WP19 065rev4 23 21
WP19 065rev4 23 21
WP19 065rev4 23 21
Ioannis Ioannou
George Serafeim
George Serafeim
Harvard Business School
Abstract: We explore the conditions under which firms maintain their competitive advantage
through sustainability-based differentiation when faced with imitation pressures by industry
peers. We document growing intraindustry convergence on sustainability actions over time
for almost all industries in our sample and show that interindustry heterogeneity in the rates
of intraindustry convergence is associated with (a) the importance of environmental and
social issues relative to governance issues, and (b) the tone and volume of feedback received
from stakeholders. Further, we find that actions characterized by low regulatory uncertainty
are more likely to be imitated whereas those characterized by high novelty are less likely to
be imitated. We then distinguish between common (i.e., imitated) and unique sustainability
actions and evidence that the adoption of unique actions is significantly and positively
associated with multiple measures of performance, whereas the adoption of common actions
is not. Overall, we explore the role of sustainability as a long-term strategy under conditions
of strong imitative forces, contributing to both the sustainability and the imitation literatures.
Ioannis Ioannou is an Associate Professor of Strategy and Entrepreneurship at London Business School. George
Serafeim is a Professor of Business Administration at Harvard Business School. George Serafeim is grateful for
financial support from the Division of Faculty and Research Development at Harvard Business School. Contact
details for Ioannis Ioannou: London Business School, Sussex Place, Regent’s Park, London, NW1 4SA, United
Kingdom, Tel: +44 (0) 20 7000 8748, e-mail: [email protected]. Contact details for George Serafeim:
Harvard Business School, 381 Morgan Hall, Boston, MA, 02163, United States, Tel: +1 (617) 495 6548, e-mail:
[email protected].
1
1 INTRODUCTION
Sustainability is becoming a central issue for the practice of strategic management globally
(Ioannou & Hawn, 2019).1 A growing number of companies voluntarily undertake a wide
advantage by integrating sustainability into their strategy, and their organizational processes
and structures (Eccles, Ioannou, & Serafeim, 2014; Khan, Serafeim, & Yoon, 2016). How
sustainable could such an advantage be though over the long(er) term, given that profitable
strategies are more likely to attract the attention of industry peers and to be imitated (e.g.,
Barney, 1991; Schmalensee, 1985) and that, if successful, the wider diffusion of these
strategies will erode the profits of firms that are being imitated (Posen, Lee, & Yi, 2013)?
The literature finds that value creation can materialize through a number of channels
Cheng, Ioannou, & Serafeim, 2014; Ioannou & Serafeim, 2015), more and better innovation
outcomes (e.g., Flammer & Kacperczyk, 2016), enhanced access to human capital and
employee engagement (e.g., Dutton, Dukerich, & Harquail, 1994; Edmans, 2011; Flammer &
Luo, 2017), improved reputation and increased sales (e.g., Luo & Bhattacharya, 2006; Du,
Bhattacharya, & Sen, 2011), favorable access to international markets (Hawn, 2020),
footprint (e.g., Bansal & Roth, 2000; Delmas & Pekovic, 2013; Hart, 1995; Russo & Fouts,
1
Indicatively, we note that 93% of the largest 250 companies in the world issue a corporate sustainability report,
and that, even more crucially, 78% already include and/or integrate sustainability information in their annual
financial reports (KPMG, 2017), which implies that financial and ESG considerations are seen as the two sides
of the same (strategic) coin.
2
For the purposes of this study, we define sustainability actions as the set of actions that a company undertakes,
in the form of adoption of policies, practices, management, and governance systems or investments, with the
aim of improving its environmental, social, and governance performance.
2
1997), and the pre-emption of regulatory intervention (e.g., Baron, 2001; Maxwell, Lyon, &
Hackett, 2000).
Even though this large body of work provides evidence that companies can establish a
competitive advantage through sustainability, the literature has not yet sufficiently explored
the conditions under which companies are able to maintain such an advantage, especially
when faced with imitation pressures. Moreover, we still lack an understanding of which
actions are more likely to diffuse via imitation, and why, and which are more likely to remain
unique and therefore, potentially valuable in the long term. From a strategic management
point of view, addressing this gap is important because it would allow us to better understand
heterogeneity across and within industries in terms of how companies decide which
sustainability actions to pursue, and would also allow us to develop a more nuanced
Anecdotal evidence suggests that sustainability actions may merely become a cost of
doing business because of imitation. For example, adopting a carbon emissions reduction
target of 20% in the early 2000s could have resulted in a differentiation advantage (e.g., due
to better environmental management); yet, by 2021 such targets have become commonplace,
even in hard to decarbonize industries like the oil and gas industry.3 This could arguably be
why in an attempt to differentiate themselves, companies like Microsoft have more recently
moved beyond net-zero commitments and instead, committed to being carbon negative by
2030.4
Currently, there are two useful approaches for understanding the adoption of
sustainability actions. On the one hand, one could argue that sustainability is rapidly
3
See BP’s recent announcement here: https://www.bloomberg.com/news/articles/2020-02-12/new-bp-ceo-sets-
out-plan-to-eliminate-its-co2-emissions-by-2050?sref=hZPO2dEb (accessed September 10, 2020).
4
See Microsoft’s announcement here: https://blogs.microsoft.com/blog/2020/01/16/microsoft-will-be-carbon-
negative-by-2030/ (accessed September 10, 2020).
3
becoming a “common practice,” i.e., a necessary cost of doing business, yet it remains
insufficient for strategic differentiation. For example, in some industries, the adoption of eco-
reducing costs and achieving cost parity with industry peers who may have already adopted
them. In fact, the literature finds that when an imitator copies a firm’s actions, these actions
make the imitator and the firm more similar, and, as a result, their profitability is likely to
converge (Posen & Martignoni, 2018), unless market leaders establish effective isolating
mechanisms that undermine imitation (e.g., Lippman & Rumelt, 1982; Posen et al., 2013;
On the other hand, companies may commit to sustainability to exploit a valuable but
superior or unique manner). For example, companies that adopt innovative circular-economy
business models or that embrace a unique corporate purpose that enhances employee
recruitment, engagement, and retention (Gartenberg, Prat, & Serafeim, 2019) are arguably
adopting a unique and difficult-to-imitate strategy that is “rooted in systems of activities that
are much more difficult to match” (Porter, 1996, p. 64). If successful, these efforts could even
could remain unmatched even when it is open to public scrutiny, and it can be slow to diffuse
imitation and the apparent tension between the two approaches. Sustainability conceptually
encapsulates a wide range of actions across the environmental, social, and governance
domains, generating high levels of variation across industries given that the same action may
be associated with differentiation advantage in some industries but not others. Accordingly,
there are valid reasons to expect that the rate of imitation will likely differ across industries
4
due to important differences in industries’ structural and competitive features (e.g., Romeo,
likely to determine how easy or how difficult they are to be imitated, especially when such
In this study, we use data from MSCI ESG Ratings, the largest provider of ESG data
globally, and find that for most industries, intraindustry convergence increases over time,
sustainability actions during the sample period. Moreover, we find that interindustry variation
in the rate of convergence is significantly associated with (a) the relative importance of
environmental and social issues relative to governance issues and (b) the level of stakeholder
scrutiny that the industry is facing, as measured by the tone and volume of media coverage.
Because of the granular nature of the MSCI ESG dataset, we also show that sustainability
practices are more likely to be imitated under conditions of low regulatory uncertainty, i.e.,
when policies and regulations are identified as the main drivers of the focal action. And we
find that actions with a high degree of novelty are less likely to be imitated, because they
require superior firm capabilities in the use of novel knowledge and technologies.
Within industries, we identify the set of sustainability actions that have been imitated
over the sample period—which we term “common practices”—and those that have not—
which we term “unique practices.” Unique actions, we argue, are not necessarily valuable; it
could be that managers are making an error in diverging from common industry practices.
Our findings show though that unique sustainability actions are significantly and positively
associated with both accounting measures of performance and market valuations, even after
controlling for past performance. Consequently, our work provides evidence that unique
sustainability actions become and remain strategic even when companies are faced with
5
strong imitation pressures. In contrast, common sustainability actions are not significantly
We make three novel contributions to the literature. First, to the best of our
knowledge, ours is the first study to adopt a dynamic approach towards understanding how
advantage when faced with imitation pressures. We offer empirical evidence characterizing
the convergence of sustainability actions within industries, the factors associated with
interindustry variation in the convergence rates, and the characteristics of actions associated
with the rate of imitation. We therefore contribute to the voluminous literature that explores
the link between sustainability and financial performance (see Aguinis and Glavas, 2012 for a
review). Second, we contribute to the stream of work that distinguishes between different
types of sustainability actions (e.g., Hawn & Ioannou, 2016; Durand, Hawn, & Ioannou,
2019) by not only proposing a new typology that links the dynamics of imitation with firm-
level decision-making but also, by identifying action-level characteristics that explain why
some actions are more likely to remain hard to imitate and thus, potentially valuable, forming
Third, although several studies have explored the dynamics of imitation within a
particular industry, we make a novel contribution to the imitation literature by exploring the
factors that affect heterogeneity in the rate of imitation across industries, an issue that has not
been sufficiently explored to date. We also contribute to this literature by identifying action-
level characteristics associated with the rate of imitation. We are able to do so precisely
because the sustainability context enables us to explore an entire set of corporate actions that,
in recent years, is strategically relevant for all industries globally (Ioannou & Hawn, 2019).
6
2 DATA AND SAMPLE CONSTRUCTION
We obtain data on sustainability actions from MSCI ESG Ratings as a proxy for the capital
market’s view on sustainability, given that these ratings are the ones most widely used by the
investment community.5 Of the 50 largest asset managers in the world ranked by assets under
management, 46 are clients of MSCI ESG ratings according to MSCI, with the total number
of clients reaching more than 1,200 investment firms globally. MSCI defines the purpose of
their ratings as “to help investors to understand ESG risks and opportunities and integrate
these factors into their portfolio construction and management process.” MSCI’s coverage
universe is based on major MSCI indices (e.g., MSCI World Index, MSCI Emerging Markets
Index, MSCI country-specific Investible Market Indices) that include the world’s largest and
MSCI ESG ratings are based on 37 “Key Issues,” each of which corresponds to one of
10 macro themes that MSCI identifies as being of concern to investors: climate change,
natural capital, pollution and waste, environmental opportunities, human capital, product
behavior. Key issues are annually selected for each of the 156 GICS subindustries and
weighted according to MSCI’s materiality mapping framework. Each Key Issue score
each material issue. For a given Key Issue score, the required risk management component
score is conditional on the risk exposure faced by the company; that is, a company with a
greater risk exposure to an issue would be required to have stronger risk management
practices in place. Conversely, a company with minimal actions on a low exposure risk issue
5
The full methodology for MSCI ESG ratings is here: https://www.msci.com/documents/10199/123a2b2b-
1395-4aa2-a121-ea14de6d708a (last accessed April 12, 2021).
7
would not be penalized. For Key Issues that measure opportunity (e.g., Opportunities in
Access to Health Care), the exposure signifies the relevance of this opportunity to the
information about company-specific actions with Key Issue–relevant macro-level data related
are sourced from corporate reporting, such as annual reports, investor presentations, and
financial and regulatory filings, and macro-level data are sourced from a wide variety of
related data come from corporate documents, government data, news media, relevant
organizations and professionals, and an assortment of popular, trade, and academic journals.
As part of their data verification process, MSCI communicates with companies directly and
invites them to participate in a data review process, which includes commenting on the
accuracy of company data for all MSCI ESG Research reports. MSCI aggregates corporate
actions to an overall score (Susty) whereby each Key Issue is weighted according to its
assessed materiality for each industry. The score ranges from 0 to 10, with 0 (10) being the
evolves over time, we first restrict our sample to companies that appear in the MSCI dataset
consistently across all years6 and obtain a final sample comprising 2,095 firms for the period
2012–2019. We choose 2012 as the starting period of our sample for multiple reasons. First,
6
We use a firm’s ISIN as the indicator to track presence over time. If a firm’s ISIN changes over time, then we
drop that firm from the sample even though it is covered by MSCI. However, in most cases these are firms that
have undergone significant corporate restructuring, such as a merger, and therefore their sustainability actions
might be changing because of that restructuring. Therefore, we would drop these firms from the sample either
way, as the change in sustainability actions might be driven by changes in their business operations, industrial
membership, and so on.
8
to avoid any influence from entry and exit of firms on the convergence measures, we require
them to appear consistently over the years covered in our sample. Starting in 2012, MSCI
coverage includes more than 6,000 companies; yet, for 2011, the database covers only about
1,200 firms, while for earlier years the coverage is merely 1,000 firms. Hence, imposing the
requirement of a stable sample of firms while including years before 2012 would have left us
with a very limited sample of about 200 to 300 firms. Nevertheless, we also report results
using an unbalanced panel of firms and expand our sample to cover the period 2006–2019.7
2013; Csaszar & Siggelkow, 2010; Ethiraj & Zhu, 2008; Nelson & Winter, 1982; Rivkin,
2000). A number of theoretical traditions have suggested different processes to describe what
Hirshleifer, & Welch, 1992) and mimetic isomorphism (DiMaggio & Powell, 1983). Indeed,
companies imitate the practices of others within the same social reference group,8 including
In this study, we predict a high degree of imitation of sustainability actions over time
and, hence, an increase in intraindustry convergence, for several reasons. First, existing work
finds that suppliers, customers, alliance partners, and the public collectively exert pressures
Husted, Jamali, & Saffar, 2016). Relatedly, a recent stream of research has produced
evidence for the “business case for sustainability” (e.g., Eccles et al., 2014; Flammer, 2015),
while other studies have shown that, at the very least, sustainability actions can mitigate risk
7
We note that for the financial performance analysis, including years before 2012 is not possible because 2012
is the earliest year for which MSCI provides historical data at the action level, which is required for the
construction of our key independent variables. Before 2012, MSCI provides historical data for an overall ESG
score and for separate environmental, social, and governance scores.
8
For a comprehensive review, see Lieberman and Asaba (2006).
9
by acting as an insurance-like protection for the relationship-based intangible assets of a
company (Godfrey, 2005; Ioannou & Serafeim, 2015). As a result, we argue that companies
are more likely to imitate peers’ sustainability actions because they perceive them to be
performance enhancing.
This argument is consistent with the information economics literature (e.g., Banerjee,
1992) that conceptualizes imitation as herding resulting from observational learning, which in
turn is defined as “the influence resulting from rational processing of information gained by
observing others” (Bikhchandani et al., 1998, p. 153). Imitation occurs because economic
actors place more weight on the information that others (e.g., industry peers) have gathered
and subsequently revealed through the actions they chose to undertake. To the extent that
industry peers are perceived to have better information (e.g., as revealed through superior
performance), companies will opt to imitate them rather than make their choices based
exclusively on their own, limited information (Gaba & Terlaak, 2013; Wu & Salomon, 2016).
cultural-cognitive pressures (i.e., isomorphism; Meyer & Rowan, 1977; Oliver, 1991, 1997).
The main idea here is that companies face coercive, normative, and mimetic pressures to
engage in isomorphism (DiMaggio & Powell, 1983). We suggest that all three types of
pressures exist in the sustainability domain. First, as discussed just above, the emergence of
the business case generates mimetic pressures on companies to undertake more sustainability
actions. Second, the literature shows that sustainability is associated with the emergence of a
new institutional logic that has gradually weakened the dominant agency logic in public
equity markets, impacting analysts’ recommendations (Ioannou & Serafeim, 2015) and even
lowering capital constraints for sustainable companies (Cheng et al., 2014). These normative
pressures are also reinforced by the broader momentum within capital markets toward
10
integrating sustainability data in investment decisions, in combination with the emergence of
numerous information intermediaries (MSCI, among others) that rate and rank companies
Third, companies face coercive pressures given that powerful social actors, including
governments, global organizations, and activist investors, are pushing companies to take
more actions to address sustainability challenges (e.g., via laws, regulations, and/or listing
requirements). For example, the number of environmental and social issues that are the
subject of shareholder resolutions in the United States has been significantly increasing
(Carroll, Lipartito, Post, & Werhane, 2012; Glac, 2010), and these resolutions have become
the wide range of environmental and social actions that companies adopt across industries.
Thus, unlike prior studies that focused on the diffusion of specific innovations (e.g.,
particular product or process innovations) in one industry, our focus on sustainability not
only allows us to study imitation at the industry-action level but also, to explore interindustry
variation in the rate of imitation; two questions that have not been sufficiently addressed to
date. In sum, for all the above reasons, we predict a high degree of imitation of sustainability
We follow prior literature (e.g., Cheng, Ioannou and Serafeim, 2014; Hawn and Ioannou,
2016; Hawn, Chatterji and Mitchell, 2018) and identify the industry as the relevant social
9
Furthermore, the signing of the Paris Agreement, in 2015, in combination with the adoption of the UN
Sustainable Development Goals have resulted in a powerful message and a renewed momentum toward the
adoption of socially and environmentally oriented laws and regulations by a large number of countries around
the world. A number of national governments (e.g., South Africa, Malaysia, China, EU, Brazil) have already
mandated that large companies disclose not only financial data but also data on their sustainability actions (i.e.,
ESG/nonfinancial data).
11
reference group because (a) companies typically benchmark their sustainability actions and
performance against competitors in the same industry, (b) accounting standard setters such as
standards, (c) investors use “best-in-class” approaches, whereby they seek to understand a
firm’s portfolio of sustainability actions relative to the industry’s practices, and (d) practically
all ESG data providers produce industry-adjusted sustainability scores (Amel-Zadeh &
Serafeim, 2018; Khan et al., 2016). This is also consistent with existing literature that has
used the same product market or industry as the relevant reference group (e.g., Lieberman &
Asaba, 2006; Marquis & Tilcsik, 2016; Sharkey & Bromley, 2015).
construct the coefficient of variation. Specifically, for each industry j in year t, we calculate
the coefficient of variation as CVjt = jt / jt, where jt is the standard deviation of Susty
across all firms in year t and industry j and jt is the average of Susty across all firms in year t
and industry j.10 To formally test whether there has been convergence over time, we estimate
where Timetrend is a variable that increases by 1 for each year after the first year in the
sample. The estimated coefficient b is our estimate of convergence whereby the more
negative b is, the faster the CV is declining over time. We include industry fixed effects to
allow the intercept to vary by industry, and, as a result, we effectively estimate b using only
within-industry variation. Standard errors are robust to heteroscedasticity and clustered at the
industry level.
10
As a reminder, Susty is the overall sustainability actions score generated by MSCI, as described in the Data
and Sample Construction section.
12
Table 1, Panel A presents the results obtained through five different models. In
Models 1 and 2, CV is constructed using a balanced set of firms across all years, thereby
mitigating any changes in CV over time that may be caused by changes in the underlying
sample. In Model 1, both the independent and the dependent variables are log-transformed;
thus, the estimated coefficient b is interpreted as the elasticity of CV with respect to time. In
Model 2, both the dependent and the independent variables are linear; therefore, the estimated
coefficient b is the change in CV for a one-year change. Models 3, 4, and 5 are variations of
Model 1 using an unbalanced sample of firms and extending the starting year of our sample
from 2012 back to 2009 and then even further back to 2006. We present these results to
complement the findings of the main Model 1, as they include more firms and more years.
This is why the sample size increases from 509 industry-year observations to 523, 685, and
826 in Models 3, 4, and 5 respectively. However, these models also have weaknesses in that
sample.11 Also, we note that before 2012, the sample of firms with available data declines
fast. Whereas in 2012 we have data on 6,085 firms, in 2009 we have data on 1,049 firms, and
The estimated coefficient b is negative and significant across all models. This
suggests that, on average, the CV has systematically declined across most industries. The
estimates in Models 1 and 3 suggest that the elasticity is close to -0.3. In Model 2 the
coefficient can be interpreted as a decline by 1.72 in the CV for every additional year that
passes. The coefficient is still negative and significant but reduced in magnitude in Models 4
and 5, either because adding earlier years to the sample is essentially adding noise to the
11
To mitigate this effect, we require a firm in our unbalanced sample to be covered by MSCI for more than four
years during the sample period.
13
estimation (given that the sample declines significantly), and/or because most of the observed
convergence occurred after 2012. In untabulated analysis, we split the sample to roughly
three equal subsamples to better understand the role of home country characteristics: US
firms, firms from other developed markets, and firms from emerging markets. We find
significant convergence in all three subsamples with the US being the highest, followed by
emerging markets and then other developed markets. Moreover, we confirm that the decline
is driven by a decrease in the standard deviation rather than an increase in the mean (i.e., the
distribution of sustainability actions becomes more concentrated over the sample period).
Despite this remarkable convergence in sustainability actions within industries, Table 1 Panel
B also shows that there is significant variation in the rate of convergence across them. This
observation resonates with early studies in the diffusion of innovation literature (e.g., Romeo,
propose that such variation is associated with structural and competitive features of industries
considering that different industries also face different sustainability pressures. Accordingly,
in this section, we explore the industry-level factors associated with interindustry variation in
the level of convergence. Specifically, we investigate three potential factors: (a) the adoption
of sustainability actions by an industry’s market leaders, (b) the importance of the underlying
environmental and social issues relative to governance issues, and (c) the degree of
Companies are more likely to imitate the actions of industry peers when they perceive such
actions to enhance performance (e.g., Banerjee, 1992; Bikhchandani et al., 1992; DiMaggio
& Powell, 1983; Oliver, 1991). When firms observe their peers achieving superior
14
performance, they assume that they have an informational advantage regarding the
profitability of a given strategy. Given that gathering information through observing peers’
actions may be relatively more cost effective than deploying limited organizational resources
to evaluate the strategy itself (e.g., Conlisk, 1980; Lieberman & Asaba, 2006), companies are
more likely to imitate the actions of the most successful companies in their industry (e.g.,
proportion of market leaders choose to adopt sustainability actions because the undertaking
Empirically, the variable Leader captures the percentage of large firms in an industry
(i.e., top quartile in market capitalization) that are also sustainability leaders (i.e., top decile
in terms of Susty in each industry).12 The variable ranges from 0 to 1 and has a mean of 30%
and a standard deviation of 27% (Table 2 Panel A), indicating that across all industries, on
average, 30% of large companies are also leaders in undertaking sustainability actions.
The type of underlying environmental and social issues that an industry faces could also
explain part of the interindustry variation in the convergence rate given that different
sustainability issues may be relatively more or less important as a function of the resources
and the impact of an industry on society at large. For example, while climate change is a
universal societal challenge, scope 1 carbon emissions in particular are a key issue for
12
Defining Leader using a quartile or quintile threshold for both market capitalization and Susty yields
qualitatively similar results.
15
companies in the electric utilities industry but relatively less important for companies in the
financial industry. Through such an approach, SASB has classified sustainability issues as
“material” by applying the SEC definition of materiality as interpreted by the U.S. Supreme
Court, whereby the standard for materiality reflects “a substantial likelihood that the
disclosure of the omitted fact would have been viewed by the reasonable investor as having
significantly altered the ‘total mix’ of information made available.” 13 In line with this
definition of materiality, data and rating providers, including MSCI, estimate and apply
different weights on environmental, social, and governance issues for different industries.
environmental and social issues are assessed as more material and thus, are assigned greater
long stream of work documents that throughout most of the 1990s and 2000s, investors,
regulators, and stock exchanges strongly pressured companies to adopt a wide range of best
governance practices (Bebchuk, Cohen, & Ferrell, 2009; Gompers, Ishii, & Metrick, 2003) as
convergence on governance actions has likely occurred prior to the sample period.
On the other hand, institutional pressures associated with environmental and social
issues, and the resulting demands for the adoption of environmentally and socially
responsible actions, are a more recent phenomenon (Amel-Zadeh & Serafeim, 2018).
Similarly, while investor activism on governance issues through private dialogues and
shareholder proposals has been a frequent phenomenon for decades—and a long literature
13
TSC Industries v. Northway, Inc., 426 U.S. 438, 449 (1976). See also Basic, Inc. v. Levinson, 485 U.S. 224
(1988).
16
environmental and social issues is also a more recent phenomenon (e.g., Dimson, Karakaş, &
scores are taken directly from MSCI and they are the weights placed annually on
environmental, social, and governance issues respectively, for each industry. The sum of the
three weights is always 100%; therefore, including all three in the models generates perfect
multicollinearity. For this reason, and to be consistent with our theoretical expectations, the
Governance Materiality score is the omitted category. These weights vary little over time
within the period of our study therefore, we calculate the average weight across the sample
period. Environmental Materiality has a mean of 35.60% and a standard deviation of 18.12%,
and Social Materiality has a mean of 40.84% and a standard deviation of 14.07% (Table 2
Panel A). This suggests that the omitted category, Governance Materiality, would have a
mean of approximately 24% given that the three categories always add up to 100%.
Not only are companies more likely to respond to sustainability issues of greater salience
(e.g., Bundy, Shropshire, & Buchholtz, 2013; Darnall, Henriques, & Sadorsky, 2010;
Kassinis & Vafeas, 2006) but they are also more likely to do so under conditions of increased
scrutiny by the media and other important stakeholders (e.g., Bartley, 2007; King, 2008; Luo,
Zhang, & Marquis, 2016). However, whether companies will converge or diverge on the
actions they decide to take arguably remains an open question. On the one hand, if companies
Marcus & Vogel, 2020 for a review) and in so doing, to promote the adoption of the same
14
Meanwhile, institutional dependencies will likely restrict further intraindustry convergence on governance
issues—provided that governance is a cultural as well as a legal construct (Aggarwal, Erel, Ferreira, & Matos,
2011).
17
practices throughout the industry, and if these self-regulation efforts are effective, then
cognitive assessments, and stakeholder relationships and reputations, then their actions are
likely to converge by less (or even diverge) (e.g., Durand et al., 2019; Lee, 2011). Further,
responses to environmental and social issues and could also generate divergent responses
(Pedersen & Gwozdz, 2014). Meanwhile, perceived pressures by primary versus secondary
stakeholders could also differentially influence firms’ responses (Buysse & Verbeke, 2003).
Another key dimension that is relevant to consider is the tone of such coverage of the
industry. We expect that external reinforcement in the form of positive coverage will likely
because sustainability actions are more likely to be legitimized through positive (stakeholder)
feedback (e.g., Marquis, Glynn, & Davis, 2007; Sharma & Vredenburg, 1998) thus
distance themselves from the criticism that the industry, and their peers, are facing (Hawn,
To measure stakeholder attention and the tone of media coverage, we use data from
TruValue Labs which is used by some of the largest financial institutions (e.g., Citi) and asset
owners (e.g., Global Pension Investment Fund of Japan) around the world (Serafeim, 2020).
TruValue Labs employs big data and artificial intelligence approaches to capture and analyze
18
for each company, from a wide variety of sources, including reports by analysts, various
media outlets, advocacy groups, and government regulators.15 Using natural language
processing, TruValue Labs interpret semantic content and generate analytics scoring data
points using a 0–100 scale. A score of 50 represents a neutral impact. Scores above 50
indicate positive sentiment, and scores below 50 reflect negative sentiment.16 The sentiment
analysis is capable of codifying not only positive versus negative sentiment in a binary way
but also, degrees of positivity or negativity. For example, the algorithms assign a relatively
more negative score to an accident affecting several workers and communities and a less
negative score to a workplace incident that leads to an injury for one worker (Serafeim,
2020). We construct the variable Attention as the average number of ESG-related articles
across all companies within the same industry, and we construct the variable Tone as the
average tone of ESG-related articles across all companies within the same industry, whereby
Convergence
We estimate the following cross-sectional model using ordinary least squares (OLS):
where Convergence is the estimated coefficient b on the Timetrend variable from equation
(1), but model (2) is run separately, for each industry, allowing us to estimate an industry-
15
TruValue Labs emphasizes that its measures focus on vetted, reputable, and credible sources that are likely to
generate new information and therefore insights for investors. To increase transparency and to validate the data,
the TruValue Labs platform allows a user to track the original source of the articles and events that inform the
sentiment analysis for each specific issue (Serafeim, 2020). The platform aggregates unstructured data from
over 100,000 sources for monitored companies.
16
For example, Ingersoll Rand had positive sentiment following news on the firm’s investments to improve
waste and hazardous materials management, materials sourcing, and product safety. In contrast, Facebook had
negative sentiment following news on the firm’s data privacy issues, concerns about regulatory pressure, and
user rights.
17
To be consistent with our sample composition we use data for all firms that have coverage in MSCI from the
unbalanced sample and TruValue Labs.
19
specific coefficient b as tabulated in Table 1 Panel B. We include all the industry
characteristics that we are interested in—and discussed above—as explanatory variables and
industry characteristics that might be correlated both with the rate of Convergence and with
Table 2 Panel A presents summary statistics, and Panel B presents the correlation
matrix for all the variables we use in the specifications. Convergence % (estimated from the
log-log timetrend model) and Convergence Linear (estimated from the linear timetrend
model) exhibit a high positive correlation of 0.76, as expected. We report the main results
check. We include variables controlling for average size, profitability, and valuation
multiples across firms within each industry and we control for the level of competition using
the Herfindahl index. We also control for the geographic composition of all the firms within
each industry. Convergence % has an average of −0.28 and a standard deviation of 0.15.
we use to construct our industry-level variables. As expected, most firms originate from the
U.S., Japan, EU excluding UK, and the UK. In Panel B, we observe Convergence %
exhibiting a negative correlation with both Environmental and Social Weight, Leader, and
Table 3 Panel A presents OLS models where the dependent variable is Convergence
% (Models 1–4 and 6) or Convergence Linear (Model 4). In models 1-6 the coefficient of
variation is constructed from a balanced sample over the years between 2012 and 2019.
Model 1 does not include any control variables to assess their influence on the estimated
18
These control variables are constructed using the unbalanced sample of MSCI coverage.
20
coefficients of interest given the small number of observations in the model. Models 4
through 6 also include the geographic composition variables tabulated in Table 2 as control
variables. These variables collectively cover approximately 93% of the sample and we
include them to account for the possibility that different industries might be composed of
companies from different home countries. Furthermore, in all the models of Table 3 Panel B,
the coefficient of variation is constructed from an unbalanced sample of firms with MSCI
coverage that varies across years. Specifically, Models 1-2, 3-4 and 5-6 estimate the
coefficient on the timetrend variable based on the periods 2012 to 2019, 2009 to 2019 and
2006 to 2019, respectively. Models 2, 4 and 6 include the geographic composition controls.
The results based on the balanced sample regressions can be summarized as follows.
The rate of convergence is higher in industries in which environmental and social issues are
more material, consistent with our theoretical predictions. The rate of intraindustry
convergence is also higher for industries that receive less attention from stakeholders. We
conjecture that this could be because industries that receive more stakeholder attention are
less effective at enforcing self-regulation across member companies and thus, are less likely
to adopt similar sustainability actions. For example, more stakeholder attention may come
with increased complexity as well as conflicts within and across stakeholders, impeding
The estimated coefficient on Tone and Leader do not achieve significance in Model 1
whereas Tone obtains significance in Model 2. An interesting question arises about the
interaction between these two variables. It could be that the effect of more positive tone on
21
do obtain a significant coefficient on an interaction term between Leader and Tone.19 This
result implies that an industry experiences a higher rate of convergence when more of its
market leaders become sustainability leaders and they are externally supported by positive
media coverage. We cautiously interpret this finding as follows: sustainability actions are
characterized by higher levels of uncertainty and can sometimes even be seen as contested
practices (Moon et al., 2005). Therefore, observing the most successful industry peers
performing well may be insufficient for companies to infer that sustainability actions are
performance-enhancing and thus, on their own, they are insufficient to trigger imitation. On
the other hand, when they are further legitimized by stakeholder endorsement in the form of
positive media coverage, then companies could make more reliable inferences about the link
between sustainability actions and the performance of the leaders, making it more likely that
We interpret the results of Table 3 Panel B with increased caution because they are
based on unbalanced panel data and consequently, the convergence measure could be biased
due to entry and exit of firms over time. Broadly speaking, the models confirm the findings
of Table 3 Panel A, and this is particularly true for Models 1 and 2, which are the most
comparable to the models of Panel A given that they are based on the same 2012-2019 time
period. We also note that as we add more years (and arguably more noise to the data),
19
We thank an anonymous reviewer for this suggestion. Relatedly, we note that we empirically explored other
interaction terms, none of which achieved significance.
20
In untabulated analysis we examined whether the stringency of the institutional environment of a Leader’s
home country yields different insights. For example, if leaders are coming from an industry from a stringent
environment, other firms might not be influenced as they might perceive the leader’s actions arising from
coercion rather than value creation. We reached out to the United Nations Principles for Responsible Investment
(PRI) and obtained access to their Global Regulation Database, which allowed us to construct a measure that
captures institutional variation across countries. Thus, countries were classified as Most Stringent if, during the
sample period, they had in place more than 4 regulations (top quartile) that were imposed by the government (as
opposed to being a listing requirement), that had been issued (i.e., not planned) and that refer to mandatory (as
opposed to voluntary) ESG provisions. The estimated coefficients on the main effect of Leader from stringent or
non-stringent institutional environments were both insignificant. The estimated coefficients on the interaction
effects with Tone for both Leader variables were negative and significant.
22
moving through Models 3 to 6, the sign of the estimated coefficients of interest remains
To explore how the higher levels of intraindustry convergence may be associated with
performance outcomes, we first focus on the industry-action level and investigate why is it
the case that some sustainability actions are less likely to be imitated than others over time.
The literature finds that when strategies are complex (Ethiraj & Levinthal, 2004; Levinthal,
1997; Rivkin, 2000), when knowledge is tacit (Zander & Kogut, 1995), when evaluation is
uncertain (Ethiraj & Zhu 2008; Greve, 2009), or when causal ambiguity exists (Lippman &
Rumelt, 1982; Ryall, 2009), imitation may be partial or prone to errors. As a result,
Specifically, we posit that when market uncertainty and regulatory uncertainty surrounding a
sustainability action is low, then companies are more likely to adopt them (i.e., a higher
degree of convergence). Importantly, we posit that actions with a high degree of novelty are
less likely to be imitated because they require superior firm capabilities in the use of novel
knowledge and technologies, which in turn, strengthen the isolating mechanisms through
which leading companies prevent imitation (e.g., Lippman & Rumelt, 1982; Posen, Lee, &
industry-action level to construct three indicator variables: (a) Low Market Uncertainty takes
the value of one if the across firm and years average of the weight on an action for all firms
23
in the industry is above the median;21 (b) Low Regulatory Uncertainty takes the value of 1 for
actions where policies and regulations are identified as drivers of the practice (e.g. anti-
financial product safety, employee health and safety, product safety and supply chain labor);
(c) High Practice Novelty takes the value of one for practices which require the use of novel
total number of pairs is 846, given that not all sustainability actions are applicable to every
industry according to the MSCI data. Table 4 Panel A presents summary statistics for
Convergence % across all actions and for each action individually. Within our sample, we
observe very low frequency of convergence on actions oriented toward data privacy, water
stress, human capital development, and product carbon footprint issues. In contrast, we
opportunities, labor-related supply chain, and employee health and safety issues.
Table 4 Panel B presents the results of models with and without industry fixed effects
in which the dependent variable is Convergence %. Consistent with our predictions, we find
that sustainability actions characterized by low regulatory uncertainty are more likely to be
imitated yet those characterized by a high degree of novelty are less likely to be imitated. The
coefficient on market uncertainty obtains a negative coefficient, which is consistent with our
expectations, but it fails to achieve significance.22 Taken together, these estimates suggest
interesting action-level heterogeneity, but they are also pointing towards the possibility that
21
Defining market uncertainty as a continuous measure yields similarly insignificant results.
22
Clustering standard errors at the industry rather than at the practice level makes the coefficient on Low Market
Uncertainty significant at the 10% level.
24
actions that are less likely to be imitated, even if valuable, may be the reason why companies
actions: specifically, we define common sustainability actions as those actions that were
imitated over time at the industry level—that is, the set of actions on which an industry has
converged—and we define as unique those on which the industry has not converged. We
conjecture that common (i.e., imitated) actions will not be associated with superior
performance because, more generally, when actions diffuse via imitation then performance
heterogeneity within an industry diminishes (Dierickx & Cool, 1989; Lippman & Rumelt,
1982), narrowing the performance gap between market leaders and their imitators (Shenkar,
2010; Knott, Posen, & Wu, 2009) by eroding the profits of firms that are being imitated
(Posen et al., 2013). For example, a commitment to reduce carbon emissions at the beginning
of our sample could have translated into eco-efficiency initiatives that reduced cost for
companies and allowed them to maintain a cost advantage for as long as industry peers
remained unable or unwilling to adopt a similar commitment. Over time, however, peers may
have observed or inferred that such initiatives have a positive impact on the bottom line,
triggering a process of imitation. As the industry converged on these actions, companies that
failed to undertake them would be faced with a cost disadvantage. It is precisely because of
such a process, then, that we argue that common actions are not likely to be associated with
outperformance.
In fact, we expect that the mere adoption of common actions will likely not be
associated with a range of potential benefits: for example, imitated actions are unlikely to
generate product market benefits because customer loyalty and satisfaction are typically
based on differentiating practices (Bhattacharya & Sen, 2003; Luo & Bhattacharya, 2006).
25
Similarly, labor market benefits derived from employee satisfaction, recruitment, and
engagement (Bode et al., 2015; Burbano, 2016) and capital market benefits in the form of
cost of capital effects (Cheng et al., 2014) are also based on the adoption of differentiating
actions. In other words, common sustainability actions will likely fail to lead to
differentiation and therefore, decision-makers (e.g., labor, product or capital market actors)
On the other hand, we maintain stronger priors that unique actions can be associated
with superior performance. This is because a unique strategy may be based on the ex-ante
and which, in turn, is protected via actions that are distinct, enabling a firm to enjoy superior
performance due to lower levels of competition or even a local monopoly (Baum & Mezias,
1992; Baum & Singh, 1994; Porter, 1980, 1991). In our setting, however, it is an open
position with superior profitability potential. It may well be the case that such actions do not
correspond to valuable underlying positions; they might even be strategic errors. Yet the
findings in Table 4 Panel B do imply that it is novel sustainability actions that remain hard to
imitate over time and therefore, to the extent that the use of novel knowledge and
starting with our primary metric, a frequently used performance metric in strategic
management (e.g., Obloj and Sengul 2020), i.e., return-on-capital (ROC). We also use return-
on-assets (ROA) and its decomposition to return-on-sales (ROS) and asset turnover and we
26
run our models using a market valuation-based value creation metric, that is also frequently
used in strategic management research (e.g., Hawn and Ioannou, 2016), i.e., Tobin’s Q.23
The two independent variables of interest are measures of firm-level common and unique
more negative than −5%, which is the elasticity of change in the coefficient of variation for a
percentage change in years. Therefore, we classify them as common if they have converged
by at least 5% each year; else we classify them as unique. The last column of Table 4 Panel A
shows the number of industries for which the focal action is classified as unique using this
indeed a significant extent of convergence during the sample period rather than a small, and
potentially, random decline. We also increase and reduce the threshold to −10% or 0%. We
note that a sustainability action, for example human capital management, may be classified as
common in one industry and as unique in another, given that the way we construct the
We then create indicator variables that take the value of one or zero based on whether
sustainability action m for industry k is common (Ckm) or unique (Ukm). We measure the
undertaking of common sustainability actions for firm i as the weighted24 score of common
actions in year t:
Common Actionsit =
23
To eliminate the influence of large outliers we windsorize the dependent variables at the 1 and 99% level or at
the closest level in the presence of extreme outliers. In our sample, those values are for ROC at 50% and -30%,
ROS at 50% and -40%, asset turnover at 3 and 0, and Tobin’s Q at 9 and 0.5.
24
We weigh each practice using the MSCI weights given to each practice for each firm-year. To preserve the
relative scaling of both variables from 0 to 10 consistent with the MSCI scores, we rescale the weights based on
the total weight allocated to the common or unique actions. Therefore, the total weights within the set of
common actions sums up to 100%, and the same is the case for unique actions.
27
Equivalently, for unique actions the variable is calculated as:
Unique Actionsit =
Both variables range between zero and ten (Table 5 Panel A) and interestingly, there is very
little correlation between the two (0.045) (Table 5 Panel B). Our final estimation model is:
Effects (5)
We include year, country, and industry fixed effects as well as the lagged level of the
dependent variable, measured as the average between 2010 and 2011. Doing so allows us to
control for a firm’s financial performance prior to the period for which we measure the
dependent variables and to control for a set of stable factors that may impact a firm’s
also include firm fixed effects, by demeaning all variables using the average of each variable
at the firm-level, in addition to country and industry fixed effects, to assess the robustness of
our main results. We further include a set of time-varying firm characteristics as controls, as
they might be correlated with either sustainability actions or financial performance. This
includes research and development expenditures scaled by sales, capital expenditures scaled
measured as total debt over total assets.25 If some of these control variables are themselves
affected by the adoption of sustainability actions, we would essentially be controlling for the
actions towards zero. The coefficients of interest in equation (5) are d and e. The sample
25
Consistent with prior studies (Khan, Serafeim and Yoon 2016) if research and development expenditures are
missing then we assume that they are zero. Capex are scaled to range from 0 to 50% of assets, R&D from 0 to
100% of sales, and debt from zero to 100% of assets.
28
includes all firms with available data in MSCI and Worldscope, producing a final sample of
close to 39,000 observations; the number of observations varies somewhat depending on the
Table 6 reports results of OLS models using accounting measures of performance, and Table
7 reports results using Tobin’s Q. Models 1–4 include industry and country fixed effects,
time-varying firm controls, and the lagged dependent variable.26 Across all specifications, the
Common Actions is insignificant. A 4-point increase typically moves a firm from being a
laggard (i.e., a score of 3 or below) to a leader (i.e., a score of at least 7) on the MSCI scale.
Such a move translates to approximately 1%, 0.46%, 1.2%, and 2.6% higher ROC, ROA,
ROS, and asset turnover respectively; all are economically meaningful changes in operating
performance.
the dependent variable is the coefficient of variation of actions across firms, is less (more)
than −5%. In Models 5 and 6, actions are classified as unique (common) if the estimated
coefficient is less (more) than 0 or −10% respectively. Comparing Model 1 to Models 5 and 6
confirms that the choice of the exact threshold makes little difference to the results.
1-year lag of the dependent variable instead of the average between 2010 and 2011. This is a
more conservative specification because it arguably controls for some of the effect we are
26
As expected, the lagged dependent variable is very significantly correlated with the dependent variable and
explains a large part of the variation; this finding is well documented in the literature on the persistence in
accounting profitability (Healy, Serafeim, Srinivasan, & Yu, 2014)
29
attempting to identify, given that the lagged level of the dependent variable now incorporates
some of the likely performance effects that may be attributed to unique actions.27 As
expected, the estimated coefficient on Unique Actions declines from 0.251 to 0.163 but
remains significant. The estimated coefficient on Lag Operating Performance jumps from
0.418 to 0.675 which is to be expected given that in this specification, Lag Operating
Performance is measured at the immediately preceding year and that accounting performance
persistence is higher for first rather than higher-order autocorrelation. Model 8 identifies the
coefficients on sustainability actions using only within-firm variation, and Model 9 includes
the time-varying firm controls. In both models, the estimated coefficient on Unique Actions is
specific models, where the dependent variable is the coefficient of variation of actions across
firms, is less (more) than −5%. In models 3 and 4, practices are classified as unique
(common) if the estimated coefficient is less (more) than 0 or −10% respectively. Lag
Tobin’s Q is measured as the average of the dependent variable of each model in years 2010
and 2011 for models 1 through 4. Lag Tobin’s Q is measured as the 1-year lag of the
dependent variable in Model 5. These specifications closely resemble Table 6. Across all of
them, we find a positive and significant association between Unique Actions and Tobin’s Q.29
27
For example, when the dependent variable is return-on-capital in 2018, the lagged dependent variable is
measured in 2017, thereby picking up any positive effects from unique actions undertaken between 2017 and
2018.
28
When we run the same analysis for ROC on a sample of only US firms, we find significant positive estimated
coefficient on Unique Actions both using across-firm and within-firm variation estimates. In a sample of firms
only from developed markets outside the US we find significant positive estimate using within-firm variation,
while for emerging markets we find significant positive estimate using across-firm variation. In all
specifications the estimated coefficient on Common Actions is insignificant.
29
We log-transform Tobin’s Q in the models of Table 7 to mitigate skewness in the variable, consistent with
past research (Gartenberg et al., 2019), but our results are similar using the raw variable.
30
contrast, the coefficient on Common Actions is insignificant in most of the specifications and
unstable, as it takes a negative value in Model 2 and a positive value in Model 5. As in Table
6, controlling for time-varying firm characteristics turns the coefficient on Common Actions
negative and significant. This is primarily driven by the positive association with firm size,
which is positively correlated with Tobin’s Q. Arguably, controlling for firm size could be
problematic when the dependent variable is Tobin’s Q, as market value is in the numerator of
This study is not without its limitations, but we do hope that these limitations will inspire
future research. First, admittedly, our estimation process does not involve a natural
experiment with random assignment. Therefore, our results provide evidence of association
but not necessarily causality. Nevertheless, we do attempt to control for several variables that
may threaten the validity of our inferences. The first threat would arise if a time-invariant
firm characteristic is correlated with both the sustainability actions variables and performance
at the end of the sample period. For example, it could be that firms that are more profitable
and have higher market valuation multiples choose to pursue unique sustainability actions.
Yet our models explicitly control for the beginning of sample-period financial performance,
thereby controlling for time-invariant firm characteristics that are influencing the level of
performance at the beginning of the sample period too. Therefore, this concern is mitigated
A further limitation is the threat to the validity of our estimates caused by a time-
varying firm characteristic that operates as a correlated omitted variable. Yet it is not easy to
identify such a variable that would also have a differential impact across unique and common
sustainability actions. Since these are defined at the industry level, as industry–action pairs –
and they are thus outside the control of any individual firm – they posit a higher hurdle for
31
correlated omitted variable bias. Essentially, to impact our findings, it should be the case that
industries where these actions are not undertaken by other firms. At the same time, this same
time-varying firm characteristic should not be correlated with firms’ sustainability actions in
industries in which these actions are undertaken by other firms. When we do introduce
additional time-varying control variables in our models however, our results are robust.
Another limitation is that the adoption of unique but not common sustainability
are not yet reflected in measures of accounting performance or market valuation multiples.
That is, firms may be adopting sustainability actions that other firms in the same industry do
not because their performance will improve over time. We note this argument could be valid
for the adoption of unique sustainability actions (i.e., as a ‘luxury good’), but less so for
common actions, that are arguably less sensitive to a firm’s performance. While certainly
plausible, this explanation requires that management has visibility over future increases in
financial performance over the next few years and that they are sufficiently certain so as to
pursue unique sustainability actions. However, evidence that management has difficulty even
predicting near-term earnings raises significant doubt that managers would adopt unique
Similar to many studies in the sustainability space, our work is also limited by its use
of particular datasets. Thus, even though the MSCI ESG ratings are widely used in the asset
management industry, the data is available for the set of firms that MSCI covers, which are
typically larger, public companies around the world. In addition, there are some apparent
discrepancies across ESG data providers in the construction of ESG datasets that should be
taken into consideration (e.g., Chatterji, Durand, Levine and Touboul, 2016). We hope that in
32
future work, scholars will seek to extend our findings by using other data sources, and in fact,
sources that would cover other types of firms, including private ones, and small to medium-
sized enterprises.
Finally, in our study, we propose and empirically test a set of measures to capture
convergence at the industry- and at the industry-action level, as well as measures to capture
how unique (or common) the overall strategy of a focal firm is. These measures are likely to
be imperfect and may not fully capture the real-world environmental and social impacts of
companies (Freiberg et al. 2020). As such, our results should be interpreted with caution in
In this paper, we explore the adoption of sustainability actions in recent years and pose a
fundamental question for strategic management: under what conditions are companies able to
pressures by industry peers? Our findings reveal that globally, within most industries,
companies undertook an increasingly similar set of sustainability actions over the past
decade. Moreover, we find that the interindustry variation in the degree of imitation of
environmental and social issues and variation in the level and tone of stakeholder attention
that industries face. By considering variation at the industry-action level, we find that actions
are more likely to be imitated when they are characterized by low regulatory uncertainty and
that practices with a high degree of novelty are less likely to be imitated. We then utilize our
the firm level and provide evidence that, in fact, unique actions become and remain strategic
during the sample period: they are reliably, consistently, and significantly associated with
financial performance. Therefore, we show that even though sustainability actions are
33
increasingly imitated within industries, some companies can maintain their competitive
With our work we make three novel contributions to the literature. First, to the best of
our knowledge, this is the first study to empirically explore the intraindustry convergence of
sustainability actions over time for a large global sample and the antecedents to interindustry
variation, while also exploring variation at the level of the action and then investigating
actions to adopt and how their choices may influence their ability to maintain a competitive
advantage. These findings could hopefully facilitate future theoretical work seeking to
understand how the dynamics of imitation at the industry level affect strategic choices at the
firm level and characterize the process through which companies decide whether and how to
respond to increasing environmental and social pressures (e.g., Durand et al., 2019).
Incorporating the heterogeneity that exists across countries in terms of institutional structures
would be particularly useful for understanding how such variation may be impacting the
Second, we contribute to the stream of research that identifies the need to distinguish
between different types of actions rather than treating sustainability as a monolithic construct
(e.g., Durand et al., 2019; Hawn & Ioannou, 2016; McWilliams & Siegel, 2001). Thus, in
addition to current distinctions such as between symbolic and substantive actions (Hawn &
Ioannou, 2016), implicit and explicit social responsibility (Matten & Moon, 2008), and
material and immaterial ESG issues (Khan et al., 2016), we propose a novel distinction
between common and unique sustainability actions and provide evidence that the latter set
could be also strategic. We also offer an explanation about why unique and valuable actions
34
may remain hard to imitate: because they are characterized by higher market uncertainty,
higher legal and regulatory uncertainty but also, they require a higher degree of novelty.
though several studies explore the process of imitation and the factors that might accelerate
or hinder it within an industry, our study proposes industry-level attributes that can (partially)
explain interindustry variation in the rate of imitation (e.g., Romeo, 1975). Relatedly, our
work enriches the stream of research that explores why unique and valuable strategies may
remain hard to imitate (e.g., Lippman & Rumelt, 1982; Zander & Kogut, 1995; Ethiraj &
Levinthal, 2004; Levinthal, 1997; Rivkin, 2000; Greve, 2009; Ryall, 2009) by showing that
when a certain strategy can be decomposed into elements that become common versus
elements that remain more unique, then imitation at the strategy level can also remain partial,
thus enabling firms to maintain their competitive advantage. Future work may therefore seek
to explore the decomposability of strategies and how the three practice-level characteristics
Finally, for executives, our findings imply that it is important to understand how
sustainability actions diffuse over time within their industries when developing and
that remain difficult to imitate in the long-run due to the novel knowledge or technology on
which they depend thus allowing for a differentiation advantage to persist. In turn, such
differentiation strategies should account for the materiality of the underlying sustainability
issues and the stakeholder scrutiny that the industry faces. Above all, it is pivotal to realize
how critical the understanding and integration of sustainability actions has become for the
35
REFERENCES
Aggarwal, R., Erel, I., Ferreira, M. and Matos, P., 2011. Does governance travel around the world?
Evidence from institutional investors. Journal of Financial Economics, 100(1), pp.154–181.
Aguilera, R.V. and Cuervo-Cazurra, A., 2004. Codes of good governance worldwide: what is the trigger?
Organization studies, 25(3), pp.415–443.
Aguinis, H. and Glavas, A., 2012. What we know and don’t know about corporate social responsibility: A
review and research agenda. Journal of Management, 38(4), pp.932-968.
Amel-Zadeh, A. and Serafeim, G., 2018. Why and how investors use ESG information: Evidence from a
global survey. Financial Analysts Journal, 74(3), pp.87–103.
Aragòn-Correa, J.A., Marcus, A.A. and Vogel, D., 2020. The effects of mandatory and voluntary
regulatory pressures on firms’ environmental strategies: A review and recommendations for future
research. Academy of Management Annals, 14(1), pp.339-365.
Barney, J., 1991. Firm resources and sustained competitive advantage. Journal of management, 17(1),
pp.99–120.
Banerjee, A.V., 1992. A simple model of herd behavior. The quarterly journal of economics, 107(3),
pp.797–817.
Bansal, P. and Roth, K., 2000. Why companies go green: A model of ecological responsiveness. Academy
of management journal, 43(4), pp.717–736.
Baron, D.P., 2001. Private politics, corporate social responsibility, and integrated strategy. Journal of
Economics & Management Strategy, 10(1), pp.7–45.
Bartley, T., 2007. Institutional emergence in an era of globalization: The rise of transnational private
regulation of labor and environmental conditions. American journal of sociology, 113(2), pp.297-351.
Baum, J.A. and Mezias, S.J., 1992. Localized competition and organizational failure in the Manhattan
hotel industry, 1898–1990. Administrative science quarterly, pp.580–604.
Baum, J.A. and Singh, J.V., 1994. Organizational niches and the dynamics of organizational
mortality. American journal of sociology, 100(2), pp.346–380.
Bebchuk, L.A. and Weisbach, M.S., 2010. The state of corporate governance research. The review of
financial studies, 23(3), pp.939–961.
Bebchuk, L., Cohen, A. and Ferrell, A., 2009. What matters in corporate governance? The Review of
financial studies, 22(2), pp.783–827.
Bhattacharya, C.B. and Sen, S., 2003. Consumer–company identification: A framework for understanding
consumers’ relationships with companies. Journal of marketing, 67(2), pp.76–88.
Bikhchandani, S., Hirshleifer, D. and Welch, I., 1992. A theory of fads, fashion, custom, and cultural
change as informational cascades. Journal of political Economy, 100(5), pp.992–1026.
Bikhchandani, S., Hirshleifer, D. and Welch, I., 1998. Learning from the behavior of others: Conformity,
fads, and informational cascades. Journal of economic perspectives, 12(3), pp.151–170.
Briscoe, F. and Safford, S., 2008. The Nixon-in-China effect: Activism, imitation, and the
institutionalization of contentious practices. Administrative science quarterly, 53(3), pp.460–491.
36
Bode, C., Singh, J. and Rogan, M., 2015. Corporate social initiatives and employee retention. Organization
Science, 26(6), pp.1702–1720.
Bundy, J., Shropshire, C. and Buchholtz, A.K., 2013. Strategic cognition and issue salience: Toward an
explanation of firm responsiveness to stakeholder concerns. Academy of management review, 38(3),
pp.352–376.
Burbano, V.C., 2016. Social responsibility messages and worker wage requirements: Field experimental
evidence from online labor marketplaces. Organization Science, 27(4), pp.1010–1028.
Burns, L.R. and Wholey, D.R., 1993. Adoption and abandonment of matrix management programs:
Effects of organizational characteristics and interorganizational networks. Academy of Management
Journal, 36(1), pp.106–138.
Buysse, K. and Verbeke, A., 2003. Proactive environmental strategies: A stakeholder management
perspective. Strategic Management Journal, 24(5), pp.453–470.
Carroll AB, Lipartito KJ, Post JE, Werhane PH. 2012. Corporate Responsibility: The American
Experience. Cambridge University Press: Cambridge, UK.
Casadesus-Masanell, R. and Zhu, F., 2013. Business model innovation and competitive imitation: The case
of sponsor‐based business models. Strategic management journal, 34(4), pp.464–482.
Cheng, B., Ioannou, I. and Serafeim, G., 2014. Corporate social responsibility and access to finance.
Strategic management journal, 35(1), pp.1–23.
Chatterji, A.K., Durand, R., Levine, D.I. and Touboul, S., 2016. Do ratings of firms converge?
Implications for managers, investors and strategy researchers. Strategic Management Journal, 37(8),
pp.1597-1614.
Conlisk, J., 1980. Costly optimizers versus cheap imitators. Journal of Economic Behavior &
Organization, 1(3), pp.275–293.
Csaszar, F.A. and Siggelkow, N., 2010. How much to copy? Determinants of effective imitation breadth.
Organization Science, 21(3), pp.661–676.
Darnall, N., Henriques, I. and Sadorsky, P., 2010. Adopting proactive environmental strategy: The
influence of stakeholders and firm size. Journal of management studies, 47(6), pp.1072–1094.
Delmas, M.A. and Pekovic, S., 2013. Environmental standards and labor productivity: Understanding the
mechanisms that sustain sustainability. Journal of Organizational Behavior, 34(2), pp.230–252.
Dierickx, I. and Cool, K., 1989. Asset stock accumulation and sustainability of competitive advantage.
Management science, 35(12), pp.1504–1511.
DiMaggio PJ, Powell WW. 1983. The iron cage revisited: institutional isomorphism and collective
rationality in organizational fields. American Sociological Review 48(2): 147–160.
Dimson, E., Karakaş, O. and Li, X., 2015. Active ownership. The Review of Financial Studies, 28(12),
pp.3225–3268.
Du, S., Bhattacharya, C.B. and Sen, S., 2011. Corporate social responsibility and competitive advantage:
Overcoming the trust barrier. Management Science, 57(9), pp.1528–1545.
Durand, R., Hawn, O. and Ioannou, I., 2019. Willing and able: A general model of organizational
responses to normative pressures. Academy of Management Review, 44(2), pp.299–320.
37
Dutton, J., Dukerich, J. and Harquail, C.V., 1994. Organizational images and membership commitment.
Administrative science quarterly, 39(2), pp.239–263.
Eccles, R.G., Ioannou, I. and Serafeim, G., 2014. The impact of corporate sustainability on organizational
processes and performance. Management Science, 60(11), pp.2835–2857.
Edmans, A., 2011. Does the stock market fully value intangibles? Employee satisfaction and equity prices.
Journal of Financial economics, 101(3), pp.621–640.
Ethiraj, S.K. and Levinthal, D., 2004. Modularity and innovation in complex systems. Management
science, 50(2), pp.159–173.
Ethiraj, S.K. and Zhu, D.H., 2008. Performance effects of imitative entry. Strategic Management Journal,
29(8), pp.797–817.
Flammer, C., 2015. Does corporate social responsibility lead to superior financial performance? A
regression discontinuity approach. Management Science, 61(11), pp.2549–2568.
Flammer, C. and Kacperczyk, A., 2016. The impact of stakeholder orientation on innovation: Evidence
from a natural experiment. Management Science, 62(7), pp.1982–2001.
Flammer, C. and Luo, J., 2017. Corporate social responsibility as an employee governance tool: Evidence
from a quasi‐experiment. Strategic Management Journal, 38(2), pp.163–183.
Flammer, C., M. W. Toffel, and K. Viswanathan. “Shareholder Activism and Firms’ Voluntary Disclosure
of Climate Change Risks.” Strategic Management Journal, forthcoming.
Freiberg, D., Park, D.G., Serafeim, G. and Zochowski, R., 2020. Corporate Environmental Impact:
Measurement, Data and Information. Harvard Business School Accounting & Management Unit
Working Paper, (20-098).
Gaba, V. and Terlaak, A., 2013. Decomposing uncertainty and its effects on imitation in firm exit
decisions. Organization Science, 24(6), pp.1847–1869.
Gartenberg, C., Prat, A. and Serafeim, G., 2019. Corporate purpose and financial performance.
Organization Science, 30(1), pp.1–18.
Glac K. 2010. The influence of shareholders on corporate social responsibility. History of Corporate
Responsibility Project Working paper no. 2, Center for Ethical Business Cultures, University of
St.Thomas: Minneapolis, MN.
Godfrey, P.C., 2005. The relationship between corporate philanthropy and shareholder wealth: A risk
management perspective. Academy of management review, 30(4), pp.777–798.
Gompers, P., Ishii, J. and Metrick, A., 2003. Corporate governance and equity prices. The Quarterly
Journal of Economics, 118(1), pp.107–156.
Greve, H.R., 2009. Bigger and safer: The diffusion of competitive advantage. Strategic Management
Journal, 30(1), pp.1–23.
Hart, S.L., 1995. A natural-resource-based view of the firm. Academy of management review, 20(4),
pp.986–1014.
Haunschild, P.R. and Miner, A.S., 1997. Modes of interorganizational imitation: The effects of outcome
salience and uncertainty. Administrative Science Quarterly, pp.472–500.
38
Hawn, O., 2020. How media coverage of corporate social responsibility and irresponsibility influences
cross‐border acquisitions. Strategic Management Journal, forthcoming.
Hawn, O. and Ioannou, I., 2016. Mind the gap: The interplay between external and internal actions in the
case of corporate social responsibility. Strategic Management Journal, 37(13), pp.2569–2588.
Hawn, O., Durand, R., and Ioannou, I., 2020. Organizational Responses to Environmental, Social and
Governance Issues. Working paper.
Hawn, O., Chatterji, A.K. and Mitchell, W., 2018. Do investors actually value sustainability? New
evidence from investor reactions to the Dow Jones Sustainability Index (DJSI). Strategic Management
Journal, 39(4), pp.949-976.
Healy, P., Serafeim, G., Srinivasan, S. and Yu, G., 2014. Market competition, earnings management, and
persistence in accounting profitability around the world. Review of Accounting Studies, 19(4), pp.1281–
1308.
Husted, B.W., Jamali, D. and Saffar, W., 2016. Near and dear? The role of location in CSR engagement.
Strategic management journal, 37(10), pp.2050–2070.
Ioannou, I. and Hawn, O., 2019. Redefining Strategy in the Age of Sustainability and Social
Responsibility, forthcoming in Oxford Handbook of Corporate Social Responsibility: Psychological and
Organizational Perspectives 2018/19
Ioannou, I. and Serafeim, G., 2015. The impact of corporate social responsibility on investment
recommendations: Analysts’ perceptions and shifting institutional logics. Strategic Management Journal,
36(7), pp.1053–1081.
Kassinis, G. and Vafeas, N., 2006. Stakeholder pressures and environmental performance. Academy of
Management Journal, 49(1), pp.145–159.
Khan, M., Serafeim, G. and Yoon, A., 2016. Corporate sustainability: First evidence on materiality. The
accounting review, 91(6), pp.1697–1724.
King, B.G., 2008. A political mediation model of corporate response to social movement
activism. Administrative Science Quarterly, 53(3), pp.395-421.
Knott, A.M., Posen, H.E. and Wu, B., 2009. Spillover asymmetry and why it matters. Management
Science, 55(3), pp.373-388.
KPMG, 2017. The KPMG Survey of Corporate Responsibility Reporting 2017. Available at:
https://home.kpmg/xx/en/home/insights/2017/10/the-kpmg-survey-of-corporate-responsibility-reporting-
2017.html
Lee, M.D.P., 2011. Configuration of external influences: The combined effects of institutions and
stakeholders on corporate social responsibility strategies. Journal of Business Ethics, 102(2), pp.281–298.
Lee S, Matsunaga SR, Park CW. 2012. Management forecast accuracy and CEO turnover. The Accounting
Review. 2012 Nov;87(6):2095–122.
Levinthal, D.A., 1997. Adaptation on rugged landscapes. Management science, 43(7), pp.934–950.
Lieberman, M.B. and Asaba, S., 2006. Why do firms imitate each other? Academy of management review,
31(2), pp.366–385.
Lippman, S.A. and Rumelt, R.P., 1982. Uncertain imitability: An analysis of interfirm differences in
efficiency under competition. The bell journal of Economics, pp.418–438.
39
Luo, X. and Bhattacharya, C.B., 2006. Corporate social responsibility, customer satisfaction, and market
value. Journal of marketing, 70(4), pp.1–18.
Luo, X.R., Zhang, J. and Marquis, C., 2016. Mobilization in the Internet age: Internet activism and
corporate response. Academy of Management Journal, 59(6), pp.2045-2068.
Marquis, C. and Tilcsik, A., 2016. Institutional equivalence: How industry and community peers influence
corporate philanthropy. Organization Science, 27(5), pp.1325–1341.
Marquis, C., Glynn, M.A. and Davis, G.F., 2007. Community isomorphism and corporate social action.
Academy of management review, 32(3), pp.925–945.
Mathiasen C, Mell E, Gallimore A. 2012. U.S. Proxy Season Review: Environmental & Social Issues.
Institutional Shareholder Services Inc. Available at: http:/ / www.issgovernance.com/ files/ private/
2012USProxy SeasonReviewES.pdf.
Matten, D. and Moon, J., 2008. “Implicit” and “explicit” CSR: A conceptual framework for a comparative
understanding of corporate social responsibility. Academy of management Review, 33(2), pp.404–424.
Maxwell, J.W., Lyon, T.P. and Hackett, S.C., 2000. Self-regulation and social welfare: The political
economy of corporate environmentalism. The Journal of Law and Economics, 43(2), pp.583–618.
McWilliams, A. and Siegel, D., 2001. Profit maximizing corporate social responsibility. Academy of
Management Review, 26(4), pp.504–505.
Meyer J. W., Rowan B. 1977. Institutionalized organizations: formal structure as myth and ceremony.
American Journal of Sociology 83(2): 340–363.
Moon, J., Crane, A. and Matten, D., 2005. Can corporations be citizens? Corporate citizenship as a
metaphor for business participation in society. Business Ethics Quarterly, 15(3), pp.429–453.
Nelson, R.R. and Winter, S.G., 1982. The Schumpeterian tradeoff revisited. The American Economic
Review, 72(1), pp.114–132.
Obloj, T. and Sengul, M., 2020. What do multiple objectives really mean for performance? Empirical
evidence from the French manufacturing sector. Strategic Management Journal, 41(13), pp.2518-2547.
Oliver C. 1991. Strategic responses to institutional processes. Academy of Management Review, 16(1):
145–179.
Oliver C. 1997. Sustainable competitive advantage: combining institutional and resource-based view.
Strategic Management Journal, 18(9): 697–713.
Pedersen, E.R.G. and Gwozdz, W., 2014. From resistance to opportunity-seeking: Strategic responses to
institutional pressures for corporate social responsibility in the Nordic fashion industry. Journal of
business ethics, 119(2), pp.245–264.
Porter, M.E., 1991. Towards a dynamic theory of strategy. Strategic Management Journal, 12(S2), pp.95–
117.
Posen, H.E. and Martignoni, D., 2018. Revisiting the imitation assumption: Why imitation may increase,
rather than decrease, performance heterogeneity. Strategic Management Journal, 39(5), pp.1350–1369.
40
Posen, H.E., Lee, J. and Yi, S., 2013. The power of imperfect imitation. Strategic Management Journal,
34(2), pp.149–164.
Reed, R. and DeFillippi, R.J., 1990. Causal ambiguity, barriers to imitation, and sustainable competitive
advantage. Academy of management review, 15(1), pp.88–102.
Rivkin, J.W., 2000. Imitation of complex strategies. Management science, 46(6), pp.824–844.
Romeo, A.A., 1975. Interindustry and Interfirm Differences in the Rate of Diffusion of an Innovation. The
review of economics and statistics, pp.311–319.
Russo, M.V. and Fouts, P.A., 1997. A resource-based perspective on corporate environmental performance
and profitability. Academy of management Journal, 40(3), pp.534–559.
Ryall, M.D., 2009. Causal ambiguity, complexity, and capability-based advantage. Management Science,
55(3), pp.389–403.
Sharkey, A.J. and Bromley, P., 2015. Can ratings have indirect effects? Evidence from the organizational
response to peers’ environmental ratings. American Sociological Review, 80(1), pp.63–91.
Schmalensee, R., 1985. Do markets differ much? The American economic review, 75(3), pp.341–351.
Sharma, S. and Vredenburg, H., 1998. Proactive corporate environmental strategy and the development of
competitively valuable organizational capabilities. Strategic management journal, 19(8), pp.729–753.
Shenkar, O., 2010. Copycats: How smart companies use imitation to gain a strategic edge. Strategic
Direction.
Wu, Z. and Salomon, R., 2016. Does imitation reduce the liability of foreignness? L inking distance,
isomorphism, and performance. Strategic Management Journal, 37(12), pp.2441–2462.
Zander, U. and Kogut, B., 1995. Knowledge and the speed of the transfer and imitation of organizational
capabilities: An empirical test. Organization science, 6(1), pp.76–92.
41
TABLE 1 PANEL A: Convergence of sustainability actions within industries over time
This table presents estimated coefficients and p-values from ordinary least squares models using panel data at the industry-year level. For models 1 and 3-5 the dependent
(independent) variable is the natural logarithm of the coefficient of variation of practices across firms (the log transformed time variable starting with one in the first year and
increasing by one every year after). For model 2 dependent variable is the coefficient of variation of practices across firms (the time variable starting with one in the first year
and increasing by one every year after). In models 1 and 2 the coefficient of variation is constructed from a balanced sample with MSCI coverage that includes the same firms
over all years between 2012 and 2019. In models 3-5 the coefficient of variation is constructed from an unbalanced sample of firms with MSCI coverage that varies across
years. Model 3, 4 and 5 estimate the coefficient on the timetrend variable from years 2012 to 2019, 2009 to 2019 and 2006 to 2019 respectively. Standard errors are
heteroscedasticity robust and clustered at the industry level.
42
TABLE 1 PANEL B: Estimated convergence rate for each industry
Convergence
Industry Convergence %
Linear
Containers & Packaging -0.62 -3.11
Tobacco -0.62 -8.37
Airlines -0.57 -2.53
Biotechnology -0.53 -1.88
Consumer Finance -0.52 -2.05
Semiconductors & Semiconductor Equipment -0.50 -2.85
Paper & Forest Products -0.49 -2.36
Diversified Chemicals -0.48 -3.30
Air Freight & Logistics -0.46 -2.34
Technology Hardware, Storage & Peripherals -0.46 -2.09
Building Products -0.45 -2.68
Electronic Equipment, Instruments & Components -0.45 -1.87
Casinos & Gaming -0.41 -2.93
Household Durables -0.41 -1.73
Construction & Farm Machinery & Heavy Trucks -0.41 -2.39
Energy Equipment & Services -0.40 -3.26
Auto Components -0.40 -1.39
Asset Management & Custody Banks -0.39 -2.72
Restaurants -0.38 -1.16
Road & Rail Transport -0.35 -1.75
Household & Personal Products -0.32 -2.24
Specialty Chemicals -0.31 -2.69
Utilities -0.31 -2.37
Electrical Equipment -0.31 -1.59
Commodity Chemicals -0.30 -2.44
Diversified Financials -0.30 -1.75
Hotels & Travel -0.29 -1.63
Oil & Gas Exploration & Production -0.28 -2.32
Real Estate -0.27 -2.15
Banks -0.27 -1.02
Life & Health Insurance -0.27 -1.76
Textiles, Apparel & Luxury Goods -0.27 -2.11
Professional Services -0.25 -1.42
Metals and Mining - Precious Metals -0.25 -1.59
Multi-Line Insurance & Brokerage -0.24 -1.82
Marine Transport -0.24 -1.70
Wireless Telecommunication Services -0.23 -1.06
Software & Services -0.23 -1.37
Construction & Engineering -0.21 -1.79
Industrial Conglomerates -0.21 -1.35
Metals and Mining - Non-Precious Metals -0.21 -2.15
Integrated Telecommunication Services -0.21 -1.55
Transportation Infrastructure -0.20 -1.37
Retail - Consumer Discretionary -0.20 -1.50
Oil & Gas Refining, Marketing, Transportation & Storage -0.19 -1.71
43
Aerospace & Defense -0.19 -0.67
Beverages -0.17 -1.16
Health Care Providers & Services -0.16 -0.93
Trading Companies & Distributors -0.15 -0.34
Food Products -0.15 -1.00
Construction Materials -0.14 -1.46
Commercial Services & Supplies -0.14 -0.97
Property & Casualty Insurance -0.13 -0.63
Retail - Food & Staples -0.13 -0.65
Leisure Products -0.12 -0.73
Integrated Oil & Gas -0.12 -1.10
Investment Banking & Brokerage -0.10 -0.51
Steel -0.09 -0.76
Health Care Equipment & Supplies -0.08 -0.48
Media & Entertainment -0.08 -0.51
Automobiles -0.07 -0.37
Broadcasting, Cable & Satellite 0.00 0.45
Pharmaceuticals 0.00 0.07
Industrial Machinery 0.02 0.05
This table presents estimated coefficients for models where the dependent (independent) variable is the natural
logarithm of the coefficient of variation of actions across firms (the log-transformed time variable starting with 1
in the first year and increasing by 1 every year after) or the dependent variable is the coefficient of variation of
actions across firms (the time variable starting with 1 in the first year and increasing by 1 every year after). The
former estimated coefficient from log-log model is defined as Convergence % and the latter as Convergence
Linear.
44
Korea 64 0.029 0.044 0.000 0.190
China 64 0.021 0.032 0.000 0.118
India 64 0.024 0.060 0.000 0.340
Mexico 64 0.005 0.015 0.000 0.074
Malaysia 64 0.013 0.033 0.000 0.150
Russia 64 0.007 0.032 0.000 0.242
Singapore 64 0.010 0.037 0.000 0.271
South Africa 64 0.029 0.044 0.000 0.219
45
TABLE 2 PANEL B: Industry-level correlation matrix
Variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14)
Convergence % (1) 1.00
Convergence Linear (2) 0.76 1.00
Environmental Weight (3) -0.16 -0.23 1.00
Social Weight (4) -0.05 0.09 -0.83 1.00
Leader (5) -0.07 0.08 0.03 0.07 1.00
Attention (6) 0.16 0.22 0.00 0.14 0.07 1.00
Tone (7) -0.13 0.15 0.28 -0.23 0.24 0.02 1.00
Size (8) 0.13 0.06 0.25 -0.11 0.04 0.37 0.04 1.00
ROC (9) 0.00 -0.34 -0.06 0.13 -0.09 0.09 -0.17 0.30 1.00
PTE (10) 0.16 0.07 -0.03 0.05 0.07 0.06 0.24 0.05 0.66 1.00
Herfindahl (11) -0.38 -0.23 -0.01 0.07 0.26 0.16 0.05 -0.07 -0.06 -0.11 1.00
USA (12) -0.23 0.00 0.00 0.03 0.14 -0.16 0.33 -0.37 -0.08 0.15 0.09 1.00
JP (13) -0.03 0.00 0.05 0.10 0.12 -0.01 0.22 0.20 0.07 0.12 0.08 -0.22 1.00
UK (14) 0.03 -0.05 -0.27 0.25 -0.34 -0.02 -0.22 -0.29 0.16 0.08 -0.17 0.05 -0.24 1.00
Australia (15) 0.11 -0.04 0.03 -0.16 -0.22 -0.15 -0.30 -0.22 -0.16 -0.07 -0.14 -0.33 -0.25 0.08
Canada (16) 0.11 0.04 0.08 -0.16 -0.32 -0.07 -0.32 0.00 -0.09 -0.21 -0.21 -0.27 -0.31 0.09
EU non-UK (17) 0.29 0.19 -0.14 -0.01 0.12 -0.02 -0.05 0.20 0.06 0.05 -0.06 -0.40 -0.14 -0.04
Brazil (18) 0.07 0.02 0.15 -0.03 0.04 0.03 -0.04 0.42 0.02 -0.01 -0.20 -0.19 -0.07 0.00
Korea (19) -0.13 -0.08 0.22 -0.18 0.29 0.01 0.11 0.42 0.01 -0.07 0.19 -0.10 0.16 -0.31
China (20) 0.10 0.07 0.10 -0.09 -0.12 0.36 -0.11 0.39 -0.11 -0.25 -0.18 -0.30 -0.08 -0.22
India (21) -0.03 -0.13 -0.10 0.20 -0.05 0.29 -0.18 0.06 0.15 -0.09 0.17 -0.21 0.17 -0.08
Mexico (22) 0.25 0.17 0.00 -0.11 -0.08 -0.06 -0.24 -0.13 0.01 0.06 0.09 -0.27 -0.24 0.12
Malaysia (23) 0.00 -0.09 0.01 -0.15 -0.13 -0.08 -0.06 0.04 0.01 -0.03 0.02 -0.24 -0.16 -0.08
Russia (24) 0.17 0.07 0.16 -0.23 -0.21 0.12 -0.24 0.35 -0.04 -0.22 -0.08 -0.31 -0.16 -0.15
Singapore (25) -0.01 0.02 -0.04 -0.07 -0.12 -0.11 0.02 -0.09 -0.06 0.00 -0.08 -0.28 0.23 -0.14
South Africa (26) 0.06 0.06 0.02 -0.06 -0.22 -0.06 -0.04 -0.02 -0.14 -0.02 -0.05 -0.30 -0.11 0.11
46
TABLE 2 PANEL B: (Continued)
Variable (15) (16) (17) (18) (19) (20) (21) (22) (23) (24) (25)
Australia (15) 1.00
Canada (16) 0.23 1.00
EU non-UK (17) 0.01 0.03 1.00
Brazil (18) -0.12 0.03 0.13 1.00
Korea (19) -0.05 -0.23 -0.20 0.14 1.00
China (20) 0.13 0.13 0.09 -0.04 -0.02 1.00
India (21) -0.17 -0.13 -0.14 0.00 -0.07 0.13 1.00
Mexico (22) 0.31 0.33 0.11 -0.08 -0.09 0.15 -0.02 1.00
Malaysia (23) 0.20 -0.09 0.08 0.00 0.19 -0.08 -0.07 0.00 1.00
Russia (24) 0.09 0.40 0.17 0.29 -0.06 0.31 0.02 0.01 0.00 1.00
Singapore (25) 0.16 -0.11 0.26 -0.07 -0.03 0.06 -0.09 -0.06 -0.04 -0.04 1.00
South Africa (26) 0.23 0.50 -0.03 0.10 -0.12 -0.03 -0.15 0.38 -0.03 0.01 -0.05
47
TABLE 3 Panel A: Industry-level variation in convergence of sustainability actions within industry over time (balanced panel)
This table presents estimated coefficients and p-values from ordinary least squares models using cross sectional data at the industry level. For models 1-4 and 6 the dependent
variable is the estimated coefficient on the logarithm transformed timetrend variable in industry specific models where the dependent variable is the natural logarithm of the
coefficient of variation of practices across firms. For model 5 dependent variable is the estimated coefficient on the timetrend variable in industry specific models where the
dependent variable is the coefficient of variation of practices across firms. In models 1-6 the coefficient of variation is constructed from a balanced sample that includes the
same firms over all years between 2012 and 2019. Environmental (social) weight is the across years mean of the average weight across all firms within an industry assigned by
MSCI to all environmental (social) issues when calculating the overall ESG score. Leader is the across years mean for each industry of the percentage of large firms (top decile
of market capitalization in all models except for firm sales in model 6), that score at the top quartile of MSCI's ESG rating. Attention is the volume of ESG-related articles as
measured by TruValue Labs averaged across companies and years in an industry. Tone is a measure of how positive or negative is the sentiment across the volume of ESG-
related articles as measured by TruValue Labs averaged across companies and years in an industry. Size is the average natural logarithm of the market capitalization of all
sample firms in an industry and then across years. ROC is the average Return on Capital of all sample firms in an industry and then across years. PE is the average Price to
Earnings Ratio of all sample firms in an industry and then across years. Geographic controls include 15 variables that measure for each industry the percentage of firms coming
from each one of the following geographies: USA, UK, EU x UK, Canada, Australia, China, Brazil, Mexico, India, Malaysia, South Korea, Japan, Singapore, Russia and South
Africa. Herfindahl index is the sum of the squares of the market shares of the firms within the industry where the market shares are expressed as fractions of all sample firms
in an industry and then averaged across years. All independent variables are constructed using the full unbalanced sample. Standard errors are heteroscedasticity robust.
48
TABLE 3 Panel B: Industry-level variation in convergence of sustainability actions within industry over time (unbalanced panel)
Dependent Variable Convergence %
Model (1) (2) (3) (4) (5) (6)
Estimate p-value Estimate p-value Estimate p-value Estimate p-value Estimate p-value Estimate p-value
Intercept -1.6578 0.0001 -2.9030 0.0000 -2.3219 0.0164 -3.7418 0.0067 -2.7124 0.0205 -1.8138 0.2581
Environmental Weight -0.0041 0.0015 -0.0037 0.0035 -0.0038 0.1290 -0.0021 0.4666 -0.0012 0.6583 -0.0030 0.4640
Social Weight -0.0057 0.0000 -0.0043 0.0080 -0.0073 0.0233 -0.0040 0.3301 -0.0052 0.1901 -0.0069 0.2271
Leader 0.0440 0.0000 0.0578 0.0000 0.0740 0.0038 0.0862 0.0068 0.0505 0.0790 0.0736 0.0346
Leader x Tone -0.0007 0.0000 -0.0010 0.0000 -0.0012 0.0050 -0.0014 0.0121 -0.0008 0.0999 -0.0012 0.0439
Attention 0.0024 0.0000 0.0026 0.0000 0.0033 0.0016 0.0033 0.0143 0.0012 0.3897 0.0025 0.1442
Tone 0.0163 0.0052 0.0303 0.0000 0.0257 0.1065 0.0371 0.1176 0.0288 0.1366 0.0445 0.0660
Size 0.0352 0.0087 0.0576 0.0232 0.0393 0.1909 0.0510 0.4421 0.0437 0.3478 -0.0440 0.5927
ROC -0.0018 0.6807 -0.0012 0.8201 0.0197 0.0909 0.0270 0.0815 -0.0003 0.9832 0.0225 0.2248
PE 0.0022 0.5215 0.0008 0.8379 -0.0087 0.3419 -0.0138 0.2427 -0.0009 0.9358 -0.0117 0.4105
Herfindal -0.0053 0.0667 -0.0079 0.0174 -0.0032 0.6304 -0.0059 0.5533 -0.0017 0.8411 0.0049 0.6031
Geographic Controls No Yes No Yes No Yes
Adjusted R-squared 52.1% 60.3% 26.7% 18.9% 5.2% 1.0%
N 66 66 66 66 66 66
This table presents estimated coefficients and p-values from ordinary least squares models using cross sectional data at the industry level. The dependent variable is the estimated
coefficient on the logarithm transformed timetrend variable in industry specific models where the dependent variable is the natural logarithm of the coefficient of variation of
practices across firms. In models 1-6 the coefficient of variation is constructed from an unbalanced sample of firms with MSCI coverage that varies across years. Model 1-2,
3-4 and 5-6 estimate the coefficient on the timetrend variable from years 2012 to 2019, 2009 to 2019 and 2006 to 2019 respectively. Environmental (social) weight is the across
years mean of the average weight across all firms within an industry assigned by MSCI to all environmental (social) issues when calculating the overall ESG score. Leader is
the across years mean for each industry of the percentage of large firms (top decile of market capitalization), that score at the top quartile of MSCI's ESG rating. Attention is
the volume of ESG-related articles as measured by TruValue Labs averaged across companies and years in an industry. Tone is a measure of how positive or negative is the
sentiment across the volume of ESG-related articles as measured by TruValue Labs averaged across companies and years in an industry. Size is the average natural logarithm
of the market capitalization of all sample firms in an industry and then across years. ROC is the average Return on Capital of all sample firms in an industry and then across
years. PE is the average Price to Earnings Ratio of all sample firms in an industry and then across years. Geographic controls include 15 variables that measure for each industry
the percentage of firms coming from each one of the following geographies: USA, UK, EU x UK, Canada, Australia, China, Brazil, Mexico, India, Malaysia, South Korea,
Japan, Singapore, Russia and South Africa. Herfindahl index is the sum of the squares of the market shares of the firms within the industry where the market shares are expressed
as fractions of all sample firms in an industry and then averaged across years. All independent variables are constructed using the full unbalanced sample. Standard errors are
heteroscedasticity robust.
49
TABLE 4 Panel A: Summary statistics for convergence at the industry–action level
# of
Standard 3rd 1st
Actions N Average Maximum Minimum Industries
Deviation Quartile Quartile
Unique
All Practices 846 -0.08 0.37 0.08 -0.25 2.81 -2.89 N/A
Access to communication 4 0.15 0.32 0.40 -0.11 0.57 -0.12 2
Access to finance 10 0.09 0.25 0.13 -0.05 0.68 -0.22 7
Access to healthcare 3 -0.28 0.28 0.02 -0.55 0.02 -0.55 1
Anticompetitive practices 63 -0.24 0.34 -0.04 -0.43 0.71 -0.89 19
Biodiversity and land use 14 -0.03 0.22 0.05 -0.13 0.57 -0.35 7
Business ethics and fraud 66 -0.01 0.36 0.13 -0.22 1.01 -0.82 33
Carbon emissions 66 -0.06 0.14 0.02 -0.17 0.25 -0.38 32
Chemicals safety 18 0.03 0.32 0.20 -0.14 0.75 -0.58 10
Controversial sourcing 7 -0.16 0.59 0.25 -0.63 0.98 -0.67 2
Corporate governance 66 -0.21 0.15 -0.13 -0.32 0.22 -0.47 10
Corruption 44 0.07 0.32 0.18 -0.03 1.01 -0.67 34
Energy efficiency 19 -0.03 0.44 0.18 -0.20 0.77 -1.32 13
Electronic waste 5 -0.13 0.22 0.06 -0.17 0.07 -0.48 2
Financial system instability 8 -0.27 0.25 -0.14 -0.46 0.18 -0.60 1
Financing environmental
impact 7 -0.09 0.17 0.07 -0.24 0.09 -0.32 3
Financial products safety 8 -0.17 0.59 -0.02 -0.54 1.03 -0.92 2
Health and safety 66 -0.25 0.27 -0.07 -0.45 0.36 -0.97 16
Human capital development 23 0.09 0.35 0.18 -0.14 1.43 -0.32 14
Insurance climate change risk 5 -0.07 0.47 0.21 0.00 0.32 -0.88 4
Insurance health demographic
risk 6 -0.22 0.53 0.00 -0.19 0.34 -1.23 2
Labor management 66 0.11 0.20 0.19 -0.03 0.73 -0.19 51
Opportunities clean tech 26 -0.08 0.16 0.03 -0.17 0.22 -0.47 12
Opportunities green buildings 5 -0.16 0.65 -0.23 -0.51 0.97 -0.58 1
Opportunities nutrition health 5 -0.19 0.13 -0.16 -0.29 -0.01 -0.32 1
Opportunities renewable
energy 1 -0.16 0.00 -0.16 -0.16 -0.16 -0.16 0
Packing materials and waste 3 -0.09 0.19 0.06 -0.31 0.06 -0.31 2
Privacy and data security 26 0.15 0.79 0.50 -0.32 2.81 -1.01 15
Product carbon footprint 9 0.03 0.17 0.09 -0.04 0.27 -0.29 7
Product safety and quality 29 -0.18 0.41 0.07 -0.35 0.42 -1.54 9
Raw material sourcing 14 -0.26 0.46 -0.01 -0.54 0.48 -1.25 6
Responsible investing 10 0.23 0.60 0.20 -0.06 1.85 -0.09 6
Supply chain labor 12 -0.55 0.88 -0.08 -0.80 0.41 -2.89 3
Toxic emission waste 66 -0.19 0.35 -0.02 -0.34 0.69 -1.97 19
Water stress 66 0.03 0.19 0.13 -0.08 0.56 -0.51 45
This table shows the magnitude of convergence across different sustainability actions estimated at the industry–
action pair level. Convergence is the estimated coefficient from a model where the dependent variable is the
natural logarithm of the coefficient of variation of actions across firms and the independent variable is the
logarithm-transformed timetrend variable. The coefficient of variation is constructed from a balanced sample that
includes the same firms over all years between 2012 and 2019. # of Industries Unique is a measure counting the
number of industries that this practice would be classified as unique using an on average annual 5% convergence
rate as the threshold to identify a unique action.
50
Table 4 Panel B: Industry-action level variation in convergence and action characteristics
This table presents estimated coefficients and p-values from ordinary least squares models using panel data at the
industry-practice level. The dependent variable is the estimated coefficient on the logarithm transformed timetrend
variable in industry specific models where the dependent variable is the natural logarithm of the coefficient of
variation of practices across firms. Low Market Uncertainty is an indicator variable taking the value of one if the
across firm and years average of the weight on a practice for all firms in the industry is above the median. Low
Regulatory Uncertainty is an indicator variable taking the value of one for practices where policies and regulations
are identified as drivers of the practice (e.g., anti-competitive practices, toxic emissions, chemical safety,
corruption, financial stability, financial product safety, employee health and safety, product safety and supply
chain labor). High Practice Novelty is an indicator variable taking the value of one for practices where they require
the use of novel knowledge and technologies (e.g., access to finance, access to communications, financing of
environmental impact, data privacy, product carbon footprint, responsible investing). Standard errors are clustered
at the practice level and are heteroscedasticity robust.
51
TABLE 5 PANEL A: Summary statistics for financial performance analysis
Variable N Average Standard Deviation Minimum Maximum
Unique Actions 39,424 4.98 1.54 0.00 10.00
Common Actions 39,424 4.14 1.78 0.00 10.00
ROC 39,040 7.47 11.95 -30.00 50.00
ROA 38,728 4.71 8.94 -30.00 35.00
ROS 39,118 8.63 16.50 -40.00 50.00
Asset Turnover 39,012 72.64 63.94 0.00 300.00
Log Tobin’s Q 38,949 0.45 0.55 -0.69 2.20
R&D/Sales 39,424 3.52 12.85 0.00 100.00
CAPEX/Assets 37,216 5.19 6.46 0.00 50.00
Log Market Cap 39,064 21.70 1.51 17.00 28.26
Debt/Assets 38,895 25.10 20.07 0.00 100.00
Variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Unique Actions (1) 1.000
Common Actions (2) 0.045 1.000
ROC (3) 0.047 0.062 1.000
ROA (4) 0.027 0.051 0.944 1.000
ROS (5) 0.066 -0.002 0.615 0.666 1.000
Asset Turnover (6) -0.004 0.082 0.278 0.254 -0.206 1.000
Log Tobin’s Q (7) 0.014 -0.005 0.304 0.265 -0.014 0.265 1.000
R&D/Sales (8) -0.028 -0.047 -0.350 -0.401 -0.364 -0.112 0.324 1.000
CAPEX/Assets (9) -0.110 -0.034 0.039 0.073 -0.022 0.025 0.047 -0.084 1.000
Log Market Cap (10) 0.003 0.138 0.225 0.224 0.210 -0.097 0.157 -0.072 -0.017 1.000
Debt/Assets (11) -0.033 0.029 -0.126 -0.057 -0.049 -0.129 -0.127 -0.106 0.101 0.033
52
TABLE 6: Operating performance analysis
53
TABLE 6: operating performance analysis (Continued)
54
TABLE 7: Market valuation analysis
55