The Association between Voluntary Carbon Disclosure and Accounting Comparability: Examining the Moderating Effect of Korean Business Groups

Abstract

This study investigates whether voluntary carbon disclosure to the Carbon Disclosure Project (CDP) is positively related to accounting comparability, one of the dimensions for evaluating the quality of financial reporting. It especially questions whether Korean business groups moderate the relationship between carbon disclosure and accounting comparability. It finds that accounting comparability is greater when firms voluntarily disclose carbon emissions to the CDP. It also finds a moderating effect of Korean business groups on the relationship between carbon disclosure and accounting comparability. These findings suggest that ethical environmental management in non-business groups encourages managers to provide comparable financial reports, whereas opportunistic environmental management in business groups results in managers producing less comparable financial reports. Therefore, this study is meaningful in finding that the voluntary disclosure of carbon emissions is a factor that increases the quality of financial reporting and in proposing that the environmental commissions provide sufficient guidance to prevent opportunistic disclosure of carbon emissions.

1. Introduction

The issues surrounding climate change have received considerable attention worldwide. With the ratification of 196 countries, the Paris Agreement was signed in 2015. The goal of this agreement is to limit global warming to well below 2 (preferably to 1.5) degrees Celsius, compared to pre-industrial levels. To achieve this long-term temperature goal, countries then submitted their plans for climate-based action to reduce their greenhouse gas emissions [1]. After the Agreement, the European Commission announced its carbon-neutrality target by 2050 through the European Green Deal [2], and the Korean government also announced its carbon-neutrality target by 2050 through the Korean Long-Term Low Greenhouse Gas Emission Development Strategies (LEDS) [3].

The Carbon Disclosure Project (CDP) is an international non-profit organization based in the UK, Germany and the USA that helps companies and cities disclose their environmental impact. It aims to encourage investors, companies and cities to take action to build a truly sustainable economy by both measuring and understanding their environmental impact [4]. Several studies have tested whether voluntary carbon disclosure to the CDP affects firm valuation [5,6], firm performance [7], carbon performance [8,9,10] and earnings management [11,12,13].

This study investigates the relationship between voluntary carbon disclosure and accounting comparability. Based on stakeholder theory, firms that voluntarily disclose carbon emissions are likely to provide more comparable financial reports because they have ethical obligations to provide transparent information to various stakeholders [14,15,16,17]. On the other hand, agency theoretical arguments state that firms are likely to voluntarily disclose carbon emissions in order to disguise their opportunistic earnings management [18,19]; thus, they are likely to provide less comparable financial reports. Herein, ethical managers are willing to do the right things for their stakeholders, thereby providing more comparable financial statements, while opportunistic ones are more likely to advance their own personal agendas, thereby providing less comparable financial statements [20].

This study also examines the moderating effect of Korean business groups on the relationship between carbon disclosure and accounting comparability for the following reasons. Agency problems between controlling and minority shareholders are prevalent in Korean business groups [21,22,23]. Controlling shareholders in business groups have more opportunities to divert member-firm resources [24,25]. Lee et al. [26] and Oh et al. [27] also argue that Korean business groups exploit corporate social responsibility (CSR) activities to satisfy controlling shareholders’ benefit at the expense of minority ones. Therefore, business groups are likely to voluntarily disclose carbon emissions in order to conceal opportunistic earnings management and less comparable financial reports.

The results of this study are as follows. First, firms that voluntarily disclose carbon emissions provide more comparable financial reports, which is in line with stakeholder theory. Second, Korean business groups have a moderating effect on the positive relationship between carbon disclosure and accounting comparability, which is consistent with agency theory. Third, carbon transparency is positively related to accounting comparability; however, Korean business groups have no moderating effect on the relationship between carbon transparency and accounting comparability. Fourth, firms that voluntarily disclose carbon emissions engage in less earnings management, and Korean business groups have a moderating effect on the negative relationship between carbon disclosure and earnings management. The results suggest that ethical management in non-business groups encourages managers to provide comparable and transparent financial reports, whereas opportunistic management in business groups results in managers producing less comparable and transparent financial reports.

Recently, environmental management, including carbon disclosure, has become more important following the Paris Agreement. To my knowledge, this is the first empirical study that tests the relationship between voluntary carbon disclosure to the CDP and accounting comparability. These findings contribute to the literature in several ways. First, this study highlights that voluntary carbon disclosure to the CDP is positively related to comparable financial reports, implying that firms with ethical obligations provide both transparent environmental disclosure and comparable financial reports. Second, this study also documents the moderating effect of Korean business groups on the positive relationship between carbon disclosure and accounting comparability, suggesting that firms in business groups disclose carbon emissions to conceal their less transparent and comparable financial reports. Finally, the novel results will help both the academic and practical fields to better understand whether voluntary carbon disclosure is related to accounting comparability as well as why firms voluntarily disclose carbon emissions or for what purpose those in business groups disclose carbon emissions.

The remainder of this study is organized as follows. In Section 2, prior research is summarized and hypotheses are developed. Then the research design and methods, including the samples and empirical models, are discussed in Section 3, and the results are reported in Section 4. Finally, the study’s conclusions and limitations are presented in Section 5.

2. Literature Review and Hypotheses Development

2.1. Why Do Firms Voluntarily Disclose Non-Financial Activities?

Several studies examine the relationship between earnings management and corporate social/environmental disclosure and argue that managers use these disclosures either for ethical purposes within a stakeholder theoretical framework or for opportunistic purposes within an agency theoretical framework. Based on stakeholder theory, transparent financial and non-financial reporting behaviors are primarily motivated by ethical obligations to satisfy the interests of various stakeholders. Prior studies document that CSR performance is negatively related to both financial restatements [14] and earnings management [16], and it is positively related to both management forecast accuracy [15] and disclosure levels [17]. Regarding the relationship between environmental disclosure and earnings management, environmental performance is negatively related to discretionary accruals [12,13]. Lemma et al. [11] found that carbon emissions were positively related to discretionary accruals and that voluntary carbon disclosure mitigated the positive relationship between carbon emissions and earnings management. These authors argued that firms with higher carbon risk exposure tended to motivate managers to engage in opportunistic earnings management in order to deflect the attention of external stakeholders, whereas firms that voluntarily disclosed carbon emissions for ethical purposes were likely to motivate managers to refrain from engaging in earnings management.

Based on agency theory, managers tend to demonstrate transparent financial and non-financial reporting behaviors to satisfy their own interests. Prior studies document that CSR performance is positively related to complex disclosures [28], corporate tax payments [29] and discretionary accruals [18]. Petrovits [19] found that small earnings-increasing firms opportunistically made charitable foundation-funding choices and argued that they used these charitable foundations as an earnings management tool. In addition, Velte [30] found that carbon performance was negatively related to discretionary accruals and was positively related to real earnings management. This author argues that managers use carbon performance opportunistically to shift from discretionary accruals to real earnings management, which is hardly likely to be detected by stakeholders.

Meanwhile, comparability is the qualitative characteristic that enables users to identify and understand similarities in and differences among items [31,32]. Wang et al. [20] argued that firms with positive CSR performance tended to provide more comparable financial statements based on stakeholder theory, whereas firms with adverse CSR performance tended to have minimal compliance incentives to follow accounting standards based on agency theory.

Within the stakeholder theoretical framework, transparent financial and non-financial reporting behaviors are ethical obligations of managers to satisfy the interests of various stakeholders [14,15,16,17]. Luo and Wu [12] argued that firms with greater financial transparency disclosed more carbon information to satisfy the needs of stakeholders. Litt et al. [13] also argued that increased monitoring by stakeholders and ethical corporate cultures prohibited managers from engaging in earnings management. Furthermore, Wang et al. [20] argued that the positive relationship between positive CSR performance and accounting comparability arose from the ethical concerns surrounding stakeholder engagement. Hence, if firms voluntarily disclosed carbon emissions in the context of an ethical obligation, then they were likely to exert greater efforts in providing comparable financial reports, leading to the following hypothesis:

Hypothesis 1.

Voluntary carbon disclosure is positively associated with accounting comparability.

2.2. Why Do Firms in Business Groups Disclose Non-Financial Activities?

A “chaebol” Korean business group is a large conglomerate that has a unique ownership structure. The owners and their families (i.e., controlling shareholders) are both directly and indirectly involved in the management of member firms in business groups [33]. Agency problems between controlling and minority shareholders are prevalent in business groups wherein the controlling shareholders exercise complete and arbitrary control over businesses of which they have only low levels of ownership [21,22,23]. Controlling shareholders in business groups tend to tunnel profit from member firms with low cash flow rights to those with high cash flow rights [34] and have more opportunities to divert member-firm resources because of business groups’ complex ownership structure [24,25].

The disparity between actual ownership and control rights may enhance the controlling shareholders’ ability and incentives to expropriate the wealth of the minority shareholders [35]. Muttakin et al. [25] and Beuselinck and Deloof [36] found that firms in business groups engaged in more earnings management. Kim and Yi [24] also found that firms in Korean business groups engaged in more earning management and that controlling shareholders with larger control-ownership disparity further engaged in earnings management. They here argued that controlling shareholders with larger control-ownership disparity in business groups had more incentives for engaging in earnings management to maximize their own wealth.

Business groups are likely to opportunistically disclose non-financial activities within the agency theoretical framework. Oh et al. [27] found that CSR performance is positively related to earnings transparency while business groups mitigate this positive relationship. Lee et al. [26] also found that voluntary CSR disclosure in sustainability reports increased firm value, while business groups served to weaken this relationship. These researchers argued that business groups strategically exploited CSR activities to satisfy controlling shareholders’ benefit at the expense of minority ones. Hence, if firms in business groups opportunistically disclose carbon emissions and have minimal compliance incentives to follow accounting standards, then they are likely to exert less effort in providing comparable financial reports, leading to the following hypothesis:

Hypothesis 2.

Business groups mitigate the positive relationship between carbon disclosure and accounting comparability.

3. Research Design

3.1. Models

To investigate the relationship between carbon disclosure and accounting comparability, the following Equation (1) is estimated:

$$\begin{array}{l}ACOM{P}_{it}={\alpha }_0+{\alpha }_{1}CDPDI{S}_{it}+{\alpha }_{2}SIZ{E}_{it}+{\alpha }_{3}LE{V}_{it}+{\alpha }_{4}MT{B}_{it} \\ +{\alpha }_{5}CF{O}_{it}+{\alpha }_{6}LOSSPro{b}_{it}+{\alpha }_{7}StdSALE{S}_{it}+{\alpha }_{8}StdCF{O}_{it}\\ +{\alpha }_{9}StdGRO{W}_{it}+{\alpha }_{10}BIG{4}_{it}+{\alpha }_{11}\%MAJO{R}_{it}+{\alpha }_{12}\%FO{R}_{it}\\ +IND\_Dumm{y}_{it}+{\epsilon }_{it}\end{array}$$

(1)

To capture the moderating effect of Korean business groups on the relationship between carbon disclosure and accounting comparability, the following Equation (2) is also estimated:

$$\begin{array}{l}ACOM{P}_{it}={\alpha }_0+{\alpha }_{1}CDPDI{S}_{it}+{\alpha }_{2}CHAEBO{L}_{it}\\ +{\alpha }_{3}CDPDI{S}_{it}\times CHAEBO{L}_{it}+{\alpha }_{4}SIZ{E}_{it}+{\alpha }_{5}LE{V}_{it}+{\alpha }_{6}MT{B}_{it}+{\alpha }_{7}CF{O}_{it}\\ +{\alpha }_{8}LOSSPro{b}_{it}+{\alpha }_{9}StdSALE{S}_{it}+{\alpha }_{10}StdCF{O}_{it}+{\alpha }_{11}StdGRO{W}_{it} \\ +{\alpha }_{12}BIG{4}_{it}+{\alpha }_{13}\%MAJO{R}_{it}+{\alpha }_{14}\%FO{R}_{it} \\ +IND\_Dumm{y}_{it}+{\epsilon }_{it}\end{array}$$

(2)

First, $ACOM{P}_{it}$ is accounting comparability as per [37]’s method whose concept is that the same economic events lead to comparable financial statements when the accounting systems of firm i and firm j are similar. The following Equation (3) incorporates a relation between the previous 16 quarters of net accounting income and stock returns:

$$Earning{s}_{it}={\alpha }_i+{\beta }_{i}Retur{n}_{it}+{\epsilon }_{it}$$

(3)

where $Earning{s}_{it}$ is quarterly net accounting income divided by the beginning market value of equity and $Retur{n}_{it}$ is the stock return during the same quarter. I estimate ${\widehat{\alpha }}_i$ and ${\widehat{\beta }}_i$ for firm i and ${\widehat{\alpha }}_j$ and ${\widehat{\beta }}_j$ for firm j. I then calculate the expected earnings of firm i and firm j, assuming they had the same return, $Retur{n}_{it}$ as follows:

$$E{\left(Earnings\right)}_{iit}={\widehat{\alpha }}_i+{\widehat{\beta }}_{i}Retur{n}_{it}$$

(4)

$$E{\left(Earnings\right)}_{ijt}={\widehat{\alpha }}_j+{\widehat{\beta }}_{j}Retur{n}_{it}$$

(5)

where $E{\left(Earnings\right)}_{iit}$ is the expected earnings produced by the accounting system of firm i with $Retur{n}_{it}$, and $E{\left(Earnings\right)}_{ijt}$ is the expected earnings produced by the accounting system of firm j with $Retur{n}_{it}$ within the same SIC two-digit industry classification. De Franco et al. [37] defined accounting comparability between firm i and firm j as the negative average value of the absolute difference between the expected earnings of firm i and firm j as follows:

$$ACOM{P}_{ijt}=-1/16\times {\displaystyle \sum }_{t-15}^t\left|E{\left(Earnings\right)}_{iit}-E{\left(Earnings\right)}_{ijt}\right|$$

(6)

where $ACOM{P}_{ijt}$ is the proxy of accounting comparability, which is greater when the value is higher. $ACOM{P}_{it}$ is the median $ACOM{P}_{ijt}$ for all firms j in the same industry as firm i during period t.

$CDPDI{S}_{it}$ is the dummy variable that is coded 1 if firm i discloses carbon emissions to the CDP; otherwise it is coded 0. $CHAEBO{L}_{it}$ is the dummy variable that is coded 1 if firm i is a member firm in a business group; otherwise it is coded 0. Business groups are annually identified by the Korea Fair Trade Commissions board based on the size of the total assets of chaebol-affiliated firms.

Regarding the control variables, Equations (1) and (2) include firm size ($SIZE$), financial leverage ($LEV$), market-to-book ratio ($MTB$), cash flows from operations ($CFO$), profitability ($ROA$), loss probability ($LOSSProb$), sales volatility ($StdSALES$), operating cash flows volatility ($StdCFO$) and sales growth volatility ($StdGROW$). Francis et al. [38] argued that firm size ($SIZE$) captures many unobservable firm-specific charateristics in the regressions that explain accounting comparability. They measured $SIZE$ as the natural log of total assets to prevent distortion of results due to significant differences in total assets. Francis et al. [38] also argued that either the economic fundamentals or the propensities around earnings management affected the similarities of the earnings between two firms. Therefore, Equations (1) and (2) include the volatility of operations (i.e., sales volatility ($StdSALES$), operating cash flows volatility ($StdCFO$) and sales growth volatility ($StdGROW$)) to control for the economic fundamentals, as well as the financial position and performance (i.e., financial leverage ($LEV$), market-to-book ratio ($MTB$), cash flows from operations ($CFO$), profitability ($ROA$) and loss probability ($LOSSProb$)) to control for the propensities around accounting comparability. Finally, Francis et al. [38] argued that the same audit style of each big 4 auditor increased accounting comparability, so equations (1) and (2) include big 4 auditors ($BIG4$) to control for the same audit style of each big 4 auditor. Wang et al. [20] argued that the governance factors related to ownerships affected accounting comparability, so equations (1) and (2) include the ownerships of the largest shareholders ($\%MAJOR$) and foreigners ($\%FOR$) to control for the governance factors.

All analyses related to Equations (1) and (2) are tested using pooled OLS regressions with corrected standard errors for both firm and year levels clustering.

I hand-collected carbon disclosure and carbon emissions data from the CDP database (www.cdp.net (accessed on 24 January 2023)) and collected financial and stock data from the FN-GUIDE database (www.fnguide.com (accessed on 24 January 2023)).

3.2. Sample Selection

The sample consists of companies listed on the Korean Stock Exchange (KSE) between 2014 and 2018 because the IFRS was introduced in South Korea in 2011, and 16 consecutive quarterly datasets are required to calculate accounting comparability. It is restricted to non-financial firms with fiscal year-ends in December to ensure homogeneity. It then removes firms with negative net assets because they are likely to experience financial distress, as well as firms without sufficient financial and stock data. The final sample consists of 2,784 firm-year observations. The sample that voluntarily discloses carbon emissions includes 230 firm-year observations out of 619 that are subject to disclosure to the CDP. I winsorized all the variables at 1 and 99 percent, with the exception of dummy variables. Panel A of

Table 1. Sample description.

Panel A: Sample Selection Procedure
Description		Number of Firms
Firms listed on the KOSPI between 2014 and 2018		3651
Less: Financial industry		247
Fiscal year-end not in December		75
Negative net assets		21
Data shortage		524
Final sample		2784
Sample subject to disclosure to the CDP		619
Panel B: Sample by sector
Sector	Frequency	Responses	Percentage
Industrial	155	72	46.5%
Consumer Discretionary	132	38	28.8%
Energy	12	3	25.0%
Materials	117	40	34.2%
Utilities	16	10	62.5%
Health Care	22	1	4.5%
Telecommunication Services	15	15	100.0%
Consumer Staples	100	22	22.0%
IT	50	29	58.0%
Sample subject to disclosure to the CDP	619	230	37.2%

presents a summary of the sample selection, and panel B presents the sample by sector based on the CDP reports. Panel B shows that 37.2% of the sample subject to disclosure to the CDP voluntarily disclosed carbon emissions.

4. Results

4.1. Descriptive Statistics

Table 2. Descriptive statistics (N = 619).

Variables	Mean	S.D.	Min	Median	Max
$ACOMP$	−0.018	0.012	−0.070	−0.014	−0.007
$CDPDIS$	0.372	0.484	0.000	0.000	1.000
$CHAEBOL$	0.664	0.473	0.000	1.000	1.000
$SIZE$	28.938	1.277	26.332	28.836	32.298
$LEV$	0.424	0.194	0.054	0.431	0.900
$MTB$	1.821	1.890	0.343	1.171	10.867
$CFO$	0.074	0.065	−0.068	0.065	0.290
$LOSSProb$	0.188	0.200	0.000	0.125	0.716
$StdSALES$	0.033	0.026	0.006	0.026	0.143
$StdCFO$	0.024	0.014	0.006	0.021	0.082
$StdGROW$	0.246	0.477	0.032	0.125	3.394
$BIG4$	0.950	0.218	0.000	1.000	1.000
$\%MAJOR$	0.416	0.162	0.091	0.390	0.787
$\%FOR$	0.221	0.159	0.010	0.184	0.753

Notes: All variables are defined in Appendix A.

presents the descriptive statistics of the variables used in this study. The mean (median) of accounting comparability ($ACOMP$) is −0.018 (−0.014), which is similar to that reported by [37]. The mean of carbon disclosure ($CDPDIS$) is 0.372, suggesting that approximately 37.2 percent of the sample firms voluntarily disclose carbon emissions. The mean of member firms in business groups ($CHAEBOL$) is 0.664, indicating that approximately two-thirds of the sample are chaebol-affiliated firms.

For the control variables, the mean of loss probability ($LOSSProb$) is 0.188, indicating that the sample firms experience the negative earnings three times over the 16 quarters examined. The means (medians) of sales volatility ($StdSALES$), operating cash flows volatility ($StdCFO$) and sales growth volatility ($StdGROWTH$) are 0.033 (0.026), 0.024 (0.021) and 0.246 (0.125), respectively, which suggests that sales growth volatility is much greater than the other volatilities. Finally, the means (medians) of the largest shareholders’ ($\%MAJOR$) and foreigners’ ($\%FOR$) ownerships are 0.416 (0.390) and 0.221 (0.184), respectively, which suggests that the largest shareholders may have a greater influence on management decision-making, including in terms of both financial and non-financial disclosures.

4.2. Correlations

Table 3. Pearson correlations (N = 619).

Variable	$\mathit{C}\mathit{D}\mathit{P}\mathit{D}\mathit{I}\mathit{S}$	$\mathit{C}\mathit{H}\mathit{A}\mathit{E}\mathit{B}\mathit{O}\mathit{L}$	$\mathit{S}\mathit{I}\mathit{Z}\mathit{E}$	$\mathit{L}\mathit{E}\mathit{V}$	$\mathit{M}\mathit{T}\mathit{B}$	$\mathit{C}\mathit{F}\mathit{O}$	$\mathit{L}\mathit{O}\mathit{S}\mathit{S}\mathit{P}\mathit{r}\mathit{o}\mathit{b}$	$\mathit{S}\mathit{t}\mathit{d}\mathit{S}\mathit{A}\mathit{L}\mathit{E}\mathit{S}$	$\mathit{S}\mathit{t}\mathit{d}\mathit{C}\mathit{F}\mathit{O}$	$\mathit{S}\mathit{t}\mathit{d}\mathit{G}\mathit{R}\mathit{O}\mathit{W}$	$\mathit{B}\mathit{I}\mathit{G}4$	$\mathit{\%}\mathit{M}\mathit{A}\mathit{J}\mathit{O}\mathit{R}$	$\mathit{\%}\mathit{F}\mathit{O}\mathit{R}$
$ACOMP$	−0.075 * (0.064)	−0.119 *** (0.003)	−0.204 *** (0.000)	−0.445 *** (0.000)	0.180 *** (0.000)	0.181 *** (0.000)	−0.537 *** (0.000)	−0.049 (0.223)	−0.063 (0.115)	0.022 (0.593)	0.021 (0.598)	0.054 (0.182)	0.121 *** (0.003)
$CDPDIS$	1	0.321 *** (0.000)	0.525 *** (0.000)	0.169 *** (0.000)	−0.085 ** (0.034)	0.082 ** (0.042)	0.073 * (0.071)	−0.111 *** (0.006)	−0.124 *** (0.002)	0.002 (0.959)	0.069 * (0.085)	−0.294 *** (0.000)	0.259 *** (0.000)
$CHAEBOL$		1	0.547 *** (0.000)	0.222 *** (0.000)	−0.244 *** (0.000)	−0.098 ** (0.015)	0.158 *** (0.000)	−0.076 * (0.058)	−0.270 *** (0.000)	−0.016 (0.695)	0.244 *** (0.000)	−0.286 *** (0.000)	0.136 *** (0.001)
$SIZE$			1	0.201 *** (0.000)	−0.353 *** (0.000)	0.026(0.522)	0.167 *** (0.000)	−0.194 *** (0.000)	−0.286 *** (0.000)	−0.102 ** (0.011)	0.191 *** (0.000)	−0.333 *** (0.000)	0.430 *** (0.000)
$LEV$				1	−0.078 * (0.051)	−0.229 *** (0.000)	0.476 *** (0.000)	0.176 *** (0.000)	0.145 *** (0.000)	−0.190 *** (0.000)	0.065 (0.109)	0.000 (1.000)	−0.201 *** (0.000)
$MTB$					1	0.221 *** (0.000)	−0.208 *** (0.000)	0.124 *** (0.002)	0.225 *** (0.000)	0.159 *** (0.000)	−0.054 (0.183)	0.056 (0.168)	0.073 * (0.071)
$CFO$						1	−0.332 *** (0.000)	−0.089 ** (0.027)	−0.052 (0.196)	−0.137 *** (0.001)	−0.097 ** (0.016)	−0.165 *** (0.000)	0.409 *** (0.000)
$LOSSProb$							1	0.076 * (0.058)	−0.006 (0.881)	0.134 *** (0.001)	0.081 ** (0.043)	−0.048 (0.236)	−0.157 *** (0.000)
$StdSALES$								1	0.569 *** (0.000)	0.229 *** (0.000)	−0.108 *** (0.007)	0.153 *** (0.000)	−0.124 *** (0.002)
$StdCFO$									1	0.021 (0.611)	−0.065 (0.105)	0.119 *** (0.003)	−0.077 * (0.055)
$StdGROW$										1	−0.004 (0.922)	0.120 *** (0.003)	−0.121 *** (0.003)
$BIG4$											1	−0.012 (0.767)	0.072 * (0.073)
$\%MAJOR$												1	−0.412 *** (0.000)
$\%FOR$													1

Notes: All variables are defined in Appendix A. Values in parentheses are p-values; ***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively.

presents the Pearson correlations of the main variables used in this study. The association between $ACOMP$ and $CHAEBOL$ is significantly negative (p = 0.003), indicating that member firms in business groups provide less comparable financial reports. However, unlike the prediction, the association between $ACOMP$ and $CDPDIS$ is only marginally significantly negative (p = 0.064). In untabulated results, it finds a positive association between $ACOMP$ and $CDPDIS$, but this is not significant when it tests the Pearson correlations, only with firms in non-business groups.

With respect to the correlation between $ACOMP$ and control variables, $ACOMP$ is negatively correlated to $SIZE$, $LEV$ and $LOSSProb$. $ACOMP$ is also positively correlated to $MTB$, $CFO$ and $\%FOR$. However, $ACOMP$ is not significantly correlated to $StdSALES$, $StdCFO$, $StdGROW$, $BIG4$ and $\%MAJOR$.

4.3. The Relationship between Carbon Disclosure and Accounting Comparability

Table 4. The relationship between carbon disclosure and accounting comparability.

Variables	Expected Sign	$\mathit{A}\mathit{C}\mathit{O}\mathit{M}\mathit{P}$
		Coefficient (t-Value)
$Constant$		0.030 ** (2.27)
$CDPDIS$	(+)	0.002 ** (2.11)
$SIZE$	(−)	−0.001 *** (−3.25)
$LEV$	(−)	−0.005 ** (−2.12)
$MTB$	(+)	0.000 (0.34)
$CFO$	(+)	−0.012 (−1.60)
$LOSSProb$	(−)	−0.024 *** (−7.51)
$StdSALES$	(−)	0.023 (1.16)
$StdCFO$	(−)	−0.075 ** (−1.99)
$StdGROW$	(−)	−0.000 (−0.15)
$BIG4$	(+)	0.005 *** (2.68)
$\%MAJOR$	(−)	−0.003 (−1.09)
$\%FOR$	(+)	0.000 (0.12)
$IND\_Dummy$		Include
F value		17.560 ***
Adjusted$R^2$		0.443
−2ln(L), AIC		−3916.6, −3914.6
Durbin-Watson		1.917
N		619

Notes: All variables are defined in Appendix A. Values in parentheses are t-values; ***, ** indicate significance at the 1%, 5% levels, respectively.

presents the results on the relationship between carbon disclosure and accounting comparability. The coefficient $CDPDIS$ is significantly positive (p = 0.036), which supports Hypothesis 1. This result is consistent with the stakeholder theory, which states that firms tend to exert greater efforts in providing comparable financial reports if they voluntarily disclose carbon emissions in the context of an ethical obligation to satisfy the interests of various stakeholders.

With respect to the control variables, $SIZE$, $LEV$, $LOSSProb$ and $StdCFO$ are negatively related to accounting comparability, and $BIG4$ is positively related to accounting comparability. These results indicate that a higher financial risk lowers accounting comparability, while a higher audit quality increases it.

With respect to the model, F value is 17.560 (p < 0.000); adjusted $R^2$ is 0.443; and the difference between −2ln(L) and AIC is -2, which means that the whole independent variables can explain 44.3% of the variation in the dependent variable and the model is well-fitting. Regarding the residual analysis, Durbin-Watson is 1.917, which means that there is no concern about autocorrelation in the residuals because the value is close to 2.

Additionally, it divides the sample into business groups and non-business groups and conducts OLS regressions with each sub-sample. In untabulated results, the coefficient $CDPDIS$ is insignificantly positive in business groups, and it is significantly positive in non-business groups, suggesting that firms in non-business groups are more likely to voluntarily disclose carbon emissions in the context of an ethical obligation compared to those in business groups.

4.4. The Moderating Effect of Business Groups

Table 5. The moderating effect of business groups.

Variables	Expected Sign	$\mathit{A}\mathit{C}\mathit{O}\mathit{M}\mathit{P}$
		Coefficient (t-Value)
$Constant$		0.041 *** (2.81)
$CDPDIS$	(+)	0.005 *** (3.95)
$CHAEBOL$	?	0.003 *** (3.15)
$CDPDIS\times CHAEBOL$	(−)	−0.004 ** (−2.52)
$SIZE$	(−)	−0.002 *** (−3.70)
$LEV$	(−)	−0.006 ** (−2.40)
$MTB$	(+)	0.000 (0.09)
$CFO$	(+)	−0.011 (−1.53)
$LOSSProb$	(−)	−0.023 *** (−7.55)
$StdSALES$	(−)	0.016 (0.79)
$StdCFO$	(−)	−0.048 (−1.24)
$StdGROW$	(−)	−0.000 (−0.30)
$BIG4$	(+)	0.004 ** (2.23)
$\%MAJOR$	(−)	−0.002 (−0.60)
$\%FOR$	(+)	0.002 (0.58)
$IND\_Dummy$		Include
F value		16.000 ***
Adjusted$R^2$		0.451
−2ln(L), AIC		−3904.4, −3902.4
Durbin-Watson		1.909
N		619

Notes: All variables are defined in Appendix A. Values in parentheses are t-values; ***, ** indicate significance at the 1%, 5% levels, respectively.

presents the results of the moderating effect of business groups on the relationship between carbon disclosure and accounting comparability. The coefficient $CPDDIS$ is significantly positive (p < 0.001), which is the same result as that presented in Table 4. Importantly, the coefficient $CPDDIS\times CHAEBOL$ is significantly negative (p = 0.012), which supports Hypothesis 2. This result is consistent with agency theory that states that firms in business groups tend to exert less effort in providing comparable financial reports if they only disclose carbon emissions opportunistically in order to satisfy controlling shareholders’ benefit at the expense of minority ones.

With respect to the model, F value is 16.000 (p < 0.000); adjusted $R^2$ is 0.451; and the difference between −2ln(L) and AIC is −2, which means that the whole independent variables can explain 45.1% of the variation in the dependent variable and the model is well-fitting. Regarding the residual analysis, Durbin-Watson is 1.909, which means that there is no concern about autocorrelation in the residuals because the value is close to 2.

As mentioned, $ACOM{P}_{it}$ is the median $ACOM{P}_{ijt}$ for all firms j in the same industry as firm i during period t in Table 4 and Table 5. It also tests with the average $ACOM{P}_{ijt}$ of either the four firms j or the ten firms j with the highest comparability to firm i during period t as per [37]. In untabulated results, it finds that the results for the average $ACOM{P}_{ijt}$ are similar to those for the median $ACOM{P}_{ijt}$.

4.5. Robustness Tests

4.5.1. Self-Selection Bias

To correct for self-selection bias from managers’ decisions to disclose carbon emissions, [39]’s two-stage model is used. As shown in Models 1 and 2 of

Table 6. Robustness tests: Self-selection bias.

Variables	Expected Sign	$\mathbf{Model} 1: \mathit{A}\mathit{C}\mathit{O}\mathit{M}\mathit{P}$	$\mathbf{Model} 2: \mathit{A}\mathit{C}\mathit{O}\mathit{M}\mathit{P}$
		Coefficient (t-Value)	Coefficient (t-Value)
$Constant$		−0.001 (−0.02)	0.003 (0.11)
$CDPDIS$	(+)	0.002 ** (2.24)	0.005 *** (3.99)
$CHAEBOL$	?	-	0.004 *** (3.26)
$CDPDIS\times CHAEBOL$	(−)	-	−0.004 ** (−2.47)
$SIZE$	(−)	−0.000 (−0.53)	−0.001 (−0.76)
$LEV$	(−)	−0.005 * (−1.89)	−0.006 ** (−2.12)
$MTB$	(+)	0.000 (1.01)	0.000 (1.01)
$CFO$	(+)	−0.013 * (−1.69)	−0.012 (−1.62)
$LOSSProb$	(−)	−0.023 *** (−7.53)	−0.023 *** (−7.57)
$StdSALES$	(−)	0.024 (1.23)	0.017 (0.84)
$StdCFO$	(−)	−0.077 ** (−2.03)	−0.049 (−1.27)
$StdGROW$	(−)	−0.000 (−0.27)	−0.000 (−0.47)
$BIG4$	(+)	0.006 *** (2.76)	0.004 ** (2.31)
$\%MAJOR$	(-)	−0.003 (−1.08)	−0.002 (−0.56)
$\%FOR$	(+)	0.000 (0.10)	0.002 (0.57)
$IMR$	?	0.002 (1.26)	0.003 (1.55)
$IND\_Dummy$		Include	Include
F value		16.870 ***	15.480 ***
Adjusted$R^2$		0.443	0.452
N		619	619

Notes: All variables are defined in Appendix A. Values in parentheses are t-values; ***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively.

, these results are similar to those reported in Table 4 and Table 5, which alleviates the concerns about self-selection bias. It performs regressions using [39]’s two-stage model as follows: In the first stage, it runs a probit model using equation (7), as per [6,8], and obtains the inverse Mills ratio. In the second stage, it includes the inverse Mills ratio as a control variable in Equations (1) and (2).

$$CDPDI{S}_{it}={\beta }_0+{\beta }_{1}SIZ{E}_{it}+{\beta }_{2}LE{V}_{it}+{\beta }_{3}RO{A}_{it}+{\beta }_{4}MT{B}_{it}+YEAR\_Dumm{y}_{it}+{\epsilon }_{it}$$

(7)

4.5.2. Total Sample

Annually, the top 250 Korean companies, based on market capitalization, are subject to disclosure to the CDP; thus, it is necessary to test the hypotheses using the total sample to investigate the relationship between carbon disclosure and accounting comparability, including involuntary non-disclosing firms. As shown in Models 1 and 2 of

Table 7. Robustness tests: Total sample.

Variables	Expected Sign	$\mathbf{Model} 1: \mathit{A}\mathit{C}\mathit{O}\mathit{M}\mathit{P}$	$\mathbf{Model} 2: \mathit{A}\mathit{C}\mathit{O}\mathit{M}\mathit{P}$
		Coefficient (t-Value)	Coefficient (t-Value)
$Constant$		0.018 (1.28)	0.045 *** (2.75)
$CDPDIS$	(+)	0.007 *** (4.07)	0.009 *** (4.55)
$CHAEBOL$	?	-	0.007 *** (5.09)
$CDPDIS\times CHAEBOL$	(−)	-	−0.005 ** (−2.22)
$SIZE$	(−)	−0.001 (−1.00)	−0.002 ** (−2.51)
$LEV$	(−)	−0.026 *** (−7.49)	−0.025 *** (−7.39)
$MTB$	(+)	−0.001 ** (−2.33)	−0.001 ** (−2.37)
$CFO$	(+)	0.006 (0.72)	0.007 (0.80)
$LOSSProb$	(−)	−0.050 *** (−17.34)	−0.051 *** (−17.43)
$StdSALES$	(−)	−0.005 (−0.28)	−0.009 (−0.48)
$StdCFO$	(−)	0.042 (1.12)	0.042 (1.12)
$StdGROW$	(−)	−0.000 (−0.39)	−0.000 (−0.30)
$BIG4$	(+)	0.003 ** (2.53)	0.002 * (1.80)
$\%MAJOR$	(−)	−0.001 (−0.20)	−0.001 (−0.24)
$\%FOR$	(+)	−0.003 (−0.70)	−0.001 (−0.27)
$IND\_Dummy$		Include	Include
F value		29.890 ***	27.860 ***
Adjusted$R^2$		0.392	0.397
N		2.784	2.784

Notes: All variables are defined in Appendix A. Values in parentheses are t-values; ***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively.

, these results are similar to those reported in Table 4 and Table 5, indicating that the findings are robust.

4.5.3. Alternative Measures

Prior research documents that firm-specific news is incorporated into stock prices before accounting earnings announcements. De Franco et al. [37], therefore, included lagged price changes into the accounting model as follows:

$$Earning{s}_{it}={\alpha }_i+{\beta }_{1i}Retur{n}_{it}+{\beta }_{2i}Retur{n}_{it-1}+{\epsilon }_{it}$$

(8)

where $Retur{n}_{it-1}$ is the stock return during the prior quarter. It then uses the variable $ACOM{P}^{PLE}$ instead of $ACOMP$ in Models 1 and 2 of

Table 8. Robustness tests: Alternative measures.

Variables	Expected Sign	$\mathbf{Model} 1: \mathit{A}\mathit{C}\mathit{O}\mathit{M}{\mathit{P}}^{\mathit{P}\mathit{L}\mathit{E}}$	$\mathbf{Model} 2: \mathit{A}\mathit{C}\mathit{O}\mathit{M}{\mathit{P}}^{\mathit{P}\mathit{L}\mathit{E}}$
		Coefficient (t-Value)	Coefficient (t-Value)
$Constant$		0.028 * (1.74)	0.040 ** (2.23)
$CDPDIS$	(+)	0.002 * (1.76)	0.004 *** (3.48)
$CHAEBOL$	?	-	0.004 *** (3.14)
$CDPDIS\times CHAEBOL$	(−)	-	−0.003 ** (−2.03)
$SIZE$	(−)	−0.001 *** (−2.61)	−0.002 *** (−3.02)
$LEV$	(−)	−0.007 ** (−2.50)	−0.008 *** (−2.77)
$MTB$	(+)	0.000 (0.26)	0.000 (0.02)
$CFO$	(+)	−0.012 (−1.49)	−0.011 (−1.39)
$LOSSProb$	(−)	−0.027 *** (−7.93)	−0.026 *** (−7.96)
$StdSALES$	(−)	0.022 (1.02)	0.014 (0.63)
$StdCFO$	(−)	−0.049 (−1.20)	−0.020 (−0.49)
$StdGROW$	(−)	−0.000 (−0.18)	−0.000 (−0.32)
$BIG4$	(+)	0.006 *** (2.89)	0.005 ** (2.39)
$\%MAJOR$	(−)	−0.005 (−1.49)	−0.003 (−0.99)
$\%FOR$	(+)	0.001 (0.37)	0.003 (0.79)
$IND\_Dummy$		Include	Include
F value		18.600 ***	17.190 ***
Adjusted$R^2$		0.451	0.458
N		619	619

Notes: All variables are defined in Appendix A. Values in parentheses are t-values; ***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively.

, and the results are similar to those reported in Table 4 and Table 5, which alleviates the concern that “prices lead earnings”.

4.6. Additional Analyses

4.6.1. Carbon Transparency

Lemma et al. [11] found that carbon emissions levels were positively related to earnings management. As such, it conducts OLS regressions with carbon emissions as a proxy for carbon transparency to investigate the relationship between carbon transparency and accounting comparability. As shown in Models 1 and 2 of

Table 9. Additional analyses: Carbon transparency.

Variables	Expected Sign	$\mathbf{Model} 1: \mathit{A}\mathit{C}\mathit{O}\mathit{M}\mathit{P}$	$\mathbf{Model} 2: \mathit{A}\mathit{C}\mathit{O}\mathit{M}\mathit{P}$
		Coefficient (t-Value)	Coefficient (t-Value)
$Constant$		0.104 *** (4.28)	0.106 *** (4.10)
$CO2EMS$	(−)	−0.004 *** (−2.67)	−0.003 ** (−2.21)
$CHAEBOL$	?	-	0.001 (0.79)
$CO2EMS\times CHAEBOL$	?	-	−0.001 (−0.29)
$SIZE$	(−)	−0.003 *** (−4.23)	−0.004 *** (−4.00)
$LEV$	(−)	−0.014 *** (−3.35)	−0.015 *** (−3.40)
$MTB$	(+)	−0.001 * (−1.70)	−0.001 * (−1.67)
$CFO$	(+)	−0.001 (−0.04)	0.001 (0.07)
$LOSSProb$	(−)	−0.015 *** (−3.16)	−0.015 *** (−3.12)
$StdSALES$	(−)	0.100 *** (2.76)	0.095 ** (2.55)
$StdCFO$	(−)	−0.269 *** (−3.63)	−0.262 *** (−3.47)
$StdGROW$	(−)	−0.003 ** (−1.99)	−0.003 * (−1.92)
$BIG4$	(+)	0.001 (0.52)	0.001 (0.44)
$\%MAJOR$	(−)	−0.012 * (−1.67)	−0.012 (−1.62)
$\%FOR$	(+)	−0.003 (−0.30)	−0.003 (−0.34)
$IND\_Dummy$		Include	Include
F value		40.260 ***	33.670 ***
Adjusted$R^2$		0.551	0.547
N		230	230

Notes: All variables are defined in Appendix A. Values in parentheses are t-values; ***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively.

, both the coefficients $CO2EMS$ are significantly negative (p = 0.008 and p = 0.028, respectively), indicating that carbon transparency is positively related to accounting comparability. However, the coefficient $CO2EMS\times ACOMP$ is not significant, suggesting that business groups have no moderating effect on the relationship between carbon transparency and accounting comparability. These results could be interpreted to indicate that it is difficult to manage carbon transparency even if firms in business groups opportunistically disclose carbon emissions.

4.6.2. Earnings Management

Luo and Wu [12] argued that voluntary carbon disclosure was negatively related to earnings management. Therefore, it conducts OLS regressions using the following Equation (9) with discretionary accruals as a dependent variable to confirm the moderating effect of business groups on the relationship between carbon disclosure and earnings management. In

Table 10. Additional analyses: Earnings management.

Variables	Expected Sign	$\mathbf{Model} 1: \mathit{D}\mathit{A}$	$\mathbf{Model} 2: \mathit{D}\mathit{A}$
		Coefficient (t-Value)	Coefficient (t-Value)
$Constant$		−0.144 **(−2.56)	−0.161 ***(−2.60)
$CDPDIS$	(−)	−0.009 **(−2.10)	−0.021 ***(−3.34)
$CHAEBOL$	?	-	−0.007(−1.25)
$CDPDIS\times CHAEBOL$	(+)	-	0.016 **(2.07)
$SIZE$	(+)	0.006 ***(3.25)	0.007 ***(3.21)
$LEV$	(−)	−0.026 **(−2.09)	−0.025 **(−2.00)
$GROW$	(+)	0.011(0.56)	0.012(0.65)
$T{A}_{t-1}$	(−)	0.112 **(2.09)	0.112 **(2.12)
$LOSS$	(−)	−0.041 ***(−6.10)	−0.042 ***(−6.16)
$BIG4$	(−)	−0.015(−1.44)	−0.013(−1.18)
$\%MAJOR$	(+)	−0.007(−0.53)	−0.009(−0.66)
$\%FOR$	(−)	−0.049 ***(−3.24)	−0.052 ***(−3.34)
$IND\_Dummy$		Include	Include
F value		10.540 ***	9.680 ***
Adjusted$R^2$		0.206	0.207
N		619	619

Notes: All variables are defined in Appendix A. Values in parentheses are t-values; ***, ** indicate significance at the 1%, 5% levels, respectively.

, I use discretionary accruals as per the modified-Jones model [40] in the accrual estimation equation. As shown in Models 1 and 2 of Table 10, both the coefficients $CDPDIS$ are significantly negative (p = 0.036 and p = 0.001, respectively), and the coefficient $CDPDIS\times CHAEBOL$ in Model 2 is significantly positive (p = 0.039), suggesting that voluntary disclosure firms are more likely to provide transparent financial reports and business groups are likely to mitigate the negative relationship between carbon disclosure and earnings management. These results strongly support the argument that ethical environmental management in non-business groups encourages managers to provide comparable and transparent financial reports, whereas opportunistic environmental management in business groups results in managers producing less comparable and transparent financial reports. In untabulated results, it uses discretionary accruals as per the performance-matched modified-Jones model [41] in the accrual estimation equation and finds the similar results. Additionally, it divides the sample into business groups and non-business groups and conducts OLS regressions with each sub-sample. In untabulated results, the coefficient $CDPDIS$ is insignificantly negative in business groups, and it is significantly negative in non-business groups, indicating that firms in non-business groups are more likely to voluntarily disclose carbon emissions to provide transparent information to various stakeholders compared to those in business groups.

$$\begin{array}{l} D{A}_{it}={\alpha }_0+{\alpha }_{1}CDPDI{S}_{it}+{\alpha }_{2}CHAEBO{L}_{it}+{\alpha }_{3}CDPDI{S}_{it}\times CHAEBO{L}_{it}\\ +{\alpha }_{4}SIZ{E}_{it}+{\alpha }_{5}LE{V}_{it}+{\alpha }_{6}T{A}_{it-1}+{\alpha }_{7}LOS{S}_{it}+{\alpha }_{8}BIG{4}_{it}+{\alpha }_9\%MAJO{R}_{it}\\ +{\alpha }_{10}\%FOR+IND\_Dumm{y}_{it}+{\epsilon }_{it}\end{array}$$

(9)

where discretionary accruals ($DA$) are calculated as per the modified-Jones model [40] as follows:

$$D{A}_{it}=\frac{T{A}_{it}}{A_{it-1}}-\left({\beta }_1\frac{1}{A_{it-1}}+{\beta }_2\frac{△S_{it}-△A{R}_{it}}{A_{it-1}}+{\beta }_3\frac{PP{E}_{it}}{A_{it-1}}\right)$$

(10)

where I calculate $DA$ using observations of at least 20 firms for each industry and year according to the same SIC two-digit industry classification.

5. Discussion of the Implications

5.1. Business Implications

The results in the paper imply that firms in non-business groups are likely to disclose non-financial activities including voluntary disclosure of carbon emissions to satisfy the interests of various stakeholders. Ethical environmental management encourages managers to provide comparable financial reports and, thus, increases the quality of financial reporting.

The quality of financial reporting is determined by various factors that are invisible and hard to find. Therefore, this study is meaningful for finding voluntary disclosure of carbon emissions to be a factor that increases the quality of financial reporting, and investors can easily identify the quality of financial reporting through voluntary disclosure of carbon emissions.

5.2. Regulatory Implications

On the other hand, the results in the paper show that business groups are likely to opportunistically disclose non-financial activities including the voluntary disclosure of carbon emissions to satisfy controlling shareholders’ benefit. Opportunistic environmental management results in managers producing less comparable financial reports and, thus, decreases the quality of financial reporting.

As mentioned above, the European Commission and the Korean government announced the achievement of their carbon-neutrality target by 2050. Therefore, a two-track strategy should be established. One is that managers should be encouraged to voluntarily disclose carbon emissions, and the other is that they should be provided sufficient guidance to prevent opportunistic disclosure of carbon emissions.

6. Conclusions

This study investigates whether firms provide more comparable financial reports when they voluntarily disclose carbon emissions to the CDP and whether business groups mitigate the relationship between carbon disclosure and accounting comparability. The sample consists of companies listed on the KSE between 2014 and 2018.

The findings support the first hypothesis that states that voluntary carbon disclosure is positively related to accounting comparability. I also find evidence that supports the second hypothesis that states that business groups mitigate the positive relationship between carbon disclosure and accounting comparability.

Three robust tests are conducted. First, the [39]’s two-stage model to correct for self-selection bias is used. Second, the hypotheses using the total sample to investigate the relationship between carbon disclosure and accounting comparability, including involuntary non-disclosing firms, are tested. Third, alternative measures for accounting comparability are used. Herein, these tests confirm that the results are similar to the main ones, indicating that the findings are robust.

Two additional analyses are performed. One is to test whether carbon transparency is related to accounting comparability, and it finds that carbon transparency is positively related to accounting comparability and that business groups have no moderating effect on this relationship. The other is to test the relationship between carbon disclosure and earnings management, and it finds that carbon disclosure is negatively related to earnings management and that business groups have a moderating effect on this negative relationship.

The main contribution of this study is that it finds empirical evidence that voluntary carbon disclosure to the CDP is positively related to accounting comparability and that business groups mitigate the positive relationship. Recently, environmental management, including carbon disclosure, has received greater attention following the Paris Agreement, and accounting comparability is considered as a critical qualitative characteristic under the IFRS and the US GAAP. Therefore, the novel results will help both academic and practical fields to better understand the relationship between voluntary carbon disclosure and accounting comparability within either a stakeholder or an agency theoretical framework.

A limitation of this study is that it may include inherent errors due to its measurements, endogeneity, and omitted variables, which could affect the results; however, I do attempt to alleviate these effects by conducting various analyses. Furthermore, future research should expand upon the findings using private data on carbon emissions through surveys because the public data obtained from the CDP reports limit the number of samples.