Environmental Science and Pollution Research (2023) 30:85154–85169

https://doi.org/10.1007/s11356-023-28376-1

RESEARCH ARTICLE

ESG investing and the financial performance: a panel data analysis

of developed REIT markets
Isil Erol1   · Umut Unal2   · Yener Coskun3 

Received: 7 February 2023 / Accepted: 18 June 2023 / Published online: 28 June 2023
© The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2023

Abstract
This study investigates the empirical link between the social and financial performance of the Real Estate Investment Trusts
(REITs) by utilizing the PVAR-Granger causality model and a fixed-effects panel data model with a rich dataset comprising
234 ESG-rated REITs across five developed economies from 2003 to 2019. The results suggest that investors pay atten-
tion to individual E/S/G metrics and price each component of ESG investing differently, with E-investing and S-investing
practices being the significant financial performance factors of REITs. This study is the first attempt to test the social impact
and risk mitigation hypotheses of the stakeholder theory of the corporation and the neoclassic trade-off argument to explore
the association between corporate social responsibility and the market valuation of REITs. The full sample results strongly
support the trade-off hypothesis, indicating that REITs’ environmental policies involve high financial costs that may drain
off capital and lead to decreasing market returns. On the contrary, investors have attached a higher value to S-investing
performance, especially in the post-GFC period from 2011 to 2019. A positive premium for S-investing supports the stake-
holder theory as the social impact could be monetarized into a higher return and a lower systematic risk and give rise to a
competitive advantage.

Keywords  Corporate social performance · ESG investing · REITs · PVAR Granger causality test · Stakeholder theory of
corporation · Trade-off hypothesis

JEL classification  G11 · G15 · G32 · M14

Introduction

The demand for investment products with substantial

environmental, social, and governance (ESG) credentials
Responsible Editor: Arshian Sharif boomed in 2020 as the COVID-19 pandemic put the need
for a sustainable world in the spotlight. Investors are increas-
* Umut Unal ingly conscious of total wealth maximization for all stake-
umut.unal@rilsa.cz
holders, not only shareholder value. Real Estate Investment
Isil Erol Trusts (REITs) are often overlooked to gain exposure to
isil.erol@ozyegin.edu.tr
socially responsible investments, in part because numer-
Yener Coskun ous open-ended funds specifically publicize themselves as
ycoskun@spk.gov.tr
socially responsible. Nevertheless, ESG investing in the
1
Department of International Finance, Faculty of Business, REIT industry has notably increased since the mid-2000s—
Ozyegin University, Nişantepe, Orman Sk. No:13, i.e., the number of ESG-rated REITs rose from 43 in 2005 to
34794 Istanbul, Turkey 217 in 2018 in the developed countries, including the USA,
2
Macroeconomic Analyses Department, Research Institute the UK, Australia, Canada, and Japan. At the same time, the
for Labour and Social Affairs (RILSA), Dělnická 213/12, average environmental score of ESG-rated REITs in these
170 00 Prague 7, Czech Republic
countries has noticeably increased by 142.7%, from 13.1 to
3
Department of Business Administration, TED University, 31.8, whereas the average social and governance scores have
Ziya Gokalp caddesi, 48/A, 06420 Ankara, Turkey

13
Vol:.(1234567890)
Environmental Science and Pollution Research (2023) 30:85154–85169 85155

increased by 21.8% (39 to 47.5) and 12.7% (44.1 to 49.7), from the social performance to the financial performance.
respectively (Thomson Reuters DataStream database). Specifically, the results suggest a significant negative asso-
It is presumably because, over the past decade or so, ciation between E-investing and the financial performance
investors have been increasingly driven by ESG factors, of REITs, providing evidence for the trade-off hypothesis—
which, in turn, influence future returns. Hence, in line with i.e., the environmental policies and activities involve high
the expectations of the stakeholders, corporations are likely financial costs that may drain off capital and other company
to carry out more and better socially responsible investments resources and lead to declining market returns. The results
that could improve the ESG metrics. A considerable body further indicate a strong positive (negative) relationship
of research has focused on the impact of corporate social between S-investing and the REIT stock return (systematic
performance (CSP) on the corporate financial performance risk), which supports the social impact hypothesis of the
(CFP) of non-REIT listed companies and documented con- stakeholder theory of the corporation.
troversial results (Salzmann 2013). Existing meta-studies, in Conflicting empirical findings on the CSP-CFP nexus inher-
contrast, have suggested unambiguous evidence for a rather ently imply that the relationship between REITs’ financial per-
positive association between CSP and CFP. Similarly, the formance and their success/failure in ESG commitments may
relevant literature on REITs has mainly found a positive largely depend on sample size, single/cross-country setting,
association between CSP and CFP although some studies and social performance metrics used. Our paper contributes to
reported mixed empirical findings.1 These puzzling results the literature in several ways. First, unlike the previous REIT
are possibly due to different databases, period, sample sizes, research that has utilized a single-country setting (e.g., Newell
model specifications, and social performance criteria used and Lin Lee (2012); Brounen and Marcato (2018); and Eichholtz
in these studies. Despite the increase in REITs’ ESG com- et al. (2012)), we employ a cross-country panel data analysis
mitments and a growing body of research on the impact of to explore the nexus. Our sample period goes back to 2003,
socially responsible investment on their performance, the which is relatively longer than prior studies (Morri et al., 2021
association, if any, between social and financial performance and Fuerst 2015), which also employed cross-country data but
has not been fully established for REITs. used only a 4- or 5-year time span. None of these papers has
This study uses the Thomson Reuters Worldscope and explored the nexus in both aggregated (ESG total score) and
DataStream databases to construct a rich dataset compris- disaggregated (individual E/S/G scores) frameworks to under-
ing 234 ESG-rated REITs in the USA, the UK, Australia, stand how well REITs have implemented their ESG investments
Canada, and Japan, which together accounted for 82% of in relation to their stock market performance. Moreover, in order
$1899.5 million of the total market cap of global REITs to mitigate potential sample selection bias, a geographically
in 2019.2 The study employs the PVAR-Granger causality diverse sample of REITs spanning 17 years was utilized in a
model and a fixed-effects panel data model for cross-country cross-country analysis. Such an approach aims to examine the
analysis of the sign of the relationship between ESG invest- causal relationship and determine the direction of the association
ing and the market-based financial performance of REITs, between Corporate Social Performance (CSP) and Corporate
respectively, between 2003 and 2019. We find strong evi- Financial Performance (CFP) within the REIT industry.
dence that REIT investors pay attention to individual E/S/G Second, we manually construct a new, rich dataset that
metrics and price each component of ESG investing differ- includes market-based financial performance measures (excess
ently, with environmental (E-investing) and social (S-invest- return, the Sharpe ratio, and the beta). Our paper is one of the
ing) practices being the significant CSP factors influencing initial attempts to employ systematic firm risk as a depend-
the financial performance of REITs. The CSP-CFP relation- ent variable to explore the association between REITs’ ESG
ship for our sample is best explained by the stakeholder the- investing and stock performance. Third, our study, for the first
ory of corporation and the trade-off hypothesis as the PVAR time, tests the social impact hypothesis and risk mitigation view
Granger causality analysis suggests a direction of causality of the stakeholder theory of corporation and the neoclassic
trade-off argument to explore the impact of socially respon-
sible investing on the market valuation of REITs. Finally, our
1
study provides the first empirical evidence of the causal positive
  Eichholtz et al. (2012); Cajias et al. (2014); Fuerst (2015), Ooi and
relationship between S-investing and financial performance of
Dung (2019), and Morri et al. (2021) reported a positive association,
whereas Mariani et  al. (2018); Coën et  al. (2018); and Westermann REIT market, indicating that S-investing is a crucial component
et  al. 2022) provided evidence for a negative CSP-CFP relationship of ESG investing that could be monetarised into a higher excess
for the REITs. (and risk-adjusted) return and a lower systematic risk. From a
2
 We selected five countries because according to EPRA Global business ethics perspective, one can argue that socially ethical
Real Estate Total Markets Table, the number of REITs reporting
behaviors of REITs may generate a positive premium for their
ESG scores publicly is very limited with a short history in France
(7), Spain (3), Belgium (8), Singapore (7), the Netherlands (3), Hong stocks and enable them to achieve competitive advantage due to
Kong (2), and Germany (2). their productive relationships with their stakeholders.

13

85156 Environmental Science and Pollution Research (2023) 30:85154–85169

The following section provides a detailed literature Australia, Canada, and Japan (PVAR Granger causality
review that consists of theoretical framework with hypoth- analysis will be discussed later in the “Unit root tests and
eses development and empirical studies of CSP-CFP nexus PVAR Granger causality test” section).
for REIT market with a conceptual discussion on the concept The second issue is the scarcity of simultaneous analysis
of CSP in the public real estate sector. The “Data descrip- of aggregated and disaggregated ESG scores in relation to
tion” section describes data and section 4 covers empirical CFP: Does each component of the E/S/G measure provides
specification including unit root tests, PVAR Granger cau- different insights into the corporate’s financial performance
sality test, and panel data methods. The “Results and discus- beyond those that the aggregated ESG score demonstrates?
sion” section presents a discussion of the findings. Finally, Bouslah et al. (2013) and Brounen and Marcato (2018) have
the “Conclusion” section concludes the paper. documented that the individual E/S/G components consoli-
dated in the overall ESG score might have different impacts
on corporate risk and return. This is partly because hetero-
Literature review geneity among corporate stakeholders might create a mis-
match between the ESG components—e.g., employees and
Theoretical literature and hypothesis development Greenpeace put different emphasis on the issues of labor
conditions (S-score) and environmental pollution (E-score).
There is no emerging agreement on the most appropriate As we expect, each individual E/S/G measure has a different
classification of theories in corporate social responsibility effect on the financial performance compared to the overall
(CSR) research (Frynas and Yamahaki 2016). The theoreti- ESG rating. Our second hypothesis is the following:
cal and empirical research surrounding the nexus remains
REIT investors price each component of ESG invest-
inconclusive but highlights the non-static nature of this rela-
ing differently; therefore, E-investing, S-investing, and
tionship (Qureshi et al. 2021). This paper attempts to investi-
G-investing have different impacts on the financial per-
gate the CSP-CFP nexus for REITs by testing the validity of
formance of REITs.
key theories such as stakeholder theory and trade-off theory.
The ongoing debate on the CSP-CFP relationship The third issue is the sign of the relationship: Are social
involves three important empirical issues. The first issue is and financial performances positively or negatively associ-
the direction of causation: Does social performance affect ated, or not associated at all? Theoretical and/or empirical
the financial performance of corporations or the opposite, research has remained inconclusive regarding the sign of the
does financial performance affect social performance, or is CSP-CFP relationship (Wang et al. 2016). Thus, we do not
there a bilateral relationship between the two? Corporate have a priori expectation on the sign of association between
social responsibility theories have conflicting views on the E/S/G scores and REIT’s financial performance and follow
CSP-CFP relationship. For instance, according to the stake- the main theoretical arguments and empirical outcomes
holder theory and trade-off theory, there is a unidirectional below.
causality running from CSP to CFP, whereas available funds The social impact hypothesis and risk mitigation view
and managerial opportunism hypotheses imply a causality of the stakeholder theory of corporation predict that social
from CFP to CSP (Preston and O’bannon 1997). In the REIT and financial performance tend to be positively associ-
literature, there is a lack of empirical work on the causal ated over the long term (Freeman 1984). The theory states
relationship between social and financial performance. To that stakeholders have different interests in a corporation
our knowledge, Cajias et al. (2014) are the only study inves- and have different impacts upon it, and the corporation is
tigating the Granger causality in the CSP-CFP relationship responsible for meeting their interests. A firm that attempts
for the US real estate companies, which found no evidence to lower its implicit costs by socially irresponsible actions
for causality. Because of the conflicting views in theory and will, as a result, incur higher explicit costs, giving rise to a
lack of empirical evidence, we have no a priori expectation competitive disadvantage. On the contrary, an open-minded
for the direction of causality. Thus, our hypothesis for REIT employee relations policy may cost less. Indeed, it can result
markets is the following: in “substantial gains in morale and productivity, yielding a
competitive advantage compared to less responsible firms”
The direction of causality between corporate social
(Waddock and Graves 1997, p. 306). Moreover, the failure to
performance (ESG score) and financial performance
meet the expectations of various non-shareowner constituen-
is unclear.
cies will generate market fears, which, in turn, will increase
By employing the panel VAR-Granger causality model, a company’s risk premium and result in higher costs. The
our study finds that the direction of causality is from the risk mitigation argument based on the stakeholder theory
social performance (ESG score) to the financial perfor- predicts that CSP is negatively related to firm risk—i.e.,
mance of REITs in the developed markets of the USA, UK, higher social performance can generate moral capital or

13
Environmental Science and Pollution Research (2023) 30:85154–85169 85157

goodwill among stakeholders, which provides insurance- the equity REITs, from health care to lodging/resorts or
like protection that reduces a firm’s risk exposure (Godfrey industrial/office buildings. The potential benefits of ESG rat-
et al. 2009). Moral capital creates relational wealth in dif- ings are far beyond firm-level performance or public image
ferent forms among different stakeholder groups, including considerations and offer a wide range of positive externali-
affective commitment among employees, legitimacy among ties concerning environmental and social responsibility and
communities and regulators, trust among partners, cred- corporate governance quality.
ibility and enhanced brand among customers, and higher REITs can contribute to decarbonization by improving
attractiveness for investors (Godfrey 2005). This relational their operational efficiency for energy and water use and
wealth reduces uncertainty about a firm’s future cash flows developing environmental management systems. Examples
and, therefore, reduces its risk. of environmental policies include seeking green building
Hence, our third hypothesis is the follwing: certifications for their properties, adopting biodiversity, land
conservation, and eco-friendly building design techniques,
Higher corporate social performance (E-score/S-
issuing green bonds to fund sustainability projects, reducing
score/G-score) leads to better financial performance
emissions at buildings, and encouraging sustainable com-
for REITs—i.e., higher return and lower risk, all other
muting (IEA 2019). REITs can also improve their social
things being equal.
performance by supporting and contributing to community
The trade-off hypothesis, in contrast, asserts that socially organizations while ensuring their workforces are inclusive
responsive activities involve financial costs, which may and diverse, and providing a safe working environment
steal capital and other resources from the firm and result (NAREIT 2019). Finally, REITs can boost their ESG perfor-
in declining stock prices relative to the market average, and mance in terms of good governance policies and practices.
may put corporations at a relative disadvantage compared to ESG metrics developed by global data providers such as
less socially active firms. This hypothesis reflects the classic MSCI, S&P, and Morningstar generally suggest that better
statement of Friedman (1970) and other neoclassical econo- governance could be achieved by establishing a high-qual-
mists’ arguments that there are few readily measurable eco- ity reporting-disclosure framework, minimizing potential
nomic benefits to socially responsible behavior while numer- conflicts of interest, avoiding fraud and bribery, supporting
ous costs exist (Waddock and Graves 1997). According to diversity and independence in the board, and developing fair
Friedman (1970, reprinted in 2007, p.178): “there is one compensation policy for executives. While there is no single
and only one social responsibility of business—to use its best corporate governance structure, ESG-minded investors
resources and engage in activities designed to increase its prefer democratic, transparent, and equitable fund manage-
profits so long as it stays within the rules of the game, …” ment strategies, focusing on long-term growth.
Managerial attention to interests other than those of inves-
tors is a breach of trust that inevitably reduces the welfare Empirical studies of CSP‑CFP nexus in REITs
of shareowners (Preston and O’bannon 1997). The trade-off
hypothesis may have a relation with the agency theory and Albeit a large body of research has documented a positive
overinvestment hypothesis. According to this hypothesis, relationship between different sustainability measures and
based on an over-incentivized behaviors of managers, over- corporate financial performance, knowledge of the financial
investment in ESG may eventually reduce shareholder value effects of corporate social investing through ESG criteria
(Barnea and Rubin 2010; Chacon et al. 2022). Hence, an remains fragmented (Friede et al. 2015) and very limited
alternative hypothesis to be tested against the third hypothe- for the REIT industry. Previous research on REITs has pri-
sis is REITs’ higher CSP ((E-score/S-score/G-score) leads to marily focused on corporate governance (e.g., Hartzell et al.
worse financial performance—i.e., lower return and higher (2006), Bianco et al. (2007); Bauer et al. (2010); Camp-
risk, all other things being equal. bell et al. (2011)) and reported its weak relationship with
the CFP. This is mainly due to the strongly regulated busi-
Empirical literature ness environment as stated by Ghosh and Petrova (2021)
and Bauer et al. (2010). The literature on the REIT’s risk-
CSP in REITs: conceptual framework return characteristics regarding socially responsible invest-
ments is scarce. Several researchers suggest that higher CSP
The real estate industry has a special responsibility for may lower volatility and a market risk premium (Eichholtz
decarbonization because of its 40% share of global carbon et al. 2013; Westermann et al. 2022) and lead to additional
dioxide emissions (United Nations Environment Program diversification benefits (Newell et al. 2011). Some empirical
2019). Intangible benefits of decarbonizing the built envi- studies, on the contrary, suggest a lack of abnormal return
ronment through REITs have a significant benefit spectrum related to portfolio greenness (Eichholtz et al. 2012; Ooi and
in the physical market depending on the property focus of Dung 2019) and indicate no risk-adjusted return from CSR

13

85158 Environmental Science and Pollution Research (2023) 30:85154–85169

practices (Westermann et al. 2022). On the other hand, Fan performed the Granger causality test and employed panel
et al. (2022) argue that costly ESG investments may dete- regressions, we could argue that lack of causality analysis
riorate the firm’s fundamentals and increase the company’s is an ongoing modeling challenge in this line of research for
risk. REITs. The scarcity of simultaneous analysis of aggregated
The studies investigating the CSP-CFP empirical relation- and disaggregated ESG scores in relation to CFP is another
ship employing ESG metrics are of the primary concern research gap in the literature. Brounen and Marcato (2018)
of our study; therefore, we summarize this line of research highlighted this knowledge gap and incorporated total ESG
regarding the data sources, single/cross-country samples, scores and disaggregated E/S/G scores using different sub-
time periods, variable selection, and modeling strategies as periods from 2002 to 2016. The present study differs from
well as the empirical evidence provided. Previous studies Brounen and Marcato (2018) and Fan et al. (2022) by using
have employed various data sources and country samples. a cross-country panel data analysis and having the system-
For instance, Newell and Lin Lee (2012) used the CSR rat- atic firm risk (or beta) as the dependent variable in an ESG
ings of Corporate Monitor for E, S, and G rating factors score-based empirical specification in this sample. This
for 16 Australian REITs. Examining the USA REIT mar- study also uses a more extended period for disaggregated
ket, Cajias et al. (2014) employed the MSCI ESG (formerly level analysis to minimize possible modeling problems aris-
KLD) database for 341 publicly traded real estate compa- ing from short time spans.
nies, whereas Brounen and Marcato (2018) utilized Global In summary, there are conflicting theoretical and empiri-
Real Estate Sustainability Benchmark (GRESB), Thomson cal views in the literature focusing on the CSP-CFP nexus.
Reuters, and KLD datasets for the aggregate ESG scores. Empirically driven nature of the investigations adopting
Among the cross-country case studies, Morri et al. (2021) various databases, sample sizes, time periods, variable selec-
and Fuerst (2015) used the GRESB database to study 50 tion, and modeling strategies might contribute to different
European REITs and approximately 400 international REITs results across different studies on this topic. Although previ-
in North America, Asia, and Europe. Chacon et al. (2022) ous research has mainly suggested a decline in risk and an
also employ the GRESB database for a global sample of 15 increase in a firm’s financial performance with the ESG or
countries in a 3-year sample of 2019 and 2021. GRESB metrics, systematic comparative work on the CSP-
Existing literature has predominantly focused on single- CFP empirical nexus employing ESG ratings in a cross-
country analysis, including Australia (Newell and Lin Lee country REITs market setting is still at an embryonic stage.
2012; Westermann et al. 2022) and the USA (Cajias et al. Our understanding of the association between individual
2014; Brounen and Marcato 2018; Aroul et al. 2022). Only E/S/G metrics and CFP, namely, insights from disaggregated
three studies so far (Fuerst 2015; Morri et al., 2021, Chacon analysis of ESG factors, is less apparent.
et al. 2022) have investigated the CSP-CFP relationship in a
cross-country setting. Furthermore, these studies have uti-
lized short-time periods, ranging from 3 to 9 years, in their Data description
analysis, except for Brounen and Marcato (2018), who used
an 18-year sample period for the US REITs. Regarding the This study uses a comprehensive dataset provided by the
variable selection, prior research has investigated the nexus Thomson Reuters Asset4 database, where ESG scores are
by using the operating performance indicators such as ROA mainly classified under the following quartiles.3 ESG score
and ROE (Morri et al., 2021), financial performance indi- within the first quartile (0 to 25) indicates poor relative per-
cators of the total return and excess or risk-adjusted return formance and insufficient degree of transparency in report-
(Newell and Lin Lee 2012; Brounen and Marcato 2018; ing data publicly; ESG scores up to the median (> 25 to 50)
Westermann et al. 2022) or both indicators (Fuerst 2015) as indicate satisfactory relative performance and a moderate
the dependent variable. Although some studies employed a degree of transparency; ESG scores in the third quartile
risk factor as an independent variable (Newell and Lin Lee (from > 75 to 100) indicate good (excellent) relative perfor-
2012; Cajias et al. 2014; Aroul et al. 2022), to our knowl- mance and above average (high) degree of transparency in
edge, no prior research has employed a risk factor as the reporting data.
dependent variable in the long-term analysis in a cross- Our sample covers 234 ESG-rated REITs in five devel-
country setting. As the only exception, Fan et al. (2022) oped markets, where 163(69.7%) REITs publicly report-
employ the overall volatility and systematic risk (Beta) as ing ESG scores are from the USA, 22(9.4%), 20(8.5%),
the dependent to measure the risk of US REITs during the and 19(8.1%) REITs from Australia, the UK, and Canada,
period of 2007 and 2021.
As the modeling approach, existing studies have generally
utilized panel regression analysis to explore the CSP-CFP 3
 Available at: https://​www.​refin​itiv.​com/​en/​susta​inable-​finan​ce/​esg-​
nexus. Given that only a single study (Cajias et al. 2014) scores (accessed on: 5 July, 2021).

13
Environmental Science and Pollution Research (2023) 30:85154–85169 85159

respectively. The number of Japanese REITs is 10 with a Table 1  Pesaran (2015)weak cross section dependence test results
4.3% share in the sample. According to our calculations Variable CD-test p-value
based on EPRA and Thomson Reuters databases, the share
of the equity market cap of ESG-rated REITs to the total ESG 252.10 0.00
market cap of public REITs in each country is 81% for Aus- Excess return 1 53.30 0.00
tralia, 83% for the UK, 82% for Canada, 28% for Japan, and Excess return 2 53.53 0.00
65% for the USA in 2019. To better grasp the representative- Sharpe ratio 1 48.93 0.00
ness of our sample, we further calculate the ratio of the total Sharpe ratio 2 49.11 0.00
market cap of ESG-rated REITs (our sample) to the total Beta 109.34 0.00
market cap of overall public REITs in five countries in 2018 E-score 163.47 0.00
and 2019, which are 82% and 63%, respectively.4 S-score 243.18 0.00
The preliminary analysis of the ESG scores across G-score 205.30 0.00
the sample reveals that the UK and the Australian REITs
recorded the highest ESG total scores of 54.5 and 45.5, on
average, respectively, during the period 2003–2019. The words, if the data is non-stationary in the level form, it needs
US and Canadian REITs have performed moderately well, to be transformed into stationary form in order to be used for
separately reporting overall ESG scores of 38.7 and 34.9. the Granger causality test (Huang 1995; Feige and Pearce
Japanese REITs have experienced the worst ESG perfor- 1979). Therefore, before we run the causality tests, we carry
mance, with an overall score of 21.9. Concerning the indi- out panel data unit root tests so as to investigate the station-
vidual E/S/G scores, the UK REITs have by far the high- arity of our variables.
est environmental score of 61.1, whereas the other country The utilization of panel data unit root tests has become
REITs have displayed notably lower E-scores, ranging from widely available among empirical scientists owing to advances
13.0 (Canada) to 39.1 (Australia). The governance score in time series econometrics and panel data analysis during the
is more equally distributed than the environment score— last 3 decades that is initiated by Levin and Lin (19921993).
e.g., Japanese REITs have the lowest G-score of 31.5, Currently, a variety of tests for unit-roots (or stationarity) in
whereas Canadian REITs have the highest score of 48.9. panel datasets are available.5 There are two generations of
The S-score across countries indicates the UK and the Aus- tests available among the panel unit-root testing framework:
tralian REITs have experienced the best social performances first generation and second generation. The former category of
as they reported 55.5 and 51.1 scores, respectively. Once tests assumes that they are convenient if there is no correlation
again, Japanese REITs have recorded the lowest S-score, between cross-sectional units, whereas the latter category of
16.9, on average. Overall, REITs in the UK and Australia tests is characterized by the rejection of the cross-sectional
have recorded superior ESG performance, particularly in independence hypothesis (Hurlin and Mignon 2007). In this
terms of environmental and social scores. Japanese REITs, regard, a simple test of weak cross-section dependence (CD)
in contrast, have been unsuccessful in their ESG-investing, proposed by Pesaran (2015) is applicable before employing
especially in social responsibility and governance quality. the appropriate unit-root test to our variables.
Table 1 presents the results for the Pesaran weak CD
test for all our variables, indicating that we reject the null
Empirical methodology hypothesis of weak CD. Therefore, we have to employ second
generation panel unit root tests that allow for cross-sectional
Unit root tests and PVAR Granger causality test dependence. Among those, due to the unbalanced structure6
of our dataset, the only available option is the Fisher test.
To test whether corporate social performance (or ESG score) Table 2 summarizes the results for the Fisher type test
affects the corporate financial performance or financial per- performs unit-root tests on each panel by combining the
formance affects social performance, or if there is a recipro- p-values from the panel-specific unit-root tests using the four
cal causality between the two, the Granger causality test is methods proposed by Choi (2001). For all our variables,
used. It should be noted that Granger causality test (Granger we strongly reject the null hypothesis being tested that all
1969) requires the use of stationary time series data. In other panels contain a unit root. Therefore, we can conclude that

4
  The number of REITs in our sample is 217 and 181 for 2018 and 5
2019, respectively. For the year 2018 (2019), our sampled ESG-rated   For more information, see Baltagi (2008).
6
REITs had $1162 ($977) billion market cap when the total market cap   Im-Pesaran-Shin (IPS) test would be another option; however, our
for all public REITs in five developed countries was $1416.3 (1553) dataset does not provide sufficient number of observations for some
billion, indicating 82% (63%) share. panel-units, and therefore this test cannot be carried out.

13

85160 Environmental Science and Pollution Research (2023) 30:85154–85169

Table 2  Panel unit root test results (Fisher PP)

Variable Panels Chi2 p-value Inverse normal p-value Inverse logit p-value Modified p-value
inv ­Chi2

ESG 234 1236.15 0.00  −17.07 0.00  −6.54 0.00 29.56 0.00
Excess return 1 234 2461.94 0.00  −46.75 0.00  −29.12 0.00 72.90 0.00
Excess return 2 234 2463.87 0.00  −46.80 0.00  −29.14 0.00 72.97 0.00
Sharpe ratio 1 234 2138.70 0.00  −40.02 0.00  −25.80 0.00 61.47 0.00
Sharpe ratio 2 234 2136.28 0.00  −39.97 0.00  −25.77 0.00 61.39 0.00
Beta 234 1242.50 0.00  −16.04 0.00  −7.69 0.00 29.79 0.00
E-score 234 987.55 0.00  −10.51 0.00  −4.19 0.00 20.77 0.00
S-score 234 1769.56 0.00  −29.04 0.00  −13.79 0.00 48.42 0.00
G-score 234 1179.53 0.00  −13.85 0.00  −4.21 0.00 27.56 0.00

AR parameter is panel specific. Time trend term is not included, cross sectional means are removed. Newey-West based lag length is chosen as 1.
The results are insensitive to further lags

our series are stationary at levels, and we may proceed with interested in finding the correct lag length, i.e., correct order,
the Granger-causality test. by using the order selection criteria. In this regard, CD: R2
In this regard, such a causal relationship is represented: criterion, which indicates the overall coefficient of determi-
m m nation for GMM models capturing the proportion of vari-
ation explained by the relevant panel VAR model, is used.
∑ ∑
Yit = 𝛼0 + 𝛽l Xi,t−l + 𝛾l Yi,t−l + 𝜇i + uit (1)
l=1 l=1 Panel data methods: fixed effects vs. random effects.
This study employs linear regressions with panel data
m m and estimates the following empirical model to investigate
� �
∑ ∑
Xit = 𝛼0 + 𝜃l Xi,t−l + 𝜆l Yi,t−l + 𝜇i + uit (2) the sign of the relationship between CSP and the financial
l=1 l=1 performance of REITs.
where it is assumed that the disturbances uit and uit are
′
∑t−1
Financial Performanceit = 𝛼 + 𝛽1 CSPit + Xit� 𝜃 + t=1 𝛿t Year t + 𝜀it
uncorrelated. Equation (1) indicates that variable X Granger- 𝜀it = 𝜇i + 𝜗it i = 1, … , N;t = 1, … , T
causes Y provided that 𝛽l ’s are statistically different from (3)
zero as a group, whereas 𝛾l ’s are not statistically different
where i represents each REIT company denoting the cross-
from zero as a group. Similarly, Y Granger-causes X given
section dimension, and t represents the time-series dimen-
that 𝛽l ’s are not statistically different from zero in Eq. (1),
sion. Financial performance is the dependent variable, and
while the set of the lagged X coefficients in Eq. (2),𝜆l’s, are
CSP is the variable of interest and measured either by an
statistically different from zero. Feedback or bilateral cau-
equally-weighted average of or individual E, S, and G scores.
sality is indicated when the sets of X and Y coefficients are
Xit is the K-dimensional vector of firm-specific variables
statistically different from zero in both equations. The most
that changes over time without a constant term. 𝜃 is a Kx1
important feature that distinguishes the panel VAR model
matrix, and 𝜗it represents the effects of the omitted variables
from the VAR model in the time series is the individual
that will change across the individual firms and time peri-
effects ( 𝜇i ) in the model.
ods. 𝜇i is a 1 × 1 scalar intercept representing the unobserved
Table 3 shows the overall results from Granger causal-
effects, which are constant over time. The random error term
ity Wald tests with the suggested lags for each equation of
is assumed to be distributed independently identically with
the underlying panel VAR model which is estimated using
mean zero and constant variance.
the generalized method of moments (GMM) developed by
Two different panel data methods, namely random effects
Abrigo and Love 2016).7As the Granger-causality test is
or fixed effects, can be used to estimate our model. Based
sensitive to the chosen lag in the causality model, we are
on the Hausman (1978) test results, we find that the speci-
fication test is strongly rejected in all our specifications,
7
  Stata 16 is not able to run the Granger-causality tests with the exist- implying that the random effects estimator would lead to
ing embedded commands for an unbalanced panel dataset. Therefore, inconsistent results. The fixed effects estimator is, therefore,
throughout the analysis held in this section, we utilized from Abrigo the appropriate methodology for our data to obtain unbiased
and Love (2016) which contributed user-written codes to Stata for
and consistent results. Although the most common draw-
these advanced panel data techniques. The relevant paper is also a
source of discussion on the theoretical background. back of fixed-effects models is the impossibility of including

13
Environmental Science and Pollution Research (2023) 30:85154–85169 85161

Table 3  Granger causality test Direction of causality Chi2 Lags Direction of causality Chi2 Lags

ESG → excess return 8.81* 4 E-score → Sharpe ratio 13.79*** 4

(0.07) (0.01)
Excess return → ESG 3.45 Sharpe ratio → E-score 1.81
(0.49) (0.77)
ESG → Sharpe ratio 13.42*** 4 S-score → Sharpe ratio 19.95*** 4
(0.01) (0.00)
Sharpe ratio → ESG 5.57 Sharpe ratio → S-score 5.07
(0.23) (0.28)
ESG → beta 11.16** 4 G-score → Sharpe ratio 8.89** 4
(0.03) (0.03)
Beta → ESG 7.58 Sharpe ratio → G-score 4.38
(0.11) (0.22)
E-score → Excess return 15.60*** 4 E-score → beta 8.96** 3
(0.00) (0.03)
Excess return → E-score 1.49 Beta → E-score 0.94
(0.83) (0.82)
S-score → Excess return 14.15*** 4 S-score → beta 13.88*** 4
(0.01) (0.01)
Excess return → S-score 4.61 Beta → S-score 18.28***
(0.33) (0.00)
G-score → Excess return 6.75 4 G-score → Beta 4.38** 1
(0.15) (0.04)
Excess return → G-score 4.76 Beta → G-score 2.66*
(0.31) (0.10)

Cluster-robust standard errors at the firm level are used to mitigate possible heterogeneity and autocor-
relation concerns. Excess return and Sharpe ratio are also calculated based on the overnight indexed swap
rates (alternative to 3-month LIBOR) as a proxy for risk-free rate of interest. Results are insensitive to these
alternative calculations, which are available upon request. As the Granger-causality test is sensitive to the
chosen lag, we find the correct lag length, i.e., correct order, using the order selection criteria. We use CD:
R2 criterion, which indicates the overall coefficient of determination for GMM models capturing the pro-
portion of variation explained by the relevant panel-VAR model

time-constant explanatory variables (Sassen et al. 2016), this subtracting 3-month interbank offered rates from the end-of-
is not the case for our model as our explanatory variables are year stock return. The risk-adjusted performance of a REIT
time-variant. Besides, we use year-fixed effects to control portfolio is measured by its Sharpe ratio—i.e., the ratio of
for changing macroeconomic conditions denoted by Year t , the annual excess return over risk-free rate to the volatil-
mainly to capture the impact of the Global Financial Crisis ity of excess returns—where REIT stock volatility is calcu-
(GFC) in 2008 and 2009. Finally, we adopt cluster-robust lated by using the annualized standard deviation of weekly
standard errors at the firm level8 to mitigate the concerns stock returns over the previous 12 months (see Auer and
about cross-sectional and time-series dependence. Schuhmacher 2016; Bouslah et al. 2013). The beta factor is
Using a panel dataset of 1408 firm-year observations over obtained from the beta index of REIT companies, consider-
the period 2003–2019,9 we examine the effect of ESG scores ing a timeframe of 60 months.10
on three market-based measures of financial performance: The model variables are listed in Table 4. We retrieved all
(1) excess return, (2) risk-adjusted performance of a portfo- financial data from the Thomson Reuters Worldscope and
lio by its Sharpe ratio, and (3) systematic firm risk or beta DataStream databases and utilized two risk-free rate prox-
factor. Excess return over a risk-free rate is calculated by ies obtained from Bloomberg, namely the 3-month LIBOR
and overnight indexed swap rate.11 ESG scores are obtained
8
 Alternative approach would be bootstrapping the standard errors.
from the Thomson Reuters Asset4 database that evaluates
Our results were insensitive in computing the variance of all esti- CSP based on three pillars of environmental, social, and
mates with 200 replications.
9
 Our initial sample comprised 1667 firm-year observations. Miss- 10
ing control variables have reduced the final sample to an unbalanced  Sassen et  al. (2016) also used the historical beta index obtained
panel of 1408 firm-year observations. The sample period starts from from DataStream database.
11
2003 as Thomson Reuters Asset4 database publishes ESG scores   We use the Overnight Indexed Swap rate data to reflect the impact
since 2002. of counterparty credit risk into risk-free rate variable.

13

85162 Environmental Science and Pollution Research (2023) 30:85154–85169

Table 4  Model variables
Variables and acronyms Indicator Vector of variables

Dependent variable
  Excess return (Ri-rf) Excess return Financial performance (CFP)
  Sharpe Ratio [(Ri-rf) /volatility] Risk-adjusted excess return
  Beta Systematic risk (historical local index)
Independent variables
  Environment pillar score (E-score) Corporate’s impact on its natural living and non-living environ- Corporate social performance (CSP)
ment—e.g., air, land, and water
  Governance pillar score (G-score) Corporate’s management commitment and effectiveness in imple-
menting good governance principles
  Social pillar score (S-score) Corporate’s ability to build trust and credibility with its employ-
ees, investors/customers, and society
  ESG total score An equally weighted average of individual E/S/G scores
Control Variables
  Market capitalisation (in USD) Size Firm characteristics
  Turnover ratio Liquidity
  Total debt ratio Leverage
  Price-to-book value ratio Future growth opportunities
  Total risk Firm Risk
  Operating expenses Operating performance
  Dividend payment t-1 Future earnings certainty
  GDP Size of the economy Country-specific factors
  M3 index Money supply
  Inflation rate Average consumer price changes

corporate governance performance. Among the country-spe- total operating expenses and used to explore if REITs with
cific variables, M3 index is derived from the OECD statistics lower operating costs would have better financial performance.
and GDP, and the inflation rate is obtained from the World Finally, we controlled for the dividend pay-out ratio, which
Economic Outlook, IMF. could be interpreted as a signal for managers’ perception of
We use the following control variables commonly certainty of future earnings. Pay-out ratio is calculated as the
adopted in the relevant literature, including Westermann ratio of dividends per share to the price per share with a time
et al. (2022), Sassen et al. (2016), Auer and Schuhmacher lag of one year since dividend cash flows are time-lagged (Sas-
(2016), and Newell and Lin Lee (2012). Firm size meas- sen et al. 2016). For all other control variables, we employed
ured as the natural log of market capitalization in US Dollar current values. Country-specific variables are employed to
accounts for the size effect on REIT’s financial performance. control macroeconomic conditions. In this respect, we use
Leverage controls the impact of REIT’s capital structure on average consumer price changes to control inflation rates of
the firm’s market risk and return and is calculated as the our sample countries. Additionally, we use M3 and GDP to
total debt-to-total assets ratio. We included REIT’s stock control variations in broad money supply or excess liquidity
market liquidity as a possible influencing factor on market and the size of the sample economies. Table 5 presents basic
return and risk, which is measured as the volume of shares descriptive statistics for model variables.
traded divided by the number of shares outstanding at the
company’s year-end. We used the price-to-book ratio to
capture different risk characteristics for growth and value Results and discussion
companies, calculated as the stock price per share to book
value per share. We also included the company total risk Full sample regression analysis results
and operating expenses to consider REIT’s market risk and
operating performance, respectively. The sign of association between ESG total and individual
As the indicator of return uncertainty, total risk reflects the scores and REIT financial performance, estimated by using
firm’s stock volatility and is measured by using the annualized Eq. (3) for the full sample period, are shown in Tables 6, 7.
standard deviation of weekly stock returns over the previous ESG total score has a negative link with REIT’s Sharpe ratio
12 months. Operating cost is measured as the natural log of (with an estimated coefficient of − 0.52 and − 0.62) and excess

13
Environmental Science and Pollution Research (2023) 30:85154–85169 85163

Table 5  Descriptive statistics Variable Mean Std. dev Min Max Observation

Excess return Ri − rf 0.03 0.30  −2.64 1.38 1667

[ ]

Sharpe ratio (Ri − rf )∕𝝈 i 0.13 1.11  −6.96 3.50 1667

[ ]

ESG total score 0.4 0.19 0.02 0.91 1667

Beta 0.85 0.59  −0.87 6.40 1667
E-score 0.26 0.30 0.00 0.98 1667
S-score 0.46 0.20 0.02 0.96 1667
G-score 0.48 0.22 0.00 0.95 1667
Log (market capitalisation) 15.00 1.09 12.06 18.44 1666
Log (turnover ratio) 0.23 0.84  −4.27 2.83 1667
Total debt ratio 0.46 0.15 0.00 1.38 1667
Log (price-to-book value ratio) 0.57 0.67  −2.30 5.03 1634
Total risk 0.26 0.17 0.08 1.71 1667
Log (operating expenses) 13.14 1.51 8.83 17.74 1647
Log (GDP) 29.97 1.34 27.88 33.92 1667
Inflation rate 0.02 0.01  −0.01 0.05 1667
Log (M3 index) 4.54 0.25 3.55 4.87 1667

return (estimated coefficient of − 0.13 and − 0.16), whereas it performance as anticipated. Country-specific macroeconomic
has no association with REIT beta during the overall study indicators do not have any statistically significant association
period. Coefficient values for the Sharpe ratio are more promi- with financial performance (excess return and beta) of REITs.
nent in magnitude than those for the excess return measure. To understand which E/S/G factor affects financial perfor-
The estimated coefficients for company-level control variables mance, we regressed our return and risk measures on the indi-
indicate that REIT size (market cap), stock market liquid- vidual E/S/G scores rather than the aggregated ESG score (see
ity (turnover ratio), price-to-book value ratio, and dividend Table 7) to test our first hypothesis. Table 7 displays a strong
pay-out ratio are all positively related to REIT excess return negative relationship between E-score and return measures—
and the Sharpe ratio. REIT total risk and operating expenses i.e., a negative association with Sharpe ratio and excess return
are, in contrast, negatively associated with their financial at 1% and 5% significance levels, respectively. In contrast, the

Table 6  The ESG total score and the financial performance of REITs from 2003 to 2019—Panel data fixed effects regression
(1) (2) (3) (4) (5) (6) (7) (8)
Variables Sharpe ratio Sharpe ratio Sharpe ratio Excess return Excess return Excess return Beta Beta

ESG total score  −0.62**  −0.52**  −0.52*  −0.16**  −0.13**  −0.13*  −0.04 0.09
Log (market capitalisation) 0.26** 0.43*** 0.43** 0.08** 0.11*** 0.11**  −0.21**  −0.23***
Log (turnover ratio) 0.17*** 0.17*** 0.17** 0.06** 0.06** 0.06** 0.02 0.03
(Dividend per price) t-1 2.67* 2.85* 2.84* 1.13** 1.22** 1.22** 0.88* 0.81*
Total debt ratio  −1.97***  −1.40**  −1.40**  −0.50***  −0.43**  −0.44** 0.17 0.10
Log (price-to-book value ratio) 0.80*** 0.70*** 0.70*** 0.21*** 0.20*** 0.20***  −0.13**  −0.13**
Total risk  −0.89*  −0.71  −0.71 -0.37**  −0.29*  −0.30*
Log (operating expenses)  −0.29**  −0.28**  −0.05  −0.04
Excess return 0.20*** 0.19***
Log (GDP) 0.35 0.41  −2.55
Log (M3 index)  −0.12  −0.16  −0.12
Inflation rate  −0.71  −0.13  −4.95
Time fixed effects Yes Yes Yes Yes Yes Yes Yes Yes
Observations 1408 1391 1391 1408 1391 1391 1408 1408
R-squared 0.57 0.57 0.57 0.65 0.65 0.65 0.56 0.57
Hausman test p value 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

13

85164 Environmental Science and Pollution Research (2023) 30:85154–85169

Table 7  The environmental, social, and governance individual scores and the financial performance of REITs from 2003 to 2019
(1) (2) (3) (4) (5) (6) (7) (8)
Variables Sharpe ratio Sharpe ratio Sharpe ratio Excess return Excess return Excess return Beta Beta

E-score  −0.57***  −0.48**  −0.48**  −0.13**  −0.11**  −0.11** 0.29** 0.31**

S-score 0.23 0.23 0.23 0.04 0.04 0.04  −0.31*  −0.28*
G-score  −0.11  −0.12  −0.12  −0.03  −0.03  −0.02  −0.13  −0.06
Log (market capitalisation) 0.28** 0.44*** 0.43** 0.08*** 0.11*** 0.11**  −0.23***  −0.24***
Log (turnover ratio) 0.17*** 0.17*** 0.16** 0.06** 0.06** 0.06** 0.03 0.03
(Dividend per price) t-1 2.73* 2.91** 2.91* 1.14** 1.24** 1.23** 0.83* 0.76*
Total debt ratio  −1.88***  −1.37**  −1.38**  −0.48***  −0.43**  −0.43** 0.09 0.03
Log (price-to-book value ratio) 0.80*** 0.71*** 0.71*** 0.21*** 0.20*** 0.20***  −0.13**  −0.13**
Total risk  −0.89*  −0.72  −0.72  − 0.37**  − 0.29*  − 0.30*
Log (operating expenses)  − 0.27**  − 0.26**  − 0.04  − 0.04
Sharpe ratio
Excess return 0.21*** 0.20***
Log (GDP) 0.08 0.35  − 2.21
Log (M3 index)  − 0.06  − 0.15  − 0.17
Inflation rate  − 1.14  − 0.22  − 4.69
Time fixed effects Yes Yes Yes Yes Yes Yes Yes Yes
Observations 1408 1391 1391 1408 1391 1391 1408 1408
R-squared 0.57 0.57 0.57 0.65 0.65 0.65 0.57 0.58
Hausman Test p value 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

Constant term is included but suppressed. Standard errors are clustered at the firm level. ***p < 0.01, **p < 0.05, *p < 0.1. The variable of
interest(s) is (are) ESG total score (E, S, and G individual scores) in Table  4a (Table  4b). The dependent variables are Sharpe ratio, excess
return, and beta. For each dependent variable the first two models involve firm specific characteristics whereas the third model controls for the
country specific controls, of log GDP, log M3 index and inflation rate. All models include time-fixed effects and firm-fixed effects

estimated coefficients of the S-score and G-score are insig- our study is one of the initial attempts to analyze REITs’
nificant with a positive and negative sign, correspondingly. CSP-CFP nexus by investigating the association between
Noticeably, the negative relationship between E-score and financial performance and (dis-)aggregated ESG scores in
REIT return measures has a dominating impact on REIT’s cross-country setting, the only comparable evidence that uses
CSP (ESG total score) and CFP nexus. Once again, the com- disaggregated ESG scores is provided by Brounen and Mar-
pany- and country-specific control variables have similar esti- cato (2018) and Fan et al. (2022). In this respect, studying
mated coefficient values with the same signs. It is important a sample of 194 USA REITs, Brounen and Marcato (2018)
to highlight that a higher S-score is negatively associated with found that a higher E-score is associated with a significant
REIT beta, indicating that S-investing reduces firm’s sys- negative excess return. Using a sample of listed US equity
tematic risk; a higher E-score, in contrast, is associated with REITs, Fan et al. (2022) also document that environmental
higher systematic risk. Furthermore, no relationship between impact is negatively associated with future stock returns, and
corporate governance (G-score) and financial performance is social and governance ratings are positively related to future
found. These results provide sufficient evidence to accept the stock returns. The authors further argue that the overall ESG
first hypothesis of each component of E/S/G that investing score cannot significantly predict future stock returns in the
has different impacts on the financial performance of REITs’. REITs, consistent with the findings in Pedersen et al. (2021).
Overall, results from the full-sample analysis indicate We also find that neither the social nor governance com-
that the environmental component of ESG investing is nega- ponent of ESG investing has a significant association with
tively (positively) associated with REITs’ return (systematic REIT market return measures, whereas the former has a
risk). These estimates provide evidence against our second weak negative relationship with beta. Previous studies
hypothesis that a higher E-score (or environmental perfor- (Bauer et al. 2010; Bianco et al. 2007; and Hartzell et al.
mance) leads to a lower financial performance, which the 2006) also found no significant relationship between corpo-
trade-off hypothesis might explain. Specifically, REITs’ rate governance and USA REIT performance and explained
environmental policies and practices involve high financial this weak impact of G-investing on REIT’s performance by
costs that may drain off capital and other company resources the strongly regulated business environment in the REIT
and result in declining stock prices and market returns. As industry (Ghosh and Petrova 2021; Bauer et al. 2010).

13
Environmental Science and Pollution Research (2023) 30:85154–85169 85165

Table 8  The ESG total score and the financial performance of REITs after the GFC, from 2011 to 2019
(1) (2) (3) (4) (5) (6) (7) (8)
Variables Sharpe ratio Sharpe ratio Sharpe ratio Excess return Excess return Excess return Beta Beta

ESG total score  −0.54  −0.55  −0.51  −0.09  −0.09  −0.09  −0.32**  −0.25
Log (market capitalisation) 0.34* 0.52*** 0.63*** 0.07* 0.09** 0.12***  −0.07  −0.08
Log (turnover ratio) 0.20** 0.22** 0.18** 0.06*** 0.06*** 0.06***  −0.03  −0.04
(Dividend per price) t-1 5.10** 5.38** 5.58** 0.88** 0.92** 0.97**  −0.49  −0.41
Total debt ratio  −1.30**  −0.69  −0.88  −0.29**  −0.22  −0.25* 0.92*** 0.94***
Log (price-to-book value ratio) 1.17*** 1.03*** 1.05*** 0.25*** 0.24*** 0.24***  −0.14**  −0.12**
Total risk  −1.69*  −1.53*  −0.96  −0.56***  −0.54***  −0.44**
Log (operating expenses)  −0.31**  −0.38**  −0.04  −0.05
Excess return 0.24*** 0.20***
Country specific controls No No Yes No No Yes No Yes
Time fixed effects Yes Yes Yes Yes Yes Yes Yes Yes
Observations 1093 1086 1086 1093 1086 1086 1093 1093
R-squared 0.41 0.42 0.44 0.47 0.47 0.49 0.58 0.62
Hausman test C ­ hi2 119.87 123.88 146.48 135.99 144.22 161.51 138.37 132.15
Hausman Test p value 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

Sub‑period regression analysis results impact argument, a corporation that attempts to lower its
implicit costs by reducing or ignoring socially responsible
Table 7 presents regression analysis results for the same actions will, therefore, incurs higher explicit costs, giving
range of models shown in Table  6, after the GFC from rise to a competitive disadvantage. In contrast, a broad-
2011 to 2019. Considering that both the financial and other minded employee relations policy may have a lower cost,
company-specific indicators still exhibited poor perfor- resulting in substantial gains in morale, productivity, and
mance in 2010, we define the post-GFC subperiod from competitive advantage compared to less responsible firms.
2011 onwards. We find no statistically significant asso- The risk mitigation argument predicts that the performance
ciation between ESG total score and REIT excess return of S-investing is negatively related to firm risk. It might be
and the Sharpe ratio (Table 8), although ESG total score is possible to claim that successful social investments of REITs
negatively related to REIT beta (model 7). An insignificant generate a reservoir of positive moral capital or goodwill
relationship between ESG total score and REIT return meas- that creates relational wealth among different stakeholders,
ures is because the dominating negative association between which in turn reduces financial risk (McGuire et al. 1988).
E-score and REIT returns is now offset by the strong posi- A considerable number of non-REIT studies have supported
tive relationship between S-score and REIT return measures the risk mitigation view by documenting a negative relation
(Table 9). Moreover, S-score has a solid negative relation- between the performance of social-investing (S-score) and sys-
ship with REIT beta with the estimated coefficients of − 0.48 tematic risk for S&P 500 firms (Oikonomou et al. 2012), for
(model 7) and −0.43 (model 8), which is already reflected in Fortune 1000 companies (Luo and Bhattacharya 2009) and a
the ESG total score-beta relationship in model 7 in Table 5a. sample of UK firms (Salama et al. 2011). The current study
In line with the full sample case, G-score has no significant provides the first empirical support for the risk mitigation argu-
association with REIT financial performance either through ment of the stakeholder theory of corporation in explaining the
excess return or systematic risk measure. In the post-GFC negative relationship between the performance of S-investing
period, stock market investors have attached a higher value and systematic risk for the developed REIT markets.
to REITs’ S-investing; S-score exhibits a strong positive Our results suggest that the negative relationship between
association with CFP through not only higher excess return E-investing and REIT returns is persistent as the analysis
and the Sharpe ratio but also a lower beta. involves both long-term (full sample, 2003–2019) and mid-
A positive premium for S-investments supports our sec- term (post-GFC sub-sample 2011–2019) periods. This find-
ond hypothesis of “higher social performance (S-score) ing seems rather counter-intuitive because a higher environ-
leads to a higher financial performance for REITs through mental score is positively related to cost-effectiveness and
higher return and lower risk,” which might be explained better financial performance (e.g., Guenster et al. (2011)).
by the social impact and risk mitigation hypotheses of the Nevertheless, Eichholtz et al. (2012) and Coën et al. (2018)
stakeholder theory of corporation. According to the social reported insignificant results for portfolio greenness and

13

85166 Environmental Science and Pollution Research (2023) 30:85154–85169

Table 9  The E/S/G individual scores and the financial performance of REITs after the GFC, from 2011 to 2019
(1) (2) (3) (4) (5) (6) (7) (8)
Variables Sharpe ratio Sharpe ratio Sharpe ratio Excess return Excess return Excess return Beta Beta

E-score  −0.54**  −0.52**  −0.60***  −0.11**  −0.11**  −0.12*** 0.14 0.07

S-score 0.57* 0.57* 0.57* 0.14** 0.14** 0.14**  −0.48**  −0.43**
G-score  −0.38  −0.40  −0.29  −0.07  −0.08  −0.06  −0.08 0.02
Log (market capitalisation) 0.34* 0.51*** 0.62*** 0.07* 0.09** 0.11***  −0.08  −0.08
Log (turnover ratio) 0.20** 0.22** 0.18* 0.06*** 0.06*** 0.06***  −0.03  −0.04
(Dividend per price) t-1 5.37** 5.64** 5.78** 0.94** 0.98** 1.02**  −0.60  −0.55
Total debt ratio  −1.22**  −0.64  −0.81  −0.27**  −0.20  −0.24 0.87*** 0.92***
Log (price-to-book value ratio) 1.16*** 1.03*** 1.05*** 0.25*** 0.24*** 0.24***  −0.15**  −0.13**
Total risk  −1.77*  −1.62*  −1.05  −0.58***  −0.56***  −0.46**
Log (operating expenses)  −0.30**  −0.36**  −0.04  −0.05
Sharpe ratio
Excess return 0.26*** 0.22***
Country specific controls No No Yes No No Yes No Yes
Time fixed effects Yes Yes Yes Yes Yes Yes Yes Yes
Observations 1093 1086 1086 1093 1086 1086 1093 1093
R-squared 0.42 0.42 0.44 0.48 0.48 0.49 0.58 0.62
Hausman Test C ­ hi2 118.58 121.81 143.83 135.07 142.24 159.16 143.92 89.53
Hausman Test p value 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

Constant term is included but suppressed. Standard errors are clustered at the firm level. ***p < 0.01, **p < 0.05, *p < 0.1. The variable of
interest(s) is (are) ESG total score (E, S, and G individual scores) in Table  4a (Table  4b). The dependent variables are Sharpe ratio, excess
return, and beta. For each dependent variable the first two models involve firm specific characteristics whereas the third model controls for the
country specific controls namely log GDP, log M3 index and inflation rate. All models include time-fixed effects and firm-fixed effects

financial performance of REITs, whereas several research- Conclusion

ers found a negative relationship between high costs of
certification and firm value (Mariani et al 2018) as well as Now, more than ever, investors understand the value of
environmental regulation and firm value (Brounen and Mar- integrating ESG metrics into their investment decisions
cato 2018; Rassier and Earnhart 2010). As discussed previ- to reduce risks, discover opportunities, and consequently
ously, the negative results provide evidence for the trade- influence corporates’ resource-allocation decisions. Exist-
off hypothesis. Empirical literature reveals some plausible ing knowledge on the financial outcomes of corporate social
explanations for this negative relation, as well. For instance, investing through ESG criteria remains fragmented and
Guenster et al. (2011) claimed that eco-efficiency is value- scarce in the REIT industry. Using a rich dataset comprising
relevant but is incorporated slowly into a company’s stock 234 ESG-rated REITs across five developed economies from
price. In the same line of time-effect argument, Derwall et al. 2003 to 2019, this paper examines the statistical relation-
(2011), Brounen and Marcato (2018), and Brounen et al. ship between ESG investing and the financial performance
(2021) argued that environmental awareness needs longer of REITs.
to gain any positive performance impact. While costs have Our paper adds to the corporate social performance (CSP)
a clear footprint in the short-term profit and loss accounts, and corporate financial performance (CFP) nexus in the REIT
the benefits of ESG investment are intangible, difficult to market in several ways. From a practical perspective, the
quantify and materialize in the short term. Hence, it could be main contributions of the present study can be summarized
argued that high costs of certification and environmental reg- as follows: (1) our study involves a longer period analysis in
ulation support the trade-off hypothesis, and the time-effect a cross-country setting; (2) it attempts to employ systematic
argument may play a role in this negative relationship.12 firm risk as a new dependent variable in this setting; (3) it is
also the first study in a cross country sample to test (i) the
nexus for both aggregated and disaggregated levels, and (ii)
12
social impact hypothesis and risk mitigation view of the stake-
 We re-ran our regression models with alternative operating per-
formance measures, including return on equity, return on investment,
holder theory for REITs; (4) finally, the present study is the
and return on total assets. Our estimation results are robust and insen- first application of causality analysis for a cross-country REIT
sitive to the exercises, which are available upon request. sample and also a rare study to fill the existing knowledge gap

13
Environmental Science and Pollution Research (2023) 30:85154–85169 85167

about the lack of causality analysis for a REITs sample (as the of the ESG rating at the expense of its moral and ethical
only exception, see, Cajias et al. 2014). dimensions (Fleischman et al 2019) in the corporate sector.
We have four main findings. Firstly, the PVAR Granger Literature reveals that previous studies mostly focus on
causality analysis suggests a direction of causality from the environmental sustainability due to the challenging nature of
CSP to the CFP, and the full sample regression results for the measuring social or governance issues for real property (see,
aggregated ESG scores suggest a negative CSP-CFP relation- Chacon et al. 2022). Our paper provides mixed evidence for
ship for the REITs. This result is in line with those of Mari- the relationship between risk and socially responsible investing
ani et al. (2018), Coën et al. (2018), and Westermann et al. from the challenging S&G investing in REITs. In this respect,
(2022). Moreover, our results provide mixed evidence of fourthly, while S-investing reduces a firm’s systematic risk; a
the statistical relationship between individual E/S/G invest- higher E-score, increases the systematic risk. We are cautious
ing and the financial performance of REITs. This evidence in interpreting the results due to inconsistencies in the evidence
set suggests that REIT investors pay attention to individual set. In this respect, on the one hand, S-investing may reduce
E/S/G metrics and price each component of ESG investing financial risk and encourage firms to be more sustainable (see,
differently, and E/S/G investments have different implications Chollet and Sandwidi 2018), but, on the other hand, the nega-
for the financial performance of REITs. For example, the tive impact of environmental investment on the firm risk and
dataset shows that the UK (Japanese) REITs have the high- market return may discourage REITs environmental invest-
est (the lowest) environmental (social) sustainability score. ments. We may interpret the latter evidence with the overinvest-
This may be the main reason for the component-based vari- ment hypothesis arguing overinvestment in environmental sus-
ations in the linkage between E-S-G investing and firm-level tainability (and hence ESG) may eventually detriment REITs
financial performance in a cross-country setting. This finding market value in terms of both declining return and increasing
implies that firm-level ESG policies may prioritize based on risk perspectives (see, Chacon et al. 2022; Fan et al 2022).
the industry-specific ESG performance component. Overall, in light of the above evidence and implication
Secondly, empirical evidence from the full-sample analy- sets, we argue that the financial return of ESG investing in
sis suggests while E-investing is negatively associated with REITs looks rather fragmented. This picture brings into
REITs’ excess return, S-investing and G-investing components consideration of potential financial performance problem
have no significant association with the financial performance of adopting ESG practices into REITs (see, Fan et al. 2022).
of REITs. From the perspective of environmental sustain- We mainly suggest that company boards and sustainability
ability, this evidence suggests strong support for the trade-off (and/or ESG) departments of REITs may essentially focus
hypothesis—i.e., REITs’ environmental activities such as own- on long-term value creation for their companies or funds but
ing green building certifications, adopting land conservation also carefully manage conflicted performance components.
and eco-friendly building design techniques, and reducing In this respect, as also suggested by our evidence set, they
emissions at buildings involve high financial costs that may may be specifically careful on the cost side of environmental
drain off company resources and result in diminishing market investments in their ESG framework.
returns. This evidence supports the belief of the high invest- This study focuses on five developed REIT markets to
ment costs of E-investing may discourage further development investigate the CSP-CFP relationship due to the limited
in sustainable REITs and result in a decline in the positive number of REITs with publicly reported ESG metrics
environmental impact of securitized real estate assets. in other countries. Further research may cover all ESG-
Thirdly, stock market investors have attached a higher rated REITs globally to improve the generalization of the
value to REITs’ social investment performance in the post- research outcomes and explore cross-country comparisons
GFC period, from 2011 to 2019; a positive premium for over a more extended period. For future research, it would
S-investing supports the social impact hypothesis and risk also be interesting to study the effect of REIT-market-
mitigation views of the stakeholder theory of the corpora- specific factors (e.g., ownership structure, involvement
tion. Therefore, it is possible to argue that the socially ethi- in development activities, and accounting standards for
cal behavior of REITs (S-investing) may generate a posi- property valuation) and to explore whether and to what
tive premium for the REIT stocks and create a competitive extent the COVID-19 pandemic will change the existing
advantage through their productive relationships with their CSP-CFP relationship for real estate corporations.
stakeholders, especially in the expansion period of the post-
Abbreviations  CFP: Corporate financial performance; CSP: Corporate
GFC era. On the other hand, governance practices have social performance; CSR: Corporate social responsibility; ESG: Envi-
not improved the financial performance of REITs over the ronmental social performance; GDP:  Gross domestic product;
17-year study period. We may speculate that ESG investing GFC: Global financial crisis; GRESB: Global Real Estate Sustainability
may provide less information on its inherent ethical impact, Benchmark; IMF: International Monetary Fund; LIBOR: London Inter-
bank Offered Rate; OECD: Organization for Economic Cooperation
namely, the governance investment quality (G scores). This and Development; REITs: Real Estate Investment Trusts
is presumably because of the growing commercial character

13

85168 Environmental Science and Pollution Research (2023) 30:85154–85169

Author contributions  Isil Erol collected and constructed the data, Chollet P, Sandwidi BW (2018) CSR engagement and financial risk:
interpreted the empirical findings, and supervised the paper. Umut a virtuous circle? International evidence. Glob Financ J 38:65–
Unal conducted the quantitative analysis. Yener Coskun developed 81. https://​doi.​org/​10.​1016/j.​gfj.​2018.​03.​004
a literature review section, edited related parts, and developed the Coën A, Lecomte P, Abdelmoula D (2018) The financial perfor-
final revision of the paper. All the authors read and approved the final mance of green REITs revisited. J Real Estate Portf Manag
manuscript. 24(1):95–105
Derwall J, Koedijk K, Ter Horst J (2011) A tale of values-driven and
Data availability  The datasets used and/or analyzed during the cur- profit-seeking social investors. J Bank Finance 35(8):2137–2147
rent study are available from the corresponding author on a reasonable Eichholtz P, Kok N, Yonder E (2012) Portfolio greenness and the finan-
request. cial performance of REITs. J Int Money Financ 31(7):1911–1929
Eichholtz P, Kok N, Quigley JM (2013) The economics of green build-
Declarations  ing. Rev Econ Stat 95(1):50–63
Fan KY, Shen J, Hui ECM (2022) ESG Materiality and responsible
Ethical approval  Not applicable investment: evidence from Real Estate Investment Trusts. Avail-
able at: https://​l24.​im/​RI5 (accessed on: 4 May 2023)
Consent to participate  Not applicable Feige EL, Pearce DK (1979) The casual causal relationship between
money and income: Some caveats for time series analysis. Rev
Consent for publication  Not applicable Econ Stat 521–533
Fleischman GM, Johnson EN, Walker KB, Valentine SR (2019) Ethics
Competing interests  The authors declare no competing interests. versus outcomes: managerial responses to incentive-driven and
goal-induced employee behavior. J Bus Ethics 158(4):951–967
Freeman RE (1984) Strategic management: a stakeholder approach.
Pitman, Boston
Friede G, Busch T, Bassen A (2015) ESG and financial performance:
aggregated evidence from more than 2000 empirical studies. J
References Sustain Financ Invest 5(4):210–233
Friedman M (1970) A friedman doctrine: The social responsibility of
Abrigo MR, Love I (2016) Estimation of panel vector autoregression business is to increase its profits. The New York Times Magazine
in Stata. Stand Genomic Sci 16(3):778–804 13(1970):32–33
Aroul RR, Sabherwal S, Villupuram SV (2022) ESG, operational Frynas JG, Yamahaki C (2016) Corporate social responsibility: review
efficiency and operational performance: evidence from Real and roadmap of theoretical perspectives. Business Ethics: a Euro-
Estate Investment Trusts. Manag Financ 48(8):1206–1220. pean Review 25(3):258–285
https://​doi.​org/​10.​1108/​MF-​12-​2021-​0593 Fuerst F (2015) The financial rewards of sustainability: a global perfor-
Auer BR, Schuhmacher F (2016) Do socially (ir)responsible invest- mance study of real estate investment trusts. Available at SSRN
ments pay? New evidence from international ESG data. Q Rev 2619434
Econ Finance 59:51–62 Ghosh C, Petrova M (2021) The effect of legal environment and regula-
Barnea A, Rubin A (2010) Corporate social responsibility as a con- tory structure on performance: crosscountry evidence from REITs.
flict between shareholders. J Bus Ethics 97(1):71–86 J Real Estate Financ Econ 63:40–81
Bauer R, Eichholtz PMA, Kok N (2010) Corporate governance and Godfrey PC (2005) The relationship between corporate philanthropy
performance: the REIT effect. Real Estate Econ 36:1–29 and shareholder wealth: a risk management perspective. Acad
Bianco C, Ghosh C, Sirmans CF (2007) The impact of corporate Manag Rev 30:777–798
governance on the performance of REITs. J Portf Manag Godfrey PC, Merrill CB, Hansen JM (2009) The relationship between
33(5):175–191 corporate social responsibility and shareholder value: an empirical
Bouslah K, Kryzanowski L, M’Zali B (2013) The impact of the test of the risk management hypothesis. Strateg Manag J 30:425–445
dimensions of social performance on firm risk. J Bank Finance Granger CW (1969) Investigating causal relations by econometric
37:1258–1273 models and cross-spectral methods. Econometrica: Journal of
Brounen D, Marcato G (2018) Sustainable insights in public real the Econometric Society 1:424–38
estate performance: ESG scores and effects in REIT markets. Guenster N, Bauer R, Derwall J, Koedijk K (2011) The economic value
Berkeley Lab, Berkeley of corporate eco-efficiency. Eur Financ Manag 17(4):679–704
Brounen D, Marcato G, Op’t Veld H (2021) Pricing ESG equity Hartzell J, Sun L, Titman S (2006) The effect of corporate govern-
ratings and underlying data in listed real estate securities. Sus- ance on investment: evidence from REITs. Real Estate Econ
tainability 13(4):2037 34(3):343–376
Cajias M, Fuerst F, McAllister P, Nanda A (2014) Do responsible Hausman JA (1978) Specification tests in econometrics. Econometrica:
real estate companies outperform their peers? Int J Strateg Prop Journal of the Econometric Society 1251–1271
Manag 18(1):11–27 Huang BN (1995) Do Asian stock market prices follow random
Campbell RD, Ghosh C, Petrova M, Sirmans CF (2011) Corporate walks? Evidence from the variance ratio test. Appl Financ Econ
governance and performance in the market for corporate control: 5:251–256
the case of REITs. J Real Estate Financ Econ 42(4):451–480 Hurlin C, Mignon V (2007) Second generation panel unit root tests.
Chacon R, Feng Z, Wu Z (2022) Does investing in ESG pay off? Evi- HAL Id: halshs-00159842
dence from REITs around the COVID-19 pandemic. Evidence IEA (2019) The 2019 Global status report for buildings and construc-
from REITs around the COVID-19 pandemic (June 27, 2022). tion. https://​www.​iea.​org/​repor​ts/​global-​status-​report-​for-​build​
Available at SSRN: https://​ssrn.​com/​abstr​act=​41477​49 ings-​and-​const​ructi​on-​2019 (accessed on 12 April 2021)
Choi I (2001) Unit root tests for panel data. J Int Money Financ Levin A, Lin CF (1992) Unit root tests in panel data: asymptotic and
20:249–272 finite-sample properties. University of California. San Diego, Dis-
cussion Paper, (92–93)

13
Environmental Science and Pollution Research (2023) 30:85154–85169 85169

Levin A, Lin CF (1993) Unit root tests in panel data: new results. Univer- performance? A case of 100 best corporate citizens. SAGE Open
sity of California at San Diego, Economics Working Paper Series 11(2):21582440211021600
Luo X, Bhattacharya CB (2009) The debate over doing good: corporate Rassier DG, Earnhart D (2010) Does the porter hypothesis explain
social performance, strategic marketing levers, and firm-idiosyn- expected future financial performance? The effect of clean water
cratic risk. J Mark 73:198–213 regulation on chemical manufacturing firms. Environ Resource
Mariani M, Amoruso P, Caragnano A, Zito M (2018) Green real estate: Econ 45(3):353–377
does it create value? Financial and sustainability analysis on Euro- Salama A, Anderson K, Toms JS (2011) Does community and envi-
pean green REITs. Int J Bus Manag 13:80–92 ronmental responsibility affect firm risk? Evidence from UK
McGuire JB, Sundgren A, Schneeweis T (1988) Corporate social panel data 1994–2006. Business Ethics: A European Review
responsibility and firm financial performance. Acad Manag J 20(2):192–204
31:854–872 Salzmann AJ (2013) The integration of sustainability into the theory
Morri G, Anconetani R, Benfari L (2021) Greenness and financial per- and practice of finance: an overview of the state of the art and
formance of European REITs. J Eur Real Estate Res 14(1):40–61 outline of future developments. J Bus Econ 83(6):555–576
NAREIT (2019) Nareit guide to ESG reporting frameworks. https://​www.​ Sassen R, Hinze AK, Hardeck I (2016) Impact of ESG factors on firm
reit.​com/​sites/​defau​lt/​files/​media/​PDFs/​Resea​rch/​Nareit_​Guide_​to%​ risk in Europe. J Bus Econ 86(8):867–904
20ESG_​Repor​ting_2_​21_​19.​pdf (accessed on 8 April 2021) United Nations Environment Program (2019) 2019 Global status report
Newell G, Lin Lee C (2012) Influence of the corporate social respon- for buildings and construction. https://​wedocs.​unep.​org (accessed
sibility factors and financial factors on REIT performance in Aus- on 10 April 2023).
tralia. J Prop Invest Financ 30(4):389–403 Waddock SA, Graves SB (1997) The corporate social performance-
Newell G, Wen Peng H, Yam S (2011) Assessing the linkages between financial performance link. Strateg Manag J 18(4):303–319
corporate social responsibility and A-Reit performance. Pac Rim Wang Q, Dou J, Jia S (2016) A meta-analytic review of corporate social
Prop Res J 17(3):370–387 responsibility and corporate financial performance: the moderat-
Oikonomou I, Brooks C, Pavelin S (2012) The impact of corporate ing effect of contextual factors. Bus Soc 55(8):1083–1121
social performance on financial risk and utility: a longitudinal Westermann S, Niblock SJ, Kortt MA (2022) Does it pay to be respon-
analysis. Financ Manage 41:483–515 sible? Evidence on corporate social responsibility and the invest-
Ooi JT, Dung DD (2019) Finding superior returns in green portfo- ment performance of Australian REITs. Asia-Pac J Account Econ
lios: evidence from Singapore REITs. J Sustain Real Estate 29(4):1102–1119
11(1):191–215
Pedersen LH, Fitzgibbons S, Pomorski L (2021) Responsible investing: Publisher's note Springer Nature remains neutral with regard to
the ESG efficient frontier. J Financ Econ 142(2):572–597 jurisdictional claims in published maps and institutional affiliations.
Pesaran MH (2015) Testing weak cross-sectional dependence in large
panels. Economet Rev 34(6–10):1089–1117 Springer Nature or its licensor (e.g. a society or other partner) holds
Preston LE, O’bannon DP (1997) The corporate social-financial exclusive rights to this article under a publishing agreement with the
performance relationship: a typology and analysis. Bus Soc author(s) or other rightsholder(s); author self-archiving of the accepted
36(4):419–429 manuscript version of this article is solely governed by the terms of
Qureshi MA, Akbar M, Akbar A, Poulova P (2021) Do ESG such publishing agreement and applicable law.
endeavors assist firms in achieving superior financial

13