Empirical evidence on the role of proxy advisors in European capital markets*

Transcription

1 Empirical evidence on the role of proxy advisors in European capital markets* Joerg-Markus HITZ Chair of Accounting and Auditing Faculty of Economic Sciences Göttingen University Nico LEHMANN** Faculty of Economic Sciences Göttingen University This version: November 2016 *Acknowledgments: For valuable comments, we are grateful to Yuping Jia (discussant), Oksana Pryshchepa (discussant), Christophe Volonté, and participants at the 11 th Workshop on Corporate Governance in St. Gallen, Switzerland (October 2014), the Annual Meeting of the European Accounting Association in Glasgow, UK (April 2015), and the Annual Meeting of the European Financial Management Association in Breukelen / Amsterdam, Netherlands (June 2015). We also thank Kristof Ho Tiu (Institutional Shareholder Services) and Carla Topino (Glass Lewis) for kindly providing data on ISS coverage and GL voting reports / GL coverage, respectively. ** Corresponding Author. I

2 Empirical evidence on the role of proxy advisors in European capital markets ABSTRACT Responding to regulators informational requests and filling a void in the academic literature, this paper provides comprehensive empirical evidence on the role of proxy advisors in 14 European countries. Exploiting coverage data and using content analysis of proxy voting reports by Institutional Shareholder Services and Glass Lewis, we provide descriptive analyses of proxy advisors firm coverage, the variation and determinants of voting recommendations, the relation between voting recommendations and shareholder voting at AGMs, and market reactions to the release of voting recommendations. Overall, this evidence is in line with an economically important role of proxy advisors in European markets. Throughout our analyses, we document that this role varies with governance- and ownershiprelated firm characteristics, and with country-level measures of institutional strength. JEL Classification: C21, G34, M41 Keywords: Information intermediaries, proxy voting advisory, corporate governance II

3 1. Introduction The European Commission and the European Securities and Markets Authority (ESMA) have recently raised concerns about the role of proxy advisors (PAs) in European capital markets, in particular with respect to their influence on voting outcomes at annual general meetings (AGMs) (ESMA, 2012). PAs are firms that advise shareholders, most importantly institutional investors, on how to vote on AGM items by providing voting recommendations. Hence, PAs aim to facilitate and support one of shareholders key governance instruments, namely their vote at AGMs ( voice ) (ESMA, 2012). From an economic perspective, PAs represent information intermediaries, i.e., institutions that collect, process, and disseminate information, mostly on the quality of firms corporate governance institutions and mechanisms, such as the independence of board members, or the effectiveness of the compensation system. Particularly for institutional investors covering firms from different countries with different governance frameworks, the services provided by PAs potentially reduce information and monitoring costs (ESMA, 2012). At the same time, the activities of these so far unregulated institutions attract the attention of European regulators. A discussion paper published by ESMA in 2012 outlines key issues of regulatory concern with the proxy advisory business, and centers on the perceived growing influence of these institutions on shareholder voting and hence on governing large European corporations. However, ESMA cautions that at the same time, it does not have sufficient information, e.g., data on the presence and market share of PAs, and their factual influence at AGMs, to reliably assess the importance and role of PAs in European capital markets (ESMA, 2012) as a prerequisite to reaching final regulatory conclusions. While there is a growing academic literature on the economic role of PAs, this literature does only partly speak to the informational demands of European regulators, as the respective papers are typically framed in the context of the U.S. capital market. This extant literature investigates various aspects of the role and influence of PAs, such as the determinants of PAs 1

4 voting recommendations, PAs influence at AGMs, and the market reaction to the release of voting recommendations (e.g., Alexander, Chen, Seppi, & Spatt, 2010; Bethel & Gillan, 2002; Cai, Garner, & Walking, 2009; Ertimur, Ferri, & Oesch, 2013, 2015; Larcker, McCall, & Ormazabal, 2013, Malenko & Shen, 2016). Consistent with an information intermediation role of PAs, this literature provides ample evidence that the activities of PAs significantly shape voting outcomes at AGMs, and are economically important. However, it is not obvious whether and how these findings translate to other markets and jurisdictions, in particular in Europe, and whether the economic role of PAs differs with the respective institutional setting they operate in. For one thing, the PA business has only recently evolved in European markets, and there is no empirical data on the presence or the role of these institutions other than anecdotal evidence (ESMA, 2012). Second, to varying degrees, the governance frameworks in European countries differ to the U.S. These differences, among other things, pertain to the legal framework, e.g., the rule of law, the effectiveness of public and private enforcement, as well as investor protection (Djankov, La Porta, Lopez-de-Silanes, and Shleifer, 2008). In addition, such differences also bear down on governance mechanisms on the firm-level, e.g., with respect to the role of large shareholders and bank financing, which are more pronounced in Europe than in the U.S. (e.g., Goergen, Manjon, & Renneboog, 2008). We conclude that there is demand both in the regulatory sphere as well as in the academic sphere for information on the role that PAs assume in European markets. Our paper directly speaks to this demand, by providing descriptive evidence on this role. In particular we address four questions of interest: (1) How present are PAs in European markets, and what are the characteristics of firms (e.g., ownership structure, governance quality) they choose to cover? (2) What are the characteristics of PAs voting recommendations, and what kind of firm characteristics are associated with negative voting recommendations? (3) To what extent are voting outcomes at European AGMs associated with PA voting recommendations, 2

5 consistent with an influential role of PAs? (4) What is the economic significance of PA voting recommendations, measured by market reactions to their publication? Finally, to shed light on the role of governance frameworks, for any of these four questions we are interested in the extent to which the pertinent findings vary on the country level, and are hence associated with institutional differences between European countries, in particular with respect to shareholder protection. We address our four questions using four sets of empirical analyses. These analyses are based on a comprehensive, manually-constructed sample of coverage information (from 2008 to 2010) and voting reports (for 2010) of two leading PAs, Institutional Shareholder Services (ISS) and Glass Lewis (GL), across 14 European countries. Our first set of analyses reveals that around 60% (34%) of European listed firms are covered by ISS (GL) during the years We demonstrate that the relative market share of PAs varies significantly on the country-level, and is negatively associated with the quality of investor protection. This finding is consistent with a higher supply and hence, potentially higher demand for PA services in countries with a greater likelihood of outside shareholder expropriation. On the firm-level, the PA coverage decision correlates with ownership structure in a way that suggests that the demand for proxy services is associated with outside ownership. Our second set of analyses provides detailed descriptive data on the 2010 voting reports, revealing that in 14.9% (17.5%) of the cases ISS (GL) issues recommendations to vote against the respective management proposal. For a subset of non-routine proposals addressing compensation or shareholder rights issues, the proportion of ISS (GL) vote against recommendations increases to 28.8% (36.3%) and 22.9% (18.8%), respectively. Using multivariate analyses, we show that the likelihood for a PA to arrive at a vote against recommendation correlates with firm-level inside ownership, governance quality, and country-level investor protection standards. These findings are consistent with documented preferences of U.S. institutional investors suggesting that PAs assume a more skeptical perspective when outside shareholders fear expropriation 3

6 (Iliev, Lins, Miller, and Roth, 2015). In line with this, we find that the PAs agreement that is, the overlap of recommendations by ISS and GL on controversial cases is strongest for firms with weak governance and for firms from countries with weak investor protection. Addressing the relation between PAs recommendations and actual shareholder voting at AGMs, our third set of analyses reveals that PAs vote against recommendations are negatively associated with voting outcomes. On average, ISS (GL) vote against recommendations correlate with 9% (6%) voting dissent. This correlation is weaker (higher) for firms in countries with a low (high) level of investor protection standards, with a low (high) level of corporate governance, with high (low) insider control, and with low (high) free float. The latter findings on the moderating role of ownership are in line with prior U.S. studies (e.g., Ertimur et al., 2013; Iliev and Lowry, 2015; Malenko & Shen, 2016). These studies find that the relationship between PA recommendations and voting outcome is stronger for firms with a small fraction of blockholdings and more dispersed (professional) ownership, consistent with these investors having weaker incentives to engage in own research, and hence being more likely to follow PAs. Also, on a more general level, our findings on the moderating role of governance and institutional differences can be linked to earlier country-level analyses such as La Porta, Lopez-De-Silanes, Shleifer, and Vishny (1998, 2002) suggesting that low investor protection standards and entrenched management render it harder for outside shareholders to translate voting dissent into effective monitoring. Our fourth and final set of analyses investigates market reactions around the release of ISS and GL proxy reports. Our findings reveal significant negative capital market reactions particularly for ISS reports around the release of negative ( vote against ) recommendations. These reactions are more pronounced for firms with weak information environment, concentrated ownership, high insider control, and especially for firms from weak investor protection countries. These findings suggest that investors attribute information content to the 4

7 news that ISS disagrees with management proposals, which is especially the case for firms with poor information environment and weak incentives to signal their governance quality. Taken together, our findings provide a comprehensive view on the role of PAs in European markets. As such, they can be relevant and informative for European regulators, who observe that European evidence on the role and influence of PAs on voting outcomes is an important prerequisite to evaluating potential policy options for the PA industry (ESMA, 2012, p. 33). For one thing, our findings shed light on the market presence of two leading PAs across Europe and document the market leadership position of ISS. This directly responds to ESMA s information request for information on market characteristics of the European proxy voting industry (ESMA, 2012). In addition, our analyses of factors shaping PAs recommendations provide insights into the voting guidelines and evaluation approaches of these firms. This also relates to discussions and concerns that PAs originally domiciled in the U.S. may not sufficiently take into account local European market conditions when developing their proxy recommendations within pre-defined voting guidelines, potentially applying what is referred to as a one-size-fits-all or a tick-the-box approach (ESMA, 2012). Although it is beyond the scope of our study to analyze the voting guidelines that underlie PAs actual recommendations, our findings document substantial firm- and countrylevel differences in the disapproval rate for both PAs. At least on the surface, these findings are not consistent with a mechanical application of presumably U.S.-flavored voting guidelines to European settings. Finally, our cross-country findings provide important insights into the question of whether the role and presence of PAs differs across Europe. This is important because EU regulation needs to take into consideration the supra-national aspects of the subject matter. Our findings provide such a supra-national perspective on the proxy voting industry. Our paper also fills a void in the academic literature on the economic role of PAs, which we supplement by a cross-country, European capital markets angle. To our knowledge, ours is 5

8 the first paper to provide comprehensive descriptive evidence on the economic role of the two leading PAs outside the U.S., and to document how this role varies with respect to crosscountry differences (e.g., differences in country-level strength of investor protection and outside shareholder expropriation). In particular, our findings reveal a similar market structure in Europe and the U.S. with respect to PAs market share, and the role of ISS as the market leader. In terms of PAs dissent, our findings show that ISS rejection rates are markedly higher in Europe compared to those reported in U.S. studies, whereas the frequency of GL rejection rates differ less between Europe and the U.S. and are higher than the ISS rates. In terms of voting guidelines and recommendations, our findings show that, consistently across Europe and the U.S., ISS s vote against recommendations are mostly relate to board and compensation-related governance issues, whereas GL focuses primarily on audit and board issues when opposing the management. At the same time, while we document on average higher PA rejection rates in Europe, our findings reveal that the correlation between PA recommendations and voting outcomes is considerably lower in economic terms as the correlation documented in prior U.S. studies. Finally, we believe that our paper also provides implications for international research on shareholder voting and corporate governance (e.g., Iliev et al., 2015). While this literature highlights cross-country differences in the voting dissent of U.S. institutional ownership, little attention has been paid to cross-country determinants of the actual voting outcome and the role of governance-related information intermediaries. Our paper adds to this literature by providing initial insights into the question of whether outside shareholders utilize and demand information intermediation services such as proxy voting advisory in the light of institutional differences. In addition, it sheds light on the question of whether actual voting outcome and outside shareholder dissent as potentially reflected in PAs recommendations varies with institutional characteristics. With respect to the relation between shareholder dissent and PA dissent, Ertimur et al. (2015) detail that extensive consultations and roundtables with PAs 6

9 clients primarily institutional investors commonly precede the issue of voting guidelines, and thus presumably form and guide PAs voting recommendations at AGMs. Consequently, our findings regarding factors and consequences associated with PAs dissent at AGMs, especially in the light of cross-country differences, should be of interest to this literature. The remainder of the paper is structured as follows. Section two provides background information on the economic role of PAs and the European proxy advisory industry and its regulation. Sections three, four, five, and six address our four questions of interest, providing detailed analyses on the role of PAs in the European context. The last section concludes. 2. Background 2.1 The Economic Role of Proxy Advisors PAs such as ISS and GL provide advisory services to institutional investors. Their main business is to develop recommendations on how to vote on AGMs agenda items. In addition, PAs may also offer a range of AGM-related services, such as governance research, customized voting guidelines for investors, or providing voting logistics. In some cases, they even exercise the voting decision in their own discretion on behalf of the investors (Choi, Fisch, & Kahan, 2010). From an economic perspective, PAs represent information intermediaries, i.e., institutions which collect, process (i.e., aggregate and enhance), and disseminate governancerelated information. Information intermediaries potentially create value in at least two ways. For one thing, PAs may use their expertise to efficiently gather available information, e.g., disclosures of director compensation, hence creating transaction cost savings for investors (information aggregation role of PAs). Second, PAs may command a superior analytical technology, e.g., for evaluating the firm-specific suitability of director compensation schemes, hence helping investors to make improved decisions when casting their votes at AGMs 7

10 (information production role of PAs). Prior literature primarily provides evidence in favor of the former role (e.g., Alexander et al., 2010; Ertimur et al., 2013). More recent studies provide additional perspectives on the information aggregation role of PAs, arguing that PAs voting recommendations aggregate, synthesize, and coordinate institutional investors preferences (Calluzzo & Dudley, 2015; Ertimur et al., 2015). For example, Calluzzo and Dudley (2015) develop a model of outside shareholder coordination in the context of proxy voting advisory, which suggests that the ability of PA s to aggregate private information in the process of shareholder coordination may improve the quality of contracting within a firm. In a related vein, Ertimur et al. (2015) shed light on the PAs agenda setting role by showing that the shareholders voting dissent tends to focus only on those issues and problems that have previously been highlighted by PAs. The underlying idea is that PAs recommendations reflect the consensus of their clients preferences, synthesized and aggregated in the process of developing voting guidelines prior to the proxy season. 2.2 The Proxy Advisory Industry The first PA firms (i.e., ISS in the U.S. and PIRC in the UK) were founded in the 1980s (ESMA, 2012, pp ). In subsequent years, and especially in the 2000s, many new PAs evolved both in the U.S. (e.g., GL, Proxy Governance, or Egan-Jones Proxy Services) and in Europe (e.g., Proxinvest in France, Manifest in the UK, or IVOX in Germany). Nowadays, ISS, a former subsidiary of RiskMetrics Group and MSCI, is considered to be the leading PA in the world, followed by GL (ESMA, 2012). According to Daines et al. (2010), ISS provides proxy voting services for over 1,700 institutional investors managing $ 26 trillion in assets, including, for example, 24 of the top 25 mutual funds. 1 In the U.S., business press, policy makers, and corporate managers perceive PAs, especially ISS and GL, as influential and powerful (Choi et al., 2010). Delaware s former 1 In contrast to other PAs, ISS also markets commercial governance ratings. These ratings, which were introduced in 2002, cover more than 8,000 firms across 31 countries (RiskMetrics, 2007). 8

11 Vice-Chancellor Leo Strine Jr., for example, observed that: [ ] powerful CEOs come on bended knee to Rockville, Maryland, where ISS resides, to persuade the managers of ISS of the merits of their views about issues like proposed mergers, executive compensation, and poison pills. They do so because the CEOs recognize that some institutional investors will simply follow ISS s advice rather than do any thinking of their own. ISS has been so successful that it now has a California rival, Glass Lewis. (Strine, 2005, p. 688) Despite the lack of empirical data, ESMA holds that the PA industry in Europe is small compared to the U.S. (in terms of coverage and turnover), and still developing (ESMA, 2012). However, over recent years, the role and influence of proxy advisors has increasingly been the subject of discussions by the business press, academia, and policy makers in Europe, potentially indicating an increasing economic relevance Regulatory Initiatives in Europe Given the growing perceived influence of PAs in European markets, European regulators, notably ESMA, recently added proxy advisory to their agendas. 3 Following up on a Green Paper in 2011 by the European Commission, ESMA in 2012 released for consultation a discussion paper on potential market failures related to the PA industry. The purpose of this discussion paper was to outline the developments of the European PA industry, to raise key issues for consultation with respect to potential market failures within the PA business, and to discuss potential policy options. Voiced regulatory concern relates to industry concentration, for one thing, as only few PAs dominate the business. Also, ESMA observes that PAs operate 2 Appendix 3 summarizes the main players in the market (ESMA, 2012, pp. 9-15). 3 There exist no European-wide regulatory measures which address the proxy voting industry. However, on member state level, there are some policy recommendations which address rather indirectly PAs, e.g., the UK FRC Stewardship Code Principle 1 & 6 from 2012 and the French AMF Recommendation No (ESMA, 2012, pp ). U.S. PAs are commonly regulated under the Investment Adviser Act from 1940 (ESMA, 2012, p. 27). Under this regulation, PAs have to comply with certain fiduciary obligations and have to meet minimum disclosure standards. Depending on the services provided, not all advisors, however, are required to register as investment advisors under the Adviser Act. In 2010, the SEC released a concept paper on the U.S. proxy voting system to review the role and influence of PAs in the U.S. and to discuss potential policy options. In June 2014, the SEC released a Staff Legal Bulletin No. 20. The main objective of the bulletin is to provide guidance on investors use of PAs (Malenko & Shen, 2016). 9

12 in a virtually unregulated environment, and that some investors appear to blindly follow the recommendations issued by them. Additional concerns relate to potential conflicts of interests (e.g., if PAs also provide services to corporate issuers) and overall low level of transparency with PAs (ESMA, 2012). As the consultation process did not provide compelling evidence of severe market failures related to the PA business, ESMA decided not to push any binding regulatory measures at the time. Instead, ESMA required PAs to develop a European code of conduct aimed at increasing integrity and transparency (ESMA, 2013). In March 2014, the Best Practice Principles for Providers of Shareholder Voting Research and Analysis were published by a group comprising an independent chairman, and six proxy voting advisors (i.e., GL, ISS, IVOX, Manifest, PIRC, and Proxinvest). At the end of 2015, ESMA published a first follow-up report on the development of the Best Practice Principles noting areas for improvements (ESMA, 2015). It is important to note that during the respective consultation processes (i.e., EC 2011; ESMA 2012, 2013, 2015), one key issue has been frequently raised, namely the influence of PAs at European AGMs. In particular, ESMA (2012, p. 39) has invited comments on the questions whether PAs have a significant impact on the voting results at European AGMs. Overall, the feedback during the consultation period has highlighted that beyond anecdotal evidence empirical data on the role and influence of PAs in European markets is not available (ESMA, 2013). In addition, ESMA cautioned that it likewise lacks more basic information, for example, regarding the presence and market share of PAs. In particular, ESMA (2012, p. 11) notes that in Europe, market shares have not been measured on a European level so far [and that even ESMA] cannot provide exact data regarding market structure and concentration. Our paper aims to address this gap in the literature and thus directly responds to ESMA s feedback request. To that end, we structure our paper and analyses along key issues 10

13 of proxy voting advisory, starting with the market structure and coverage decision and concluding with the relevance of PAs voting recommendations. In sum, in the next four sections, we provide descriptive evidence on the following four key issues: (1) the factors associated with PAs market share and coverage decision, (2) the factors associated with PAs voting guidelines, (3) the relation between PAs recommendations and shareholder votes, and (4) the market reaction to the release of PAs recommendations. 3. ISS and GL Proxy Voting Coverage in Europe 3.1 Distribution of ISS and GL Coverage To address our first question of interest How present are PAs in European markets, and what are the characteristics of firms (e.g., ownership structure, governance quality) they choose to cover? our initial set of analyses examines firm and market coverage characteristics of PAs. We exploit comprehensive coverage data provided by ISS and GL for the years between 2008 and 2010 across 14 European countries. As detailed in Appendix 1 (Panels A and D), our first sample, as a result of merging ISS data, GL data, and the underlying Worldscope dataset, comprises all AGMs by publicly listed firms covered by either ISS or GL with a minimum of Worldscope data available (e.g., ISIN, nation code, and basic accounting and ownership data). [Table 1 near here] Panel A of Table 1 reports the distribution of PAs coverage rates across our 14 European countries. In essence, we observe that around 60% of listed firms in the Worldscope universe are covered by ISS proxy services, while GL proxy services covers around 34%. In absolute terms, both PAs cover the highest number of firms in the UK, followed by France and Germany. These numbers underscore the market leadership position of ISS in Europe. Moreover, they map into frequently reported estimates of the U.S. PA market structure. Belinfanti (2009), for example, reports that ISS controls 61% of the U.S. market shares (in 11

14 terms of turnover), whereas GL, as the number two in the US market, is only counted for 36% of the U.S. market shares. The development of market share over time suggests that while ISS coverage is fairly stable between 2008 and 2010, GL seems to be expanding business during that period of time (by 7.4 percentage points). Panel B of Table 1 gives details of differences and overlap in coverage. Accordingly, around 65% of all listed firms are either covered by ISS or GL, whereas less than 33% are jointly covered by both PAs. Underlining the market leadership of ISS, we observe that around 30% of all listed firms are exclusively covered by ISS, while this number is only 5% for GL. Untabulated results reveal that especially large and profitable firms with high free float are covered by both PAs. Panel C of Table 1 provides insights into the relative market share of both PAs on the country level. We define a PA s relative market share as the absolute number of firms covered per country divided by the number of firms in the respective Worldscope firm universe. To document institutional differences across our sample countries, we follow prior literature (e.g., Iliev et al., 2015) and group the countries along their legal origin classification as proposed by La Porta et al. (1998). For each country, we report two widely used proxies for the quality of the legal system and investor protection that is, the Anti-Self-Dealing Index developed by Djankov et al. (2008) and the Regulatory Quality Index developed by Kaufman, Kraay, and Mastruzzi (2009). To document country-level investor protection and governance quality from an accounting perspective, we also use the Accounting Enforcement Index by Brown, Preiato, and Tarca (2014), which aggregates different aspects of the country s enforcement and auditing regulation. Collectively, all three measures are expected to capture country-level investor protection strength and the probability of outside shareholder expropriation. Accordingly, the data reported in Panel C of Table 1 shows that French legal origin countries with on average weaker investor protection standards exhibit the highest PA market share. Using multivariate analyses, the next section 3.2 sheds more light on these 12

15 country-level factors of the relative PA market share, while section 3.3 addresses the role of firm-level characteristics with respect to the PA s coverage decision. 3.2 Proxy advisors market share and investor protection To analyze how PA market share varies with country-level institutional features, we estimate the following country-level OLS regression based on 42 country-year observations for the years : PA_MARKET_SHARE = α+ γ COUNTRY_INST + γ MARKET_SIZE +ε (1) As detailed in Section 3.1, the dependent variable PA_MARKET_SHARE measures the relative market coverage of the respective PA (either ISS or GL). Our independent variable COUNTRY_INST measures the strength of investor protection (probability of shareholder expropriation) on the country level, which we capture using the three alternative proxies introduced in section 3.1 (Regulatory Quality, Anti-Self-Dealing, and Accounting Enforcement). We follow Djankov et al. (2008) and control for the general level of stock market size and development (MARKET_SIZE) by including stock market capitalization scaled by GDP (provided by the World Bank on a yearly base). To address cross-sectional dependence in the dataset, we use standard errors which are one-way clustered at year level. Appendix 2 summarizes the respective variable definitions. [Table 2 near here] Panel A of Table 2 reports the results from estimating equation (1). Consistent with the descriptive findings (Panel C of Table 1), we find (weak) evidence that PA market coverage, in particular by ISS, varies systematically with country-level institutional strength. Specifically, two of three coefficient estimates (Regulatory Quality, Accounting Enforcement) obtain negative signs and become significant at conventional levels for ISS s market coverage, while GL market share is negatively correlated only with the Accounting Enforcement measure. Collectively, these findings show that the relative supply of PA s 13

16 services across European countries is higher when country-level investor protection standards are relatively weak. This is consistent with a higher demand for PA in those countries Proxy advisors Coverage and Firm Characteristics To analyze PAs coverage decision, we estimate the following firm-level logit model for a sample of more than 14,800 firm-year observations for the period : PA_COVERAGE = α+ γ FIRM_CHAR + γ IND + γ COUNTRY (2) + γ YEAR +ε The dependent variable PA_COVERAGE takes the value of one if a firm is covered by the respective PA, and zero otherwise. To assess firm-level factors (FIRM_CHAR), we employ log of market value (for firm size), leverage (for financing structure), analyst following (for the information environment), and the following three ownership measures: free float (% of shares in free float), stake of investment firms (% of shares held by investment firms), and insider control (% of closely held shares by insiders of the firm). Since institutional investors represent the principal clients of PAs (e.g., ESMA, 2012; Rose, 2007), we expect that ownership structure plays an important factor for PAs firm coverage decision. To control for industry-, year-, country-wide factors of PAs firm coverage, we include indicators for first-digit SIC industry classification, years, and countries. 5 To control for timeseries dependency in the panel dataset, we use standard errors which are one-way clustered at firm-level. Panel B of Table 2 reports the corresponding results for regression model (2). Across all model specifications, we observe that firms covered by PAs are on average larger and attract 4 Untabulated sensitivity analyses indicate that the inclusion or exclusion of single sample countries (e.g., Greece) do not alter our main findings. 5 In accounting research, it is commonplace to include indicators (e.g., for industry or year) in Logit models to control for indicator specific effects (e.g., Guo, Huang, Zhang, & Zhou, 2016; Srinivasan, Wahid, & Yu, 2015). 14

17 more analyst following. With respect to ownership characteristics, we find that firms with high free float, high ownership by investment companies, and low insider control are more likely to be covered by PAs. Collectively, the findings from section 3.2 and 3.3 underline the endogenous nature of PAs coverage decisions. Specifically, country-level investor protection standards, ownership structure, and the demand by outside investors seem to correlate with PAs coverage decision and market share. Consistent with Iliev and Lowry (2015), these findings suggest that the demand for PA services varies with the investor s cost and benefits of engaging in own and independent research. 6 At face value, large shareholders with significant stakes in the firm might have higher incentives to engage in independent research (resulting in less demand for PA services), whereas institutional differences might induce additional information acquisition costs to outside investors (resulting in higher demand for PA services). In the next sections, we turn to the implications of PA coverage, by investigating its principal outcome, PAs reports and recommendations on how to vote on AGM agenda items. 4. ISS and GL Proxy Voting Recommendations 4.1 ISS and GL Reports Our second set of analyses turns to the actual reports issued by PAs, to address the second question of interest: What are the characteristics of PA s voting recommendations, and what kind of firm characteristics are associated with negative voting recommendations? We employ a comprehensive, manually-constructed dataset of ISS and GL proxy voting recommendations for the proxy season 2010 across 14 European countries. Each individual proxy voting recommendation is manually retrieved from the underlying reports. To organize and align our sample across countries and PAs, we adopt ISS s AGM item classification, a numerical classification system that group similar agenda items consistently across different 6 Iliev and Lowry (2015) show that U.S. mutual funds with higher benefits and lower cost of independent research (measured by fund size, fund turnover, and investment size) are less likely to rely on ISS proxy voting. 15

18 countries. We use this classification to manually match each AGM item individually to (1) GL voting recommendations and (2) to the firm s disclosed item-based voting results. As further detailed in Appendix 1, we use the ISS and GL coverage datasets (Section 3.1) to arrive at a final sample of 7,116 ISS voting recommendations (and underlying voting items) for 573 unique firms or AGMs, and 5,649 GL voting recommendations (and underlying voting items) for 432 unique firms or AGMs. As a result of our sample selection and manually coding process, the GL sample constitutes a subsample of the ISS sample. 7 Both ISS and GL reports start with a short summary of general information on the company (e.g., firm identifier, industry segment, meeting and record date) and report AGM agenda items, plus the corresponding management and PA recommendations. In almost all cases of our sample reports, the individual names of the respective PA s analysts are included on the front page of the reports. ISS additionally highlights upfront any key engagement activities with the firm by summarizing details such as the date of the engagement, name of the initiating party, topics discussed, and outcomes reached. After the initial summary of the agenda items and the recommendations, both PAs usually provide detailed information on the company s performance and governance structures (e.g., board and compensation details, or information on the firm s ownership structure). In the main part of the reports, both PAs discuss each of the agenda item in detail and outline their respective vote recommendation. In terms of report size and analyst staffing, we observe in our ISS sample 58 unique ISS lead analysts and an average ISS report size of 17 pages. 8 In contrast, our GL sample reveals 15 unique GL lead analysts and an average GL report size of 24 pages. In particular, the sample differences in analyst staffing seem to mirror our findings regarding the market shares of both PAs in Europe (see Section 3.1). 7 This sample selection restriction is less of a concern as the exclusive GL coverage is limited in our underlying sample (Panel B, Table 1). 8 We define lead analysts as the analyst reported first in the respective report. 16

19 4.2 Distribution of ISS and GL Recommendations Panel A of Table 3 gives a breakdown of all voting recommendations by PA and country into supportive vote for recommendations, and vote against recommendations. On average, ISS recommends their clients to vote against a meeting item in 14.9% of the cases. In comparison, GL recommends to vote against the management proposal in 17.5% of all cases. This ratio shows substantial variation between countries, and between the two PAs. For example, ISS issues a higher proportion of vote against recommendations in French legal origin countries compared to Common legal origin countries (25.4% vs. 4.2%). Also, GL, for example, issues vote against recommendation in the UK market in 18.9% of the cases, whereas ISS opposes the proposals of UK managers in only 3.7% of the cases. A similar distinct pattern is observable for Switzerland (10.2% GL vote against recommendation rate vs. 1.9% ISS vote against recommendation rate). 9 Panel B of Table 3 breaks down voting recommendations according to their type or subject area, i.e., into four distinct categories (Board, Compensation, Shareholder Rights, and Audit), and other proposals. For both PAs, we observe that management proposals addressing compensation issues receive the highest proportion of vote against recommendations (28.8% for ISS, 36.6% for GL). Board- and shareholder rights-related voting items also receive above-average rejection rates by both PAs (20.6% and 22.9%, respectively, for ISS, and 18.8% and 19.3%, respectively, for GL). A comparison of these results with prior U.S. findings yields two main insights. First, ISS rejection rates on compensation and board-related proposals are markedly higher in Europe compared to those reported in U.S. studies. Choi et al. (2010) and Ertimur et al. (2013), for example, document ISS rejection rates for director elections of 6.8%, and 11.3% for say on pay votes. In contrast, GL rejection rates on board and compensation issues differ less between Europe and the U.S. Choi et al. (2010) and 9 The univariate and multivariate inferences in this section also hold when re-estimating all analyses on a fully matched sample comprising only firms which are simultaneously covered by both PAs. However, following prior research on the comparison of proxy voting and governance rating vendors (e.g., Choi et al., 2010; Daines et al., 2010), we consider the PAs individual firm universe within our baseline sample. 17

20 Ertimur et al. (2013) report GL rejection rates of 18.8% and 21.7% for director elections and say on pay votes, respectively. Second, consistently across Europe and the U.S., GL issues a higher proportion of against votes than ISS (Table 3; Ertimur et al., 2013; Choi et al., 2010). Panels C and D of Table 3 finally link the PA s rejection rates to the firm s ownership and governance structure. In particular, we see that firms with concentrated ownership and poor governance quality tend to receive a larger number of vote against recommendations by both PAs. For example, firms with poor governance quality receive in 26.2% (21.3%) of the cases an ISS (GL) vote against recommendation, whereas their counterparts with high governance quality only witness an ISS (GL) rejection rate of 2.1% (14.9%). In the next section, we address the factors of ISS and GL recommendations in a multivariate context. [Table 3 near here] 4.2 PA Recommendations, country-level institutions, and firm characteristics Firm-level analyses To identify factors that potentially play a role for PAs voting recommendations, we estimate the following firm-level OLS regression model for the proxy season 2010: PA_AGAINST_% = α+ γ FIRM_CHAR + γ COUNTRY_CHAR +ε (3) The dependent variable PA_AGAINST_% reflects the PA s rejection rate at firm-level (i.e., number of PA vote against recommendations divided by number of AGM items on the firm-level). The vector of firm-level variables (FIRM_CHAR) comprises log of market value, leverage, and the three measures of ownership structure: free float, stake of investment firms, and insider control. 10 As a measure of firm-level governance quality, we use corporate governance ratings developed by ISS. 11 As this governance proxy is only available for a 10 In contrast to our firm-level analyses in Section 3, we excluded financial analyst following from the regression model due to multicollinearity concerns (and high VIF values). 11 GRID (Governance Risk Indicator) is the commercial governance rating marketed by ISS. Based on up to 80 single governance items, ISS provides GRID as the successor of CGQ ratings for over 8,000 firms worldwide. While we are aware of general concerns with third-party summary measures of governance quality (e.g., Daines 18

21 subsample of ISS and GL reports and recommendations (see Appendix 1), we estimate regression model (3) based on the overall sample as well as for a subsample for which governance data are available. In line with Section 3.2, we measure country-level differences (COUNTRY_CHAR) using three alternative measures of investor protection and legal quality (i.e., Regulatory Quality, Accounting Enforcement, and Anti-Self-Dealing). Appendix 2 summarizes definitions of all variables used in our analyses. [Table 4 near here] Panels A and B of Table 4 report the respective findings from estimating (3). Across all model specifications for the subsample, we observe that governance quality significantly correlates with both rejection rates. Thus, firms with low governance quality appear to have a higher likelihood to receive vote against recommendations. In contrast, firm leverage and firm size remain largely unrelated to the rejection rates of both PAs. With respect to ownership and country-level factors, ISS and GL voting policies seem to differ. ISS s rejection rate increases with lower dispersed ownership and higher insider control, whereas GL s rejection rate increases with the ownership stake of investment firms. In addition, ISS s voting policy significantly correlates with all three investor protection proxies suggesting a higher ISS rejection rate in countries with poor investor protection standards. In contrast, GL s rejection rate is higher for countries with high investor protection standards. 12 et al., 2010; Hitz & Lehmann, 2015), we follow prior cross-country governance research and rely in the absence of a parsimonious alternative on ISS governance ratings (Aggarwal et al., 2009, 2011; Bruno & Claessens, 2010; Chhaochharia and Laeven, 2009; Chung et al., 2010; Doidge et al., 2007). Nevertheless, method consistency or a common governance perception across ISS proxy voting and rating products might infer with the construct validity of this proxy in our setting. However, we mitigate this problem by framing our governance analyses within additional governance-related variables, for example, insider ownership or countrylevel investor protection, and by reporting comparable results for GL, which should be unrelated to these concerns. 12 These findings coincide with our descriptive observations in Table 3 and potentially point out different competitive edges and market strategies of GL and ISS across markets. In other words, GL might position itself as a more aggressive PA in terms of rejection rates to gain market shares and expand business, especially in those countries with a more management friendly ISS voting policy (e.g., the UK market which constitutes the largest European PA market in absolute numbers of covered firms). 19

22 4.2.2 AGM-item-level analyses To gain a more nuanced understanding how governance quality relates to voting policies, we additionally adopt an AGM item-level perspective. Specifically, we are interested whether the quality of certain governance mechanisms such as board, composition or audit quality predicts the likelihood of receiving pertinent vote against recommendations by PAs. For that purpose, we estimate various versions of the following AGM item-level logit model: PA_AGAINST = α+γ GOV + γ FIRM_CHAR + γ COUNTRY +ε (4) The dependent variable PA_AGAINST iv is a 1/0 indicator variable reflecting the PA s vote against recommendation for a specific AGM voting item v of firm i. Our main test variable is either the overall governance quality of the firms, or its breakdown to specific governance areas such as the quality of the board, the compensation system, shareholder rights, or the audit process. Note that we estimate each regression specification addressing a specific governance aspect based on the corresponding subsample of AGM voting items. For example, we examine the predictive ability of board quality in the context of board-related voting items and the PA s board related voting policy. To control for firm, ownership, and country-level effects, we include log of market value, leverage, free float, and country indicators in all model specifications. For variables definitions, see Appendix 2. Panels C and D of Table 4 present the corresponding results from estimating (4). We observe that over and above the summary measure of governance quality board and audit quality consistently predict the likelihood of receiving ISS and GL vote against recommendations. These findings turn out to be economically meaningful. For example, the probability of receiving ISS (GL) vote against recommendations ranges between 15.9% and 9.8% (23.9% and 14.4%) contingent on the respective overall governance quality. 13 This translates into a decrease of 38% (40%) in the likelihood of receiving vote against 13 We estimate a decrease in the likelihood of receiving vote against recommendations as the relative change in the predictive probability of vote against recommendations between low and high governance quality firms (e.g., 38% = ( ) / 15.9). 20

23 recommendations by ISS (GL) when moving from the lowest to the highest governance quality in our sample. The economic significance is even more pronounced for board- and audit-related vote against recommendations across both PAs. Collectively, the findings in sections and emphasize the role of governance quality in the PA s process of setting voting recommendations. Our results are consistent with prior U.S. evidence suggesting that ISS s negative recommendations center on board and compensation-related issues, whereas GL focuses primarily on audit and board issues when opposing management proposals (Choi, Fisch, & Kahan, 2009). More generally, our findings are consistent with recent evidence by Iliev et al. (2015), who show that the voting dissent by U.S. institutional investors increases with managerial entrenchment and poor external shareholder protection, i.e., with circumstances where outside shareholders fear expropriation. 4.3 ISS and GL Agreement To provide a more nuanced understanding of the PAs recommendation policies, we analyze the agreement and the overlap between both, ISS and GL. Such an analysis potentially sheds light on the extent to which both voting guidelines differ, and whether potential similarities and differences vary systematically with firm- and country-level characteristics. Table 5, which documents the distribution of different agreement rates between both PAs, yields three main insights. First, while GL issues vote against recommendations more frequently than ISS (see Table 4), the set of firms with ISS vote against recommendations is not just a subset of those with negative GL recommendations. Notably, 374 voting items (6.6% of all voting cases) received a vote against recommendation by ISS, and a vote for recommendation by GL. Second, ISS and GL reach the same conclusion in 84% of all sample cases (= ( ) / 5649). Yet, most of this agreement results from vote for recommendations (75.9%). For controversial cases, that is, cases in which at least one PA 21

24 recommends to vote against the management, the ISS and GL agreement rate drops substantially to 34% (= 460 / ( )). However, both agreement rates appear to be higher in Europe than in the U.S. Ertimur et al. (2013) report an overall agreement rate and an agreement rate on controversial cases of 77% and 17.9%, respectively, for a sample of U.S. say on pay voting items. Third, Table 5 further suggests that ISS and GL agreement rates, especially the agreement on controversial topics, vary with firm- and country-level characteristics. For example, the ISS and GL agreement rate on controversial topics varies between 11% for firms from Common legal origin countries and 46% for firms from French legal origin countries. A similar pattern is observable for subsamples with high and low governance quality (10% agreement rate on controversial topics for well governance firms versus 49% for poorly governed firms). Untabulated multivariate analyses with the PAs agreement rate as the dependent variable and firm- and country-level characteristics as independent variables corroborate these findings and suggest that the overlap between ISS and GL recommendations is increasing for firms with poor corporate governance and for firms from weak investor protection countries. Overall, the findings presented in Section 4 highlight that despite differences in both PA s voting guidelines two factors are systematically associated with the outcome of PA recommendations, namely firm-level governance quality and country-level investor protection standards. At the same time and despite the apparent similarities in both PAs voting guidelines (e.g., with respect to governance issues), our descriptive findings, especially the country-level differences in the rejection rates (e.g., for the UK or Switzerland), also allow for casual inferences as to whether and how PAs adjust their voting guidelines or the application of these guidelines, e.g., on a country basis. As noted, ESMA has voiced concerns that U.S. domiciled PAs do not sufficiently consider local market conditions when developing their proxy recommendations (ESMA, 2012), resulting in industry-wide one-size-fits-all voting 22