Economic Analysis and Policy 77 (2023) 599–616

Contents lists available at ScienceDirect

Economic Analysis and Policy

journal homepage: www.elsevier.com/locate/eap

Modeling economic policy issues

Short-run and long-run effects of ESG policies on value

creation and the cost of equity of firms
∗
Javier Rojo-Suárez, Ana B. Alonso-Conde
Department of Business Administration, Rey Juan Carlos University, Paseo de los Artilleros s/n, 28032 Madrid, Spain

article info a b s t r a c t

Article history: Despite the general trend to include ESG scores in the evaluation of firm performance,
Received 9 June 2022 the effect of ESG policies on the market value of companies is currently a subject of
Received in revised form 24 November 2022 debate. In this paper we propose a dynamic version of Ohlson’s model under time-
Accepted 22 December 2022
varying discount rates consistent with the Campbell–Shiller present value identity. This
Available online 24 December 2022
enables differentiation between short term and long term implications of ESG perfor-
JEL classification: mance on value creation, as well as income and substitution effects. Our results suggest
F64 that, although ESG policies imply almost no effects in the short-run, at longer horizons,
G12 better ESG performance results in lower value creation, mainly due to substitution effects
G15 channeled to market value via higher long-term discount rates. Our results are consistent
G34
with ESG strategies implying transitory effects on the cost of equity and the market
Keywords: value, which may result from time-varying investor preferences, long-term reputational
ESG penalties, or market misvaluation.
CSR © 2022 Economic Society of Australia, Queensland. Published by Elsevier B.V. This is an
Corporate governance open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Value creation
Cost of equity
Dynamic Ohlson model

1. Introduction

Environmental issues, social policies and corporate governance concerns have experienced an extraordinary boom in
recent years given their evident importance in a wide variety of areas, ranging from sustainable economics to politics.
However, the effects of these aspects on corporate value creation are still far from clear. Furthermore, the classic
distinction between short-run and long-run effects on the value of corporations and the cost of capital, typically analyzed
in the literature on asset pricing and capital structure, is almost unexplored in the study of the impact of environmental,
social and corporate governance (ESG) policies on firm performance. The fact that such considerations depend not only
on the informational value of ESG policies on firm fundamentals, but also on investor preferences, largely explains the
mixed results provided by the recent literature on the topic (Pedersen et al., 2021). Thus, while part of the literature
concludes that ESG strategies are positively related to shareholder value (Hong and Kacperczyk, 2009; Luo and Balvers,
2017; Zerbib, 2020; Zhang and Lucey, 2022; Pástor et al., 2021, 2022), other studies conclude the opposite, stating that
ESG policies translate into lower value creation (Hassel et al., 2005; Baker et al., 2018; Tampakoudis et al., 2021) or can
produce ambiguous outcomes (Pedersen et al., 2021).
On this basis, in this paper we propose a dynamic Ohlson (1995) model, which uses economic profit to account for
abnormal earnings under a time-varying cost of equity consistent with the Campbell and Shiller (1988) model, to study the

∗ Corresponding author.
E-mail address: [email protected] (A.B. Alonso-Conde).

https://doi.org/10.1016/j.eap.2022.12.017
0313-5926/© 2022 Economic Society of Australia, Queensland. Published by Elsevier B.V. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
J. Rojo-Suárez and A.B. Alonso-Conde Economic Analysis and Policy 77 (2023) 599–616

extent to which the market value of stocks captures information resulting from the variation over time in ESG performance
at the firm level. As shown below, according to our model, the contemporary cross-sectional relationship between ESG
ratings and abnormal earnings, measured by economic profit, determines the short-run effects of ESG strategies on value
creation, while the predictive power of the ESG score to forecast the components of economic profit – specifically return
on equity (ROE) and cost of equity – captures long-run effects on the market value of shares. Therefore, based on the
Ohlson (1995) model setup, we exploit these relationships to study the extent to which ESG policies imply short-term
shocks and long-term effects on the value for shareholders.
We evaluate model performance using accounting and market data from all companies listed on the equity markets of
®
the four largest economies in the euro zone, namely, Germany, France, Italy and Spain, for which Refinitiv ESG Scores
®
provided by the Datastream database are available. Remarkably, Refinitiv provides data series that comprise one of
the largest ESG data collection among private databases, collected from different sources of information such as annual
reports, stock exchange filings, corporate social responsibility (CSR) reports or company websites. Thus, our final sample
consists of five years of historical data from 2016 to 2020 for 487 firms, sorted into ‘Banks’, ‘Industrial’, ‘Insurance’ and
‘Other financial’ industries, following the general categories defined by Datastream. Hence, our sample not only allows
us to study the effects of ESG strategies in aggregate terms, but also to analyze the differences that arise across different
countries and sectors, and their relationship with different variables, such as the price-to-book value ratio, ROE and cost
of capital.
Our paper contributes to the literature on the topic in the following terms. First, to the best of our knowledge, this is
the first paper to use economic profit, determined according to a time-varying cost of equity consistent with Campbell
and Shiller (1988), as a proxy for current abnormal earnings within the Ohlson (1995) model setup. In this regard, other
literature in the area uses the Ohlson (1995) model to study the extent to which the market value of equities accounts for
ESG information. For example, in their classic paper, Hassel et al. (2005) use the Ohlson (1995) model on Swedish listed
firms to conclude that environmental performance has a negative influence on the market value of companies. However,
to avoid an explicit specification of abnormal earnings, the authors reformulate the model to write market value increased
by dividends paid in the period as a function of lagged market value and current net income. Landau et al. (2020) follow
a similar approach to study the effect of integrated reports on the equity value of firms included in the STOXX Europe
50 index. However, both studies ignore the explicit definition of abnormal earnings and required rates of return in the
model specification. In this context, we show below that our measure of economic profit is strongly significant and with
high explanatory power for the difference between the market value and the book value of the companies under study.
Second, although previous research in the area analyzes the effect of ESG policies on value creation for specific sectors
(Ionescu et al., 2019; Miralles-Quirós et al., 2019) or countries (Hassel et al., 2005; Bofinger et al., 2022; Rodríguez-García
et al., 2022; Pástor et al., 2022), studies that simultaneously analyze the main industries and countries of the euro zone,
both on an aggregate and an individual basis, are the exception. Furthermore, our database comprises 1846 observations,
which represents a significantly larger sample than many studies on the topic. For example, Castro et al. (2021) use
an approach based in part on the Ohlson (1995) model to study the impact of environmental performance on firms’
stock prices. However, although the authors use a sample of 2638 European firms, their study covers only the effect
of environmental variables on market value, ignoring social and governance issues. Similarly, Grassmann (2021) uses a
sample of 8992 observations to study the effect of environmental expenditures and CSR on firm value. Importantly, our
results show that the effects of ESG strategies on value creation exhibit different industry- and country-specific patterns,
highlighting the relevance of our study and calling into question complete market integration at the European level.
Third, our approach allows us to distinguish short-run and long-run effects of ESG strategies on value creation, which
has rarely been studied in the related literature despite its obvious importance. For example, the results documented by
Hassel et al. (2005) allow the authors to conclude that the negative relationship between environmental performance
and equity value suggests that the ‘best’ firms in terms of environmental policies are not, in general, highly valued
by investors. However, the authors do not differentiate between short-term and long-term effects, thus overlooking
potentially offsetting relationships in their research. Moreover, this shortcoming is present in most of the related literature
(Ionescu et al., 2019; Landau et al., 2020; Castro et al., 2021; Grassmann, 2021). In contrast, the conditional nature of our
model allows us to study the effects of ESG strategies on current ROE and cost of equity, but also their impact on long-term
value creation. In fact, the conditional form in which our model is expressed represents an important advance in relation
to other asset pricing models developed to account for ESG information. For example, Pástor et al. (2021) and Pástor et al.
(2022) assume heterogeneous investor tastes for green holdings and a single-period setup to propose a two-factor model
that accounts for the effects of ESG preferences on unconditional expected returns. However, the unconditional nature of
the model developed by the authors hinders a straight evaluation of the effects of ESG policies over time.
Our results show that while in some industries ESG policies enhance value creation, in most cases a higher ESG
commitment translates into lower long-term market value relative to book value. Furthermore, with some exceptions, for
most of the countries and sectors under analysis, this lower value creation is primarily driven by long-term effects inducing
higher discount rates, which in the vast majority of cases is not compensated with a higher ROE. Hence, our results are
consistent with ESG strategies translating into higher expected returns in the long-run rather than with negative effects
on expected earnings due to costs incurred to improve environmental performance.
The remainder of the paper is organized as follows. Section 2 defines the model. Section 3 describes the data and
discusses model results. Finally, Section 4 concludes the paper.
600
J. Rojo-Suárez and A.B. Alonso-Conde Economic Analysis and Policy 77 (2023) 599–616

2. Methodology

We build on the classic Gordon dividend growth model and an accounting system that satisfies a clean surplus relation,
as follows:
∞
∑
Pt = R−τ Et (Dt +τ ) (1)
τ =1

Bt = Bt −1 + Xt − Dt (2)
where P t is the share price at time t, R is the required rate of return plus unity, E t (·) is the expectation conditional on
time t information, Dt is the dividend paid at time t, Bt is the book value per share, and X t denotes earnings per share
at time t, under the assumption that the company keeps the number of shares constant over time in order to simplify
notation. In this framework, residual income valuation models arise naturally by defining abnormal earnings per share
X at as the difference between earnings and the expected return on equity:

X at = X t − (R − 1) Bt −1 (3)
Hence, Eqs. (1) to (3) result in the following residual income valuation model:
∞
∑
R−τ E t Xta+τ
( )
P t = Bt + (4)
τ =1

Eq. (4) allows us to directly relate the Ohlson (1995) model and the Campbell and Shiller (1988) present value identity.
In particular, the Ohlson (1995) model introduces specific information dynamics for abnormal earnings, as follows:
X at+1 = ωX at + νt + εt +1 (5)

νt +1 = γ νt + ηt +1 (6)
where ν t denotes relevant information not captured by accounting, ω and γ are parameters, and ε t +1 and ηt +1 are error
terms. Based on Eqs. (4) to (6), Ohlson (1995) obtains the following pricing function:
P t = Bt + Φ1 X at + Φ2 νt (7)
where:
Φ1 = ω/ (R − ω) (8)

Φ2 = R/ [(R − ω) (R − γ )] (9)
Importantly, although Eqs. (8) and (9) show that Φ1 and Φ2 can be determined endogenously within the model setup, in
practice these coefficients are often used as regression coefficients for model evaluation purposes. Furthermore, following
the common practice of testing the Ohlson (1995) model using panel data analysis under fixed or variable effects, Φ1
and Φ2 are often assumed to be constant over time. However, this practice is inconsistent with the main results and
conclusions of the literature on the predictability of stock returns, which shows that different economic and non-economic
variables exhibit significant predictive power in forecasting expected returns (Campbell, 1987; Fama and French, 1988;
Baker and Wurgler, 2000; Lettau and Ludvigson, 2001; Lamont and Stein, 2004; Cochrane, 2011; Novy-Marx, 2014; Rojo-
Suárez et al., 2022). On this basis, Campbell and Shiller (1988) develop their widely-recognized loglinear present value
model, which overcomes the constraint of a constant discount rate in Eq. (1) to account for evidence of stock return
predictability. Hence, based on the standard definition of gross return:
Pt +1 + Dt +1
Rt +1 = (10)
Pt
Campbell and Shiller (1988) derive the following present value identity:
∞
∑ ∞
∑
pdt ≈ ρ j−1 ∆dt +j − ρ j−1 rt +j (11)
j=1 j=1

where pdt is the price–dividend ratio in logs at time t, r t +j is the log return, ∆dt +j is log dividend growth, and ρ =
exp (pd) / [1 + exp (pd)]. We can use Eq. (11) instead of Eq. (1) to derive the Ohlson (1995) model without loss of
generality, resulting in the following pricing function:
Pt = Bt + Φ1,t (R) Xta + Φ2,t (R) νt (12)
where Φj,t (R) denotes model coefficients at time t, conditional on the vector of expected returns R. Remarkably, Eq. (12)
shows that time-varying discount rates directly result in time-varying model coefficients, in which we call hereafter the
601
J. Rojo-Suárez and A.B. Alonso-Conde Economic Analysis and Policy 77 (2023) 599–616

dynamic Ohlson model. At this point, it is important to note that the predictability pattern of dividend growth and discount
rates relies heavily on the forecasting regression of the two terms in the right-hand side of Eq. (11) on the variables used
as predictors. Specifically, using the classic derivation of Campbell and Shiller (1988):
∞
∑
ρ j−1 ∆dt = bd dpt + ε d (13)
j=1
∞
∑
ρ j−1 rt +j = br dpt + ε r (14)
j=1

where dpt is the dividend yield in logs, bd and br are regression coefficients, and ε d and ε r are error terms. Furthermore,
recent research on return predictability opens the door to multivariate explanations in forecasting dividends and expected
returns, which implies that Eqs. (13) and (14) must consider forecasting variables other than the dividend yield to capture
short-run and long-run effects on predictability patterns (Cochrane, 2011). Accordingly, the dynamics represented in
Eqs. (11), (13) and (14) allow us to write abnormal earnings X at in Eq. (12) as a function of the forecasting regressions
of their main components. Specifically, scaling X at by the opening book value, Eq. (3) naturally results in the following
expressions:

X t +τ /Bt = b′X Ft + ε Xt+τ (15)

Rt +τ = b′R Ft + ε Rt+τ (16)

where τ denotes the number of lags, bX and bR are K -dimensional vectors of parameters, Ft is a K -dimensional vector of
predictors, and εtX and εtR are error terms.
In order to keep the model in Eqs. (12), (15) and (16) useful and manageable for practical applications, below we
introduce the following transformations to analyze the effect of ESG policies on value creation and cost of capital. First,
we use Eq. (12) to explain the difference between the market value and book value of shares over time instead of the
market price. Second, we proxy abnormal earnings by economic profit, that is, the product of the opening book value
and the difference between the ROE and the stock return. Third, we identify relevant information not yet captured by
accounting νt in Eq. (12) and the vector of predictors Ft in Eqs. (15) and (16) with ESG performance captured by the
variation in ESG scores. Accordingly, Eqs. (12), (15) and (16) can be rewritten as follows:

MVi,t +τ − BVi,t +τ = Φt + ΦEP ,t EPi,t + ΦESG,t ∆ESGi,t + εiMV −BV

,t +τ (17)

ESG ∆ESGi,t + εi,t +τ

ROEi,t +τ = aROE + bROE ROE
(18)

Ri,t +τ = aR + bRRM RMt +τ + bRESG ∆ESGi,t + εiR,t +τ (19)

where MV i,t denotes the market value of equity for the company i at time t, BV i,t is the book value of equity, EP i,t is the
economic profit, ∆ESGi,t is the variation rate of the ESG score, RM t is the return on the market portfolio, the coefficients
Φ , a and b are model parameters, and the variables ε are error terms. Consistent with the literature on stock return
predictability, Eq. (17) shows that the t subscripts in Φ coefficients allow the model to account for time-varying expected
returns, which means that the procedure used to estimate these parameters in Eq. (17) must be consistent with this fact.
Regarding Equation (19), consistent with the Campbell and Shiller (1988) present value identity, the returns Ri,t are
determined according to the standard definition of gross return in Eq. (10). Furthermore, these returns together with ROE i,t
in Eq. (18) determine EP i in Eq. (17), which illustrates the fact that Eqs. (17) to (19) are interrelated parts of the same
model and, consequently, must be interpreted jointly. Finally, given the strong comovement of stocks in equity markets,
in order to allow the model to isolate the fraction of returns that results from the variation of ESG scores, Eq. (19) includes
the return on the market portfolio – that is, the return on a broad-based value-weighted portfolio – as an explanatory
variable.
As noted, in the next section we use the defined model to analyze the effects of ESG strategies on value creation, ROE
and cost of equity for all companies traded on the securities exchanges of the four largest economies of the euro zone,
namely, Germany, France, Italy and Spain. In this context, our model not only allows us to isolate income and substitution
effects through the study of ROEi,t and Ri,t dynamics, respectively, but also provides a robust framework to evaluate the
effects of ESG policies at different horizons.

3. Results and discussion

We compile all accounting and market data from the Datastream database. Specifically, we use the ‘Worldscope Balance
Sheet’ and ‘Worldscope Profit & Loss Statement’ templates to compile the financial statements of all listed firms in the
®
four largest economies in the euro zone by GDP for which Refinitiv ESG Scores are available. This search totals 517
companies, of which 188 are German, 159 French, 99 Italian and 71 Spanish. However, the strong presence of missing
602
J. Rojo-Suárez and A.B. Alonso-Conde Economic Analysis and Policy 77 (2023) 599–616

Table 1
Total number of observations.
Country Sector Number of observations
All All 1846
All Banks 108
All Industrial 1526
All Insurance 68
All Other fin. 144
Germany All 653
France All 586
Italy All 345
Spain All 262

data for some companies and for the years prior to 2016 reduces our sample to 487 firms and an annual data time interval
spanning 2016 to 2020. As noted above, we use the general categories defined by Datastream to sort all companies into
four groups, namely, ‘Banks’, ‘Industrial’, ‘Insurance’ and ‘Other financial’ industries. Table 1 shows the total number of
observations in our sample by industry and by country.
In order to determine stock returns and account for the market value of the firms under study, we compile total return
and market value series from the Datastream database (RI and MV series, respectively). Importantly, total return series
includes returns resulting from price variations as well as dividend payments, as required by the model. Additionally, we
proxy the return on the market portfolio series by the cross-sectional average return of the companies under analysis,
weighted by market value.
®
Regarding ESG indicators, we use Refinitiv ESG Scores to proxy for ESGi,t in Eqs. (17) to (19). These indicators are
divided into three groups, namely, the Environmental Score, the Social Score and the Governance Score, each including
different data categories. Thus, the Environmental Score is divided into ‘Resource use’, ‘Emission’ and ‘Environmental
innovation’ scores. The Social Score includes ‘Workforce’, ‘Human rights’ and ‘Community’ scores. Finally, the Governance
Score consists of scores for ‘Management’, ‘Shareholders’ and ‘CSR strategy’. Table A.1 in Appendix A shows full details
on the variables used to determine these scores, some of which are numeric while others are Boolean. Therefore, the
Environmental Score, the Social Score and the Governance Score result from determining weighted averages of these
indicators, ranging from 0 to 100. Additionally, we estimate an integrated ESG score that is determined by the weighted
average of the three scores provided by Refinitiv, hereinafter referred to as the ESG score.
Fig. 1 shows the means and confidence intervals for different variables, namely, price-to-book value ratios, ROE, cost
of equity and ESG score. It should be noted that, although the ROE follows a downward trend over time for the period
under study, the cost of equity reaches a minimum value in 2019 to increase considerably in 2020, probably due to the
turbulence caused by the COVID-19 pandemic. For this time interval, the ESG score exhibits a U-shape with a minimum
value in 2018. Regarding industries, the industrial sector is the one that presents the highest price-to-book value ratio and
ROE among those considered, also exhibiting a high cost of equity. Conversely, banks and insurance companies have the
lowest price-to-book value and ROE, with banks also having the lowest cost of equity and the highest ESG score. Focusing
on country-specific patterns, Fig. 1 shows that Spain has the highest – albeit widely dispersed – price-to-book value ratio
and ROE, as well as the highest ESG score. On the other hand, Germany exhibits the highest cost of equity and the lowest
ESG score, as well as remarkably low ROE.
Regarding the estimation procedures followed to determine model coefficients, we use different panel data analysis
tools to adapt model estimation to the specific features of Eqs. (17) to (19). Specifically, the time-varying coefficients
in Eq. (17) require the use of panel data estimation under variable coefficients. By contrast, the constant coefficients in
Eqs. (18) and (19) allow us to use standard panel data analysis to estimate their parameters. Additionally, according to the
results provided by the Hausman test on the models under study, we assume fixed effects in the estimation of Eq. (17),
while random effects to estimate parameters in Eqs. (18) and (19). Consequently, Tables 2 to 4 show the main results
obtained for Eqs. (17) to (19), respectively, using the ESG score to account for ESG performance. In order to study short-
run and long-run effects of ESG policies, each table shows the model results assuming 0 to 2 lags in the ESG score.
Additionally, Table B.1 to Table B.9 in Appendix B show model results using the Environmental Score, the Social Score
and the Governance Score as information variables instead of the ESG score.
Table 2 documents the estimates for the slope coefficients in Eq. (17), as well as the standard errors and R2 statistics
ignoring and including ∆ESGi,t as information variable (labeled ‘R2 plain’ and ‘R2 full’ in Table 2, respectively), with the
last column showing the difference between these R2 statistics. The results in Table 2 provide us with several important
findings. First, as noted, economic profit exhibits a strongly significant explanatory power for the difference between
market value and book value, where in the vast majority of cases a higher contemporary economic profit translates into a
lower value for MV i,t − BV i,t , and vice versa. Moreover, the significance of economic profit is particularly important in the
case of the banking sector. Nevertheless, it should be noted that the negative relationship between economic profit and
value creation is a logical consequence of the dynamics represented in Eqs. (10) and (17). Indeed, a higher economic profit
at time t implies a higher value of ROEi,t relative to Ri,t , which, at least in period t, implies a stronger positive variation
in book value than in market value ceteris paribus.
603
J. Rojo-Suárez and A.B. Alonso-Conde Economic Analysis and Policy 77 (2023) 599–616

Fig. 1. Means and confidence intervals of the variables under analysis.

Second, Table 2 shows that the variation in the ESG score exhibits low explanatory power in Panels A and B, meaning
that ESG policies seem to imply negligible effects in value creation in the short-run. In fact, only for the insurance sector
in Panel A does the variation in the ESG score become statistically significant, implying in turn an increase in the R2
statistic of 7.2%. However, Table 2 also shows that, at longer horizons, the variation in the ESG score becomes statistically
604
J. Rojo-Suárez and A.B. Alonso-Conde Economic Analysis and Policy 77 (2023) 599–616

Table 2
Regression results for MVi,t − BVi,t using the ESG score as information variable.
Country Sector ΦEP σ (ΦEP ) ΦESG · 10−3 σ (ΦESG )· 10−3 R2 plain R2 full Diff.
Panel A: Number of lags = 0
All All −1.149∗∗∗ 0.117 −0.005 0.004 16.5% 17.4% 1.0%
All Banks −1.300∗∗∗ 0.066 0.016 0.013 83.2% 84.7% 1.5%
All Industrial −1.228∗∗∗ 0.310 −0.009 0.005 14.4% 16.8% 2.4%
All Insurance −2.063∗∗∗ 0.281 −0.114∗∗∗ 0.033 70.7% 78.0% 7.2%
All Other fin. −1.973∗∗∗ 0.272 0.006 0.008 51.4% 53.4% 2.1%
Germany All −0.948∗∗∗ 0.174 −0.016∗ 0.007 24.4% 26.7% 2.3%
France All −1.452∗∗∗ 0.255 −0.020 0.013 15.9% 17.5% 1.6%
Italy All −1.673∗∗∗ 0.177 0.001 0.003 43.8% 44.0% 0.2%
Spain All 3.752∗∗∗ 0.854 −0.007 0.008 22.9% 24.2% 1.4%
Panel B: Number of lags = 1
All All −1.259∗∗∗ 0.144 −0.006 0.004 16.5% 20.1% 3.6%
All Banks −1.298∗∗∗ 0.082 0.008 0.010 83.2% 90.4% 7.2%
All Industrial −1.183∗∗∗ 0.312 −0.008 0.005 14.4% 18.3% 3.8%
All Insurance −2.041∗∗∗ 0.301 0.009 0.014 70.7% 79.6% 8.9%
All Other fin. 1.809∗∗∗ 0.318 0.002 0.004 51.4% 57.1% 5.7%
Germany All −1.044∗∗∗ 0.223 −0.011 0.007 24.4% 29.2% 4.8%
France All −1.622∗∗∗ 0.302 −0.023 0.016 15.9% 20.3% 4.3%
Italy All −1.479∗∗∗ 0.226 −0.002 0.003 43.8% 34.2% −9.6%
Spain All 3.810∗∗∗ 0.896 −0.006 0.008 22.9% 28.3% 5.4%
Panel C: Number of lags = 2
All All −1.315∗∗∗ 0.155 −0.012∗∗ 0.004 16.5% 24.2% 7.7%
All Banks −1.324∗∗∗ 0.087 −0.005 0.005 83.2% 96.7% 13.5%
All Industrial −0.990∗ 0.390 −0.011∗ 0.005 14.4% 18.8% 4.4%
All Insurance −1.955∗∗∗ 0.318 0.009 0.010 70.7% 88.3% 17.6%
All Other fin. −2.124∗∗∗ 0.361 −0.009 0.006 51.4% 80.7% 29.4%
Germany All −1.132∗∗∗ 0.244 −0.012∗ 0.006 24.4% 33.3% 8.9%
France All −1.664∗∗∗ 0.316 −0.026 0.016 15.9% 25.3% 9.3%
Italy All −1.483∗∗∗ 0.259 −0.002 0.004 43.8% 39.9% −3.9%
Spain All −0.962∗∗ 0.306 −0.013 0.020 22.9% 23.7% 0.9%

Note: The table shows the slope coefficients and standard errors that result from the panel data regression of the difference between the market
value and the book value, using the ESG score as an information variable. Asterisks denote significance, where ∗ p < 0.1, ∗∗ p < 0.05, and ∗∗∗ p < 0.01.
Each panel use a different number of lags for the ESG score, ranging from 0 to 2. Columns labeled ‘R2 plain’ and ‘R2 full’ show the R2 statistics of
the regressions ignoring or including the ESG score as an information variable, respectively. The column labeled ‘Diff.’ shows the difference between
these statistics.

significant in the entire sample, implying important increases in the R2 statistics in several cases, especially for companies
in the financial sector (i.e. banks, insurance companies and other financial firms). Thus, while the 2-lag ESG score in Panel
C results in the R2 statistic for the banking sector increasing from 83.2% to 96.7%, for the insurance sector it increases from
70.7% to 88.3%. Furthermore, in the case of other financial companies, the R2 statistic rises from 51.4% to 80.7%. Across the
countries considered, the effects of firm-specific ESG policies are much smaller, with only Germany providing statistically
significant coefficients, albeit a modest increase in the R2 statistic.
As noted above, the coefficients ΦESG in Table 2 show that the variation in ESG score is, in general, inversely related
to the difference between market value and book value, meaning that the greater the increase in ESG performance, the
lower the value of MVi,t − BVi,t , and vice versa. In this regard, Eqs. (13) and (14), and their equivalents, Eqs. (18) and (19),
show that, within our model setup, the variation in the ESG score influences MVi,t − BVi,t via ROEi,t and Ri,t at potentially
infinite horizons. Accordingly, in order to study the effects of ESG policies on ROE and cost of equity at different horizons,
Table 3 shows the regression results for Eq. (18), while Table 4 does the same for Eq. (19). Specifically, Table 3 shows
the estimates for the slope coefficient, the p-value and the R2 statistic that result from the forecasting regressions of the
ROE on the variation in the ESG score (see Eq. (18)) across industries and countries. On the other hand, Table 4 has the
same structure as Table 2, showing the estimates for the slope coefficients in Eq. (19), the p-values and the R2 statistics
ignoring and including ∆ESGi,t as a predictor, with the last column showing the difference between both R2 statistics.
The results in Table 3 show that ROE is scarcely affected by the variation in the ESG score at all horizons. Furthermore,
this applies to all industries and countries under analysis, with the insurance sector in Panel C achieving the highest R2
statistic (27.3%). These results suggest that, contrary to previous literature that refers to the cost-concerned school to
explain the negative influence of environmental performance on the market value of firms (see for example Hassel et al.
(2005) and Landau et al. (2020)), cost increases tied to ESG strategies do not seem to explain the lower value creation
of companies with higher ESG scores. On the contrary, as illustrated in Tables 3 and 4, our results suggest that it is not
the effects of ESG policies on ROE (i.e. income effects), but rather the effects on discount rates (i.e. substitution effects)
that primarily drive differences in value creation across firms. Specifically, the results in Table 4 show that, although
the variation in the ESG score has a small effect on the cost of equity in the short-run, its explanatory power increases
significantly with the horizon, as it is the case with MVi,t − BVi,t in Table 2.
605
J. Rojo-Suárez and A.B. Alonso-Conde Economic Analysis and Policy 77 (2023) 599–616

Table 3
Regression results for ROEi,t using the ESG score as information variable.
( )
Country Sector bROE
ESG p bROE
ESG R2
Panel A: Number of lags = 0
All All 0.018 0.499 0.0%
All Banks 0.067 0.147 2.4%
All Industrial 0.018 0.534 0.0%
All Insurance −0.010 0.331 3.8%
All Other fin. 0.008 0.765 0.8%
Germany All 0.111∗∗ 0.041 0.9%
France All 0.128 0.165 0.3%
Italy All 0.077 0.233 0.7%
Spain All −0.005 0.936 0.0%
Panel B: Number of lags = 1
All All 0.002 0.947 0.0%
All Banks −0.002 0.960 0.4%
All Industrial 0.002 0.958 0.0%
All Insurance 0.013 0.183 9.7%
All Other fin. 0.050 0.251 4.0%
Germany All 0.029 0.615 0.2%
France All −0.227∗∗ 0.043 1.4%
Italy All 0.053 0.363 0.3%
Spain All −0.003 0.967 0.0%
Panel C: Number of lags = 2
All All −0.008 0.958 0.0%
All Banks 0.085 0.272 1.6%
All Industrial −0.018 0.929 0.0%
All Insurance −0.005 0.769 27.3%
All Other fin. −0.015 0.846 4.0%
Germany All 0.045 0.660 0.1%
France All −0.381∗ 0.082 1.9%
Italy All 0.040 0.560 6.8%
Spain All 0.221 0.840 0.1%

Note: The table shows the slope coefficient and the p-value that result from the panel data regression
of ROE, using the ESG score as an information variable. Asterisks denote significance, where ∗ p < 0.1,
∗∗
p < 0.05, and ∗∗∗ p < 0.01. The table also provides the R2 statistic of the regressions. Each panel use
a different number of lags for the ESG score, ranging from 0 to 2.

More precisely, Table 4 shows that contemporary and 1-lag ESG scores are essentially useless in explaining the cross-
sectional variation of discount rates (see Panels A and B), with only the banking sector experiencing a modest 10% increase
in the R2 statistic in Panel B. However, Panel C in Table 4 shows that the variation in the ESG score becomes highly
explanatory when forecasting the cost of equity at a 2-year horizon, especially for companies in financial sectors. In
particular, while ∆ESGi,t causes the R2 statistic for the banking sector to increase from 37.5% to 61.1%, the R2 statistic for
‘Other financial’ firms rises from 41.2% to 66.5%. Regarding the countries under study, the variation in the R2 statistic is
highest for Germany and Spain, where ∆ESGi,t leads the R2 statistic to increase by 15.9% and 12.7%, respectively.
Importantly, in most of the cases represented in Table 4, Panel C, the slope coefficient bRESG is positive, meaning that a
higher ESG performance generally implies a higher discount rate in the long-run, and vice versa. This fact is notable for
banks and ‘Other financial’ firms, for which the bRESG coefficients are strongly significant. On the other hand, Table 4 also
shows that the return on the market portfolio RMt has significant explanatory power in estimating discount rates for most
of the industries and countries under analysis, consistent with the strong comovement of stock returns. Furthermore, the
bRRM coefficients can be interpreted within the capital asset pricing model setup (Sharpe, 1964; Lintner, 1965a,b) as the
beta coefficients of stock returns on the wealth portfolio return, where Table 4 shows that for most of the industries and
countries under study betas are around 1, with the exception of the insurance sector, where the beta coefficient is below
0.5.
Summarizing the results from Tables 2 to 4 we have the following. Although the effects of ESG policies are small and
show little significance in the short-run, for longer time intervals, ESG performance is inversely related to the difference
between the market value and the book value, and generally implies a higher cost of capital in the long-run for most of
the sectors and countries under study. Conversely, the effects of the ESG strategies on ROE are almost negligible for all
horizons. Remarkably, contemporary economic profit has a significant negative effect on value creation across all horizons.
Most of the patterns illustrated in Tables 2 to 4 persist in Table B.1 to B.9 in Appendix B, where we use the
Environmental Score, Social Score and Governance Score instead of the ESG score as information variables. Remarkably,
the Governance Score provides a statistically significant 2-year ΦG coefficient using the entire sample (see Panel C in
Table B.7), with values similar to those shown in Table 2 for the ESG score. In contrast, although the Environmental Score
and Social Score in Tables B.1 and B.4 allow Equation (17) to increase the R2 statistic to the same extent as the ESG score in
606
J. Rojo-Suárez and A.B. Alonso-Conde Economic Analysis and Policy 77 (2023) 599–616

Table 4
Regression results for Ri,t using the ESG score as information variable.
( ) ( )
Country Sector bRRM p bRRM bRESG p bRESG R2 plain R2 full Diff.
Panel A: Number of lags = 0
All All 1.002∗∗∗ 0.000 0.024 0.182 27.8% 27.9% 0.1%
All Banks 0.975∗∗∗ 0.000 −0.029 0.833 37.5% 37.7% 0.2%
All Industrial 1.032∗∗∗ 0.000 0.029 0.131 27.5% 27.6% 0.2%
All Insurance 0.424∗∗∗ 0.002 −0.118∗ 0.061 17.9% 23.4% 5.5%
All Other fin. 0.978∗∗∗ 0.000 −0.074 0.410 41.2% 41.8% 0.6%
Germany All 1.246∗∗∗ 0.000 0.133∗∗ 0.047 31.0% 31.5% 0.5%
France All 0.918∗∗∗ 0.000 0.037 0.640 29.3% 29.3% 0.1%
Italy All 0.981∗∗∗ 0.000 −0.167∗∗ 0.033 39.8% 41.4% 1.5%
Spain All 0.582∗∗∗ 0.000 0.018 0.364 11.1% 11.5% 0.4%
Panel B: Number of lags = 1
All All 0.932∗∗∗ 0.000 −0.007 0.721 27.8% 27.0% −0.8%
All Banks 0.980∗∗∗ 0.000 −0.000 0.998 37.5% 47.5% 10.0%
All Industrial 0.962∗∗∗ 0.000 −0.010 0.644 27.5% 26.4% −1.1%
All Insurance 0.381∗∗∗ 0.006 0.048 0.447 17.9% 20.8% 3.0%
All Other fin. 0.839∗∗∗ 0.000 0.125 0.332 41.2% 37.5% −3.7%
Germany All 1.148∗∗∗ 0.000 0.075 0.418 31.0% 29.8% −1.2%
France All 0.838∗∗∗ 0.000 −0.055 0.595 29.3% 27.9% −1.4%
Italy All 0.890∗∗∗ 0.000 0.056 0.536 39.8% 38.1% −1.8%
Spain All 0.628∗∗∗ 0.000 −0.014 0.496 11.1% 14.5% 3.4%
Panel C: Number of lags = 2
All All 0.941∗∗∗ 0.000 0.080 0.246 27.8% 35.7% 8.0%
All Banks 0.903∗∗∗ 0.000 0.553∗∗ 0.017 37.5% 61.1% 23.6%
All Industrial 0.974∗∗∗ 0.000 0.003 0.976 27.5% 36.0% 8.6%
All Insurance 0.406∗∗ 0.012 −0.136 0.246 17.9% 29.9% 12.0%
All Other fin. 1.146∗∗∗ 0.000 0.685∗∗∗ 0.000 41.2% 66.5% 25.3%
Germany All 1.086∗∗∗ 0.000 0.051 0.626 31.0% 46.8% 15.9%
France All 0.849∗∗∗ 0.000 0.208 0.185 29.3% 38.4% 9.1%
Italy All 0.946∗∗∗ 0.000 0.377∗∗∗ 0.004 39.8% 47.5% 7.7%
Spain All 0.829∗∗∗ 0.000 −0.265 0.282 11.1% 23.9% 12.7%

Note: The table shows the slope coefficients and p-values that result from the panel data regression of the cost of equity, using the ESG score as
an information variable. Asterisks denote significance, where ∗ p < 0.1, ∗∗ p < 0.05, and ∗∗∗ p < 0.01. Each panel use a different number of lags for
the ESG score, ranging from 0 to 2. Columns labeled ‘R2 plain’ and ‘R2 full’ show the R2 statistics of the regressions ignoring or including the ESG
score as an information variable, respectively. The column labeled ‘Diff.’ shows the difference between these statistics.

Table 2, their significance is lower, generally showing a positive relationship with MVi,t − BVi,t for companies in financial
sectors. Regarding ROE, Table B.2, B.5 and B.8 in Appendix B provide similar results to those shown in Table 3, with ESG
variables providing a weak explanation for ROE across industries and countries, with the sole exception of the insurance
sector, where the Environmental Score, Social Score and Governance Score provide R2 statistics above 25% at a 2-year
horizon. Notably, the effects of ESG policies on the cost of equity vary more across ESG indicators than for MVi,t − BVi,t
and ROEi,t . Thus, while the Environmental Score and Governance Score in Tables B.3 and B.9 have lower explanatory power
than the ESG score in Table 4, the Social Score in Table B.6 has high explanatory power in forecasting discount rates at a
2-year horizon, providing negative slope coefficients for most of the sectors and countries under study.
Hence, our results show that the smaller difference between market value and book value that results in the long-
run for the best ESG performers is consistent with higher ESG scores forecasting higher long-term discount rates, rather
than higher costs stemming from ESG policies. This suggests that investors are willing to accept lower returns in the
short term – or equivalently, pay higher current prices in the stock markets – than in the long-run for those companies
committed to ESG principles. Accordingly, good ESG performance generally translates into higher long-term discount rates
and, consequently, lower market value at long horizons. Therefore, our results are consistent with ESG policies implying
transitory effects on the cost of equity, which may be a consequence of time-varying investor preferences, long-term
reputational penalties, or short-term market misvaluation.
Our findings are partially in line with those reported by Pástor et al. (2022), who show that U.S. stocks issued by firms
committed to ESG principles (i.e. green stocks) outperformed stocks of firms with little commitment to ESG principles
(brown stocks) for the period from 2012 to 2020. Moreover, based on the equilibrium model proposed by Pástor et al.
(2021), the authors explain that such outperformance is directly related to shifts in customers’ tastes for green products
and investors’ tastes for green holdings, which may partly explain the predictive power of ESG scores to forecast future
stock returns. However, according to the authors, that does not mean that green stocks have higher expected returns
than brown stocks. In fact, Pástor et al. (2022) explain that just the opposite is true, with green stocks exhibiting a lower
unconditional cost of capital than brown stocks as a consequence of investors’ green tastes and the fact that green assets
are a better hedge against climate risk. Nonetheless, the authors also highlight the complexity of disentangling ex ante
and ex post effects of ESG preferences by looking at realized returns in periods of changing ESG tastes. In this regard,
607
J. Rojo-Suárez and A.B. Alonso-Conde Economic Analysis and Policy 77 (2023) 599–616

our results are not contradictory with those provided by Pástor et al. (2022) given the conditional nature of the dynamic
Ohlson model proposed in our paper and the unconditional form of the Pástor et al. (2021) model. In fact, as noted
above, our results suggest that investors are conditionally willing to accept lower returns in the short term than in the
long-run for green stocks, which does not mean that the unconditional cost of equity of green firms is higher. Moreover,
following Cochrane (2011), our results are perfectly reconcilable with those obtained by Pástor et al. (2022) under specific
term structures of time-varying expected returns. However, a thorough empirical analysis of the relationship between the
dynamic Ohlson model and that of Pástor et al. (2021) requires considering potentially infinite horizons – or at least a
sufficiently high number of periods –, which is a difficult task with the currently available ESG information.
Our results also complement the findings provided by Bofinger et al. (2022), who find that an improvement in a
company’s CSR leads to a higher ratio of actual to true firm value, mainly due to the current global trend of sustainable
investing. Additionally, our results are consistent with the model proposed by Pedersen et al. (2021), in which the effects
of ESG performance on equilibrium prices largely depend on the presence of different types of investors who are more
or less aware of ESG policies.

4. Conclusions

Despite the general trend to include sustainability and CSR indicators to evaluate firm performance and non-financial
value creation, the effect of such policies on the market value of companies is currently the subject of a lively debate. In
this context, we propose a dynamic version of the Ohlson (1995) model that accounts for abnormal earnings using the
economic profit under time-varying discount rates consistent with the Campbell and Shiller (1988) model, in order to
differentiate between short term and long term implications of ESG performance on corporate value creation, as well as
income and substitution effects.
Our results suggest that, although ESG policies imply almost no effects on value creation in the short-run, at longer
horizons, better ESG performance results in a smaller difference between market value and book value, mainly due to
substitution effects channeled to market value via higher long-term discount rates. These effects are particularly clear for
firms in financial sectors, such as banks and other financial institutions, which are characterized by relatively low price-
to-book value ratios and cost of capital. Hence, our results are consistent with ESG strategies implying transitory effects
on the cost of equity and the market value of firms, which may result from non-separabilities in investor preferences that
include ESG factors within marginal utility, among other reasons.
Based on these results, future research should address different aspects that may provide further explanation about
the effects of ESG performance on corporate value. Regarding time horizon effects of ESG performance, our study faces the
limitations that arise from a short time series on ESG information. Future research should not only expand the sample
period studied, but also find different proxies for ESG variables for which longer time series are available. This could
include well-established procedures in the asset pricing literature, such as mimicking portfolio analysis.
On the other hand, our results are sensitive to capital structure effects. Specifically, although our results suggest that
ESG policies have small effects on ROE, return on equity is directly affected by the difference between return on invested
capital minus the cost of borrowing, which may result in offsetting effects when analyzing the impact of ESG scores on
ROE. Future research should study the extent to which other value creation measures, such as economic value added (EVA),
may result in effects not considered in our research. Furthermore, further research on the effects of ESG performance on
borrowing capacity is mandatory.

Acknowledgments

We would like to acknowledge Christine Brown of Monash University for her valuable comments and helpful
suggestions on the manuscript.

Funding

This work was supported by Programa Operativo FEDER Andalucía 2014–2020 [B-SEJ-740-UGR20].

Appendix A

See Table A.1.

Appendix B

See Tables B.1–B.9.

608
 J. Rojo-Suárez and A.B. Alonso-Conde
Table A.1
®
Indicators considered to determine Refinitiv ESG Scores.
Environmental Social Governance

Resource use Emission Environ. innov Workforce Human rights Community Product resp Management Shareholders CSR strategy

Environment Policy Emissions Environ Products Empl Satisfaction Policy Freedom of Policy Fair Competition Customer Satisfaction Board Functions Shareholder Rights CSR Sust Cmte
Management Team Targets Emissions Noise Reduction Diversity and Opp Association Policy Bribery and Policy Customer Health & Board Meeting Att Policy Integrated Strategy in
Policy Water Efficiency Biodiversity Impact Fleet Fuel Women Employees Policy Child Labor Corruption Safety Succession Plan Voting Cap Percentage MD&A
Reduction Consumption Women Managers Policy Forced Labor Policy Business Ethics Policy Data Privacy External Consult Global Compact
Policy Energy Flaring Gases Hybrid Vehicles HRC Corporate Equality Policy Human Rights Improvement Tools Busi Policy Responsible Adt Cmte Mgt Ind Director Election Signatory
Efficiency Cement CO2 Fleet CO2 Emissions Index Fund Human Rights ILO Ethics Marketing Comp Cmte Ind/Mgt Majority Req Stakeholder
Policy Sustainable Equivalents Emis Environ Assets Under Flexible Hours UN Whistleblower Protection Policy Fair Trade Nom Cmte Ind Shareholders Vote on Engagement
Packaging Ozone-Depleting Mgt Day Care Services Human Rights Contractor Policy Community Product Resp Monitoring Nom Cmte Involv Executive Pay CSR Sust Reporting
Policy Environ Supply Substances Nuclear Production Empl With Disabi Ethical Trading Initiative Involvement Product Access Low Price Board Attendance Public Availability GRI Report Guidelines
Chain NOx and SOx Labeled Wood Trade Union Repr ETI OECD Guidelines for Healthy Food or Products Board Structure Corporate Statutes CSR Sust Report
Targets Water Emissions Rd Percentage Turnover of Empl Human Rights Breaches Multinational Enterprises Embryonic Stem Cell Brd Bkgd and Skills Veto Power or Golden Global Activities
Efficiency e-Waste Reduction Organic Products Strikes Contr Extractive Industries Research Board Gender Div share CSR Sust External
Targets Energy Emissions Trading Initiatives Salary Gap Transparency Initiative Retailing Responsibility Brd Specific Skills State Owned Audit
Efficiency Environ Ptr GMO Products Net Empl Creation Community Lending and QMS Certified Percent Board Tenure Enterprise SOE
Environ Materials EMS Certified Agrochemical Products Announced Layoffs To Investments Quality Mgt Systems Non-Exec Brd Mbr Equal Shareholder
609

Sourcing Environ Restoration Total Employees Product Sales at Discount Ind Board Members Rights
Toxic Chemicals Init Animal Testing Health & Safety to Emerging Markets CEO-Chairman Sep Anti Takeover Devices
Reduction Staff Trans Impact Renewable/Clean Employees Health & Safety Diseases of the Developing Brd Member Affl Above Two
Cement Energy Use Reduction Energy Products Team World Brd Indiv Re-election Auditor Tenure
Green Buildings Climate Change Comm Water Tech Empl Health Safety Critical Country 1 Board Cultural Diversity, Litigation Expenses
Water Recycled Risks Sustainable Building Training Hours Corporate Resp Awards Percent Non-audit to Audit
Environ Supply Chain Self-Reported Environ Products Employees Health & Safety Total Donations To Executive Members Gender Fees Ratio
Mgmt Fines Real Estate Sust OHSAS Revenues Diversity
Env Supply Chain Ptr Estimated CO2 Certifications Supply Chain Health & Executive Comp
Termination Equivalents Emis Env R&D Expnd To Safety Comp Impr Tools
Land Environ Impact VOC or Particulate Revenues Occ Diseases CEO Compensation
Reduction Matter Emis Red Equator Principles or HIV-AIDS Program Total Senior Exec
Environ Supply Chain Total Waste To Env Project Fin Injuries To MM Hrs Sh Approval Stock Comp
Monitoring Revenues USD Renewable Energy Lost To Total Days Plan
Total Energy Use To Waste Recycled To Supply Trng and Dev Plcy Exec Indiv Comp

Economic Analysis and Policy 77 (2023) 599–616

Revenues Total Waste Product Impact Avg Training Hrs Highest Rem Pkg
Renewable Energy Use Total Hazardous Waste Minimization Internal Promotion Exec Comp LT Obj
Ratio To Rev Supplier ESG training Sust Comp Incentives
Water Use To Water Pollutant Emis Training Costs/Empl Int Audit Dept Report
Revenues To Revenues
Environ Expnd Inv
J. Rojo-Suárez and A.B. Alonso-Conde Economic Analysis and Policy 77 (2023) 599–616

Table B.1
Regression results for MVi,t − BVi,t using the Environmental Score as information variable.
Country Sector ΦEP σ (ΦEP ) ΦE · 10−3 σ (ΦE )· 10−3 R2 plain R2 full Diff.
Panel A: Number of lags = 0
All All −1.133∗∗∗ 0.120 −0.002 0.002 16.5% 17.3% 0.9%
All Banks −1.295∗∗∗ 0.067 0.005 0.005 83.2% 84.3% 1.0%
All Industrial −1.223∗∗∗ 0.313 −0.003 0.002 14.4% 17.0% 2.6%
All Insurance 0.740∗∗∗ 0.074 −0.086∗∗∗ 0.009 70.7% 80.6% 9.9%
All Other fin. −1.948∗∗∗ 0.306 0.000 0.000 51.4% 53.3% 1.9%
Germany All −0.905∗∗∗ 0.183 −0.005 0.006 24.4% 25.2% 0.8%
France All −1.465∗∗∗ 0.260 −0.012 0.008 15.9% 17.8% 1.9%
Italy All −1.685∗∗∗ 0.181 −0.001 0.001 43.8% 44.0% 0.2%
Spain All 3.749∗∗∗ 0.871 −0.003 0.004 22.9% 24.5% 1.6%
Panel B: Number of lags = 1
All All −1.231∗∗∗ 0.146 −0.002 0.002 16.5% 19.7% 3.2%
All Banks −1.303∗∗∗ 0.083 0.001 0.004 83.2% 90.2% 7.0%
All Industrial −1.201∗∗∗ 0.316 −0.002 0.002 14.4% 18.0% 3.6%
All Insurance −2.034∗∗∗ 0.310 0.032 0.035 70.7% 81.2% 10.5%
All Other fin. −1.891∗∗∗ 0.371 0.002 0.003 51.4% 57.5% 6.1%
Germany All −1.000∗∗∗ 0.231 −0.004 0.006 24.4% 27.6% 3.1%
France All −1.592∗∗∗ 0.302 −0.012 0.010 15.9% 20.0% 4.1%
Italy All −1.472∗∗∗ 0.231 −0.001 0.001 43.8% 34.3% −9.5%
Spain All 3.786∗∗∗ 0.901 −0.012 0.016 22.9% 28.2% 5.3%
Panel C: Number of lags = 2
All All −1.277∗∗∗ 0.159 −0.004 0.005 16.5% 23.2% 6.7%
All Banks −1.310∗∗∗ 0.084 0.015 0.010 83.2% 96.9% 13.7%
All Industrial −0.994∗ 0.392 −0.002 0.003 14.4% 18.1% 3.7%
All Insurance −1.884∗∗∗ 0.297 0.002∗ 0.001 70.7% 88.7% 18.0%
All Other fin. −2.009∗∗∗ 0.391 0.002 0.003 51.4% 78.5% 27.2%
Germany All −1.055∗∗∗ 0.255 −0.000 0.000 24.4% 30.8% 6.4%
France All −1.655∗∗∗ 0.316 −0.015 0.010 15.9% 25.3% 9.4%
Italy All −1.515∗∗∗ 0.264 −0.001 0.001 43.8% 40.4% −3.4%
Spain All −0.956∗∗ 0.302 −0.014 0.016 22.9% 23.8% 0.9%

Note: The table shows the slope coefficients and standard errors that result from the panel data regression of the difference between the market
value and the book value, using the Environmental Score as an information variable. Asterisks denote significance, where ∗ p < 0.1, ∗∗ p < 0.05,
and ∗∗∗ p < 0.01. Each panel use a different number of lags for the Environmental Score, ranging from 0 to 2. Columns labeled ‘R2 plain’ and ‘R2
full’ show the R2 statistics of the regressions ignoring or including the ESG score as an information variable, respectively. The column labeled ‘Diff.’
shows the difference between these statistics.

Table B.2
Regression results for ROEi,t using the Environmental Score as information variable.
( )
Country Sector bROE
E p bROE
E R2
Panel A: Number of lags = 0
All All 0.001 0.858 0.0%
All Banks 0.014 0.532 0.4%
All Industrial 0.004 0.746 0.0%
All Insurance −0.007 0.145 5.9%
All Other fin. 0.000 0.909 0.3%
Germany All −0.001 0.886 0.0%
France All 0.068 0.402 0.1%
Italy All 0.009 0.333 0.6%
Spain All −0.060 0.837 0.0%
Panel B: Number of lags = 1
All All −0.001 0.947 0.0%
All Banks 0.007 0.751 0.1%
All Industrial −0.005 0.877 0.0%
All Insurance 0.005 0.113 18.6%
All Other fin. 0.000 0.957 0.6%
Germany All 0.000 0.939 0.3%
France All −0.154∗ 0.094 1.0%
Italy All 0.017 0.304 0.4%
Spain All −0.301 0.649 0.2%

(continued on next page)

610
J. Rojo-Suárez and A.B. Alonso-Conde Economic Analysis and Policy 77 (2023) 599–616

Table B.2 (continued).

( )
Country Sector bROE
E p bROE
E R2
Panel C: Number of lags = 2
All All 0.000 0.994 0.0%
All Banks 0.037 0.544 0.0%
All Industrial 0.026 0.818 0.0%
All Insurance 0.001 0.884 27.8%
All Other fin. 0.000 0.864 2.1%
Germany All 0.001 0.832 0.0%
France All 0.067 0.688 0.1%
Italy All −0.004 0.861 7.3%
Spain All 0.175 0.891 0.0%

Note: The table shows the slope coefficient and the p-value that result from the panel data regression
of ROE, using the Environmental Score as an information variable. Asterisks denote significance, where
∗
p < 0.1, ∗∗ p < 0.05, and ∗∗∗ p < 0.01. The table also provides the R2 statistic of the regressions. Each
panel use a different number of lags for the Environmental Score, ranging from 0 to 2.

Table B.3
Regression results for Ri,t using the Environmental Score as information variable.
( ) ( )
Country Sector bRRM p bRRM bRE p bRE R2 plain R2 full Diff.
Panel A: Number of lags = 0
All All 0.991∗∗∗ 0.000 −0.001 0.834 27.8% 27.2% −0.5%
All Banks 0.966∗∗∗ 0.000 −0.029 0.657 37.5% 38.0% 0.4%
All Industrial 1.014∗∗∗ 0.000 −0.005 0.537 27.5% 26.9% −0.6%
All Insurance 0.416∗∗∗ 0.003 −0.045 0.127 17.9% 20.1% 2.2%
All Other fin. 1.038∗∗∗ 0.000 0.002 0.657 41.2% 45.4% 4.2%
Germany All 1.226∗∗∗ 0.000 0.002 0.773 31.0% 30.0% −1.0%
France All 0.891∗∗∗ 0.000 0.005 0.940 29.3% 28.5% −0.8%
Italy All 1.006∗∗∗ 0.000 −0.018∗ 0.063 39.8% 41.4% 1.6%
Spain All 0.599∗∗∗ 0.000 0.076 0.460 11.1% 11.6% 0.5%
Panel B: Number of lags = 1
All All 0.923∗∗∗ 0.000 0.005 0.404 27.8% 26.5% −1.3%
All Banks 0.972∗∗∗ 0.000 0.114 0.220 37.5% 49.1% 11.6%
All Industrial 0.946∗∗∗ 0.000 0.019 0.375 27.5% 25.8% −1.6%
All Insurance 0.358∗∗ 0.017 0.023 0.445 17.9% 19.2% 1.3%
All Other fin. 0.852∗∗∗ 0.000 0.003 0.506 41.2% 39.0% −2.2%
Germany All 1.150∗∗∗ 0.000 0.002 0.795 31.0% 29.2% −1.8%
France All 0.831∗∗∗ 0.000 −0.047 0.587 29.3% 27.4% −1.9%
Italy All 0.893∗∗∗ 0.000 0.049∗ 0.085 39.8% 38.7% −1.1%
Spain All 0.633∗∗∗ 0.000 −0.030 0.869 11.1% 14.4% 3.3%
Panel C: Number of lags = 2
All All 0.940∗∗∗ 0.000 0.001 0.867 27.8% 35.3% 7.5%
All Banks 0.861∗∗∗ 0.000 0.089 0.650 37.5% 54.5% 17.0%
All Industrial 0.968∗∗∗ 0.000 −0.019 0.684 27.5% 35.7% 8.2%
All Insurance 0.386∗∗ 0.035 −0.005 0.892 17.9% 23.3% 5.4%
All Other fin. 1.182∗∗∗ 0.000 0.000 0.949 41.2% 51.3% 10.1%
Germany All 1.089∗∗∗ 0.000 0.001 0.887 31.0% 46.9% 15.9%
France All 0.870∗∗∗ 0.000 0.199∗ 0.098 29.3% 38.3% 9.1%
Italy All 0.906∗∗∗ 0.000 0.045 0.249 39.8% 39.9% 0.1%
Spain All 0.792∗∗∗ 0.000 0.148 0.604 11.1% 22.9% 11.8%

Note: The table shows the slope coefficients and p-values that result from the panel data regression of the cost of equity, using the Environmental
Score as an information variable. Asterisks denote significance, where ∗ p < 0.1, ∗∗ p < 0.05, and ∗∗∗ p < 0.01. Each panel use a different number of
lags for the Environmental Score, ranging from 0 to 2. Columns labeled ‘R2 plain’ and ‘R2 full’ show the R2 statistics of the regressions ignoring or
including the ESG score as an information variable, respectively. The column labeled ‘Diff.’ shows the difference between these statistics.

611
J. Rojo-Suárez and A.B. Alonso-Conde Economic Analysis and Policy 77 (2023) 599–616

Table B.4
Regression results for MVi,t − BVi,t using the Social Score as information variable.
Country Sector ΦEP σ (ΦEP ) ΦS · 10−3 σ (ΦS )· 10−3 R2 plain R2 full Diff.
Panel A: Number of lags = 0
All All −1.144∗∗∗ 0.116 −0.002∗ 0.001 16.5% 17.4% 0.9%
All Banks −1.302∗∗∗ 0.067 0.018 0.013 83.2% 84.7% 1.5%
All Industrial −1.229∗∗∗ 0.310 −0.006 0.005 14.4% 16.7% 2.3%
All Insurance −2.073∗∗∗ 0.287 −0.128∗∗∗ 0.035 70.7% 78.0% 7.2%
All Other fin. −1.974∗∗∗ 0.272 0.003 0.006 51.4% 53.1% 1.8%
Germany All −0.943∗∗∗ 0.173 −0.008 0.005 24.4% 25.8% 1.3%
France All −1.450∗∗∗ 0.255 −0.004 0.006 15.9% 17.3% 1.3%
Italy All −1.669∗∗∗ 0.177 0.001 0.003 43.8% 43.9% 0.1%
Spain All 3.689∗∗∗ 0.856 −0.005 0.006 22.9% 24.1% 1.2%
Panel B: Number of lags = 1
All All −1.252∗∗∗ 0.144 −0.004 0.003 16.5% 19.9% 3.5%
All Banks −1.292∗∗∗ 0.081 0.021 0.013 83.2% 90.8% 7.6%
All Industrial −1.197∗∗∗ 0.312 −0.006 0.005 14.4% 18.1% 3.7%
All Insurance −2.056∗∗∗ 0.298 0.011 0.014 70.7% 79.9% 9.1%
All Other fin. 1.823∗∗∗ 0.322 0.005 0.009 51.4% 57.5% 6.1%
Germany All −1.030∗∗∗ 0.223 −0.007 0.005 24.4% 28.7% 4.3%
France All −1.609∗∗∗ 0.301 −0.011 0.012 15.9% 19.7% 3.8%
Italy All −1.474∗∗∗ 0.226 −0.001 0.002 43.8% 34.1% −9.7%
Spain All 3.813∗∗∗ 0.896 −0.006 0.006 22.9% 28.3% 5.4%
Panel C: Number of lags = 2
All All −1.310∗∗∗ 0.157 −0.005 0.004 16.5% 23.6% 7.1%
All Banks −1.336∗∗∗ 0.090 −0.006 0.006 83.2% 96.7% 13.5%
All Industrial −0.986∗ 0.390 −0.004 0.004 14.4% 18.3% 3.9%
All Insurance −1.979∗∗∗ 0.331 0.008 0.010 70.7% 88.2% 17.5%
All Other fin. −2.055∗∗∗ 0.384 0.006 0.005 51.4% 79.3% 28.0%
Germany All −1.144∗∗∗ 0.249 −0.006 0.005 24.4% 32.3% 7.8%
France All −1.638∗∗∗ 0.321 −0.014 0.022 15.9% 24.1% 8.2%
Italy All −1.494∗∗∗ 0.259 −0.003 0.004 43.8% 40.0% −3.8%
Spain All −0.933∗∗ 0.304 −0.007 0.013 22.9% 23.8% 0.9%

Note: The table shows the slope coefficients and standard errors that result from the panel data regression of the difference between the market
value and the book value, using the Social Score as an information variable. Asterisks denote significance, where ∗ p < 0.1, ∗∗ p < 0.05, and ∗∗∗ p
< 0.01. Each panel use a different number of lags for the Social Score, ranging from 0 to 2. Columns labeled ‘R2 plain’ and ‘R2 full’ show the
R2 statistics of the regressions ignoring or including the ESG score as an information variable, respectively. The column labeled ‘Diff.’ shows the
difference between these statistics.

Table B.5
Regression results for ROEi,t using the Social Score as information variable.
( )
Country Sector bROE
S p bROE
S R2
Panel A: Number of lags = 0
All All 0.010 0.579 0.0%
All Banks 0.044 0.213 1.7%
All Industrial 0.011 0.600 0.0%
All Insurance −0.002 0.564 2.7%
All Other fin. −0.009 0.558 1.1%
Germany All 0.047 0.181 0.4%
France All 0.049 0.470 0.0%
Italy All 0.053 0.326 0.5%
Spain All −0.003 0.938 0.0%

(continued on next page)

612
J. Rojo-Suárez and A.B. Alonso-Conde Economic Analysis and Policy 77 (2023) 599–616

Table B.5 (continued).

( )
Country Sector bROE
S p bROE
S R2
Panel B: Number of lags = 1
All All 0.004 0.860 0.0%
All Banks 0.012 0.713 0.0%
All Industrial 0.004 0.874 0.0%
All Insurance 0.000 0.955 5.6%
All Other fin. 0.018 0.494 2.6%
Germany All 0.005 0.887 0.2%
France All −0.137 0.146 0.8%
Italy All 0.042 0.386 0.2%
Spain All 0.000 0.992 0.0%
Panel C: Number of lags = 2
All All 0.199 0.131 0.5%
All Banks 0.128∗ 0.069 7.4%
All Industrial 0.236 0.153 0.5%
All Insurance −0.006 0.732 28.6%
All Other fin. 0.043 0.505 4.5%
Germany All −0.033 0.707 0.1%
France All −0.048 0.778 0.1%
Italy All 0.114∗ 0.056 8.7%
Spain All 1.674∗ 0.055 5.1%

Note: The table shows the slope coefficient and the p-value that result from the panel data regression
of ROE, using the Social Score as an information variable. Asterisks denote significance, where ∗ p < 0.1,
∗∗
p < 0.05, and ∗∗∗ p < 0.01. The table also provides the R2 statistic of the regressions. Each panel use
a different number of lags for the Social Score, ranging from 0 to 2.

Table B.6
Regression results for Ri,t using the Social Score as information variable.
( ) ( )
Country Sector bRRM p bRRM bRS p bRS R2 plain R2 full Diff.
Panel A: Number of lags = 0
All All 1.001∗∗∗ 0.000 0.010 0.414 27.8% 27.8% 0.0%
All Banks 0.972∗∗∗ 0.000 0.047 0.660 37.5% 37.4% −0.1%
All Industrial 1.031∗∗∗ 0.000 0.012 0.375 27.5% 27.5% 0.1%
All Insurance 0.418∗∗∗ 0.003 −0.043 0.109 17.9% 22.0% 4.1%
All Other fin. 0.972∗∗∗ 0.000 −0.025 0.653 41.2% 41.4% 0.2%
Germany All 1.239∗∗∗ 0.000 0.031 0.476 31.0% 31.1% 0.1%
France All 0.916∗∗∗ 0.000 −0.045 0.444 29.3% 29.4% 0.1%
Italy All 0.985∗∗∗ 0.000 −0.087 0.191 39.8% 40.3% 0.5%
Spain All 0.583∗∗∗ 0.000 0.011 0.402 11.1% 11.5% 0.3%
Panel B: Number of lags = 1
All All 0.933∗∗∗ 0.000 −0.009 0.523 27.8% 27.0% −0.8%
All Banks 0.980∗∗∗ 0.000 0.007 0.949 37.5% 47.7% 10.2%
All Industrial 0.963∗∗∗ 0.000 −0.010 0.496 27.5% 26.4% −1.1%
All Insurance 0.377∗∗∗ 0.007 0.020 0.433 17.9% 20.9% 3.0%
All Other fin. 0.831∗∗∗ 0.000 0.009 0.924 41.2% 36.4% −4.8%
Germany All 1.146∗∗∗ 0.000 0.026 0.632 31.0% 29.7% −1.3%
France All 0.844∗∗∗ 0.000 −0.075 0.389 29.3% 28.0% −1.3%
Italy All 0.891∗∗∗ 0.000 −0.007 0.921 39.8% 37.9% −1.9%
Spain All 0.628∗∗∗ 0.000 −0.011 0.432 11.1% 14.6% 3.5%
Panel C: Number of lags = 2
All All 0.940∗∗∗ 0.000 −0.018 0.765 27.8% 35.5% 7.7%
All Banks 0.962∗∗∗ 0.000 0.733∗∗∗ 0.001 37.5% 60.9% 23.4%
All Industrial 0.965∗∗∗ 0.000 −0.122∗ 0.075 27.5% 36.6% 9.1%
All Insurance 0.420∗∗ 0.010 −0.123 0.288 17.9% 29.1% 11.3%
All Other fin. 1.063∗∗∗ 0.000 0.533∗∗∗ 0.000 41.2% 63.3% 22.1%
Germany All 1.052∗∗∗ 0.000 −0.118 0.171 31.0% 47.4% 16.5%
France All 0.862∗∗∗ 0.000 −0.014 0.908 29.3% 37.7% 8.4%
Italy All 0.965∗∗∗ 0.000 0.385∗∗∗ 0.001 39.8% 50.2% 10.4%
Spain All 0.819∗∗∗ 0.000 −0.111 0.589 11.1% 22.9% 11.8%

Note: The table shows the slope coefficients and p-values that result from the panel data regression of the cost of equity, using the Social Score as
an information variable. Asterisks denote significance, where ∗ p < 0.1, ∗∗ p < 0.05, and ∗∗∗ p < 0.01. Each panel use a different number of lags for
the Social Score, ranging from 0 to 2. Columns labeled ‘R2 plain’ and ‘R2 full’ show the R2 statistics of the regressions ignoring or including the ESG
score as an information variable, respectively. The column labeled ‘Diff.’ shows the difference between these statistics.

613
J. Rojo-Suárez and A.B. Alonso-Conde Economic Analysis and Policy 77 (2023) 599–616

Table B.7
Regression results for MVi,t − BVi,t using the Governance Score as information variable.
Country Sector ΦEP σ (ΦEP ) ΦG · 10−3 σ (ΦG )· 10−3 R2 plain R2 full Diff.
Panel A: Number of lags = 0
All All −1.139∗∗∗ 0.117 −0.005 0.003 16.5% 17.5% 1.0%
All Banks −1.295∗∗∗ 0.064 0.011 0.008 83.2% 84.6% 1.4%
All Industrial −1.231∗∗∗ 0.310 −0.006∗ 0.003 14.4% 17.0% 2.6%
All Insurance −2.029∗∗∗ 0.250 0.016∗ 0.007 70.7% 74.9% 4.2%
All Other fin. −1.979∗∗∗ 0.272 −0.004 0.005 51.4% 53.9% 2.6%
Germany All −0.940∗∗∗ 0.174 −0.010∗ 0.004 24.4% 26.4% 1.9%
France All −1.441∗∗∗ 0.255 −0.007 0.005 15.9% 17.7% 1.7%
Italy All −1.685∗∗∗ 0.172 0.004∗ 0.002 43.8% 44.8% 1.0%
Spain All 3.804∗∗∗ 0.853 −0.003 0.003 22.9% 24.8% 2.0%
Panel B: Number of lags = 1
All All −1.245∗∗∗ 0.144 −0.005 0.003 16.5% 20.0% 3.5%
All Banks −1.304∗∗∗ 0.080 0.006 0.006 83.2% 90.6% 7.4%
All Industrial −1.217∗∗∗ 0.312 −0.006∗ 0.003 14.4% 18.2% 3.8%
All Insurance −2.022∗∗∗ 0.297 0.027 0.015 70.7% 82.3% 11.6%
All Other fin. 1.817∗∗∗ 0.315 −0.007 0.008 51.4% 58.1% 6.7%
Germany All −1.042∗∗∗ 0.223 −0.005 0.004 24.4% 28.5% 4.1%
France All −1.575∗∗∗ 0.304 −0.008 0.006 15.9% 19.9% 3.9%
Italy All −1.482∗∗∗ 0.223 −0.004 0.003 43.8% 34.7% −9.1%
Spain All 3.877∗∗∗ 0.923 −0.002 0.003 22.9% 28.1% 5.2%
Panel C: Number of lags = 2
All All −1.315∗∗∗ 0.155 −0.006∗ 0.003 16.5% 24.0% 7.5%
All Banks −1.305∗∗∗ 0.086 −0.002 0.003 83.2% 96.5% 13.3%
All Industrial −1.007∗ 0.389 −0.006∗ 0.003 14.4% 18.8% 4.4%
All Insurance −1.959∗∗∗ 0.304 0.004 0.004 70.7% 89.0% 18.3%
All Other fin. −2.084∗∗∗ 0.328 −0.010∗ 0.004 51.4% 82.2% 30.8%
Germany All −1.125∗∗∗ 0.244 −0.007 0.004 24.4% 33.2% 8.8%
France All −1.660∗∗∗ 0.315 −0.010 0.006 15.9% 25.0% 9.0%
Italy All −1.461∗∗∗ 0.259 −0.008 0.012 43.8% 40.5% −3.3%
Spain All −0.960∗∗ 0.309 −0.002 0.005 22.9% 23.5% 0.6%

Note: The table shows the slope coefficients and standard errors that result from the panel data regression of the difference between the market
value and the book value, using the Governance Score as an information variable. Asterisks denote significance, where ∗ p < 0.1, ∗∗ p < 0.05, and
∗∗∗
p < 0.01. Each panel use a different number of lags for the Governance Score, ranging from 0 to 2. Columns labeled ‘R2 plain’ and ‘R2 full’ show
the R2 statistics of the regressions ignoring or including the ESG score as an information variable, respectively. The column labeled ‘Diff.’ shows the
difference between these statistics.

Table B.8
Regression results for ROEi,t using the Governance Score as information variable.
( )
Country Sector bROE
G p bROE
G R2
Panel A: Number of lags = 0
All All 0.027 0.279 0.1%
All Banks 0.074∗∗ 0.032 5.3%
All Industrial 0.030 0.308 0.1%
All Insurance 0.005 0.515 2.6%
All Other fin. 0.006 0.744 0.8%
Germany All 0.032 0.144 0.4%
France All 0.057 0.116 0.4%
Italy All 0.062 0.108 1.1%
Spain All −0.050 0.756 0.1%

(continued on next page)

614
J. Rojo-Suárez and A.B. Alonso-Conde Economic Analysis and Policy 77 (2023) 599–616

Table B.8 (continued).

( )
Country Sector bROE
G p bROE
G R2
Panel B: Number of lags = 1
All All 0.001 0.974 0.0%
All Banks −0.028 0.321 0.8%
All Industrial 0.003 0.942 0.0%
All Insurance 0.013∗∗ 0.049 15.2%
All Other fin. −0.008 0.782 1.9%
Germany All 0.007 0.762 0.2%
France All −0.026 0.528 0.2%
Italy All −0.006 0.866 0.0%
Spain All 0.057 0.823 0.0%
Panel C: Number of lags = 2
All All −0.040 0.448 0.1%
All Banks 0.026 0.633 0.4%
All Industrial −0.043 0.479 0.1%
All Insurance −0.003 0.693 25.6%
All Other fin. −0.116∗ 0.092 12.4%
Germany All 0.001 0.979 0.0%
France All −0.076 0.280 0.7%
Italy All −0.026 0.638 7.7%
Spain All −0.243 0.532 0.6%

Note: The table shows the slope coefficient and the p-value that result from the panel data regression
of ROE, using the Governance Score as an information variable. Asterisks denote significance, where ∗ p
< 0.1, ∗∗ p < 0.05, and ∗∗∗ p < 0.01. The table also provides the R2 statistic of the regressions. Each panel
use a different number of lags for the Governance Score, ranging from 0 to 2.

Table B.9
Regression results for Ri,t using the Governance score as information variable.
( ) ( )
Country Sector bRRM p bRRM bRG p bRG R2 plain R2 full Diff.
Panel A: Number of lags = 0
All All 0.993∗∗∗ 0.000 0.075∗∗∗ 0.000 27.8% 28.6% 0.9%
All Banks 0.970∗∗∗ 0.000 −0.062 0.564 37.5% 37.9% 0.3%
All Industrial 1.023∗∗∗ 0.000 0.087∗∗∗ 0.000 27.5% 28.6% 1.1%
All Insurance 0.460∗∗∗ 0.001 −0.007 0.900 17.9% 17.9% 0.0%
All Other fin. 0.983∗∗∗ 0.000 −0.019 0.743 41.2% 41.3% 0.1%
Germany All 1.216∗∗∗ 0.000 0.107∗∗∗ 0.000 31.0% 32.9% 2.0%
France All 0.913∗∗∗ 0.000 0.051 0.130 29.3% 29.6% 0.4%
Italy All 0.989∗∗∗ 0.000 −0.019 0.701 39.8% 39.9% 0.1%
Spain All 0.577∗∗∗ 0.000 0.030 0.575 11.1% 11.3% 0.2%
Panel B: Number of lags = 1
All All 0.933∗∗∗ 0.000 0.002 0.935 27.8% 27.0% −0.8%
All Banks 0.981∗∗∗ 0.000 −0.056 0.624 37.5% 47.8% 10.3%
All Industrial 0.962∗∗∗ 0.000 0.002 0.949 27.5% 26.3% −1.1%
All Insurance 0.361∗∗ 0.011 −0.050 0.351 17.9% 21.5% 3.6%
All Other fin. 0.816∗∗∗ 0.000 −0.028 0.753 41.2% 36.5% −4.7%
Germany All 1.154∗∗∗ 0.000 0.006 0.865 31.0% 29.6% −1.4%
France All 0.848∗∗∗ 0.000 0.016 0.673 29.3% 27.8% −1.4%
Italy All 0.887∗∗∗ 0.000 −0.029 0.658 39.8% 38.0% −1.8%
Spain All 0.620∗∗∗ 0.000 −0.021 0.763 11.1% 14.2% 3.1%
Panel C: Number of lags = 2
All All 0.937∗∗∗ 0.000 0.012 0.607 27.8% 35.5% 7.8%
All Banks 0.852∗∗∗ 0.000 0.331∗∗ 0.039 37.5% 62.7% 25.2%
All Industrial 0.970∗∗∗ 0.000 0.014 0.585 27.5% 36.1% 8.6%
All Insurance 0.478∗∗∗ 0.004 −0.077 0.181 17.9% 31.5% 13.6%
All Other fin. 1.014∗∗∗ 0.000 −0.074 0.687 41.2% 41.4% 0.2%
Germany All 1.072∗∗∗ 0.000 0.032 0.307 31.0% 47.6% 16.6%
France All 0.841∗∗∗ 0.000 0.051 0.319 29.3% 38.1% 8.8%
Italy All 0.901∗∗∗ 0.000 0.091 0.414 39.8% 39.7% −0.1%
Spain All 0.828∗∗∗ 0.000 −0.142 0.100 11.1% 25.5% 14.4%

Note: The table shows the slope coefficients and p-values that result from the panel data regression of the cost of equity, using the Governance
Score as an information variable. Asterisks denote significance, where ∗ p < 0.1, ∗∗ p < 0.05, and ∗∗∗ p < 0.01. Each panel use a different number
of lags for the Governance Score, ranging from 0 to 2. Columns labeled ‘R2 plain’ and ‘R2 full’ show the R2 statistics of the regressions ignoring or
including the ESG score as an information variable, respectively. The column labeled ‘Diff.’ shows the difference between these statistics.

615
J. Rojo-Suárez and A.B. Alonso-Conde Economic Analysis and Policy 77 (2023) 599–616

References

Baker, Malcolm, Bergstresser, Daniel, Serafeim, George, Wurgler, Jeffrey, 2018. Financing the Response to Climate Change: The Pricing and Ownership
of U.S. Green Bonds. National Bureau of Economic Research Working Paper Series No. 25194, http://dx.doi.org/10.3386/w25194.
Baker, Malcolm, Wurgler, Jeffrey, 2000. The equity share in new issues and aggregate stock returns. J. Finance 55 (5), 2219–2257. http://dx.doi.org/
10.1111/0022-1082.00285.
Bofinger, Yannik, Heyden, Kim J., Rock, Björn, 2022. Corporate social responsibility and market efficiency: Evidence from ESG and misvaluation
measures. J. Bank. Financ. 134, 106322. http://dx.doi.org/10.1016/j.jbankfin.2021.106322.
Campbell, John Y., 1987. Stock returns and the term structure. J. Financ. Econ. 18 (2), 373–399. http://dx.doi.org/10.1016/0304-405X(87)90045-6.
Campbell, John Y., Shiller, Robert J., 1988. The dividend-price ratio and expectations of future dividends and discount factors. Rev. Financ. Stud. 1
(3), 195–228. http://dx.doi.org/10.1093/rfs/1.3.195.
Castro, Paula, Gutiérrez-López, Cristina, Tascón, María T., Castaño, Francisco J., 2021. The impact of environmental performance on stock prices in
the green and innovative context. J. Clean. Prod. 320, 128868. http://dx.doi.org/10.1016/j.jclepro.2021.128868.
Cochrane, John H., 2011. Presidential address: Discount rates. J. Finance 66 (4), 1047–1108. http://dx.doi.org/10.1111/j.1540-6261.2011.01671.x.
Fama, Eugene F., French, Kenneth R., 1988. Dividend yields and expected stock returns. J. Financ. Econ. 22 (1), 3–25. http://dx.doi.org/10.1016/0304-
405X(88)90020-7.
Grassmann, Michael, 2021. The relationship between corporate social responsibility expenditures and firm value: The moderating role of integrated
reporting. J. Clean. Prod. 285, 124840. http://dx.doi.org/10.1016/j.jclepro.2020.124840.
Hassel, Lars, Nilsson, Henrik, Nyquist, Siv, 2005. The value relevance of environmental performance. Eur. Account. Rev. 14 (1), 41–61. http:
//dx.doi.org/10.1080/0963818042000279722.
Hong, Harrison, Kacperczyk, Marcin, 2009. The price of sin: The effects of social norms on markets. J. Financ. Econ. 93 (1), 15–36. http:
//dx.doi.org/10.1016/j.jfineco.2008.09.001.
Ionescu, G.H., Firoiu, D., Pirvu, R., Vilag, R.D., 2019. The impact of ESG factors on market value of companies from travel and tourism industry.
Technol. Econ. Dev. Econ. 25 (5), 820–849. http://dx.doi.org/10.3846/tede.2019.10294.
Lamont, Owen A., Stein, Jeremy C., 2004. Aggregate short interest and market valuations. Amer. Econ. Rev. 94 (2), 29–32. http://dx.doi.org/10.1257/
0002828041301759.
Landau, Alexander, Rochell, Janina, Klein, Christian, Zwergel, Bernhard, 2020. Integrated reporting of environmental, social, and governance and
financial data: Does the market value integrated reports? Bus. Strategy Environ. 29 (4), 1750–1763. http://dx.doi.org/10.1002/bse.2467.
Lettau, Martin, Ludvigson, Sydney, 2001. Consumption, aggregate wealth, and expected stock returns. J. Finance 56 (3), 815–849. http://dx.doi.org/
10.1111/0022-1082.00347.
Lintner, John, 1965a. Security prices, risk, and maximal gains from diversification. J. Finance 20 (4), 587–615. http://dx.doi.org/10.1111/j.1540-
6261.1965.tb02930.x.
Lintner, John, 1965b. The valuation of risk assets and the selection of risky investments in stock portfolios and capital budgets. Rev. Econ. Stat. 47
(1), 13–37. http://dx.doi.org/10.2307/1924119.
Luo, H. Arthur, Balvers, Ronald J., 2017. Social screens and systematic investor boycott risk. J. Financ. Quant. Anal. 52 (1), 365–399. http:
//dx.doi.org/10.1017/S0022109016000910.
Miralles-Quirós, María Mar, Miralles-Quirós, José Luis, Redondo-Hernández, Jesús, 2019. The impact of environmental, social, and governance
performance on stock prices: Evidence from the banking industry. Corp. Soc. Respons. Environ. Manage. 26 (6), 1446–1456. http://dx.doi.org/10.
1002/csr.1759.
Novy-Marx, Robert, 2014. Predicting anomaly performance with politics, the weather, global waming, sunspots, and the stars. J. Financ. Econ. 112
(2), 137–146. http://dx.doi.org/10.1016/j.jfineco.2014.02.002.
Ohlson, James A., 1995. Earnings, book values, and dividends in equity valuation. Contemp. Account. Res. 11 (2), 661–687. http://dx.doi.org/10.1111/
j.1911-3846.1995.tb00461.x.
Pástor, Ľuboš, Stambaugh, Robert F., Taylor, Lucian A., 2021. Sustainable investing in equilibrium. J. Financ. Econ. 142 (2), 550–571. http://dx.doi.org/
10.1016/j.jfineco.2020.12.011.
Pástor, Ľuboš, Stambaugh, Robert F., Taylor, Lucian A., 2022. Dissecting green returns. J. Financ. Econ. 146 (2), 403–424. http://dx.doi.org/10.1016/j.
jfineco.2022.07.007.
Pedersen, Lasse Heje, Fitzgibbons, Shaun, Pomorski, Lukasz, 2021. Responsible investing: The ESG-efficient frontier. J. Financ. Econ. 142 (2), 572–597.
http://dx.doi.org/10.1016/j.jfineco.2020.11.001.
Rodríguez-García, Martha del Pilar, Galindo-Manrique, Alicia Fernanda, Cortez-Alejandro, Klender Aimer, Méndez-Sáenz, Alma Berenice, 2022. Eco-
efficiency and financial performance in latin American countries: An environmental intensity approach. Res. Int. Bus. Finance 59, 101547.
http://dx.doi.org/10.1016/j.ribaf.2021.101547.
Rojo-Suárez, Javier, Alonso-Conde, Ana Belén, Ferrero-Pozo, Ricardo, 2022. Liquidity, time-varying betas and anomalies: Is the high trading activity
enhancing the validity of the CAPM in the UK equity market? Int. J. Finance Econ. 27 (1), 45–60. http://dx.doi.org/10.1002/ijfe.2136.
Sharpe, William F., 1964. Capital asset prices: A theory of market equilibrium under conditions of risk. J. Finance 19 (3), 425–442. http://dx.doi.org/
10.1111/j.1540-6261.1964.tb02865.x.
Tampakoudis, Ioannis, Noulas, Athanasios, Kiosses, Nikolaos, Drogalas, George, 2021. The effect of ESG on value creation from mergers and acquisitions.
What changed during the COVID-19 pandemic? Corp. Govern.: Int. J. Bus. Soc. 21 (6), 1117–1141. http://dx.doi.org/10.1108/CG-10-2020-0448.
Zerbib, Olivier David, 2020. A sustainable capital asset pricing model (S-CAPM): Evidence from green investing and sin stock exclusion. In: Finance
Meeting EUROFIDAI - ESSEC.
Zhang, Dongyang, Lucey, Brian M., 2022. Sustainable behaviors and firm performance: The role of financial constraints’ alleviation. Econ. Anal. Policy
74, 220–233. http://dx.doi.org/10.1016/j.eap.2022.02.003.

616