t þ HCRi;t þ HCRi;t * DSCRi;t þ BETAi;t þ LEV i;t þ SIZEi;t þ
;
porate carbon risk, we modify our proxy to include both Scope 1 GRW i;t þ PRFT i;t þ TANGi;t þ VLT i;t þ INTENSIVEi;t
and Scope 2 GHG emissions (Chapple et al., 2013; Jung et al., 2018; (1d)
Wang et al., 2013). As reported in Table 4, our results are robust to DSCRi;t ¼ CRi;t þ LEV i;t þ SIZEi;t þ GRW i;t þ PRFT i;t
the choice of GHG emissions' categories included the computation þ INTENSIVEi;t (2)
of the corporate carbon risk variable.
Again, in our main analyses, we used the annual sales turnover as a CRi;t ¼ DSCRi;t−1 þ LEV i;t þ SIZEi;t þ GRW i;t þ PRFT i;t
scalar in measuring the corporate carbon risk variable. However, both þ INTENSIVEi;t (3)
Jung et al. (2018) and Clarkson et al. (2011) express concern that the Table 1 shows that the median cost of equity for all firms in our
annual sales turnover of a firm is, in fact, affected by the country's sample is 10.1%. It also shows that top 50% of firms (in terms of dis-
macroeconomic performance, and thus, our corporate carbon risk closure score) incur lower cost of equity capital (9.6%) compared with
proxy might be impacted by macroeconomic performance of the the bottom 50% of firms (10.4%), significant at 1% level (p
country through the scalar rather than corporate carbon risk itself. value = 0.006). On the other hand, there are no statistically significant
Thus, we alternatively use total fixed assets as the scalar, following prior differences in the cost of equity capital between high carbon risk and
studies (Clarkson et al., 2011; Jung et al., 2018) and repeat the analyses. low carbon risk firms. In terms of correlations, Table 2 shows the cost
Once again, our results remain broadly the same (see Table 5). of equity capital is negatively (not) correlated with disclosure score
(corporate carbon risk). These observations are identical with the ones
4.5 | Additional analyses reported in the main analysis.
Table 6 shows results in which the cost of equity capital (KEi,t ) is
To examine which of the component costs of capital is driving the
the dependent variable in Equations 1c and 1d. The results based on
observed interrelations among corporate carbon risk, voluntary disclo-
the revised systems of equations were generally consistent with those
sure, and the overall cost of capital, we carry out additional analysis
presented in the main analysis, except that the coefficient of DSCRi,t is
using the component costs of capital as dependent variables. First,
statistically significant (Z = −2.09, p value = 0.036) only in the system
we use the cost of equity capital (KEi,t), estimated using the CAPM
of equations with HCRi,t and HCRi,t * DSCRi,t, suggesting that cost of
described in Sharpe (1964) and Lintner (1956) and based on estimates
equity capital (KEi,t) is not as sensitive to voluntary carbon disclosure
from INET BFA database, as our dependent variable. We replace
(DSCRi,t) as the overall cost of capital. Also, although the leverage
Equation 1a with Equation 1c, and thus, the revised system of equa-
(LEVi,t) and volatility (VLT i,t) variables loaded statistically significantly
tions would be as follows:
(with p values of 0.004 and 0.081, respectively), the asset tangibility
KEi;t ¼ DSCRi;t þ BETAi;t þ LEV i;t þ SIZEi;t þ GRW i;t þ PRFT i;t (TANGi,t) variable became statistically insignificant (p value = 0.730).
þ TANGi;t þ VLT i;t þ INTENSIVEi;t (1c) The results of Equation 2 are also broadly consistent with our main
analysis, except that the marginally significant loading that the lever-
age variable (LEVi,t) had on the overall cost of capital disappeared
DSCRi;t ¼ CRi;t þ LEV i;t þ SIZEi;t þ GRW i;t þ PRFT i;t
þ INTENSIVEi;t (2) when cost of equity capital is used as the dependent variable. The
results for Equation 3 were identical.
Next, we use cost of debt (KDi,t) as the dependent variable. Thus,
CRi;t ¼ DSCRi;t−1 þ LEV i;t þ SIZEi;t þ GRW i;t þ PRFT i;t
we replace Equation 1a with Equation 1e in the system of equations:
þ INTENSIVEi;t (3)

KDi;t ¼ DSCRi;t þ ICOV i;t þ LEV i;t þ SIZEi;t þ GRW i;t þ PRFT i;t
where KEi,t is the cost of equity capital of firm i in year t estimated
þ TANGi;t þ VLT i;t þ INTENSIVEi;t (1e)
using the CAPM model, KE = r f + BETAi,t(Rm − r f ), as in Sharfman
and Fernando (2008). Pursuant to convention, we used the 10‐year DSCRi;t ¼ CRi;t þ LEV i;t þ SIZEi;t þ GRW i;t þ PRFT i;t
South African Treasury Bond as a proxy for the risk‐free rate (r f ). þ INTENSIVEi;t (2)

BETAi,t is the covariance of the market's return with the individual CRi;t ¼ DSCRi;t−1 þ LEV i;t þ SIZEi;t þ GRW i;t þ PRFT i;t
company's common stock return divided by the market's variance, þ INTENSIVEi;t (3)
which measures a market's systemic risk (Botosan, 1997; Dhaliwal where KDi,t denotes the cost of debt of firm i in year t computed by
et al., 2011; Sharfman & Fernando, 2008), and is based on estimates the ratio of interest expenses to total interest‐bearing debt, as in Jung
from INET BFA database. We set the default risk premium (Rp = Rm − r f ) et al. (2018).
at 6.0% following common practice in South Africa (Fernandez, To examine the influence of carbon risk (CRi,t) on cost of debt
Aguirreamalloa, & Avendaño, 2011; Hassan & Van Biljon, 2010). All (KDi,t) and its relationship with carbon disclosure (DSCRi,t), we intro-
the remaining variables are as defined in the earlier models. duce HCRi,t and the interaction term (HCRi,t * DSCRi,t). Thus, we replace
12 LEMMA ET AL.

TABLE 6 Three‐stage least squares regression results of cost of equity capital, carbon disclosure score, and corporate carbon risk proxied by the
inverse of scope 1 greenhouse gas emissions scaled by sales turnover

Dependent variables Dependent variables


Independent var. KEi,t DSCRi,t CRi,t KEi,t DSCRi,t CRi,t
DSCRi,t −0.047 −0.079**
(−1.45) (−2.09)
DSCRi,t−1 −455.851*** −456.662***
(−3.45) (−3.45)
CRi,t −0.001** −0.001**
(−2.50) (−2.50)
HCRi,t −7.795
(−1.21)
DSCRi,t* HCRi,t 0.104
(1.35)
BETAi,t −0.108 −0.278
(−0.21) (−0.54)
LEVi,t −4.587*** 9.281 39 065.670*** −5.060*** 9.281 39 059.950***
(−2.59) (1.64) (4.06) (−2.85) (1.64) (4.06)
SIZEi,t 1.140*** 2.975*** 2240.211 1.155*** 2.975*** 2241.997
(3.25) (2.93) (1.19) (3.33) (2.93) (1.19)
GRWi,t 2.640* −8.864* −6876.068 1.856 −8.864* −6876.864
(1.70) (−1.88) (−0.79) (1.17) (−1.88) (−0.79)
PRFT i,t −7.565*** 3.948 −5453.851 −7.811*** 3.948 −5457.311
(−3.00) (0.50) (−0.38) (−3.13) (0.50) (−0.38)
TANGi,t −0.391 −0.877
(−0.15) (−0.35)
VLT i,t 1.425** 1.114*
(2.30) (1.75)
INTENSIVE 3.135* 9.053* 12 806.910 3.754** 9.053* 12818.470
(1.70) (1.72) (1.30) (2.02) (1.72) (1.30)
CONSTANT −3.525 45.352*** 17 173.430 −0.295 45.352*** 17215.660
(−0.55) (3.04) (0.61) (−0.04) (3.04) (0.61)
Industry Indicators Yes Yes Yes Yes Yes Yes
χ2 58.45*** 35.26*** 57.02*** 63.36*** 35.25*** 57.07***

Note. All continuous variables are winsorized at the first and 99th percentiles. KEi,t is the cost of equity capital of firm i in year t estimated using the CAPM
as described in Sharfman and Fernando (2008) and is obtained from the INET BFA database. DSCRi,t is a measure of voluntary carbon disclosure of firm i in
year t, obtained from the publicly available annual CDP reports. DSCRi,t−1 is a one year lagged carbon disclosure score of firm i in year t. CRi,t is the inverse of
Scope 1 GHG emissions of firm i in year t, as reported on the annual CDP reports, scaled by its total annual sales turnover. HCRi,t is a dummy variable that
equals 1 if the firm's CRi,t in year t is higher than the sample median, and 0 otherwise. BETAi,t denotes the unleveraged beta of firm i in year t obtained from
the INET BFA database. LEVi,t is the sum of current liabilities and noncurrent liabilities of firm i in year t scaled by its total assets. SIZEi,t is the natural log-
arithm of annual sales of firm i in year t. GRWi,t is the first difference of the natural logarithm of annual sales turnover of firm i in year t. PRFT i,t is earnings
before interest and taxes of firm i in year t scaled by its total assets. TANGi,t is total fixed assets of firm i in year t scaled by its total assets. VLT i,t is a five‐
rolling standard deviation of earnings before interest and taxes of firm i in year t. INTENSIVEi,t is a dummy variable that equals 1 if the firm operates in a
carbon‐intensive industry, and 0 otherwise.
*The estimated coefficient is statistically significant at 10% level;
**The estimated coefficient is statistically significant at 5% level;
***The estimated coefficient is statistically significant at 1% level.

Equation 1d with Equation 1f, and thus, the revised system of equa- disclosure score) incur lower cost of debt (5.8%) compared with the
tions is as follows: bottom 50% of firms (6.7%), significant at 5% level (p value = 0.024).
*
KDi;t ¼ DSCRi;t þ HCRi;t þ HCRi;t DSCRi;t þ ICOV i;t þ LEV i;t On the other hand, there are no statistically significant differences in
þ SIZEi;t þ GRW i;t þ PRFT i;t þ TANGi;t þ VLT i;t the cost of debt between high carbon risk and low carbon risk firms.
þ INTENSIVEi;t (1f) In terms of correlations, Table 2 shows the cost of debt is negatively
(not) correlated with carbon disclosure score (corporate carbon risk).
DSCRi;t ¼ CRi;t þ LEV i;t þ SIZEi;t þ GRW i;t þ PRFT i;t
These observations are identical to those observed in our main analysis.
þ INTENSIVEi;t (2)
Table 7 presents results where the cost of debt (KDi,t ) is the
dependent variable in Equations 1e and 1f. The results based on the
CRi;t ¼ DSCRi;t−1 þ LEV i;t þ SIZEi;t þ GRW i;t þ PRFT i;t
revised system of equations were generally consistent with those pre-
þ INTENSIVEi;t (3)
sented in our main analysis, except that the coefficient of DSCRi,t was
Table 1 shows that the median cost of debt for all firms in our sam- statistically insignificant (Z = 0.63, p value = 0.526) suggesting that vol-
ple is 6.1%. It also shows that top 50% of firms (in terms of untary carbon disclosure (DSCRi,t) is not associated with cost of debt
LEMMA ET AL. 13

TABLE 7 Three‐stage least squares regression results of cost of debt, carbon disclosure score, and carbon risk proxied by the inverse of scope 1
greenhouse gas emissions scaled by sales turnover
Dependent variables Dependent variables
Independent var. KDi,t DSCRi,t CRi,t KDi,t DSCRi,t CRi,t

DSCRi,t 0.002 0.003


(0.36) (0.63)
DSCRi,t−1 −493.988*** −493.320***
(−3.76) (−3.75)
CRi,t −0.001*** −0.001***
(−3.12) (−3.12)
HCRi,t 0.973
(1.28)
DSCRi,t* HCRi,t −0.009
(−0.98)
ICOVi,t 0.000 −0.001
(−0.14) (−0.38)
LEVi,t −0.415 12.525** 38547.050*** −0.435* 12.526** 38549.850***
(−1.54) (2.19) (4.05) (−1.65) (2.19) (4.05)
SIZEi,t −0.062 3.332*** 4731.117*** −0.069 3.332*** 4729.653***
(−1.35) (3.48) (2.80) (−1.51) (3.48) (2.80)
GRWi,t 0.291 −8.116** −8133.450 0.255 −8.116* −8130.332
(1.36) (−1.79) (−1.01) (1.19) (−1.79) (−1.01)
PRFT i,t −0.665 −5.319 −6091.924 −0.595 −5.319 −6089.303
(−1.32) (−0.64) (−0.42) (−1.20) (−0.64) (−0.42)
TANGi,t −0.003 0.103
(−0.01) (0.30)
VLT i,t 0.208*** 0.205***
(2.57) (2.57)
INTENSIVE 0.045 10.376** 9750.519 0.007 10.376* 9740.295
(0.17) (1.92) (1.00) (0.02) (1.92) (1.00)
CONSTANT 0.728 43.100*** −14731.230 0.478 43.100*** −14767.660
(3.12) (−0.58) (0.53) (3.12) (−0.58)
Industry Indicators Yes Yes Yes Yes Yes Yes
χ2 26.35** 40.83*** 69.17*** 33.82** 40.83*** 69.12**

Note. All continuous variables are winsorized at the first and 99th percentiles. KDi,t is interest expense of firm i in year t scaled by total interest bearing debt
and is obtained from the INET BFA database. DSCRi,t is a measure of voluntary carbon disclosure of firm i in year t, obtained from the publicly available
annual CDP reports. DSCRi,t−1 is a one year lagged carbon disclosure score of firm i in year t. CRi,t is the inverse of Scope 1 GHG emissions of firm i in year
t, as reported on the annual CDP reports, scaled by its annual total sales turnover. HCRi,t is a dummy variable that equals 1 if the firm's CRi,t in year t is
higher than the sample median, and 0 otherwise. ICOVi,t is earnings before interest and tax of firm i in year t scaled by its interest expenses. LEVi,t is the
sum of current liabilities and noncurrent liabilities of firm i in year t scaled by its total assets. SIZEi,t is the natural logarithm of annual sales of firm i in year
t. GRWi,t is the first difference of the natural logarithm of annual sales turnover of firm i in year t. PRFT i,t is earnings before interest and taxes of firm i in year
t scaled by its total assets. TANGi,t is total fixed assets of firm i in year t scaled by its total assets. VLT i,t is a five‐rolling standard deviation of earnings before
interest and taxes of firm i in year t. INTENSIVEi,t is a dummy variable that equals 1 if the firm operates in a carbon‐intensive industry, and 0 otherwise.
*The estimated coefficient is statistically significant at 10% level;
**The estimated coefficient is statistically significant at 5% level;
***The estimated coefficient is statistically significant at 1% level.

(KDi,t). This is inconsistent with the results we found for the overall analyzed financial and carbon‐related information pertaining to JSE com-
(and equity) cost of capital. As in the cost of equity (KEi,t) model, panies for the period 2010 to 2015. Consistent with our expectation, we
although the leverage (LEVi,t) and volatility (VLT i,t) variables loaded find the voluntary carbon disclosure is negatively associated with a firm's
statistically significantly (with p values of 0.100 and 0.010, respec- overall (and equity) cost of capital. The results suggest that JSE companies
tively), the tangibility (TANGi,t) and GHG‐intensive sector utilize voluntary carbon disclosures to manipulate the markets, especially
(INTENSIVEi,t) variables became statistically insignificant (with p value the stock market, to reduce their overall (and equity) cost of capital.
of 0.767 and 0.981, respectively). The results for Equations 2 and 3 Like both Jung et al. (2018) and Dhaliwal et al. (2011), we docu-
were identical across the system of equations. ment a positive association between past carbon disclosures and cor-
porate carbon risk. This association is consistent with the argument
5 | C O N CL U S I O N S that a firm may have the incentive to signal the possibility of higher
carbon risk in advance to reduce the possibility of shareholder litiga-
This study was set out to investigate the interplay among corporate car- tion that might be triggered by withholding carbon information (Al‐
bon risk, voluntary disclosure, and a firm's cost of capital in a unique setting Tuwaijri et al., 2004; Hassan & Romilly, 2018). The positive association
of South Africa, one of the “rising powers” in the climate policy debate. We that we observe between voluntary carbon disclosure and corporate
14 LEMMA ET AL.

carbon risk is consistent with legitimacy theory. It corroborates the Beyer, A., Cohen, D. A., Lys, T. Z., & Walther, B. R. (2010). The financial
argument that firms with high corporate carbon risk would be willing reporting environment: Review of the recent literature. Journal of
Accounting and Economics, 50(2), 296–343.
to increase disclosure to obtain legitimacy for their activities and earn
Bharath, S. T., Sunder, J., & Sunder, S. V. (2008). Accounting quality and
other rewards such as lower cost of capital, which ultimately help
debt contracting. The Accounting Review, 83(1), 1–28.
assure their continued existence (He et al., 2013).
Botosan, C. A. (1997). Disclosure level and the cost of equity capital. The
In sum, firms making better quality carbon disclosures appear to be Accounting Review, 72(3), 323–349.
rewarded with a lower cost of capital; more transparency with respect Campbell, D. (2003). Intra‐and intersectoral effects in environmental dis-
to carbon‐related information maps into a better outcome for the firm. closures: Evidence for legitimacy theory? Business Strategy and the
Therefore, it encourages firms to be more transparent in providing Environment, 12(6), 357–371.

information about climate risk. However, from societal and planetary Chapple, L., Clarkson, P. M., & Gold, D. L. (2013). The cost of carbon:
Capital market effects of the proposed emission trading scheme
perspectives, this outcome is only positive if voluntary carbon disclo-
(ETS). Abacus, 49(1), 1–33.
sure mirrors carbon performance. That is, firms that have the best qual-
Chen, J. C., & Roberts, R. W. (2010). Toward a more coherent understand-
ity carbon disclosures are also those that have the lowest carbon risk. ing of the organization–society relationship: A theoretical
Unfortunately, that does not appear to be the case. The findings show consideration for social and environmental accounting research. Journal
of Business Ethics, 97(4), 651–665.
that corporate carbon risk and voluntary carbon disclosure are related
so that those with the highest carbon risk tend to be better disclosers. Chen, L. H., & Gao, L. S. (2012). The pricing of climate risk. Journal of Finan-
cial and Economic Practice, 12(2), 115–131.
If voluntary carbon disclosures can be utilized in a way that has neg-
Cho, C. H., Freedman, M., & Patten, D. M. (2012). Corporate disclosure of
ative social consequences, regulators mandating these disclosures need environmental capital expenditures: A test of alternative theories.
to be cognizant of this in designing their regulations. Mandated stan- Accounting, Auditing & Accountability Journal, 25(3), 486–507.
dards need to reflect corporate carbon risk or be structured so that a user Cho, C. H., & Patten, D. M. (2007). The role of environmental disclosures
has a clear means to assess that risk. Good quality carbon disclosures as tools of legitimacy: A research note. Accounting, Organizations and
Society, 32(7), 639–647.
that appropriately reflect corporate carbon risk and result in a lower cost
Clarkson, P. M., Fang, X., Li, Y., & Richardson, G. (2013). The relevance of
of capital for the firm would be a win–win for both companies as well as
environmental disclosures: Are such disclosures incrementally informa-
the society at large. Because our study considered only JSE companies, tive? Journal of Accounting and Public Policy, 32(5), 410–431.
the results are obviously skewed towards bigger firms; thus, the findings Clarkson, P. M., Li, Y., Pinnuck, M., & Richardson, G. D. (2015). The valua-
should be read with some degree of circumspection. Future research tion relevance of greenhouse gas emissions under the European Union
endeavors could focus on the impact of the recently enacted mandatory carbon emissions trading scheme. European Accounting Review, 24(3),
551–580.
reporting regime. Because our findings imply value relevance of volun-
Clarkson, P. M., Li, Y., Richardson, G. D., & Vasvari, F. P. (2008). Revisiting
tary carbon disclosure, future research endeavors might consider the
the relation between environmental performance and environmental
relation between carbon‐related information and the value of the firm. disclosure: An empirical analysis. Accounting, Organizations and Society,
33(4), 303–327.
ORCID Clarkson, P. M., Li, Y., Richardson, G. D., & Vasvari, F. P. (2011). Does it
really pay to be green? Determinants and consequences of proactive
Tesfaye T. Lemma http://orcid.org/0000-0001-9806-3170
environmental strategies. Journal of Accounting and Public Policy,
30(2), 122–144.
RE FE R ENC E S Cogan, D. G. (2006). Corporate governance and climate change: Making the
Adams, C. A. (2004). The ethical, social and environmental reporting‐per- connection. Boston, MA: Ceres.
formance portrayal gap. Accounting, Auditing & Accountability Journal, Coulson, A. B., & Monks, V. (1999). Corporate environmental performance
17(5), 731–757. https://doi.org/10.1108/09513570410567791 considerations within bank lending decisions. Corporate Social‐Respon-
sibility and Environmental Management, 6(1), 1–10.
Al‐Tuwaijri, S. A., Christensen, T. E., & Hughes, K. E. (2004). The relations
among environmental disclosure, environmental performance, and eco- de Villiers, C., Naiker, V., & van Staden, C. J. (2011). The effect of board
nomic performance: A simultaneous equations approach. Accounting, characteristics on firm environmental performance. Journal of Manage-
Organizations and Society, 29(5), 447–471. https://doi.org/10.1016/ ment, 37(6), 1636–1663.
S0361‐3682(03)00032‐1 Deegan, C. (2007). Organisational legitimacy as a motive for sustainability
reporting. In J. Unerman, J. Bebbington, & B. O'Dwyer (Eds.), Sustain-
Andreasson, S. (2017). Fossil‐fuelled development and the legacy of post‐ ability Accounting and Accountability (pp. 127–149). London: Routledge.
development theory in twenty‐first century Africa. Third World Quar-
Dhaliwal, D. S., Gleason, C. A., Heitzman, S., & Melendrez, K. D. (2008).
terly, 38(12), 2634–2649.
Auditor fees and cost of debt. Journal of Accounting, Auditing & Finance,
Baker, L., Newell, P., & Phillips, J. (2014). The political economy of energy 23(1), 1–22.
transitions: The case of South Africa. New Political Economy, 19(6), Dhaliwal, D. S., Li, O. Z., Tsang, A., & Yang, Y. G. (2011). Voluntary nonfi-
791–818. nancial disclosure and the cost of equity capital: The initiation of
corporate social responsibility reporting. The Accounting Review, 86(1),
Barth, M. E., Cahan, S. F., Chen, L., & Venter, E. R. (2017). The economic 59–100.
consequences associated with integrated report quality: Capital market
Diamond, D. W., & Verrecchia, R. E. (1991). Disclosure, liquidity, and the
and real effects. Accounting, Organizations and Society, 62(1), 43–64.
cost of capital. The Journal of Finance, 46(4), 1325–1359. https://doi.
Bewely, K., & Li, Y. (2000). Disclosure of environmental information by org/10.1111/j.1540‐6261.1991.tb04620.x
Canadian manufacturing companies: A voluntary disclosure perspec- Dutta, S., & Nezlobin, A. (2017). Information disclosure, firm growth, and
tive. Advances in Environmental Accounting and Management, 1(0), the cost of capital. Journal of Financial Economics, 123(2), 415–431.
201–226. https://doi.org/10.1016/j.jfineco.2016.04.001
LEMMA ET AL. 15

Dye, R. A. (1985). Disclosure of nonproprietary information. Journal of Hoffmann, V. H., & Busch, T. (2008). Corporate carbon performance indi-
Accounting Research, 23(1), 123–145. cators. Journal of Industrial Ecology, 12(4), 505–520.
Easley, D., & O'hara, M. (2004). Information and the cost of capital. The Hughes, K. E. (2000). The value relevance of nonfinancial measures of air
Journal of Finance, 59(4), 1553–1583. https://doi.org/10.1111/ pollution in the electric utility industry. The Accounting Review, 75(2),
j.1540‐6261.2004.00672.x 209–228.
Fernandez, P., Aguirreamalloa, J., & Avendaño, L. C. (2011). Market risk Hughes, S. B., Anderson, A., & Golden, S. (2001). Corporate environmental
premium used in 56 countries in 2011: A survey with 6,014 answers. disclosures: Are they useful in determining environmental perfor-
Unpublished manuscript, IESE Business School, Madrid. mance? Journal of Accounting and Public Policy, 20(3), 217–240.
Francis, J. R., Khurana, I. K., & Pereira, R. (2005). Disclosure incentives and Jung, J., Herbohn, K., & Clarkson, P. (2018). Carbon risk, carbon risk aware-
effects on cost of capital around the world. The Accounting Review, ness and the cost of debt financing. Journal of Business Ethics, 150(4),
80(4), 1125–1162. https://doi.org/10.2308/accr.2005.80.4.1125 1154–1171.
Freedman, M., & Jaggi, B. (1992). An investigation of the long‐run relation- Kim, J. W., & Shi, Y. (2011). Voluntary disclosure and the cost of equity
ship between pollution performance and economic performance: The capital: Evidence from management earnings forecasts. Journal of
case of pulp and paper firms. Critical Perspectives on Accounting, 3(4), Accounting and Public Policy, 30(4), 348–366.
315–336. https://doi.org/10.1016/1045‐2354(92)90024‐L
Kim, Y.‐B., An, H. T., & Kim, J. D. (2015). The effect of carbon risk on the
Freedman, M., & Patten, D. M. (2004). Evidence on the pernicious effect of cost of equity capital. Journal of Cleaner Production, 93, 279–287.
financial report environmental disclosure. Accounting Forum, 28(1),
Kolk, A., Levy, D., & Pinkse, J. (2008). Corporate responses in an emerging
27–41.
climate regime: The institutionalization and commensuration of carbon
Freedman, M., & Wasley, C. (1990). The association between environmen- disclosure. European Accounting Review, 17(4), 719–745.
tal performance and environmental disclosure in annual reports and
Labatt, S., & White, R. R. (2007). Carbon finance: The financial implications of
10Ks. Advances in Public Interest Accounting, 3(2), 183–193.
climate change ( ed., Vol. 362). Hoboken, New Jersey: John Wiley &
Garber, S., & Hammitt, J. K. (1998). Risk premiums for environmental liabil- Sons.
ity: Does superfund increase the cost of capital? Journal of
Lambert, R., Leuz, C., & Verrecchia, R. E. (2007). Accounting information,
Environmental Economics and Management, 36(3), 267–294.
disclosure, and the cost of capital. Journal of Accounting Research,
García‐Teruel, P. J., Martínez‐Solano, P., & Sánchez‐Ballesta, J. P. (2014). 45(2), 385–420.
Supplier financing and earnings quality. Journal of Business Finance &
Accounting, 41(9–10), 1193–1211. Leftwich, R. W., Watts, R. L., & Zimmerman, J. L. (1981). Voluntary corpo-
rate disclosure: The case of interim reporting. Journal of Accounting
Ge, W., & Kim, J‐B. (2010). Real earnings management and cost of debt. Research, 19(0), 50–77.
CAAA Annual Conference.
Lemma, T. T. (2015). Corruption, debt financing and corporate ownership.
Gray, P., Koh, P.‐S., & Tong, Y. H. (2009). Accruals quality, information risk Journal of Economic Studies, 42(3), 433–461.
and cost of capital: Evidence from Australia. Journal of Business Finance
& Accounting, 36(1–2), 51–72. Lemma, T. T., & Negash, M. (2011). Rethinking the antecedents of capital
structure of Johannesburg Securities Exchange listed firms. Afro‐Asian
Gray, R., Kouhy, R., & Lavers, S. (1995). Corporate social and environmental Journal of Finance and Accounting, 2(4), 299–332.
reporting: A review of the literature and a longitudinal study of UK dis-
closure. Accounting, Auditing & Accountability Journal, 8(2), 47–77. Lemma, T. T., & Negash, M. (2012). Debt maturity choice of a firm: Evi-
dence from African countries. Journal of Business and Policy Research,
Griffin, P. A., Lont, D. H., & Sun, E. Y. (2017). The relevance to investors of 7(2), 60–92.
greenhouse gas emission disclosures. Contemporary Accounting
Research, 34(2), 1265–1297. Lemma, T. T., & Negash, M. (2013). Institutional, macroeconomic and firm‐
specific determinants of capital structure: The African evidence. Man-
Guidry, R. P., & Patten, D. M. (2012). Voluntary disclosure theory and
agement Research Review, 36(10), 1–38.
financial control variables: An assessment of recent environmental dis-
closure research. Accounting Forum, 36(2), 81–90. https://doi.org/ Lemma, T. T., Negash, M., Mlilo, M., & Lulseged, A. (2018). Instututional
10.1016/j.accfor.2012.03.002 ownership, product market competition, and earnings management:
Some evidence from international data. Journal of Business Research,
Hahn, R., Reimsbach, D., & Schiemann, F. (2015). Organizations, climate
90(9), 151–163.
change, and transparency: Reviewing the literature on carbon disclo-
sure. Organization & Environment, 28(1), 80–102. https://doi.org/ Leuz, C., & Wysocki, P. D. (2016). The economics of disclosure and finan-
10.1177/1086026615575542 cial reporting regulation: Evidence and suggestions for future
research. Journal of Accounting Research, 54(2), 525–622.
Hallding, K., Olsson, M., Atteridge, A., Vihma, A., Carson, M., & Roman, M.
(2011). Together Alone: BASIC Countries and the Climate Change Li, Y., Eddie, I., & Liu, J. (2014). Carbon emissions and the cost of capital:
Conundrum (Copenhagen: Norden – Nordic Council of Ministers). Australian evidence. Review of Accounting and Finance, 13(4), 400–420.

Hart, S. L. (1995). A natural‐resource‐based view of the firm. Academy of Li, Y., Richardson, G. D., & Thornton, D. B. (1997). Corporate disclosure of
Management Review, 20(4), 986–1014. environmental liability information: Theory and evidence. Contempo-
rary Accounting Research, 14(3), 435–474.
Hassan, O. A. G., & Romilly, P. (2018). Relations between corporate
economic performance, environmental disclosure and greenhouse gas Liao, L., Luo, L., & Tang, Q. (2015). Gender diversity, board independence,
emissions: New insight. Business Strategy and the Environment. environmental committee and greenhouse gas disclosure. The British
https://doi.org/10.1002/bse.2040 Accounting Review, 47(4), 409–424.
Hassan, S., & Van Biljon, A. (2010). The equity premium and risk‐free rate Lintner, J. (1956). Distribution of incomes of corporations among divi-
in a turbulent economy: Evidence from 105 years of data from South dends, retained earnings, and taxes. The American Economic Review,
Africa. South African Journal of Economics, 78(1), 23–39. 46(2), 97–113.
He, Y., Tang, Q., & Wang, K. (2013). Carbon disclosure, carbon perfor- Lopes, A. B., & de Alencar, R. C. (2010). Disclosure and cost of equity cap-
mance, and cost of capital. China Journal of Accounting Studies, 1(3–4), ital in emerging markets: The Brazilian case. The International Journal of
190–220. Accounting, 45(4), 443–464.
Healy, P. M., & Palepu, K. G. (2001). Information asymmetry, corporate dis- Luo, L., Lan, Y.‐C., & Tang, Q. (2012). Corporate incentives to disclose car-
closure, and the capital markets: A review of the empirical disclosure bon information: Evidence from the CDP Global 500 report. Journal of
literature. Journal of Accounting and Economics, 31(1–3), 405–440. International Financial Management & Accounting, 23(2), 93–120.
16 LEMMA ET AL.

Luo, L., & Tang, Q. (2014). Does voluntary carbon disclosure reflect under- Schmitz, H. (2017). Who drives climate‐relevant policies in the rising pow-
lying carbon performance? Journal of Contemporary Accounting & ers? New Political Economy, 22(5), 521–540.
Economics, 10(3), 191–205. Shan, Y. G., & Taylor, D. (2014). Theoretical perspectives on corporate
Matsumura, E. M., Prakash, R., & Vera‐Muñoz, S. C. (2014). Firm‐value social and environmental disclosure: Evidence from China. Journal of
effects of carbon emissions and carbon disclosures. The Accounting Asia‐Pacific Business, 15(3), 260–281.
Review, 89(2), 695–724. Sharfman, M. P., & Fernando, C. S. (2008). Environmental risk management
McKendall, M., Sánchez, C., & Sicilian, P. (1999). Corporate governance and and the cost of capital. Strategic Management Journal, 29(6), 569–592.
corporate illegality: The effects of board structure on environmental Sharpe, W. F. (1964). Capital asset prices: A theory of market equilibrium
violations. The International Journal of Organizational Analysis, 7(3), under conditions of risk. The Journal of Finance, 19(3), 425–442.
201–223.
Stanny, E., & Ely, K. (2008). Corporate environmental disclosures about the
Merton, R. C. (1987). A simple model of capital market equilibrium with effects of climate change. Corporate Social Responsibility and Environ-
incomplete information. The Journal of Finance, 42(3), 483–510. mental Management, 15(6), 338–348.
Modigliani, F., & Miller, M. H. (1958). The cost of capital, corporation Subramaniam, N., Wahyuni, D., Cooper, B. J., Leung, P., & Wines, G. (2015).
finance and the theory of investment. The American Economic Review, Integration of carbon risks and opportunities in enterprise risk manage-
48(3), 261–297. ment systems: Evidence from Australian firms. Journal of Cleaner
Production, 96(0), 407–417.
Murray, A., Sinclaire, D., Power, D., & Gray, R. (2006). Do financial market
care about social and envionment disclosure? Accounting, Auditing, & Tang, Q., & Luo, L. (2014). Carbon management systems and carbon miti-
Accountability Journal, 19(2), 228–255. gation. Australian Accounting Review, 24(1), 84–98.
Never, B. (2012). Who drives change? Comparing the evolution of domes- Thompson, P. (1998). Bank lending and the environment: Policies
tic climate governance in India and South Africa. The Journal of and opportunities. International Journal of Bank Marketing, 16(6), 243–252.
Environment & Development, 21(3), 362–387. Thompson, P., & Cowton, C. J. (2004). Bringing the environment into bank
Patten, D. M. (2002). The relation between environmental performance lending: Implications for environmental reporting. The British Account-
and environmental disclosure: A research note. Accounting, Organiza- ing Review, 36(2), 197–218.
tions and Society, 27(8), 763–773. Trueman, B., & Titman, S. (1988). An explanation for accounting income
Pegels, A. (2014). The politics of South Africa's renewable energy support. smoothing. Journal of Accounting Research, 26(0), 127–139.
In A. Pegels (Ed.), Green Industrial Policy in Emerging Countries Ullmann, A. A. (1985). Data in search of a theory: A critical examination of
(pp. 126–147). London: Routledge. the relationships among social performance, social disclosure, and eco-
nomic performance of US firms. Academy of Management Review, 10(3),
Peters, G., & Romi, A. (2014). Does the voluntary adoption of corporate
540–557.
governance mechanisms improve environmental risk disclosures? Evi-
dence from greenhouse gas emission accounting. Journal of Business Verrecchia, R. E. (1983). Discretionary disclosure. Journal of Accounting and
Ethics, 125(4), 637–666. Economics, 5(1), 179–194.

Plumlee, M., Brown, D., Hayes, R. M., & Marshall, R. S. (2015). Voluntary Verrecchia, R. E. (2001). Essays on disclosure. Journal of Accounting and
environmental disclosure quality and firm value: Further evidence. Economics, 32(1), 97–180.
Journal of Accounting and Public Policy, 34(4), 336–361. Walls, J. L., Berrone, P., & Phan, P. H. (2012). Corporate governance and
Rankin, M., Windsor, C., & Wahyuni, D. (2011). An investigation of volun- environmental performance: Is there really a link? Strategic Manage-
tary corporate greenhouse gas emissions reporting in a market ment Journal, 33(8), 885–913.
governance system: Australian evidence. Accounting, Auditing & Wang, L., Li, S., & Gao, S. (2013). Do greenhouse gas emissions affect
Accountability Journal, 24(8), 1037–1070. financial performance?—An empirical examination fo Australian public
Richardson, A. J., Welker, M., & Hutchinson, I. R. (1999). Managing capital firms. Business Stratgy and the Environment, 23(8), 505–519.
market reactions to corporate social resposibility. International Journal Weber, O. (2012). Environmental credit risk management in banks and
of Management Reviews, 1(1), 17–43. financial service institutions. Business Strategy and the Environment,
21(4), 248–263.
Roberts, R. W. (1992). Determinants of corporate social responsibility dis-
closure: An application of stakeholder theory. Accounting, Organizations Wiseman, J. (1982). An evaluation of environmental disclosures made in
and Society, 17(6), 595–612. corporate annual reports. Accounting, Organizations and Society, 7(1),
53–63.
Rockness, J., Schlachter, P., & Rockness, H. (1986). Hazardous waste dis-
posal, corporate disclosure and financial performance in the chemical
industry. Advances in Public Interest Accounting, 1, 167–191.
How to cite this article: Lemma TT, Feedman M, Mlilo M,
Saka, C., & Oshika, T. (2014). Disclosure effects, carbon emissions and cor-
porate value. Sustainability Accounting, Management and Policy Journal, Park JD. Corporate carbon risk, voluntary disclosure, and cost
5(1), 22–45. of capital: South African evidence. Bus Strat Env. 2018;1–16.
Salo, J., & van Ast, L. (2009). Carbon risks and opportunities in the S&P 500. https://doi.org/10.1002/bse.2242
Boston, MA: Trucost.

You might also like