1 of 34

New Stories from Old Data:�Evidence From a Novel KLD Measure of Relative Environmental Responsibility

Gabriele Lattanzio (University of Melbourne) &

Lubomir P. Litov (University of Oklahoma)

Journal of Accounting, Auditing & Finance Conference

Hanken School of Economics

May 2023

2 of 34

Competing Through Environmental Responsibility

In this paper we propose and test that firms compete in the market for environmental responsibility ratings and that their relative position with respect to the best (worst) practices of their “green” (“toxic”) peersmay increase firm value.

Successfully competing through environmental responsibility ratings generates advantage that can lead to: higher sales; higher profit margin; higher profitability; higher firm value & higher stock returns.

Therefore, relative corporate environmental engagement (or relative environmental responsibility, i.e., RER) may contribute to firm value.

3 of 34

Is CSR engagement value relevant?

	Agency (Negative) View of CSR	“Doing Well by doing Good”(Positive) View of CSR
Theoretical Literature	Friedman (1970); Bénabou & Tirole (2010); Kitzmueller & Shimshack (2012).	Tirole (2001); Heinkel, Kraus, Zechner (2001); Magill, Quinzii, Rochet (2015); Stoughton, Wong, Li (2017); John, Lee, Oh (2018).
Empirical Literature	Margolis et al. (2009); Di Giuli & Kostovetsky (2014); Masulis and Reza (2015); Cheng et al. (2016); Hong et al. (2012); Lys et al. (2015).	Dowell et al. (2000); Edmans (2011); Flammer (2015); Fernando and Sharfman (2008); El Ghoul et al. (2011); Dhaliwal et al. (2011); Albuquerque et al. (2015); Ioannou et al. (2014); Hong and Liskovich (2016).

This noteworthy growth brings in the question as to whether corporate social responsibility induces material value implications for existing shareholders, and, if positive, through which economic channels. However, extant literature is inconclusive. On the one hand standard economic theory postulates that such investments should be irrelevant (e.g., Modigliani & Miller [1958]) or uncommon (e.g., Friedman [1970], Benabou and Tirole [2010], Kitzmueller and Shimshack [2012]). On the other hand, recent articles show within a rational framework that these socially desirable practices are consistent with profit maximization. For example, corporate social responsibility (CSR) engagement allows firms to mitigate corporate-specific risks that may have substantial impact on their profitability if passed through to existing stakeholders (e.g., consumers, employees, the environment).

Theory inconclusiveness is paired by contrasting empirical results, which further exacerbates the disagreement among both scholars and practitioners about whether CSR engagement increases shareholder wealth. In particular, studies have identified a strong positive association between CSR activities and revenues, corporate profitability and shareholder wealth (e.g., Dowell, Hart and Yeung [2000], Edmans [2011], and Flammer [2015]), as well as decreased cost of capital (El Ghoul et al. [2011], Dhaliwal et al. [2011], and Albuquerque, Durnev and Koskinen [2015]), lower financial constraints (Ioannou, Serafeim, and Cheng [2014]), and more lenient penalties in litigation (Hong and Liskovich [2016]). However, several contemporaneous studies document that CSR engagement is associated with lower firm value and negative future stock returns (Margolis, Elfenbein, and Walsh [2009], Di Giuli and Kostovetsky [2014], Lys, Naughton, Wang [2016]) and that it can help advance the self-interest of insider officers and directors at a cost to the corporation (Masulis and Reza [2015] and Cheng, Hong and Shue [2016]).

Moreover, some studies attribute any positive association between CSR and firm value to a spurious correlation. In fact, the often identified positive association between corporate social responsibility and shareholder wealth might be an empirical artifact due to the higher propensity to engage in CSR activities characterizing profitable firms (Hong, Kubik, and Scheinkman [2012], Lys, Naughton and Wang [2016]). See Tirole (2001), Heinkel, Kraus, and Zechner (2001), Magill, Quinzii, and Rochet (2015), Stoughton, Wong and Yi (2017), and John, Lee and Oh (2018).

4 of 34

Traditional Method of CSR Engagement

5 of 34

KLD Data Properties (Table 1)

Panel A: Environmental Strengths
Percentage of non-missing observations
Year	Beneficial Products	Pollution Prevention	Recycling	Clean Energy	Other Strengths	Excluded Factors
Full Sample	0.788	0.933	0.612	0.936	0.811	0.076
1998-2009	1.000	1.000	1.000	1.000	1.000	0.040
2010-2017	0.492	0.838	0.067	0.845	0.547	0.127
Percentage of “zero” observations
Year	Hazardous Waste	Regulatory Concerns	Ozone Depleting Chemicals	Substantial Emissions	Other Concerns	Excluded Factors
Full Sample	0.965	0.977	0.988	0.928	0.968	0.925
1998-2009	0.969	0.983	0.983	0.960	0.984	0.972
2010-2017	0.959	0.967	0.996	0.883	0.932	0.858

�

Panel B: Environmental Concerns
Percentage of non-missing observations
Year	Beneficial Products	Pollution Prevention	Recycling	Clean Energy	Other Strengths	Excluded Factors
Full Sample	0.788	0.933	0.612	0.936	0.811	0.454
1998-2009	1.000	1.000	1.000	1.000	1.000	0.282
2010-2017	0.492	0.838	0.067	0.845	0.546	0.694
Percentage of “zero” observations
Year	Hazardous Waste	Regulatory Concerns	Ozone Depleting Chemicals	Substantial Emissions	Other Concerns	Excluded Factors
Full Sample	0.967	0.953	0.999	0.952	0.990	0.947
1998-2009	0.944	0.935	0.998	0.951	0.985	0.945
2010-2017	1.000	0.977	1.000	0.954	0.999	0.950

6 of 34

An Inconsistent Measure

This proxy is problematic on four dimensions:

Richness:

Equally Weighted Individual Categories of Concerns & Strengths

All factors (strengths and concerns) receive equal weights.

Peer-Effect

Changes in best (worst) peer practices (due to for example change in technology or in social norms) may affect a firms’ CSR behavior (Flammer (2016), Cao et al. (2018)).

Relatedness

Each CSR factor is considered in isolation, rather than in an interconnected system.

General Electic Co.
Concerns	Strenghts
Hazardous waste	Environmental Opportunities
Regulatory compliance	Actions to deal with issues concerning climate change
Other concerns
TOTAL SCORE = 2 - 3 = -1
International Paper Co.
Hazardous waste	Effective Waste Management System
Toxic spills or releases
TOTAL SCORE = 1 - 2 = -1

Our ability to test for the existence of such competitive effects hinges on the availability of annual corporate CSR engagement data capturing both (i) changes in best (worst) CSR engagement practices in the relevant industry over time, and (ii) the time-series of a firm’s CSR engagement . In this sense, common measures of CSR engagement appear to be problematic. E.g., unitary measures of CSR built upon information provided in for example Kinder, Lydenberg, and Domini (i.e., KLD) dataset raise concerns to their interpretability, as they are based on a difference of the number of reported CSR “strengths” for a given CSR dimension and the number of CSR “concerns” for the same dimension.

A count measure such as the above is therefore unable to provide a metric of the intensity of a firm’s CSR engagement. as it would fail to distinguish between firms with different CSR positive (or negative) characteristics (i.e., “strengths” and “concerns” in KLD’s annotation), as well as between firms presenting the same number of negative and positive CSR characteristics with those involved in no CSR engagement at all.

Similarly, prior measures do not discern the various environmental concerns and strengths among themselves, as well as in terms of their importance as they equally weights strengths and concerns independently from their absolute (i.e., firm) and relative (i.e., within-industry) importance and materiality (Khan et al. [2018]). For example, following KLD’s concerns related to the correct disposal of eventual “hazardous waste” and to a firm’s “regulatory compliance” would receive equal weights in existing scores.

Third, importantly existing measures do not capture the effect of peers (i.e., the effect of best environmental practices of “green” industry peers as well as the worst environmental practices of “toxic” industry peers), whose economic relevance has been emphasized in recent studies (Cao, Liang, Zhang [2018] and Liu and Wu [2016]).

Lastly, commonly used counting scores do not take into consideration the relatedness of different CSR engagement categories appearing in for example KLD database, thus considering these social characteristics in isolation rather than within an interconnected system. Indeed, firms should strategically allocate their resources in a “portfolio” of CSR activities which should deliver the highest expected financial and/or social outcome, rather than embracing individual CSR engagements absent consideration of other possible CSR engagements.

E.g., a corporation with zero environmental concerns and one environmental strength will be viewed of the same environmental quality of CSR as a firm with 13 environmental strengths and 12 environmental concerns following existing popular metrics of CSR based on KLD data. Moreover, a firm with for example five environmental strengths and five environmental concerns will be viewed as a firm with zero environmental strengths and zero environmental concerns.

7 of 34

Environmental Responsibility

We focus on the five most common five strengths and five concerns in KLD.

For example, in the environmental dimension, we consider:

Concerns:

Hazardous waste (Env_con_A);
Regulatory compliance (Env_con_B);
Ozone depleting chemicals (Env_con_C);
Toxic spills or releases (Env_con_D);
Other concerns (Env_con_X).

Strengths:

Environmental Opportunities (Env_str_A);
Effective Waste Management System (Env_str_B);
Environmentally Friendly Packaging (Env_str_C);
Actions to deal with issues concerning climate change (Env_str_D);
Other strengths (Env_str_X).

Similar approach for employment, community, human rights, product, and governance dimensions of CSR in KLD.

8 of 34

Relative Environmental Responsibility (RER) Score

The distance to strengths (RERS) is therefore defined as: (s_i,j,t – s_*,j,t)’(s_i,j,t – s_*,j,t).

Similarly, the distance to concerns (RERC) is therefore defined as (c_i,j,t – c_*,j,t)’(c_i,j,t – c_*,j,t).

The RER score is defined as = log(1+RERC_i,j,t) – log(1+RERS_i,j,t).

9 of 34

RERS & RERC Score

We calculate the distance to strengths of

“green” peers as shown on this graph:

Industry vector

of strengths

of green peers

Firm distribution of

environmental strengths

We calculate the distance to concerns of

“toxic” peers as shown on this graph:

Industry vector of

concerns of

toxic firms

Firm distribution of

environmental concerns

10 of 34

RER Score (Continued)

We add a sixth element to the firm vector as nearly 83% of the KLD observations have neither concerns nor strengths.

We measure the distance to strengths of “green” firms and distance to concerns of “toxic” firms separately.

We define green & toxic peers within a industry-year as follows:

Green peers: peer firms within a given industry-year with zero environmental concerns and at least one strengths.

Toxic peers: peer firms within a given industry-year with zero environmental strengths and at least one concern.

We further examine alternative definitions: e.g., a “green” peer being defined as no environmental concerns, and maximum number of environmental strengths within an industry-year, etc.

We focus on the environmental dimension of RER.

11 of 34

RER Score Calculation: Example

Consider Chevron (2003) RSR Environmental Score calculation

Green & Toxic firms
Worst Practice	Best Practice
1. Burlignton Resources Inc	1. Marathon Oil Corp.
2. Lubrizol Corp	2. Murphy Oil Corp.
3. Magnum Hunter Rsrcs.	3. Nuevo Energy Co.
4. Patina Oil & Gas Corp.	4. Occidental Petroleum
5. Pogo Producing Co.	5. Premcor Inc.
6. Prima Energy Corp.	6. Quaker Chemical Corp.
7. Range Resources Corp.
8. Remington Oil & Gas Corp.
9. Spinnaker Exploration Co.
10. Vintage Petroleum Inc.
11. X T O Energy Inc.

Toxic Peers	Centroid	Chevron
Hazardous Waste	19%	25%
Regulatory Problems	21%	25%
Ozone Depleting Chemicals	1%	0%
Substantial Emissions	15%	25%
Other Concerns	5%	0%
In-Sample Indicator	39%	25%

Green Peers	Centroid	Chevron
Beneficial Products & Services	11%	0%
Pollution Prevention	4%	0%
Recycling	7%	0%
Clean Energy	25%	50%
Other Strenghts	4%	0%
Sample Indicator	49%	50%

12 of 34

Relative Environmental Responsibility Score

13 of 34

Net Environmental & Total Environmental Industry-Adjusted Scores Histograms

14 of 34

Summary Statistics

Score	N	Mean	Median	Min	Max
Relative Environmental Score (RER FF5)	20,716	0.0444	0.0284	-0.4361	0.8284
Relative Environmental Score (RER FF48)	10,180	0.0362	0.0064	-0.5108	0.9190
Relative Environmental Score (RER SIC-2)	9,063	0.0398	0.0191	-0.5174	0.9382
Relative Environmental Score (RER SIC-3)	6,021	0.0611	0.0076	-0.5108	0.9212
Total Environmental Score	22,622	0.0416	0.0000	-4.0000	4.0000
Net Environmental Score	22,340	0.0164	0.0000	-1.0000	1.0000
Industry-Adjusted Net Environmental Score FF48	22,340	0.0011	-0.0033	-1.2297	1.0052
Neutral	22,622	0.8140	1.0000	0.0000	1.0000
Green Firm	22,622	0.0951	0.0000	0.0000	1.0000
Toxic Firm	22,622	0.0542	0.0000	0.0000	1.0000

15 of 34

Correlation Matrix

TES is the Total Environmental Score, NES is the Net Environmental Score.

Our measure of environmental RSR is highly correlated with existing measures.

For that purpose we will perform robustness tests on said measure by using residuals

of a regression of RSR on TES (or NES, or Ind. Adj. TES) instead of RSR

	Total Score	Net Score	RER FF 48	RER SIC-3	RER SIC-2
Total Score	1.000
Net Score	0.904	1.000
RER FF 48	0.658	0.629	1.000
RER SIC-3	0.711	0.675	0.974	1.000
RER SIC-2	0.673	0.642	0.969	0.973	1.000

16 of 34

RER Measure Validation (Table 3, Panel A)

‘
	(1)	(2)	(3)	(4)
VARIABLES	Environmental Emissions

2-Lagged RER Environmental Score	-7.167^***
	(1.802)
2-Lagged Total Score		-2.245^***
		(0.529)
2-Lagged Net Score			-8.453^***
			(1.874)
2-Lagged Industry-Adj. Score				-8.801^***
				(2.016)
2-Lagged Environmental Emissions	0.353^***	0.318^***	0.319^***	0.319^***
	(0.0591)	(0.0604)	(0.0604)	(0.0603)
2-Lagged Size	2.603^***	1.949^***	1.947^***	1.986^***
	(0.757)	(0.624)	(0.622)	(0.623)
2-Lagged Log Sale	-0.310	0.303	0.340	0.297
	(0.713)	(0.710)	(0.710)	(0.706)

Year FE	Yes	Yes	Yes	Yes
Industry FE	FF48	FF48	FF48	FF48
Observations	2,298	3,396	3,396	3,396
R-squared	0.695	0.650	0.649	0.649

17 of 34

RER Measure Validation (Table 3, Panel B)

Panel B: Predicting Environmental Performance - Firm Estimates
	(1)	(2)	(3)	(4)
VARIABLES	Environmental Emissions

2-Lagged RER Environmental Score	-4.964^**
	(2.071)
2-Lagged Total Score		-1.059^***
		(0.384)
2-Lagged Net Score			-1.740
			(1.355)
2-Lagged Industry-Adj. Score				-0.877
				(1.439)

Additional Controls (as in Panel A)	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes
Observations	2,298	3,396	3,396	3,396
Adjusted R-squared	0.206	0.156	0.150	0.149

18 of 34

RER Measure Validation (Table 3, Panel C)

Panel C: Orthogonalization Test
	(1)	(2)	(3)
VARIABLES	Environmental Emissions

2-Lagged RER Score	-5.361^**	-5.098^**	-4.969^**
	(2.118)	(2.104)	(2.082)
2-Lagged Orth. Total Score	-0.873	-1.210
	(0.533)	(1.645)
2-Lagged Orth. Net Score			-0.0553
			(1.933)
2-Lagged Orth. Industry-Adj. Score


Additional Controls (as in Panel A)	Yes	Yes	Yes
Year FE	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes
Observations	2,298	2,298	2,298
Adjusted R-squared	0.209	0.206	0.206

19 of 34

Main Result: RER & Firm Value (Table 5)

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
VARIABLES	Tobin’s Q

RER Score	0.345^***	0.199^***			0.240^**	0.179^*	0.182^**
	(0.0752)	(0.0722)			(0.0951)	(0.103)	(0.0882)
RER Strengths			0.0412
			(0.0809)
RER Concerns				0.288^***
				(0.102)
Post-2009							-0.105
							(0.164)
Post-2009 x RER Score							0.0460
							(0.136)
Firm Level Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Industry FE & Year FE	Yes	No	No	No	No	No	No
Firm FE & Year FE	No	Yes	Yes	Yes	Yes	Yes	Yes
Observations	8,723	8,804	14,759	10,670	6,939	2,429	8,804
Adjusted R-squared	0.302	0.743	0.738	0.744	0.746	0.773	0.743

Table V reports regressions of Tobin's Q on the relative environmental responsibility score. All variables are defined in Table A.I, in the online appendix. Industry fixed effects are estimated based upon the Fama-French 48 industry classification. Standard errors are clustered at the firm level. All independent variables are lagged to avoid mechanical relations with the dependent variable. The ^*, ^**, and ^*** indicate statistical significance at the 10%, 5%, and 1% level, respectively.

Observations:

All independent variables are lagged
Standard errors are shown in parentheses below coef. estimates
RSR is always significant and economically important. Discuss its economic effect.
In specifications that uses industry fixed effects, NES, TES and ind. adj. TES do not load. That is due to the inclusion of our measure. When excluded, the NES coef and TES coef are positive and significant. These are …0.059 for NES, 0.169 for TES and 0.137 for Ind. Adj. TES.
Results appear to be stronger for neutral firms, importantly however they do not lose significance when neutral firms are controlled for – why – because the score for those might not be well defined
Results are stronger after 2009 but do not lose significance in the period preceding it. KLD retained material factors and dropped immaterial factors?
A one stand. dev. increase in RSR generate a 3.57% increase in Q. (calculated as coef. * stand. Dev. For RSR / in-sample average of Q) (for firm fixed effects)
Need to demean RES in interactions in models (5) and (6)

20 of 34

Marginal Information Content of RER (Table 6)

	(1)	(2)	(3)	(4)	(5)	(6)
VARIABLES:	Tobin's Q
RER Score	0.453^***	0.369^***	0.384^***	0.161^**	0.186^***	0.189^***
	(0.0916)	(0.0903)	(0.0900)	(0.0729)	(0.0720)	(0.0719)
Net Environmental Score	-0.103^***
	(0.0208)
Total Environmental Score		-0.254^***
		(0.0690)
Industry-Adj. Total Environmental Score FF48			-0.289^***
			(0.0692)
Residuals TS - RER FF48				-0.103^***
				(0.0208)
Residuals NS - RER FF48					-0.254^***
					(0.0690)
Residuals Ind.-Adj. Total Environmental Score – RER FF48						-0.289^***
						(0.0692)

Additional Controls	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes
Fiscal Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Observations	8,804	8,735	8,735	8,804	8,735	8,735
Adjusted R-squared	0.745	0.744	0.745	0.745	0.744	0.745

Table VI reports regressions of Tobin's Q on several environmental scores. All variables are defined in Table A.I, in the online appendix. All specifications include firm fixed effects and fiscal year fixed effects. Standard errors are clustered at the firm level. All independent variables are lagged to avoid mechanical relations with the dependent variable. Additional controls include firm size, firm age, loss, leverage, capital expenditures, R&D, institutional ownership, E-index, cash holdings, advertisement, and idiosyncratic volatility. The ^*, ^**, and ^*** indicate statistical significance at the 10%, 5%, and 1% level, respectively.

Observations:

Strong significance of the RSR residual score
Some of the existing literature scores (TES) are negative and significant, consistent with Friedman (1970)
Residual NES, TES and ind. adj. TES are all negative and statistically significant.
Economic effect magnitude? One stand. Dev. Increase in [Residuals RSR FF48 – NES] lead to 3.45% in Tobin’s Q
This nearly unchanged magnitude suggests that the value effect is nearly always captured by RSR, not NES, TES, nor ind.adj. TES.

Observations:

Strong significance of the RSR score
RSR does not change coefficient estimate significantly as compared to ind. fix. effect specification.
Traditional proxies for CSR engagement all negative and significant – consistent with the Friedman (1970) hypothesis
Results appear to be stronger for neutral firms, importantly, however they do not lose significance when neutral firms are controlled for
Among neutral firms, those with higher RSR have higher Q
Results are stronger after 2009 but do not lose significance in the period preceding it
-Economic magnitude 3.75%

21 of 34

Several Observations

Residual RER score maintains strong significance.

Residual NES, TES and industry-adjusted TES are all negative and statistically significant, consistent with Friedman (1970).

Economic effect magnitude: one stand. dev. increase in [Residuals RER FF48–NES] lead to 3.45% in Tobin’s Q.

This nearly unchanged magnitude suggests that the value effect is nearly always captured by RSR, not NES, TES, nor industry-adjusted NES.

We next focus on several robustness tests:

Varied industry definitions;
Redefined green and toxic firm definitions;
5 x 1 factor RER score;
All Factor RER (post 2009);
Distance to lagged centroids;
Time series extended through 1991-2016;
Controls for alternative CSR dimensions (employment, product, community, diversity, corporate governance).

22 of 34

Alternative RER Scores (Table 7)

	(1)	(2)	(3)	(4)	(5)
VARIABLES:	Tobin’s Q

RER 5 Factors	0.108^**
	(0.0449)
RER Full Sample (1991-2016)		0.145^**
		(0.0599)
RER Lagged Centroid			0.0915^*
			(0.0530)
RER All Factors				0.142^**
				(0.0639)
RER (Material Factors)					0.190^*
					(0.0927)

Additional Controls	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes
Fiscal Year FE	Yes	Yes	Yes	Yes	Yes
Observations	451	9,866	4,010	8,805	17,126
Adjusted R-squared	0.539	0.174	0.127	0.125	0.177

23 of 34

Robustness to Peer Definitions (Table 8)

	(1)	(2)	(3)	(4)	(5)	(6)
VARIABLES	Tobin’s Q

RER Score	0.199^***	0.140^**	0.131^**	0.106^*	0.208^***	0.207^***
	(0.0722)	(0.0642)	(0.0797)	(0.0712)	(0.0711)	(0.0733)
Green					-0.0289
					(0.0326)
Toxic					-0.0150
					(0.0412)
Neutral						0.0267
						(0.0309)

Additional Controls	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes
Fiscal Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Industry Definition	FF48	FF5	SIC 2	SIC 3	FF48	FF48
Observations	8,804	16,600	7,874	5,194	8,804	8,804
Adjusted R-squared	0.743	0.737	0.738	0.761	0.743	0.743

Table VIII reports regressions of Tobin's Q on the relative environmental responsibility score. All variables are defined in Table A.I in the online appendix. All specifications include firm fixed effects and fiscal year fixed effects. Standard errors are clustered at the firm level. All independent variables are lagged to avoid mechanical relations with the dependent variable. Additional controls include firm size, firm age, loss, leverage, capital expenditures, R&D, institutional ownership, E-index, cash holdings, advertisement, and idiosyncratic volatility. The ^*, ^**, and ^*** indicate statistical significance at the 10%, 5%, and 1% level, respectively.

Observations:

Alternative market definitions do not materially affect results
Using FF-5, SIC-3 and SIC-2 definitions produces similar results

24 of 34

Sales Growth, Profit Margin and Profitability�(Table 11)

	(1)	(2)	(3)	(4)
VARIABLES:	Sales Growth (t+1)	Sales Growth (t+1 & t+2)	ROA (t+1)	ROA (t+1 & t+2)

RER Score	0.0400^**	0.0372	0.00965^*	0.00840
	(0.0164)	(0.0317)	(0.00516)	(0.0129)

Firm FE	Yes	Yes	Yes	Yes
Fiscal Year FE	Yes	Yes	Yes	Yes
Observations	7,685	7,665	8,724	7,665
Adjusted R-squared	0.124	0.476	0.289	0.622

Economic magnitude for sales growth is 3.1%.

Economic magnitude for profit margin is 3.3%.

Economic magnitude for profitability (ROA) is 4.9%.

25 of 34

RER Score & Stock Market Performance (Table 9)

	(1)	(2)	(3)
VARIABLES	Raw Returns	Market Adjusted Returns	Size-Adjusted Returns

RER FF48	0.0509^**	0.0511^**	0.0579^**
	(0.0250)	(0.0252)	(0.0248)

Additional Controls	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes
Fiscal Year FE	Yes	Yes	Yes
Observations	8,805	8,805	8,805
Adjusted R-squared	0.355	0.137	0.125

26 of 34

Long Term Value: A Portfolio Regression Approach

If superior environmental ratings are a competitive advantage, and if this is not perfectly accounted for by investors, a hedged portfolio buying high environmental ratings firms and shorting low environmental rating firms should yield a positive alpha.

Portfolio regressions are based on the Fama-French Four-Factor Model.

Firms are sorted in decile (quintile) by their yearly score.

All portfolios are built as in January 1^st.

27 of 34

Portfolio Returns (Table 10)

Panel A: Value Weighted Portfolio - Decile
Short - Lowest Decile	Long - Highest Decile	Long - Short
0.0006	0.0027^***	0.0017^***
0.0015	0.0030^***	0.0021^***
0.0017^*	0.0030^***	0.0020^***
Panel B: Equal Weighted Portfolio – Decile
Short - Lowest Decile	Long - Highest Decile	Long - Short
0.0022	0.0047^***	0.0035^***
0.0021^*	0.0048^***	0.0040^***
0.0025^*	0.0049^***	0.0039^***
Panel C: Value Weighted Portfolio – Quintile
Short - Lowest Decile	Long - Highest Decile	Long - Short
0.005	0.0027^***	0.0017^**
0.0012	0.0030^***	0.0024^***
0.0016	0.0030^***	0.0023^***
Panel D: Equal Weighted Portfolio – Quintile
Short - Lowest Decile	Long - Highest Decile	Long - Short
0.0021	0.0047^***	0.0035^***
0.0019	0.0048^***	0.0041^***
0.0024^*	0.0049^***	0.0040^***
Panel E: Other Environmental CSR Measures
Value Weighted - Hedge Portfolio - Deciles
Net Score	Total Score	Industry-Adj. Score
-0.0014	-0.0012	0.0017
-0.0011	-0.0009	0.0017
-0.0012	-0.009	0.0015

Differences b/n VW and EW portfolios alphas.

Economic magnitude in for example VW (0;36) is annualized return of 2.55% =1.0021¹²-1.

Economic magnitude in for example EW (0;36) is annualized return of 4.91% =1.004¹²-1.

Table X reports estimates for calendar-time portfolio regressions estimated via the Fama-French 4 Factors Model. The "short portfolio" includes all firms for which RES FF48 is in the lowest decile (quintile) within a given fiscal year. Analogously, the "long portfolio" includes all firms in the highest decile (quintile) in a given fiscal year. The Long-Short portfolio is the hedged portfolio built by shorting the Short portfolio and by going long the Long portfolio. All portfolios are built on January 1st. The reported values represent Jensen's alphas. The holding period is reported in the number of months. Panel E reports Jensen's alpha based on hedged portfolios built upon three alternative measures, exclusively. All standard errors are heteroskedasticity robust. The ^*, ^**, and ^*** indicate statistical significance at the 10%, 5%, and 1% level, respectively.

Observations:

Differences b/n VW and EW portfolios
Economic magnitude in for example VW (0.36) is annualized return of 2.55% = 1.0021¹² - 1

28 of 34

Endogeneity Concerns

More profitable firms are more likely to be environmentally responsible.

Hong, Kubik, and Scheinkman (2012) and Lys, Naughton and Wang (2016).

We employ three instrumental variables to provide exogenous variation to firms’ engagement in CSR activities.

Current and lagged market-wise changes in the best practices regarding environmental engagements

State-specific outcome of the most recent U.S. presidential election

Di Giuli and Kostovetsky [2014]

Indicators whether a competitor has approved a CSR-related proposal

Flammer [2015] and Cao et al. [2018]

29 of 34

Market-wide Changes in Best (Worst) Practices

For small RERC or RERS, can approximate 1+log(RERC_i,j,t) by RERC_i,j,t.

Therefore ∆[1+log(RERC_i,j,t)] is approximately:

∆RERC_i,j,t= (c_i,j,t – c_*,j,t)’(c_i,j,t – c_*,j,t) - (c_i,j,t-1 – c_*,j,t-1)’(c_i,j,t – c_*,j,t-1)

It can be shown with some manipulation that the latter is equal to:

(c_i,j,t – c_i,j,t-1)’(c_i,j,t – c_,j,t-1) - 2(c_I,j,t – c_t,j,t-1)’ (c_i,j,t – c_i,j,t-1) –

(c_*,j,t – c_*,j,t-1)’(c_*,j,t – c_*,j,t-1) + 2(c_*,j,t – c_i,j,t-1)’ (c_*,j,t – c_*,j,t-1)

The first and second term in the above represent endogenous components, while the third and fourth term represent an exogenous

If c_i,j,t is nearly the same as c_i,j,t-1 (i.e., no change in the environmental policy at the firm level), and if c_i,j,t-1 is nearly the same as c_*,j,t-1 (i.e., convergence to the industry practices in equilibrium) this change collapses to (c_*,j,t – c_*,j,t-1)’(c_*,j,t – c_*,j,t-1).

An appealing feature of the REC, RES and RE measures is that their changes can be approximately decomposed into exogenous (peer group) and endogenous (firm) change components. For example, for small REC_i,j,t, we can approximate 1+log(REC_i,j,t) by REC_i,j,t.

Therefore ∆[1+log(REC_i,j,t)] is approximately ∆REC_i,j,t= (c_i,j,t – c_*,j,t)’(c_i,j,t – c_*,j,t) - (c_i,j,t-1 – c_*,j,t-1)’(c_i,j,t – c_*,j,t-1). It can be shown with some manipulation that the latter is equal to:

(c_i,j,t – c_i,j,t-1)’(c_i,j,t – c_,j,t-1) - 2(c_*,j,t – c_t,j,t-1)’ (c_i,j,t – c_i,j,t-1)

- (c_*,j,t – c_*,j,t-1)’(c_*,j,t – c_*,j,t-1) + 2(c_*,j,t – c_i,j,t-1)’ (c_*,j,t – c_*,j,t-1)

The first and second term in the above represent endogenous components (i.e., function of the firm vector of environmental concerns at time t), while the third and fourth term represent an exogenous (i.e., function of the industry vector of environmental concerns at time t, or function of pre-determined vector values at time t-1). Moreover, if c_i,j,t is nearly the same as c_i,j,t-1 (i.e., no change in the environmental policy at the firm level), and if c_i,j,t-1 is nearly the same as c_*,j,t-1 (i.e., convergence to the industry practices in equilibrium) this change collapses to (c_*,j,t – c_*,j,t-1)’(c_*,j,t – c_*,j,t-1). A similar decomposition can be presented for ∆RES_i,j,t. Therefore the change in the relative environmental score ∆RE_i,j,t can be approximated (under the above assumptions) by (c_*,j,t – c_*,j,t-1)’(c_*,j,t – c_*,j,t-1) - (s_*,j,t – s_*,j,t-1)’(s_*,j,t – s_*,j,t-1). Note that the latter is the exogenous variation in the worst (best) practices of the “toxic” (“green”) firms in the KLD sample. We use that variable as an instrument in our causal analysis.

30 of 34

First Stage 2SLS & GMM (Table 12)

	2SLS	GMM
VARIABLES:	RER Score
Lagged Delta Centroid	-0.0488^**	-0.0484*
	(0.0238)	(0.0366)
Lagged Peers Effect	0.0137^**	0.0306^***
	(0.00674)	(0.0108)
Democratic State (HQ level)	0.00319	0.0123
	(0.00713)	(0.00857)
Passed Environmental CSR Proposal	-0.00862	-0.00924
	(0.00900)	(0.00704)

Firm Level Control Variables	Yes	Yes
Firm and Fiscal Year FE	Yes	No
Industry and Fiscal Year FE	No	Yes
Kleibergen-Paap Wald F-Statistic	11.23
Hansen J Stat (p-value)	0.335
Kleibergen-Paap Underidentification Test (p-value)	0.0051
AR(1) and AR(2) test (p-value)		0.001 & 0.177
Hansen test of over-identification (p-value)		0.001
Diff-in-Hansen tests of exogeneity (p-value)		0.263
Observations		3,599
R-squared		0.095

31 of 34

RER & Firm Value: Endogeneity (Table 13)

	2SLS	GMM	First-Difference
VARIABLES:	Tobin's Q
RER Score	0.352^***	2.874^**	0.191^***
	(0.124)	(1.302)	(0.0432)

Firm Control Variables	Yes	Yes	Yes
Firm FE	Yes	No	No
Industry FE	No	Yes	Yes
Year FE	Yes	Yes	Yes
Kleibergen-Paap Wald F-Statistic	11.23
Hansen J Stat (p-value)	0.335
Kleibergen-Paap Underidentification Test (p-value)	0.0051
AR(1) test (p-value)		0.001
AR(2) test (p-value)		0.177
Hansen test of over-identification (p-value)		0.001
Diff-in-Hansen tests of exogeneity (p-value)		0.263
Observations	4,657	3,599	6,088
R-squared	0.153	0.095	0.144

32 of 34

The Cost of CSR Engagement (Table 14)

Panel A: First Stage
Variables:	RER Score
Lagged Delta Centroid	-0.0531^***	-0.501^***	-0.0542^***
	(0.0214)	(0.0214)	(0.0214)
Peer Effects	0.01189	0.01203	0.01412^*
	(0.0100)	(0.0094)	(0.0047)

Kleibergen-Paap Wald F-Statistic	12.31	11.99	9.87
Additional Controls	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes
Year FE	Yes	Yes	Yes
Observations	4,662	4,656	2,686
Adjusted R-squared	0.0036	0.0038	0.0034

Panel B: Second Stage
Variables:	CAPEX to Assets	Change in PPE	Change in intangible Capital

RER Score	0.150^**	1.966^**	-0.357
	(0.0728)	(0.949)	(0.261)

Additional Controls	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes
Year FE	Yes	Yes	Yes
Observations	4,662	4,656	2,686
Adjusted R-squared	0.0548	0.0976	0.0923

33 of 34

Conclusions

We introduce a measure to capture competition through environmental engagement ratings.

Relative Environmental Responsibility (RER) is positively associated with sales growth, profitability, profit margin, and ultimately Tobin's Q and shareholder returns.

Results appear to be robust to additional tests.

Results appear to be causal, or at least immune to the omitted or unobserved variable bias.

Impact of RER independent of NES and TES suggesting important incremental information present.

Future research could use RER as an additional metric for CSR to capture the value channel outlined above.

34 of 34

Appendix: Summary Statistics (Table 4)

Variable Name	N	Mean	Median	Min	Max
Institutional Ownership	22,622	0.7244	0.7560	0.0021	1.0000
R&D	22,622	0.1551	0.0000	0.2153	1163.8348
Advertisement	22,622	0.0117	0.0000	0.0000	8.0019
Idiosyncratic Volatility	22,622	0.3462	0.3049	0.0635	3.0262
E-Index	22,622	1.8936	1.0000	0.0000	6.0000
Cash Holding	21,874	0.1692	0.0934	0.0000	0.9956
CAPEX	22,622	0.0446	0.0289	0.0000	0.2764
Raw Returns	22,622	0.0625	0.9765	-3.8066	2.7776
Market Adjusted Returns	22,622	-0.0404	-0.0223	-3.1887	2.6569
Size Adjusted Returns	22,622	-0.0281	-0.0097	-3.5853	2.7862
Leverage	22,622	0.2134	0.1806	0.0000	0.8632
Loss	22,622	0.1774	0.0000	0.0000	1.0000
Firm Age	22,622	2.9791	2.9444	1.6094	4.1743
Size	22,622	4.5281	7.4002	2.3254	14.7603
Tobin's Q	22,622	1.9548	1.4760	0.7615	8.1086