Inside the ESG Ratings: (Dis)agreement and Performance 2021 PDF

Received: 1 February 2020 Revised: 6 November 2020 Accepted: 30 November 2020 DOI: 10.1002/csr.2177 RESEARCH ARTICLE Inside the ESG ratings: (Dis)agreement and performance Monica Billio1 | Michele Costola1 | Loriana Pelizzon3 1 Ca' Foscari University of Venice, Venice, Italy 2 Leibniz Instit...

Received: 1 February 2020 Revised: 6 November 2020 Accepted: 30 November 2020 DOI: 10.1002/csr.2177 RESEARCH ARTICLE Inside the ESG ratings: (Dis)agreement and performance Monica Billio1 | Michele Costola1 | Loriana Pelizzon3 1 Ca' Foscari University of Venice, Venice, Italy 2 Leibniz Institute for Financial Research SAFE, Frankfurt am Main, Germany 3 Leibniz Institute for Financial Research SAFE, Goethe University Frankfurt, Ca’ Foscari University of Venice and CEPR, Venice, Italy Iva Hristova1 | Carmelo Latino2 | Abstract We analyze the ESG rating criteria used by prominent agencies and show that there is a lack of a commonality in the definition of ESG (i) characteristics, (ii) attributes and (iii) standards in defining E, S and G components. We provide evidence that heterogeneity in rating criteria can lead agencies to have opposite opinions on the same Correspondence Monica Billio, Ca' Foscari University of Venice, Venice, Italy. Email: [email protected] Funding information Leibniz Institute for Financial Research SAFE, Frankfurt; European Investment Bank (EIB) for the project “ESG-Credit.eu - ESG Factors and Climate Change for Credit Analysis and Rating” evaluated companies and that agreement across those providers is substantially low. Those alternative definitions of ESG also affect sustainable investments leading to the identification of different investment universes and consequently to the creation of different benchmarks. This implies that in the asset management industry it is extremely difficult to measure the ability of a fund manager if financial performances are strongly conditioned by the chosen ESG benchmark. Finally, we find that the disagreement in the scores provided by the rating agencies disperses the effect of preferences of ESG investors on asset prices, to the point that even when there is agreement, it has no impact on financial performances. KEYWORDS Corporate Social Responsibility, ESG rating agencies, sustainable investments JEL CLASSIFICATION M14; G24; G11 1 | I N T RO DU CT I O N consequence, the process of asset allocation started evolving. Furthermore, a raising environmental, social and governance consciousness has According to Milton Friedman, the main responsibility of a company is been observed worldwide. This trend appeared as a result of the the maximization of the shareholders' returns. For decades, environmen- increased occurrence of extreme weather events,1 damaging infrastruc- tal, social and governance (ESG) responsibilities were not considered rel- tures and perturbing global markets, but also because of the 2008 evant by most of the companies that have been focusing on profit financial crisis, which affected both private and public sectors. Indeed, maximization. Not only were ESG responsibilities believed to merely the effect of the subprime crisis has been threefold: firstly, it highlighted have no incidence on financial performance, but they were also per- the relevance of investors' decisions and therefore their inherent role; ceived as a potential burden to the latter, being related to cost increases. secondly, it raised the public consciousness in terms of social responsi- Nevertheless, in the last twenty years, environmental, social and bility; thirdly, it stresses the importance of good governance practices. governance issues revealed their influence not only on the profitability, Thus, even though socially responsible investment (SRI) has but also on the financial viability of several firms. As a natural existed since 1920s, only recently, it has only recently experienced a This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. © 2021 The Authors. Corporate Social Responsibility and Environmental Management published by ERP Environment and John Wiley & Sons Ltd. 1426 wileyonlinelibrary.com/journal/csr Corp Soc Responsib Environ Manag. 2021;28:1426–1445. considerable surge of interest and has become a general preoccupa- 2 | T H E CU R R E N T E S G S T A T E O F T H E A R T tion rather than a niche investment practice. Consequently, we have observed the development of a new segment of the rating agencies' The term ESG was officially coined in 2004 with the publication of market and the big three credit rating agencies (Moody's, S&P, Fitch) the report “Who Cares Wins” by the UN Global Compact Initiative have started to include ESG evaluation in their ratings. (UN, 2004). It set the ambitious goal to regroup three of the main Nevertheless, in this new, insufficiently regulated and rapidly expanding sector, investors, corporate managers and policy makers ethical finance pillars: environmental, social and governance. All of them encompass different issues and present a specific assessment target. need a deeper understanding of the ESG inherent particularities and a The environmental pillar focuses on issues such as climate change, wider knowledge of the potential impacts of ESG on the real and deforestation, air and water pollution, land exploitation and biodiver- financial sectors. In response, the academic research is developing at a sity loss. Therefore, it evaluates the efforts of a company in terms of tremendous pace. energy efficiency, greenhouse gas emissions, waste, water and If a classification of this quickly expanding research was pro- resource management. A large part of the literature has thus naturally posed, four major scientific orientations could be identified. First, tried to define the relationship between environmental and financial a major part of the literature is aiming to investigate economic and performance. Derwall et al. (2004) showed that more eco-friendly financial performance of ESG stocks and portfolio. In particular, firms were benefiting from higher stock returns than their less eco- the literature concentrates on the ESG portfolio profitability,2 the friendly counterparts. These findings remain significant even after impact of ESG on firms' financial performance,3 and the effect of several methodological controls. Manrique and Martí-Ballester (2017) ESG scores on credit ratings.4 A second strand of the literature also provide similar conclusions on the basis of a sample including studies and evaluates the particularities of the ESG rating score and 2982 large firms from both developed and developing countries. the adopted methodologies highlighting their disagreement.5 The social pillar includes aspects related, for instance, to gender To our knowledge, there are no papers that relate these two policies, protection of human rights, labor standards, workplace and aspects and investigate the implications that ESG rating disagreement product safety, public health and income distribution, which are all might have on ESG portfolios performance. affecting employees’ satisfaction. According to Edmans (2011) there The present study aims to cover this void. We contribute to the exists a clear positive relationship between employee's satisfaction existing literature by (i) investigating the disagreement among ESG and long-run stock return. Namely, American companies considered rating agencies in terms of scores and its effects in the identification as proposing the best working condition have earned a 4-factor alpha of the constituents of ESG indexes in terms of constituents overlap of 3.5% per year (2.1% above the industry benchmark) in the period (ESG agreement portfolio) and (ii) studying the performance of the 1984–2009. ESG agreement with respect to non-ESG portfolios. At last, the governance pillar is related to aspects such as the inde- Firstly, we analyze the rating criteria used by nine prominent pendence of the board of administration, shareholders' rights, man- agencies and show that there is a lack of a common characteristics, agers' remuneration, control procedures and anti-competitive practices, attributes and standards in defining the E, S and G components. as well as the respect of the law. Several studies tend to stress the Secondly, we show that the heterogeneity in the ESG industry can significant positive impact of these practices, like Gompers et al. (2003), lead the agencies to have opposite opinions on the evaluated Tarmuji et al. (2016) and Velte (2017). While Tarmuji et al. (2016) con- companies and that agreement, among rating agencies, is relatively sider companies from Malaysia and Singapore, Velte (2017) focuses on low. Moreover, we show that alternative definitions of ESG also German corporations, and Gompers et al. (2003) study US firms. These affect the benchmarks identification used by the sustainable invest- three analyses highlight the positive impact of stronger governance ments' industry. In relation to this, we analyse four representative practices on companies' profitability. ESG indexes and show that the overlap of their common constitu- The relevance of ESG on companies' financial performance and ents is approximately 15% (the ESG agreement portfolio). Thirdly, profitability represents one of the major strands of the current litera- we show that the low overlap of the ESG indexes (due to the ture. A large majority of studies, concerning both industrialized coun- disagreement in the scores provided by the rating agencies) dis- tries and emerging economies, stresses the positive impact of ESG perses the effect of preferences of ESG investors on asset prices, to efforts and disclosure on firms' financial performance. the point that, even for the ESG agreement portfolio, there is no impact on the financial performances. For instance, Zhao et al. (2018) focus on Chinese listed power generation firms and find that good ESG performance can improve The paper is organized as follows. Section 2 introduces the con- financial performance. Brogi and Lagasio (2019), using MSCI ESG KLD cept of ESG investing, and presents the ESG investing landscape evo- STATS data from 2000 to 2016, provide evidence of the positive lution. Section 3 describes the industry of ESG ratings and proposes impact of ESG on US companies' profitability measured through ROA, an assessment of their reliability including a structural comparison of especially for the banking sector. Ortas et al. (2015), using the ASSET4 four ESG indexes. Section 4 investigates the implication of disagree- Database including MSCI data, obtain similar results for the cases of ment, among the rating providers, on the performance of ESG and Spain, France and Japan, stressing the significant positive impact non-ESG portfolios. Finally, the last section concludes. of ESG performance on financial performance for companies adopting 15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2177 by Cochrane Netherlands, Wiley Online Library on [25/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 1427 BILLIO ET AL. BILLIO ET AL. the United Nations Global Compact (UNGC). The findings of Aureli investment” (SRI) portfolios perform better only in Japan, while in et al. (2020) confirm the relevance of ESG disclosure on firms' market emerging Asian stock markets they are not especially rewarded. This value for 55 Dow Jones Sustainability World Index listed companies. last observation is also confirmed by Auer and Schuhmacher (2016) Similarly, Giese et al. (2019) find, on the basis of MSCI ESG data, that using the Sustainalytics ESG indicators. Namely, the authors obtain a ESG information affects positively not only companies' valuation but similar performance for the Asia-Pacific region and in the United also their performance. The authors even identify the precise channels States, while in Europe, the investment performance can even be neg- for such effects, namely through: reduced capital costs, higher valua- atively affected, for certain industries and specific ESG criteria. Never- tions, higher profitability, and lower exposure to tail risk. Lo and theless, the meta-analysis performed by Friede et al. (2015), Kwan (2017) analyze the case of Hong Kong companies using the combining the findings of over 2200 individual studies, highlights that Dow Jones Sustainability Asia Pacific Index (DJSIAP) and FTSE4Good about 90% of them show a non-negative relationship between ESG Global 100 Index (FTSE4Good) and find weak but positive evidence and corporate financial performance (CFP) with a clear positive rela- of the market reaction on ESG information. Furthermore, ESG initia- tionship on the large majority. Interestingly, Halbritter and Dorf- tives present a stronger observed effect comparatively to SRI. leitner (2015) showed that the size and directionality of the However, there are other studies leading to different conclusions, overperformance of ESG portfolios strongly depend on the rating pro- for instance, on the absence of a statistical impact (Landi & vider which highlights the considerable discrepancies among ESG rat- Sciarelli, 2019) or the presence of very weak effects of only some of ings and the necessity for a stronger harmonization.
s et al., 2019). While the first paper the ESG pillars (Miralles-Quiro The third section of the literature evaluates the effects of ESG focuses on Italian companies and exploits the FTSE MIB, the second factors on credit ratings. Attig et al. (2013) find evidence, on the basis one considers a larger panel of mainly OECD members and runs its of MSCI ESG STATS, that firms with good social performance benefit analysis on the basis of the ESG data provided by Thomson Reuters from relatively high ratings provided by credit rating agencies. Simi- Eikon. Nevertheless, given the different methodologies accounting for larly, Devalle et al. (2017) and Weber et al. (2010), confirm that firms ESG efforts across the different rating agencies and data providers, with important environmental and sustainability performances are but also given the inherent country-specific particularities, and the benefiting from higher credit rating scores. While the first study uses ESG materiality issue, it is difficult to draw a robust conclusion on ESG data from Thomson Reuters Datastream, the second one is based the topic. In this light, what seems evident, is the necessity of a fur- on a questionnaire covering 58 sustainability criteria and addressed to ther standardization of ESG accounting procedures, allowing inves- credit officers, from 40 German banks, providing credits to small and tors, policy makers and scientists to evaluate at best the effects of medium-sized firms. Kiesel and Lücke (2019) demonstrate a small but ESG performances. clear influence of ESG performance on rating decisions, especially of Another strand of the literature is focusing on the performance of the corporate governance pillar. The authors apply the LDA model ESG portfolios. Kempf and Osthoff (2007), Statman and Glushkov identifying ESG topics in 3719 Moody's credit rating reports. (2009), Nofsinger and Varma (2014), Henke (2016) confirm that According to the work of Jang et al. (2020), focusing on the case of investing in ESG-firms-based portfolios can clearly provide a perfor- South Korea and using ESG data from Korea Corporate Governance mance gain. Namely, Kempf and Osthoff (2007), using data from KLD Service (KCGS), ESG ratings are complementary to credit ratings as Research & Analytics, find that the strategy of buying stocks with high they encompass essential non-financial information and can lower the socially responsible ratings and sell stocks with low socially responsi- cost of debt financing, especially for small firms. Bhattacharya and ble ratings leads to high abnormal returns which remain significant Sharma (2019) attempt a similar type of analysis for the Indian market, despite transaction costs. Statman and Glushkov (2009), base their using ESG data from Bloomberg, and observe that ESG efforts present analysis on the same database and find that socially responsible inves- a positive effect on credit ratings only for small and middle-level com- tors benefit from a return advantage relative to conventional ones. panies. As for the previous literature strands, it is difficult to draw a Nevertheless, the authors also find that the systematic exclusion of robust conclusion on the ESG role on credit ratings. sin stocks might be penalizing. Nofsinger and Varma (2014) consider Thus, the presented above elements and the obtained contradic- and identify also an outperformance of socially responsible mutual tory results within the current literature, stress the necessity of reli- funds especially during periods of market crises. According to the able and harmonized ESG data. authors, using the Morningstar database and the Domini Social Index, Nonetheless, given the current urgent need for precise actions this effect is especially pronounced for ESG funds using positive coping with ESG issues, sustainable finance's role seems evident. In screening techniques and the obtained performances depend only on result, several ESG investment strategies have emerged. The following the socially responsible fund attributes. The work provided by subsection discusses precisely these aspects. Henke (2016), evaluating US and Eurozone funds (with data provided by US SIF and Euro SIF)6, stresses also the outperformance of ESG portfolios during crisis periods. The obtained results maintain their 2.1 | ESG investment strategies significance even after a large set of robustness checks. However, Yen et al. (2019) perform a similar analysis but for Asian stock markets, on Sustainable investing is an investment approach that integrates ESG the basis of ASSET4 ESG ratings, and find that “socially responsible characteristics with classical investment techniques in the portfolio 15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2177 by Cochrane Netherlands, Wiley Online Library on [25/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 1428 construction and management process. According to the Global Sus- assimilated to corporate risk indicators, allowing for the reduction of tainable Investment Alliance (GSIA), several strategies have been information asymmetries (see Utz, 2017). Unlike credit ratings, ESG emerged as global standards in the industry. The main sustainable measurement is somehow nebulous given the lack of a common defi- investment strategies can be classified as follows: nition, reporting standards and shared characteristics among each ESG component and across rating providers. Currently, rating agen- 1. Negative/exclusionary screening. Exclusion of specific unacceptable cies are proposing several metrics similar to the credit ratings market, or controversial sectors or companies whose activities may harm but, unlike those, ESG ratings are derived from alternative and com- the environment or society. According to the Global Sustainable peting definitions. Hence, a common standard for ESG is missing and Investment Review (GSIA, 2018), this has been the most popular makes the sustainability of a company very difficult to assess and, in strategy for open-end funds (gathering $19.8 trillion in asset under some cases “unratable.” management). This achievement might be due to the ease of Windolph (2011) identifies six causes that hinder a transparent implementation of such a procedure, based on the identification and objective rating, such as: lack of standardization, lack of credibility and exclusion of the so-called “nonESG” stocks. of information, bias, trade-offs, lack of transparency, and lack of inde- 2. Positive/best-in-class screening. Selection of the best ESG per- pendence. Moreover, during the last years, ESG rating agencies have forming companies within a specific business sector, all other con- not only integrated new criteria into their assessment models in order ditions being equal. It corresponds to the exclusion of companies to respond to new global challenges (Escrig-Olmedo et al., 2019) but not meeting certain performance thresholds. also the assessment criteria are changing rapidly, making the evaluat- 3. Sustainability themed investing. Targeted investments, including ing process even more complex. only activities related to the chosen theme (clean energy, pollution The conceptual framework is therefore very fragile. Nevertheless, reduction, low carbon emissions, water resources management, this paper offers an alternative view on three main challenges (and sustainable agricultural activities etc.). the implications that may arise): (i) agreement and disagreement in the 4. Impact/community investing. Private investments dedicated to specific projects solving social and environmental issues such as ESG ratings, (ii) agreement and disagreement in the ESG indexes and (iii) implications of (ii) on financial performance. renewable energy use, social housing investments etc. 5. ESG integration. Systematic and explicit inclusion of ESG factors into financial analysis. Given the qualitative and subjective charac- 3.1 | How agencies judge corporates: ESG ratings ter of this type of evaluation, the role of ESG rating agencies is crucial. We investigate how the ESG criteria are evaluated, and if the data 6. Corporate engagement and stock activism. Exercise of the share- offered by the agencies is reliable. In doing so, the present research holders' rights aiming to influence corporate behavior through firstly provides an overview of the most relevant issues faced by the direct dialogue with corporate management and proposal agencies. Secondly, it illustrates the used methodologies, as well as submissions. the business assessment process, in order to draw conclusions on the 7. Norm-based screening. Investing only in stocks respecting minimum thresholds of ethical business practices based on international principles. achieved consensus among all players. Rating agencies have developed their own assessment methodology to evaluate ESG engagement. Given the different methodologies, Table 1 offers an overview of the factors that major ESG As the description provided above, these investment strategies rating providers are considering within their assessments. All the are complex and required information availability and a deep analysis information has been extracted and pooled from companies' of the firms subject to these investments. Despite this complexity, in websites. The goal of this first analysis is to check whether the 2018, the global sustainable investing assets accounts for 30 trillions agencies converge in terms of methodology and if the ratings they of dollars invested through these strategies and continue to increase provide are consistent. (GSIA, 2018). This indicates a significant demand for information As Table 1 shows, the first difference is related to data sources. about the compliance of a company to the ESG properties and thus, The identification of sensitive data points is, in fact, a crucial prerequi- the need of ESG ratings. site for a good rating. The principal sources for all agencies are publicly available information, such as companies' reports and websites. However, the sources of information change from one rating agency 3 | RATING THE “UNRATABLE” AND IMPLICATIONS ON ESG INDEXES to another. For example, ISS-Oekom and Bloomberg leverage on direct contact with the company, while Thomson Reuters also considers stock market registrations. RobecoSAM, on the other hand, In response to the rising demand for reliable ESG data and ESG rat- uses a completely different approach. The ESG rating assessment for ings, the sustainability rating market has grown noticeably and is now this provider begins with the invitation of the world's largest listed in a phase of consolidation (Escrig-Olmedo et al., 2019). The observed companies to participate in a questionnaire which, along with the interest is mainly a result of the fact that ESG ratings can be provided information, will also assess all the missing information. 15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2177 by Cochrane Netherlands, Wiley Online Library on [25/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 1429 BILLIO ET AL. Human resources, Environmental Human rights, Climate change, Environment, Natural Business behavior, resources, Community, Pollution and Involvement, waste, Corporate Management, governance Environmental opportunities Social Product liability, Human capital, Stakeholder needs, Social opportunities Governance Corporate behavior, Corporate governance Main risk factors Environmental Resource use, Emission, Innovation Social Workforce, Human rights, Community, Product responsibility Governance Management, Shareholders, CSR strategy 178 37 N. Criteria 100 Publicly available information, Interview with stakeholders, information on company policies and practices, company direct contact Munich, Germany 1985 D- to A+ ISS Oekom Milan, Italy 1997 F to EEE ECPI New York, United States 2008 0 to 100 Bloomberg 74 Environmental Environmental strategy policy, Environmental Mgmt, Products, Production process Social & Governance Employees and human capital, Community relations, Markets, Corporate governance & shareholder 80/86 Environmental Carbon emissions, Climate change effect, Pollution, Waste disposal, Renewable energy, Resource depletion Social Supply chain, Political contributions, Discrimination, Diversity, Community relations, Human rights, Governance Cumulative voting, Executive compensation, Shareholders' Rights, Takeover Defense, Staggered boards, Independent directors 120 Company reports, Survey approach Company Publicly available reports, Media information, and news, Company direct Regulatory contact data, Bloomberg and Thomson Reuters, University networks Zurich, Switzerland 1995 0 to 100 RobecoSAM Industry-specific Environment About 21 Climate change strategy, indicators. IndustryEcoefficiency, Energy Factors Specific Mgmt, Env. impact of change indicators. product, Env. Mgmt, according to Three main water risk and impact the industrial dimensions: Social group to Economic Equal opportunities, which a (38/100) Freedom of company Environmental association, Health and belongs (27/100) safety, Human rights, Social Product responsibility, (35/100) Social impact of product, Supply chain mgmt, Taxes Governance Business ethics, Compliance, Independence of the board, Remuneration, Shareholder democracy and structure 155 Public disclosure, Media and news, NGO reports Amsterdam, Netherlands 1992 0 to 100 Sustainalytics Environmental Biodiversity, Climate change, Pollution & resources, Water security, Supply chain Social Labor standards, Human rights & community, Health & safety, Customer responsibility, Supply chain Governance Tax transparency, Risk management, Corporate governance, Anticorruption 300 Publicly available information, Company direct contact, Other sources (governments and NGOs) London, United Kingdom 2001 0 to 5 FTSE Russell BILLIO ET AL. 15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2177 by Cochrane Netherlands, Wiley Online Library on [25/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 38 Company disclosure, Company websites, Company Company reports, Recommendation, disclosure, NGO websites, Media Conventions 1600+ Media and news, Stock sources, 100+ exchange filings specialized dataset Toronto, Canada Sources Headquarter New York, United States 2002 1983 History Paris, France D- to A+ and 0 to 100

to ++ Rating score CCC to AAA 1990 Refinitiv MSCI Key differences between ESG rating agencies using publicly available information on agencies websites Vigeo- Eiris TABLE 1 1430 Proprietary Definition. Ratings are calculated using an Exposure weighted average. Alignment with the UN Sustainable Development Goals (SDGs) Proprietary Definition. Based on principles developed by International Bodies (e.g., GRI, CDP, SASM for three industries. FSB Task Force on Climaterelated Financial Disclosures) Bloomberg ECPI Proprietary Definition. Based on principles developed by International Bodies (e.g., UN Global Compact Initiative and UN PRI) FTSE Russell However, although sometimes sources of information tend to converge, the way they are processed may also be different. Besides, the number of assessed indicators changes among the different raters. MSCI and FTSE Russell represent the extreme cases since they assess respectively 37 and 300 ESG criteria. Other agencies, instead assess different metrics in relation to the industry the company belongs to (see Sustainalytics and RebecoSAM). The main risk factors summarize any single score from the indicators used in the previous step. At this stage, we noticed considerable structural divergences. For instance, ECPI considers two sustainable dimensions since it incorporates the social dimension in the governance sphere. RobecoSAM substitutes the governance dimension with an economic one (which includes also corporate governance). Finally, the difficulty in achieving a general definition of ESG materiality,7 eventually triggering the weighting mechanism of the ISS Oekom Sustainalytics Refinitiv Proprietary Definition. Disclosure of criteria and weighting of the 61 industries analyzed Proprietary Definition. Proprietary Proprietary Definition. Selection and weighting Definition of Standard weighting for of 5 key issues per Materiality all sector and 800 subindustry the categories detailed Industrylevel. Environmental = 34%, Specific Criteria Assessment of Social 35.5%, the potentially Governance = 30.5% material issues in the future RobecoSAM assessed criteria, may generate further divergence in the overall rating. As Table 1 shows, all the examined ESG rating agencies have developed a proprietary definition of materiality. Consequently, the weighting procedures vary considerably. Furthermore, we observe that almost all rating agencies adjust their final rating by integrating issues, specific to the considered industry, but very few of them publish the assigned weights transparently. Thus, the discrepancy across ratings may simply reside in the consideration of different components weighted differently. In order to further illustrate the existing disparity, we provide an example including four ESG providers: Sustainalytics, 8 RobecoSAM, Refinitiv and MSCI. Table 2 reports the ESG rating for four companies (Verizon Communications Inc., Nissan Motor Co., Ltd, Oracle Corp. JPN and Goodman Group). The rating scale for Sustainalytics, RobecoSAM and Refinitiv ranges from 0 to 100 while MSCI provides a classical taxonomy ranging from CCC to AAA scale. As shown in Table 2, the ratings differ considerably across the providers. For instance, in the case of Nissan Motor Company, Sustainalytics and MSCI assign a low rating while RobecoSAM and Refinitiv include it among the best-in-class. These discrepancies can be explained by the fact that both Sustainalytics and MSCI focus on the managerial perspective investigating the firm's exposure to ESG MSCI Proprietary Definition. Based on principles developed by International Bodies. Materiality Proprietary Definition. and weighting Analysis on material risks and opportunities for all the GICS subsectors TABLE 1 (Continued) Vigeo- Eiris risk and its mitigation strategy (e.g., unmanaged ESG risks and potentially higher production costs for its carbon-intensive products). On the other hand, despite giving a low ESG controversial score, Refinitiv judges Nissan positively, especially in the environmental and governance spheres. Finally, RobecoSAM overweights the environmental dimension for the automotive industry. In particular, operational ecoefficiency and climate strategy account for 20% of the overall rating. As a result, the ranking given by RobecoSAM awards Nissan's cleantech innovation capacity compared to the other analyzed competitors (32 companies). Similar considerations can be formulated for the remaining companies reported in the table and for which we observe diverging ESG ratings inherently related to the concerned metrics (i.e., Verizon Communications is very high in Sustainalytics and very low in RobecoSAM while it is in the middle range in MSCI and Refinitiv). This simple 15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2177 by Cochrane Netherlands, Wiley Online Library on [25/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 1431 BILLIO ET AL. BILLIO ET AL. Company Sustainalytics Nissan Motor Co., Ltd RobecoSAM Refinitiv MSCI 6 77 72 CCC Verizon Communications Inc. 91 20 67 BB Oracle Corp. Jpn 78 8 63 BB Goodman Group 86 21 58 AA T A B L E 3 Rank correlation (Top panel), mean absolute error (Mid panel) and the percentage of observed agreement (Bottom panel) among the considered rating providers Sustainalytics RobecoSAM Refinitiv MSCI T A B L E 2 Example of divergence in ESG ratings for Verizon Communications Inc., Nissan Motor Co., Ltd, Oracle Corp. JPN and Goodman Group describing the proportion of firms in the sample, for which the ESG rating agencies agree to provide the same ESG rating after applying the conversion common scale described above (Bottom panel in Table 3). The table shows that the agreement of the ESG ratings is on average 24% and ranges from 19% (between RobecoSAM and MSCI) to 28% Sustainalytics - RobecoSAM 0.69 - Refinitiv 0.64 0.69 - MSCI 0.53 0.45 0.43 - Sustainalytics RobecoSAM Refinitiv MSCI (between Sustainalytics and RobecoSAM). Thus, both measures confirm the existence of a strong evaluation disagreement among the different rating providers, for the same considered companies.11 Given the obvious discrepancies in terms of methodology and ESG ratings assigned to the same firms, it is important to identify the channels that might affect Sustainalytics - RobecoSAM 1.2812 - Refinitiv 1.1878 1.3937 - MSCI 1.3260 1.5872 1.1144 - implemented metrics vary among the rating providers, there will also be Sustainalytics RobecoSAM Refinitiv MSCI disagreement in the selection of the ESG indexes constituents. We Sustainalytics the ESG indexes construction. As noted also by Engle et al. (2019), heterogeneity in the ESG scores can lead to the formation of different ESG portfolios due to disagreement in the ratings. In fact, if the investigate this issue in detail in the next subsection. - RobecoSAM 28.22% Refinitiv 23.74% 20.59% - MSCI 25.36% 19.46% 27.93% 3.2 | ESG indexes agreement - Note: On average, the rank correlation, the mean absolute error and the percentage of the observed agreement are equal to 0.58%, 1.32% and 24%, respectively. The quantity of actively managed portfolios combining environmental, social and governance factors has grown considerably. According to iShares (2019), more than 1000 indexes, including ESG preferences expressed by the investors, are now available on the market. example clearly highlights that the currently applied metrics can lead to contradictory evaluations. Given the lack of a globally accepted standard methodology, we investigate the discrepancies among the ESG indexes, provided by the In order to better understand the ratings' divergence across pro- four ESG rating agencies considered above (i.e., Sustainalytics, viders, we consider a large set of 1049 companies listed in the MSCI RobecoSAM, Refinitiv and MSCI) and focusing on developed markets World Index with available ESG ratings from the four providers men- investment set. tioned above.9 Firstly, we compute the rank correlation among the Table 4 specifies the names of the four ESG indexes, their associ- ESG ratings to measure the divergence between the considered rating ated ESG Rating providers and the number of their constituents. agencies (Top panel in Table 3). Despite the different dimensions in terms of constituents, the consid- The correlation between ESG ratings is on average 0.58 and ranges ered indexes remain quite similar in terms of sectorial and geographi- from 0.43 (between Refinitiv and MSCI) to 0.69 (between Refinitiv and cal composition. As Figure 1 shows, at the sectorial level, all the four RobecoSAM). Similarly to Berg et al. (2019), our results confirm a dis- indexes have a similar composition. For each class the average differ- agreement among the different rating providers, for the same consid- ence ranges from 1% to 3%. The largest difference is 5% in consumer ered companies. Secondly, we compute the mean absolute error (MAE) discretionary (STOXX and MSCI) and Information Technology in order to evaluate the magnitude of their evaluation differences in (Refinitiv and MSCI). At the geographical level, the average difference term of classes. (Mid panel in Table 3).10 In relation to this, we harmo- is 15% (North America), 16% (Europe) and 6% (Asia Pacific). As can be nize and convert the rating classes among the four considered pro- seen from Figure 2, the difference between North America and viders by applying a common scale ranging from 1 to 7. The conversion Europe can be explained mainly by STOXX and MSCI. Based on the table is reported in the Appendix A. The MAEs of the ESG ratings are average exposition, the first index is over-(under)exposed to Europe on average 1.32 and range from 1.11 to 1.59. Given that the total num- (North America) while the second is less (more) exposed to North ber of rating classes is 7, it implies that the observed disagreement America (Asia Pacific). among rating providers is on average larger than one class. Thirdly, we The sectorial and geographical analysis of the ESG indexes com- compute the percentage of agreement among the ESG ratings, position highlight the presence of some geographical differences, 15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2177 by Cochrane Netherlands, Wiley Online Library on [25/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 1432 T A B L E 4 List of the considered four rating providers and their corresponding ESG indices including the number of their constituents (last column) Label Rating Provider ESG Index Number of Constituents STOXX Sustainalytics STOXX Global ESG Leaders 431 Dow Jones RobecoSAM Dow Jones Sustainability World 317 Refinitiv Refinitiv Refinitiv Global ESG 404 MSCI MSCI MSCI World ESG Leaders 777 20% 18% 16% 14% 12% 10% 8% 6% 4% 2% 0% STOXX DOW JONES REFINITIV MSCI F I G U R E 1 The sectorial composition for the four considered indices. STOXX (blue), Dow Jones (red), Refinitiv (green) and MSCI (purple). The sectors are: consumer services, consumer discretionary, consumer staples, energy, financials, health care, industrials, information technology, materials, real estate and utilities [Colour figure can be viewed at wileyonlinelibrary.com] despite the significant overlap in terms of geographical coverage. In disagreement (the ESG disagreement portfolio). To check if the two line with the previous section, we aim to measure the agreement portfolios are coherent, we compute the MAE and the percentage of between the rating providers also from this perspective. To investi- observed agreement on the ratings of the constituents across the gate the degree of similarity on the selection of the constituents, that four rating agencies. The MAE of the ESG agreement (disagreement) is, the agreement, we compute the overlap coefficient by using the portfolio is on average 0.73 (1.44) and ranges from 0.46 (1.18) to Szymkiewicz–Simpson coefficient. The latter corresponds to the size 0.90 (1.79). The percentage of observed agreement of the ESG of the intersection between two indexes divided by the size of the agreement (disagreement) portfolio is on average 46% (22%) and smallest one.12 ranges from 33% (17%) to 68% (27%). Results confirm that the con- Table 5 shows that the agreement rate among the evaluated indexes is low and ranges from 35% to 59%. In this respect, the low- sensus among the rating providers is stronger in the ESG agreement portfolio. est agreement rate is observed between the Refinitiv and the Dow However, the observed low agreement rate in ESG indexes might Jones indexes while the highest is observed between STOXX and be related to the differences in terms of geographical exposition MSCI indexes. The overall overlap coefficient is 15% (48 constituents), rather than to ESG ratings. For this reason, we perform an additional which confirms that the divergence is greater when considering all the analysis, by computing the overlap measures across the different geo- indexes together. graphical areas. Thus, we aim to evaluate how agreement and dis- Therefore, having an index where there is agreement between agreement rates might evolve after controlling the geographical the four ESG rating providers would imply to consider all the firms allocation of the ESG indexes constituents. It is worth noting that we that have been included in all the four ESG indexes. This is exactly are considering ESG global indexes that might be allocated with differ- the portfolio coming from the overall overlap (the ESG agreement ent intensities within the same geographical areas. This might imply portfolio). Conversely, an index composed of the firms that have that the overlap coefficient is low just because the two indexes focus been considered only by one rating agency in its index would imply on different geographical areas.13 15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2177 by Cochrane Netherlands, Wiley Online Library on [25/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 1433 BILLIO ET AL. BILLIO ET AL. 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% North America Europe STOXX DOW JONES REFINITIV Asia Pacific MSCI F I G U R E 2 The geographical composition for the four considered indices. STOXX (blue), Dow Jones (red), Refinitiv (green) and MSCI (purple). The geographical location is divided according to three main area: North America, Europe and Asia Pacific [Colour figure can be viewed at wileyonlinelibrary.com] TABLE 5 Therefore, the geographical allocation might influence the measured Overlap coefficients among four ESG indices STOXX Dow Jones Refinitiv MSCI overlap effect, but on average such impact remains largely marginal.14 Thus, we conclude that the disagreement across ESG indexes persists STOXX - Dow Jones 50% - Refinitiv 43% 35% - MSCI 59% 49% 50% even after controlling the overlap coefficient for the geographical area. The above described results highlight plausible doubts on the achievement of a consensus among ESG ratings, since it is evident that - Note: The indices include companies from the developed markets that exhibit high ESG performance according to the related provider. The overall overlap coefficient is 15%. ESG rating agencies provide heterogeneous information to the market not only in terms of ESG ratings but also in terms of ESG indexes and their inherent constituents. The lack of a globally accepted standard methodology and a minimum level of technical requirements create two major issues. From one perspective, investors face considerable difficulties in Clearly, we expect an increase in the agreement rate for those selecting ESG targets for investment, and conversely, companies encoun- indexes with a similar exposition in a given area with respect to the total ter significant difficulties in identifying the characteristics they should composition. normalize comply with, in order to be included into ESG indexes. This assessment the geographical allocation and look for the overlap. The analysis shows further motivates us to analyse how the agreement and disagreement that for the ESG indexes that have a significant difference in the geo- concerning the inclusion of ESG indexes constituents can affect the finan- graphical allocation (e.g., STOXX and MSCI) the overlap coefficient cial performance of ESG portfolios with respect to a nonESG counterpart. To further investigate this issue, we increases by 10% for North America, 5% for Europe and 1% for Asia. However, for the indexes with very similar exposition, for example, Refinitiv and MSCI in North America (difference of 4%), Dow Jones and Refinitiv in Europe (3%), and STOXX and Refinitiv (0%) and Dow Jones 4 | THE P ERFORMANC E O F ES G AGREEMENT AND nonESG PORTFOLIOS and Refinitiv (5%) in Asia Pacific, the agreement decreases for Refinitiv and MSCI in North America by
3% and for STOXX and Refinitive in In this section we aim to assess whether the agreement of the inclu- Asia Pacific by
7% while, it increases for Dow Jones and Refinitive in sion of stocks in an ESG index by the four considered ESG rating Europe, +19%, and Dow Jones and Refinitiv in Asia Pacific, +5%. agencies (ESG agreement portfolio) affects the financial performance, 15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2177 by Cochrane Netherlands, Wiley Online Library on [25/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 1434 that is, if it is generating an extra-performance after controlling for structural change in the financial performance of ESG and nonESG port- the remuneration of financial risk factors. The performance measure folios. The first period ranges from December 1999 to December 2004 that captures this feature best is the Jensen-alpha. We aim to also while the second one starts in January 2005 and ends in January 2020. check whether the performance of the ESG agreement portfolio are The total returns for the constituents have been downloaded from better than a nonESG portfolio. Eikon/Datastream at monthly frequency.16 The portfolios are equally weighted. Table 6 includes both the performance measures for the ESG 4.1 | Building the portfolios agreement and the nonESG portfolios. The second portfolio shows higher Sharpe and Sortino ratios in the first period (1.22 and 1.53) with respect to the ESG portfolio (0.25 and 0.38) while in the second The ESG and the nonESG portfolios are built as follows: period the two ratios are more similar. The same conclusions hold true for the Omega ratio and the Max Drawdown and Var at 5%.17 1. The ESG agreement. We consider the common constituents of the four indexes described in Table 4. Coherently with the previous This simple analysis indicates that the nonESG portfolio is better analysis, if a firm is included as the constituent of an index, it is con- preferable than the ESG agreement portfolio from the financial point of sidered as an ESG leader according to the corresponding rating pro- view. Nevertheless, it also suggests that there might have been a change vider. Consequently, there is agreement when a stock is included in in investors' preferences or tastes after 2005 with the introduction of the all the four indexes. This portfolio is originated from the overall ESG firm's characteristics. Indeed, the performances of the two portfolios overlap discussed in Section 3.2, and includes 48 constituents. are closer in the second period. However, the presence of a Jensen-alpha 2. The nonESG portfolio. “nonESG” stocks have a broader definition. that captures the ESG component in the market could only be investigated Indeed, they are not only related to a certain type of industry, but after controlling for the remuneration of the usual financial risk factors. they also depend on the firm's characteristics and attributes (E, S In regard to this, we estimate the Jensen-alpha of both portfolios and G components). We build the nonESG portfolio following a neg- representing ESG agreement stocks and the corresponding counter- ative screening approach which considers all the stocks that are part (nonESG portfolio) using the Carhart four-factor model excluded from the ESG investment universe. In order to identify (Carhart, 1997). The model is an extension of the Fama–French three- these nonESG firms we consider all firms that that do not comply factor model (Fama & French, 1992) and allows one to disentangle with the principles of the UNGC which is where the concept of ESG the portfolio performance measuring the impact of (1) the market risk, originated. We define the nonESG portfolio as follows. Firstly, we (2) the outperformance of small versus big companies, (3) the out- consider the constituents of the MSCI World Index which represent performance of high book/market versus small book/market compa- a good proxy of the investment universe of the developed markets. nies and (4) the momentum factor.18 After controlling for these Secondly, we consider the constituents of the MSCI World ESG factors, the difference on the alphas between ESG and its counterpart screened Index which excludes firms from the MSCI World index should reveal if an ESG impact does exist on stock performance. that are not in compliance with the UNGC principles and are The estimated results are reported in Table 7 for both considered involved in controversial or unacceptable activities which may hurt periods. The first column in the table shows the estimation for the ESG the environment or society as a whole.15 Finally, the nonESG port- agreement portfolio while the second shows the estimation for the folio is obtained as the complement of the MSCI World ESG nonESG portfolio for the period 2005–2004. In order to measure screened and includes 119 constituents. None of these 119 constit- whether the observed difference between the two portfolios is statisti- uents have been included in any of the other three ESG indexes. cally significant, we estimate a long/short portfolio which is built between the ESG agreement portfolio (long position) and the nonESG (short position). After controlling for the four-risk factors, a positive 4.2 | Results (negative) significant alpha would imply that the ESG portfolio overperforms (is outperformed by) the nonESG portfolio, while a non- Since the term ESG was first coined in 2005, it is reasonable to divide significant alpha, would imply that there is no statistical difference the analysis into two periods in order to detect the presence of a in the two portfolios. The result is reported in the third column. TABLE 6 Performance measures for the ESG agreement and the nonESG portfolios for the considered periods Sharpe ratio 2000–2004 Sortino ratio Omega ratio MaxDD VaR 5% ESG nonESG ESG nonESG ESG nonESG ESG nonESG ESG nonESG 0.254 1.224 0.377 1.535 1.209 2.441
31.9%
18.9%
7.3%
5.0% 2005–2019 0.870 1.044 1.153 1.287 1.953 2.144
47.7%
34.5%
6.7%
4.7% 2000–2019 0.691 1.092 0.945 1.361 1.694 2.217
47.7%
34.5%
6.9%
4.9% Note: The first three columns provide the annualized Sharpe Ratio, Sortino Ratio and Omega Ratio respectively. The forth column provides the Max Drawdown (MaxDD) and the last provides the Value at Risk at 5%. 15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2177 by Cochrane Netherlands, Wiley Online Library on [25/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 1435 BILLIO ET AL. BILLIO ET AL. TABLE 7 Estimates of the Carhart four-factor model for the ESG agreement and the nonESG portfolios for the two periods 2000–2004 2005–2019 ESG nonESG ESG–nonESG ESG nonESG ESG–nonESG Alpha 0.0049 (0.004) 0.0037 (0.003) 0.0012 (0.005) 0.0059*** (0.001) 0.0053*** (0.001) 0.0006 (0.001) Mkt-Rf 0.808*** (0.093) 0.6669*** (0.074) 0.1415 (0.087) 0.8552*** (0.035) 0.6838*** (0.027) 0.1713*** (0.038) SMB 0.1888 (0.182) 0.0353 (0.077) 0.1536 (0.454)
0.1364* (0.075)
0.0714 (0.070)
0.065 (0.088) HML
0.0177 (0.133) 0.5479*** (0.088)
0.5655*** (0.126) 0.0352 (0.080) 0.1326** (0.067)
0.0975 (0.092) WML
0.241*** (0.062)
0.0811 (0.056)
0.1599** (0.079)
0.1187* (0.067) 0.0721** (0.034)
0.1908*** (0.055) F-Statistic 45.63 59.04 18.97 202.5 203.5 11.37 0.77 0.752 0.54 0.821 0.831 0.252 60 60 60 180 180 180 R 2 Obs. Note: The last column (ESG-nonESG) includes the estimate on the long (ESG) – short (nonESG) portfolio. Robust standard errors (HAC) are reported within brackets. Statistical significance is denoted by ***, **, and * at the 1%, 5%, and 10% level, respectively. The following three columns describe the same type of analysis but for 1. The sectorial analysis. We perform an analysis by distinguishing the period 2005–2019. In the first period, the alphas are not significant among the different sectors that might be more or less exposed to neither for the ESG agreement nor for the nonESG portfolios. How- the impact of ESG risks. In the same manner, we repeat the long- ever, in the second period, both the ESG and nonESG portfolios pro- short portfolio exercise and investigate for any differences in the vide a similar positive and significant alpha of 0.59% and 0.53%, alphas between the sectorial ESG agreement and the nonESG respectively. Conversely, the alphas on the long-short portfolio are not portfolios. Results are included in Appendix D and show that, significant for both periods and hence, it appears that, there are no except for the communication services sector in 2005–2019, the significant portfolio performance differences after controlling for the alphas are not statistically significant. four risk factors. According to our findings, the observed results are a 2. The ESG matched portfolio. Another potential issue that might consequence of the previously detected scores' disagreement among affect our results, concerns the Carhart model that might not be the rating agencies, to the extent that even when there is agreement, able to fully capture the portfolio characteristics due to a mismatch the ESG effect is diluted and it has no impact on performances. on the sectorial and geographical composition (see Figure E1 in Nevertheless, the ESG agreement and the nonESG portfolios show the Appendix). To address this issue, we have performed an addi- some differences in terms of their exposition to the four factors. In the tional analysis using a matched ESG agreement portfolio with the first period, there is a negative and significant difference with the high nonESG portfolio through the propensity score approach minus low (HML) factor and the winners minus losers in the previous (Heckman et al., 1998). Results are included in Appendix E and 12 months (WML) factor. The first factor indicates that the nonESG confirm that the alpha in the long short portfolio is not statistically portfolio is exposed to growth stocks while the second highlights that significant. Moreover, the two portfolios are very similar in terms the ESG portfolio is negatively related to the momentum. The latter is of exposition to the four factors. confirmed in the second period where there is also a positive significant 3. The rebalanced portfolio. An additional constraint lies in the non- difference in terms of market exposition for the ESG portfolio. This is synchronized data availability, that is, the indexes and their constit- also consistent with the findings in Monasterolo and De Angelis (2020) uents' data start at different points of time.19 For the analysis, we where the green asset class shows a market exposition that is larger have selected the constituents of the indexes in 2019 as reference than the brown one. Therefore, this requires some robustness checks in order to include the pre-ESG era for all the four considered to validate the results and assess their consistency. This procedure is indexes. As a robustness check, we have considered a rebalanced especially important in order to avoid that our results being biased due ESG agreement portfolio which tracks the periodic revision in all to portfolio characteristics not related to ESG features. the four considered indexes. The commonly available period for the four indexes starts from January 2012. Results are included in the Appendix F. The analysis shows that the portfolio rebalance 4.3 | Robustness Analysis does not affect our results. 4. ESG indexes commonly available period. As highlighted above, ESG To validate our results, we perform four robustness analyses on the indexes started in different periods. To assess if the different ESG portfolio. starting dates affect our results, we run the Jensen-alpha 15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2177 by Cochrane Netherlands, Wiley Online Library on [25/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 1436 performance test for the 2012–2019 period. This complementary We provide evidence on this aspect and confirm that the test confirms our previous results: the Jensen-alpha for the long- observed ESG rating heterogeneity leads to the identification of alter- short ESG-nonESG portfolio is not statistically different from zero. native benchmarks, since the common component in terms of constit- Detailed results are reported in the Appendix F. uents is represented by a relatively small sample of firms. The quota is 5. The ESG disagreement portfolio. We also consider the disagreement 15% for the ESG index with the smallest number of constituents and among the rating providers and build an alternative portfolio: the ESG 6% of the largest one. In this respect, we further develop our analysis disagreement portfolio (i.e., the one where the stocks considered only and investigate the financial implication of ESG disagreement among have one ESG rating agency that included it in its ESG index). Also in rating agencies. Namely, we evaluate the financial performance of