Voting Matters

Voting Matters: Evidence from S&P 500 Index Additions Stuart L. Gillan Department of Finance, Terry College of Business ...

1 downloads 34 Views 736KB Size
Voting Matters: Evidence from S&P 500 Index Additions Stuart L. Gillan Department of Finance, Terry College of Business 351 Amos Hall University of Georgia Athens, GA 30602 Email: [email protected]

Florian Muenkel Finance, Information Systems and Management Science Department Sobey School of Business Saint Mary’s University Halifax NS, B3H 3C3 Email: [email protected]

Christine A. Panasian Finance, Information Systems and Management Science Department Sobey School of Business Saint Mary’s University Halifax NS, B3H 3C3 Email: [email protected]

1

Abstract Using a sample of S&P 500 Index additions, we explore the mutual fund voting on corporate governance proposals. We find that voting support for governance proposals increases considerably for addition firms, especially for shareholder-sponsored proposals. When further exploring the mechanisms leading to this outcome, we attribute the changes to i) increased ownership by mutual funds in general, ii) increased ownership by index funds in particular, and iii) an increased propensity of funds to vote in favor of governance proposals after index addition and iv) a decrease in fund propensity to vote with management. We test whether the observed results have an impact on aspects of compensation for these firms. We find that overall, S&P 500 Index firms tend to have significantly higher levels of total compensation relative to the match sample however, pay-for-performance sensitivity (delta) is significantly lower, in general, for the addition firms, post-addition, relative to match firms.

2

Introduction Both empirical and theoretical research during the last several decades provides compelling evidence that institutional investors are active in firm governance. Indeed, institutional investors are increasingly willing to express their preferences as to desirable (and undesirable) governance practices.1 Moreover, as the amount of assets under passive management have increased, the potential for large institutional investors to shape the governance landscape has become increasingly apparent. At the same time, the role of passive or indexed institutional investors in corporate governance has been questioned. Some argue that such passive investors lack the incentives and resources to monitor their large and diverse portfolios, while others provide evidence that such institutions have an important influence on governance practices. For example, Schmidt and Fahlenbrach (2017) report that exogenous increases in “passive” ownership leads to weaker governance structures of affected firms, and subsequently to negative effects on firm value. In contrast, several recent papers document that index funds play a key role in influencing firms’ governance choices, which ultimately leads to increased valuations (e.g., Appel et al, 2016; Mullins et al, 2014). The exact channel through which passive funds affect changes in corporate policy is, however, still the subject of ongoing investigation. We explore this issue by focusing on institutional investor voting behavior and subsequent governance changes for a sample of large US firms. Specifically, we focus on firms added to the S&P 500 Index. Once a firm is part of a major index such as the S&P 500, its ownership structure changes substantially (Pruitt and Wei, 1989, Biktimirov et. al. 2004, Chen et al, 2004). This quasi-exogenous shock to ownership

1

See, for example, statements by Blackrock https://www.bloomberg.com/news/articles/2017-10-11/fink-saysblackrock-exceeds-hedge-funds-at-corporate-governance 3

structure provides a basis for examining the influence of ownership structure on governance from a causal perspective. Specifically, we focus on how ownership structure and voting outcomes on governance proposals change after index additions. The literature to date has provided valuable insights into the determinants and outcomes of shareholder voting on a variety of issues. For example, papers examining aggregate shareholder voting results suggest that institutional investors are particularly influential in voting outcomes, and that voting outcomes are associated with changes in corporate policies (Gillan and Starks, 2000; Ferri and Maber, 2013; Cai, Garner and Walking, 2009; Ertimur et al., 2010). More recent work has focused on the way that individual funds vote their proxies. Of note, in 2003 the Securities and Exchange Commission (SEC) mandated that mutual funds disclose proxy voting policies along with an annual report on how they actually voted. Papers examining these disclosures report variation in fund voting based on fund governance, portfolio firm performance, and proposal characteristics (for example Chou et al., 2011; Davis and Kim, 2007, Iliev and Lowery (2014), Ng et al., 2009; Morgan et al., 2011,). More recently, Crane et al. (2016) provide evidence that there are more shareholder proposals, an increase in voting participation, and increased votes cast against management proposals following the rise of institutional ownership after Russell Index 1000/2000 reconstitutions. They further find significant increases in dividends, R&D expenses, and operating performance and decreased cash holdings and executive compensation. In a related study, Appel et al. (2016), also using the change in institutional ownership around the Russell index reconstitution, find differences in governance structures in firms with higher passive institutional ownership. Specifically, more independent boards, more poison pill removals, and lower presence of dual class shares. Additionally, they document an association between higher

4

ownership concentration by passive institutions and lower (greater) support for management (shareholder) sponsored governance proposals, suggesting that passive funds use their voting rights to exert influence on firms’ governance. Our paper complements both Appel et al. (2016) and Crane et al. (2016) in that we also study the impact of institutional ownership on firm governance and the channels of influence. However, we contribute to the literature along several dimensions. First, we focus on a different sample and use a different identification strategy to address endogeneity concerns: specifically we use plausibly exogenous changes in ownership structure around S&P 500 Index additions, a set of relatively large ad economically important firms. Second, we employ a matched sample for comparison purposes that allows us to account for potential time trends during the sample period. Third, while previous studies analyze voting results at the firm level, we go a step further and use data on the voting behavior of individual funds. This allows us to identify changes in the propensity of individual funds to vote with management, and thus affords a more in-depth investigation of ownership changes and fund voting behavior around index additions. Fourth, while we identify the so-called “passive” institutions in our sample as funds having an indexbased strategy, we further examine heterogeneity in institutional investor voting by identifying a category of “active voting funds” based on the propensity of each fund to vote against management proposals. Overall, this unique setting allows for a clear identification of the main channels through which institutions impact governance: the changes in ownership structure and changes in institutional voting behavior after a firm is added to the S&P index. We document a strong increase in voting support for governance related proposals after index addition. Specifically, we find that the post-addition years are associated with significantly higher support for governance related proposals, an effect that is concentrated in

5

shareholder-sponsored governance proposals. We also find that greater ownership by index funds is positively associated with voting support for shareholder sponsored governance proposals, while ownership of non-index funds is unrelated to proposal support for either management or shareholder proposals. The observed effects appear to be attributable to two key factors: 1) fundamental changes in firms’ ownership structures, and 2) a change in institutional investor voting behavior. With regard to the former, we find that aggregate fund ownership in a firm changes little following index inclusion. However, index funds significantly increase their stakes while nonindex funds as a group decrease their ownership. We also find a significantly higher probability of favorable votes for governance proposals (either management or shareholder) once a firm is added to the S&P 500, even after controlling for firm and proposal characteristics. This supports the findings of prior studies that institutional shareholders influence the governance of the firms they hold, and more importantly, clearly identifies one channel for this impact: voting on governance issues at annual meetings. Finally, we address whether these changes lead to governance adjustments in these firms. Given our focus on executive compensation proposals, we look at changes in aspects of compensation with emphasis on whether there are significant differences between the addition and match firms in the pre-and post addition periods. We find that, overall, S&P 500 Index firms tend to have significantly higher levels of total compensation relative to the match sample however, pay-for-performance sensitivity is significantly lower for the addition firms, postaddition compared to the match sample, pointing to a possible impact of voting on index firms.

6

The remainder of the paper is organized as follows: section 1 presents background analysis on S&P 500 Index additions, ownership changes, and a discussion of related proxy voting literature, section 2 presents our sample selection and data. In section 3, we present our multivariate models, including proposal support, ownership changes, and voting behavior. Section 4 examines the impact the increased voting support post-addition has on executive compensation and pay-performance sensitivity. Section 5 concludes.

1. Background There has been a dramatic increase in funds under passive (or indexed) strategies during the last decade. The Financial Times reports that passive funds accounted for one third of mutual fund assets at the end of 2016.2 Furthermore, in 2017, S&P Dow Jones Indices reported over $7.8 trillion benchmarked to the S&P 500 Index alone, with indexed assets comprising approximately $2.2 trillion of this total. Moreover, the S&P 500 index represents approximately 80% of the U.S. available market capitalization. As our sample is comprised of firms added to the S&P 500, these firms are large. At $11.5 billion, the average firm size in our sample is approximately 4 times larger than the average firm size for firms at the bottom of Russell 1000 (approximately $3.4 billion). As is well documented, index funds increase their stake in firms added to a major index in order to fulfill their tracking objectives. Consequently, the number and percentage of ownership by indexers increases significantly for such firms.3’4 More importantly, as the

2

Financial Times, September 11th, 2016 https://www.ft.com/content/4cdf2f88-7695-11e6-b60a-de4532d5ea35 See Chen, Noronha and Singal (2004). 4 S&P 500 additions are prompted by a deletion caused by either liquidity falling below a threshold, a delisting, acquisition, or just an evaluation of the S&P index committee that a firm has become less representative than the possible replacement, and are thus representing an exogenous shock to the ownership structure of a company. 3

7

addition to the S&P 500 Index is a plausibly exogenous shock to the ownership structure of a firm, it provides a basis for examining causal relationships between ownership and governance. S&P 500 Index additions are prompted by deletions and, as suggested by Standard and Poor’s, are not based on an evaluation of a company’s future performance (Aghion, Van Reenen and Zingales (2013), Boss and Ruotolo (2000)). Moreover, even if an index addition is associated with expected future firm performance, there is no reason to believe that these firms are more likely to improve their governance simply as a result of addition to the index. Prior work reports evidence of real changes at firms after index addition. For example, Denis, McConnell and Ovtchinnikov (2003) find that once a firm is part of an index, it takes more value enhancing decisions as monitoring by investors and analysts becomes more effective. Moreover, as noted above, Appel et al (2016) and Crane et al (2016) report real policy and govrnance changes based on ownership changes around Russell Index reconstitutions. Another strand of the literature investigates the influence of institutional investor voting on firm governance and policies. For example, Morgan et al. (2011) show that mutual funds tend to vote affirmatively more often on proposals with wealth-increasing potential, and such voting influences the likelihood of a proposal passing. Iliev and Lowry (2014) report that funds that are more “engaged” in voting earn higher alphas and affect more value-enhancing outcomes in the firms they own, while Matvos and Ostrovsky (2010) find evidence of strategic interactions between funds that affects their voting behaviour. Further supporting the monitoring role of mutual funds, Ng, Wang and Zaiats (2009) show that the way mutual funds vote is significantly related to prior firm performance, while Chou, Ng and Wang (2011) assert that funds with good governance practices tend to exercise their shareholder rights more responsibly and use their votes to positively impact the governance of portfolio firms. As noted earlier, Crane et al. (2016)

8

provide evidence that there are more shareholder proposals, an increase in voting participation, and increased votes cast against management proposals following the rise of institutional ownership after Russell Index reconstitutions. Similarly, Appel et al (2016) report less support for management proposals by passive investors, and increased support for shareholder-sponsored governance proposals. In the spirit of this prior work, we focus on the intersection of S&P Index additions and institutional shareholder voting in order to further explore the monitoring role of institutions and the channels through which institutions causally influence portfolio firm governance.

2. Sample and Data Description We start with all firms added to the S&P 500 Index from 2004 through 2014, inclusive. The sample starts with 2004 as this is the first year that the SEC required all mutual funds and other registered investment management companies to disclose their votes at portfolio companies on Form N-PX. Over the 11 years, 252 companies (254 securities) were added to the S&P 500 Index, and the distribution of these additions by year is presented in Table 1.5’6 We create a matched sample of firms by matching our additions on 2-digit NAICS codes and market value for each year. Additionally, we ensure that the match firms are neither a current S&P 500 Index constituent, nor are they in the S&P 500 in the 2 years prior or 2 years after the match. Our comparison sample comprises 239 unique firms as some of the matches are used multiple times. For each addition and matched sample firm we extract firm specific variables and financial statement information from Compustat for the year of addition to the index and 4 years either

5

We lose three firms from our addition sample as they were taken private shortly after being added to the index (e.g. Realogy in Dec 2006, Michaels in 2006) 6 Google and Discovery Communications have 2 classes of common shares listed on the S&P 500 Index. 9

side of the addition year. Thus, each firm in our sample has 9 years of observations, resulting in 2,286 firm-year observations for each addition and matched firm. We combine the addition and matched samples with institutional holdings data from the Thompson Reuters S12 mutual fund holdings database, available from Wharton Research Data Services (WRDS).7 The primary source for the mutual fund holdings data is from the semiannual SEC N-30D filings that funds are required to file with the SEC. We also incorporate mutual fund voting data from the ISS Voting Analytics Database. This includes the way that each mutual fund casts their vote on each agenda item at each portfolio firm (For, Against, Withhold, or Abstain). A detailed description and examples of proposals is presented in Appendix Table A.2. Additionally, we employ a second ISS database containing information on specific agenda items, whether the proposal was sponsored by management or shareholders, the percentage support received by each proposal, and the ultimate voting outcome (pass or fail). A. Descriptive Statistics – Firm Characteristic Variables Table 2 presents summary statistics for addition and matched firms for firm specific characteristics including total assets, total number of common shareholders, asset turnover, cash/total assets, firm age, capital expenditures to total assets, leverage ratio, market to book, ROA and Tobin’s Q. The last column reports test statistics of the difference in means for each of the variables between addition and matched firms. We see that addition firms are generally similar to matched firms, but have higher asset turnover ratios, higher cash flow to assets, higher ROA and are slightly older.

Wharton Research Data Services (WRDS) notes about S12 dataset: “The primary source for the mutual fund holdings data is SEC N-30D filings. These filings, which include semi-annual reports to shareholders, are required to be filed with the SEC twice a year by mutual fund companies. To a lesser extent, Thomson taps fund prospectus and contacts mutual fund management companies to increase update frequency.” 7

10

B. Descriptive Statistics – Ownership Characteristics In Table 3 we report averages of several measures of ownership for additions and matched firms, for the years prior to and after addition including the total number of: funds owning each firm in the sample, index funds, non-index funds. We also report the total percentage owned by: institutions overall, index funds, non-index funds, and the total percentage owned by index and non-index funds as a percentage of total institutional ownership. For each of the variable averages presented in Table 3 we report test statistics for differences in means between the years before and after addition, for both additions and matched firms. The average firm in the sample has about 542 individual fund owners over the entire sample period. The average firm has 402 fund owners before addition, a value that increased significantly after addition to 682 (a 70% increase). For matched firms there is an increase of only 28%, from 345 to 443 over a similar time frame. While aggregate institutional ownership increased slightly for both additions and matched firms (3% and 5%, respectively), not surprisingly, ownership by index funds for addition firms increased substantially compared to the matched firms (142% vs. 48%, respectively). These findings suggest that the observed changes in ownership cannot be attributed merely to a time trend. C. Descriptive Statistics – Proposals Characteristics We next focus on proposal types. We group all agenda items into 8 different categories and present a list and description of the most common types of items included in each category in Table A.2 in the Appendix. Panel A in Table 4 shows that the total number of proposals on the agenda at annual meetings increased by 85% for addition firms (from 6,198 before addition to 11,487 after addition) and by 51% for matched firms (from 6,358 before addition to 9,624 after addition). 11

While recent regulatory reforms have required that additional issues be put to a shareholder vote, the values suggest a larger increase for the addition sample over and above that of the matched sample, thus it is not just the initiation of new voting requirements that leads to the increase for addition firms. While the bulk of the proposals are for director elections (67% of the total, or 33,567 proposals) other proposals with a large frequency include those related to executive compensation (10.2% or 3,427 proposals), ratify auditors (9.9% or 3,283 proposals) and other governance related proposals (3.8% or 1,279 proposals). The subset that we will focus on are those related to Corporate Governance and Executive Compensation. Looking at governance related proposals in Table 4 panel B, the number of governance proposals for addition firms increased by more than 220%, from 146 to 472. During the same period, the matched sample exhibits an increase of only of 2.4%, from 328 to 336. Furthermore, when distinguishing between management and shareholder sponsored proposals for addition firms, management proposals increased by 188% and shareholder proposals by 298% while for matched firms, management proposals dropped by 18% while shareholder proposals increased by 256%. Panel C in Table 4 shows the number of executive compensation proposals of our sample over the coverage period, separated by sponsor and additions vs. matches. We note that management proposals more than doubled during the post period for both addition firms and matches. While shareholder proposals also doubled for the matched firms, they increased almost five-fold for addition firms.

3. Multivariate Analyses Our focus is on the channels through which institutions affect the governance of the firms they own with emphasis on governance issues in general and compensation in particular.

12

As such, we first examine the overall voting results on governance and compensation-related proposals (both management and shareholder) and test whether the voting support for these proposals differs in the post-addition period. Moreover, we study the way that different types of institutional owners cast their votes on these proposals. We then proceed by exploring two key channels that potentially drive these voting results. First, we examine ownership changes that occur after index addition, including changes in ownership by different fund types. Second, we examine the way that these different categories of institutional investors vote after a firm is added to the index. The latter analyses are designed to examine directly whether there are changes in the propensity of various types of funds to vote differently on governance proposals after index additions. Again, throughout our analysis we continue to make comparisons with the sample of matched firms. A. Proposal Support We use an OLS specification where we model the number of FOR votes each proposal (both shareholder- and management-sponsored governance sponsored) receives as a fraction of the total number of votes cast for that specific proposal and other controls as follows: FOR Votes (%) = α + β1POST + β2ADD + β3POSTADD + (vector of firm specific controls) + (proposal fixed effects) + (industry fixed effects) + ε

(1)

Where ADD, is a dichotomous variable taking a value of 1 if the firm is part of the addition sample (0 if it is a matched firm), POST take a value of 1 if the observation is postaddition (0 for pre-addition years), and an interaction of the two, our main variable of interest, POST*ADD which captures the marginal effect on voting for addition firms after they become part of the index.

13

Table 5 reports the results from several different model specifications using all governance proposals. All models include firm size, leverage, the capital expenditure divided by total assets, cash flow to assets ratio, ROA and firm age, along with proposal fixed effects to account for between proposal variation. Model (1) reports results for the entire sample, as per equation (1). Given our focus on the link between ownership and voting outcomes we include the percentage ownership by index funds (Idx Fnd Own, %). Model (2) includes an indicator variable to capture the Institutional Shareholder Services (ISS) voting recommendation for each proposal (For = 1, Against = 0). In Model (3) we replace the Idx Fnd Own (%) measure with Non-Idx Fd Own (%), capturing the ownership percentage held by institutions other than index funds. Model 4 adds the ISS Recommendation to the Model (3) specification. The results suggest that Additions tend to receive a lower percentage of votes in favor of governance proposals, as shown by the negative significant coefficients of the ADD indicator in most of our models. However, the POST*ADD interaction terms in Models (1) and (3) of Table 5 have positive and significant coefficients, indicating that after a firm joins the S&P 500 Index, governance proposals receive significantly more favorable votes. Once we introduce the ISS recommendation variable (in models 2 and 4), the significance of the interaction term either declines (Model (4)) or disappears (Model (2)). That is, the increased FOR votes post addition is associated with ISS recommendations. When focusing on the different ownership classifications, comparing Models (1) and (2) with Models (3) and (4), we find that the coefficients of Idx Fd Own (%) are positive and significant at 1% level, while those for Non-Idx Fd Own (%) are not significant. That is, increased index fund ownership is associated with a stronger support FOR governance proposals, while ownership by non-index funds is unrelated to governance proposal support. We note that 14

in all models ISS Recommendation has a strong positive association with the percentage of FOR votes, a result consistent with prior work. In terms of economic significance, using the coefficients from Model (1), an addition firm during the post-addition years, receives on average 6.8% more FOR votes on its governance proposals. Furthermore, a 10% increase in Index fund ownership increases the support a proposal receives, on average, by 0.14%. The same ownership increase, after accounting for the ISS recommendation, increases proposal support by 0.12%, all else equal. When we split the sample into those sponsored by management or shareholders, in Table 6, we see that the positive coefficients for ownership by indexers are driven exclusively by the subset of shareholder sponsored governance proposals. The positive and significant coefficients for Idx Fd Own (Model 3) suggest that firms with greater index fund ownership receive higher support for shareholder sponsored governance proposals, while the same coefficient in management sponsored proposals (Model 1) is insignificant. Interpreting this result in terms of economic significance, a 10% increase in ownership in the firm by indexers leads to a 0.25% increase in support for governance proposals. In contrast, ownership by non-indexers is unrelated to the support a proposal receives for both management and shareholder proposals. Taken together, the results in tables 5 and 6 suggest that index inclusions are followed by a significant increase in voting support for governance proposals, but this increase is concentrated primarily in higher percentages of FOR votes on shareholder-sponsored governance proposals. We also explore the support for executive compensation proposals using similar specifications and report the results in Table 7. These tests include the compensation proposals submitted by management and shareholders combined. This is due to the very low number of shareholder-sponsored proposals that does not support a separate test for each of these groups. 15

We find that the interaction term POST*ADD is again positive and significant in 3 of the 4 models suggesting that firms joining an index receive a significantly higher proportion of FOR votes on executive compensation related proposals compared with the control group, all else equal. Looking at the coefficients of the two sub-groups of institutional investors we follow, we note that three of our four models exhibit a negative relation between their ownership stake and the support the proposal receives. This result is similar for ownership of index funds (Model 2) and non-index funds (Models 3 and 4) and indicates that the higher the stakes institutions8 hold in a firm the lower the support compensation proposals receive. Given the fact that the largest majority of these proposals are management sponsored, this result is not surprising. B. Ownership Changes We showed in the previous section that addition to the S&P 500 Index leads to higher support for governance proposals (and compensation proposals). As index additions lead to structural changes in a firms’ ownership, we posit that firm-level governance after index addition will be influenced by the change in institutional investor types and their ownership stakes in firms. This external shock to ownership structure allows us to address endogeneity challenges present in studies investigating the causal relationships between institutional ownership, voting outcomes, and ultimately governance changes. The analysis in this section explores in more detail how index addition affects the ownership composition of firms. In particular, we focus on changes in ownership by both index and non-index funds during the post-addition years. This is particularly important as institutions

16

provide one avenue for overall changes in tone or practice of governance, as driven by their underlying objectives and strategies. As reported in the univariate tests in Table 3, ownership structure changes significantly once a firm becomes part of the index. Specifically, the number of index funds rises considerably, as does the total number of funds owning a firm post-addition. More importantly, the total ownership held by index funds surges dramatically after the index addition. In order to explore the nature of ownership changes, we begin with an OLS regression to examine the variation in ownership structure. Specifically, we estimate a model of the following form: Own (%) = α + β1ADD + β2POST + β3POSTADD + (vector of firm specific controls) + (industry fixed effects) + ε

(2)

Own (%) takes a different value depending on the specific category of institutional ownership that we focus on (as detailed below). However, in all cases Own (%) captures the fraction of total shares outstanding held by the investor classification of interest: Model (1) uses the total percentage held by all institutions, Model (2) and Model (3) use the percentage held by the subsamples of Index and Non-Index Funds, respectively. We also include a set of firm specific variables identified in the literature to be related to ownership structure: size, leverage, CapEx/Assets, CF/Assets, ROA, firm age. Size, as measured by the firm’s market capitalization, is directly related to ownership dispersion: larger firms are known to be more widely held. Demsetz and Lehn (1985) and Demsetz and Villalonga (2001) show that leverage and firm risk are important determinants of ownership. Based on these studies, we expect size to be negatively related, and leverage to be positively related to percentage of ownership held by institutions.

17

Table 8 reports results of these analyses. Our primary variables of interest in all three models are the dichotomous variables: ADD, POST, and the interaction POST*ADD. All models include industry fixed effects. The firm specific controls are winsorized at the 1% level. The results show that addition firms have significantly higher ownership by mutual funds overall (Model 1), Index Funds (Model 2), and Non-Index Funds (Model 3). Moreover, during the postaddition years, mutual funds in general and both sub-groups have higher stakes relative to preaddition years. As our sample covers the years 2004 to 2014, a period in which mutual fund ownership in U.S. markets has grown significantly, the positive significant coefficient of POST indicator clearly captures the trend of increased institutional ownership trend during this period. However, looking at the coefficients for the interaction terms, capturing the incremental ownership for the addition sample in the post-addition years, we see differential effects based on the classifications of institutional investors. In Model (1), the coefficient for POST*ADD is significant and negative, suggesting that during the later years of the sample, index-additon firms have lower overall ownership by institutions compared to the matched sample. However, addition firms have significantly higher Index Funds ownership post-addition (Model 2) compared to the matched firms, and considerably lower ownership by Non-Index Funds (Model 3). In terms of economic significance, using the interaction term POST*ADD, the coefficients show that addition firms have on average 1.33% more Index and 3.5% less Non-Index Funds ownership relative to the control sample. This initial analysis corroborates the earlier findings of the ownership changes present in our panel data set, and further substantiates findings of extant studies investigating the impact of index additions on ownership structure. While aggregate ownership by institutions stays relatively constant, index funds increase their stakes during the post-addition years. Of note, this 18

highlights the changing ownership clientele with different types of mutual funds having, arguably, distinctive objectives and methods of achieving them.9 C. Voting Behavior After Addition This section looks at the voting behavior of funds in our sample firms, as a second channel of influence on governance structures. Given that funds differ in their objectives, priorities, and ability to commit resources to monitoring portfolio firms, one would expect variation across funds with regard to their voting policies (see Iliev and Lowry, 2014). It has also been argued that the actions, voting patterns, and conduct of institutional owners changes following the shift of balancing forces at play when a firm is added to a major index.10 Large institutional investors, particularly index funds, have become increasingly dominant players in corporate ownership. Coupled with their inability to “vote with their feet,” they have an increasingly important role when it comes to voting. Furthermore, in order to succeed in their proxy contests or other campaigns, activist shareholders increasingly need to rely on support from indexed institutions (see Appel et al, 2016, 2016b). 11 We first focus on the set of governance proposals sponsored by both management and shareholders and the votes cast in favor of those proposals (as listed in Appendix Table A.2). We analyze the propensity of FOR votes for each governance and executive compensation proposal at each fund-firm-year level. In other words, we identify each individual fund vote for

9

We repeat the analysis using Indexers and Non-Indexers as percentage of total institutional ownership (Own%) and results are similar and available upon request. 10

Institutional Investors: Power and Responsibility, speech by Commissioner Luis A. Aguilar, U.S. Securities and Exchange Commission, April 19, 2013. https://www.google.ca/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact=8&ved=0ahUKEwixnsfkwt 3YAhXl7IMKHU4iAwoQFggpMAA&url=https%3A%2F%2Fwww.sec.gov%2Fnews%2Fspeech%2F2013spch041913laahtm&usg=AOvVaw3fethkOGsZCCYJVAZoEO4u 11 2016 U.S. Shareholder Activism Review and Analysis, Sullivan & Cromwell LLP, November 28, 2016. 19

each firm and in each year of the sample as a 1 if vote cast is FOR and 0 otherwise (that includes votes registered as AGAINST, ABSTAIN OR WITHHOLD). Based on results reported by prior literature, we expect that this effect will be stronger for post-addition years for the addition sample than for the matching firms. Although our main goal is to investigate voting behavior of index funds and ultimately its impact on firm governance, for comparison purposes in this setting we also examine a second sub-group: “Active-Voting Funds.” To identify these funds, we divide the voting data into deciles and define Active-Voting Funds as those that that vote against management recommendations more than 50% of the time. This flags approximately 10% of the funds in our data set. We keep our original identification of Index Funds as defined in the previous sections and, by default, we classify the remaining funds as General Funds. We include the same set of firm level controls used in prior tests: firm size, leverage, CapEx/Assets, profitability as captured by ROA, ratio of cash flow to assets and firm age. We also include the ISS Recommendation, used previously, an indicator variable that takes the value of 1 if ISS has issued a FOR recommendation for the proposal, and zero if it issued a recommendation AGAINST. Similarly, we include an indicator equal to 1 if management issues a FOR recommendation, and zero if they recommend against (Management Recommendation). These variables are included in order to capture the link between the advice of the proxy advisory service and of management and the odds of a FOR vote. The general specification is: Logit (FOR vote) = α + β1POST + β2ADD + β3POSTADD + β4SponsorShareholder + β5 ISS Recommendation + (vector of firm specific controls) + (industry fixed effects) + ε

(3)

20

Table 9 reports results for governance proposal votes. In all specifications the dependent variable is an indicator variable equal to one if the proposal received a FOR vote, and zero otherwise.12 In Model (1) we report the coefficients of a general model, for our entire sample as specified in equation (3), above. Models (2), (3) and (4) split the sample into the FOR votes of Indexers, General Funds and Active-Voting Funds, respectively. In Model (1) we find that governance proposals for the addition sample have 0.30 lower odds of a FOR vote than the matched sample, and that during the post-addition years governance proposals have 0.42 lower odds of a FOR vote than the years before addition.13 However, the effect of post-addition years is stronger for addition firms relative to the matched sample, as seen in the positive and significant coefficient of the interaction term for all the 4 models in Table 9. The economic interpretation of the POST*ADD term in Model 1 tells us that addition firms in the post-addition years have 3.5 higher odds of a FOR, for the entire sample, and 4.4 higher odds for the sample restricted to active-voting funds only. In the remaining three specifications the sign and significance of the coefficients for the main variables remains the same. Of note, the likelihood of FOR votes increases significantly for proposals with a FOR recommendation by management, and for proposals having a FOR recommendation from ISS. The magnitudes of these range from 4.75 to 16.29 higher odds for a FOR recommendation from management and from 2.6 to 18.28 higher odds for a ISS FOR recommendation. These results hold for the entire sample and for all three sub-samples (Indexers, General Funds and Active Voting Funds). Of note, governance related proposals sponsored by shareholders have significantly higher odds of FOR votes in our overall sample and

12 13

The options for voting are: FOR, AGAINST, ABSTAIN OR WITHHOLD. For example, for the coefficient of ADD, -1.206, we calculate the odds ration as e-1.206 = 0.3

21

the effect seems to come exclusively from the sub-sample of Active Voting Funds (Model 4). This effect is absent for the Indexers and General Funds sub-samples tests (Models 2 and 3). In Table 10, we focus on the same probability of a fund voting FOR a governance proposal, but split the sample into management and shareholder proposals. Once again, we focus on Indexers (Models 1 and 2), General Indexers (Models 3 and 4) and Active Voting Funds (Models 5 and 6). The coefficients for the addition indicator (ADD), post addition years (POST) and interaction of the two (POST*ADD), have the same sign and significance levels as in the prior analysis. Specifically, there is a lower FOR vote propensity for addition firms, it is lower in the post-addition years, but differentially higher for addition firms in the post addition period. These findings hold for both sub-samples of management and shareholder proposals. In all six models, a FOR recommendation from the proxy advisor leads to significantly higher odds of a FOR vote for a governance proposal. While the results are similar for the management recommendation for the sub-sample of indexers and general funds, it is insignificant for Active Voting Funds. This suggests that Active Voting Funds do not, on average, simply follow management recommendations. Taken together, the results for voting on governance proposals are consistent with a higher overall probability of a FOR vote once a firm is added to the S&P 500 Index. This effect is present for both management and shareholder sponsored proposals. Moreover, the odds of a FOR vote increase with a FOR recommendation from both management and ISS. Therefore, the results are supportive of a second channel of governance influence: voting behavior changes for funds that own index included firms. This increased support for governance is not solely driven by the increased frequency of governance proposals in the post-addition years, as our tests in Tables 9 and 10 use the number of FOR votes scaled by the total number of votes cast. 22

Table 11 replicates the analysis in Tables 9 using the set of executive compensation proposals, again both management and shareholder sponsored. While results are generally similar to the analysis of governance proposals, while General and Active-voting sub-groups exhibit lower propensity of FOR votes during the post-addition years for the entire sample (Model 1 of Table 11) for executive compensation proposals Index Funds show a significantly higher proportion (Model 2 of Table 11), all else equal. This shows that indexers increased their support for proposals related to executive compensation during the second half of our coverage period relative to the prior years, pointing to their active involvement in compensation issues. The post-addition years show significantly higher odds of FOR votes for the entire sample and all individual sub samples (odds ranging from 1.16 to 1.42 higher) and these odds increase significantly for shareholder sponsored proposals, and management or ISS FOR recommendations. 4. Impact of Voting on Executive Compensation Our analyses thus far demonstrate that overall voting support at the firm level increases significantly for both governance and executive compensation proposals during the post-addition years. Moreover, we show that two main channels for this effect are, first, the changes in the institutional ownership make-up of the firms as they are added to the index and, second, changes in funds’ propensity to vote in support of these proposals (governance and compensation). The final question we address is whether these changes lead to governance adjustments in these firms. Given our focus on executive compensation proposals, we look at changes in aspects of compensation with emphasis on whether there are significant differences between the addition and match firms in the pre-and post addition periods.

23

We focus first on total executive compensation, and model it as a function of firm specific controls, while including our indicator variables capturing the differential effects of addition firms, post addition years, and the two combined. Our model is as follows: Total Compensation = α + β1ADD + β2POST + β3POSTADD + (vector of firm specific controls) + (industry fixed effects) + ε

(4)

Total compensation is defined as the sum of salary, bonus and incentive based compensation, as reported in Execucomp (TDC1), in millions. Table 12 presents the results of specification (4) above, Model 1 examines aggregate pay for the top 5 executives as a group, Model 2 uses the median value of the top 5, and Model 3 uses only CEO total compensation. The results in Table 12 suggest that, overall, addition firms tend to have significantly higher levels of total compensation relative to the match sample. However, we find no statistical difference in total compensation for the addition firms during the post-addition years. We next focus on the pay-for performance sensitivity (delta) and report results in Table 14. We find that, post-addition, the median pay-for-performance average sensitivity of the top 5 executives is lower, while it is higher for the CEO. Moreover, in all 3 specifications we see a negative coefficient for the interaction of POST*ADD suggesting that the PPS (pay-forperformance sensitivity) is generally lower for index firms after addition compared with the matched sample. 5. Conclusions We use a sample of firms added to the S&P 500 Index along with a size and industry matched sample from 2004 through 2014 in order to further investigate the mechanisms through which institutions can influence governance changes in the firms they own. As an S&P 500 24

Index addition leads to a plausibly exogenous change in firm ownership structure, our analysis provides a basis for inferring causal relationships between ownership concentration and governance outcomes. We find that governance proposals receive significantly higher approval rates after index addition, and this effect is present even after controlling for possible time trends. When we look at sub-samples based on the proposal sponsor, the effect is attributable solely to higher support for shareholder-sponsored proposals. Moreover, we observe a higher propensity of FOR votes for proposals at firms with greater ownership by index funds, supporting the view that indexers are increasingly influential in governance practices at portfolio firms. We also explore two key avenues leading to the observed increased support of governance related proposals in corporate boards, after additions to the index. The first is the change in ownership structure prompted by index inclusion. Our results suggest that while the ownership of institutions in aggregate does not change significantly, ownership shifts among the different types of institution. Specifically, indexers increase their stakes, as reported in prior studies, while non-indexers reduce their holdings. Therefore, the shift in voting power among the different institutions is one direct way in which firm monitoring changes. The second mechanism we posit affects firm governance structures is the voting behavior of institutional investors (classified into different categories) once a firm becomes part of the S&P 500 Index. We find higher propensities of FOR votes on governance proposals during the post-addition years, an effect that is concentrated in the set of shareholder proposals. In addition, we document a decrease in overall propensity of funds to vote in line with management recommendations on governance proposals after index addition.

25

Finally, we look at the impact these voting results, ownership and voting propensity have on the compensation structure. We find that overall, S&P 500 Index firms tend to have significantly higher levels of total compensation relative to the match sample however, PPS is significantly lower, in general for the addition firms, post-addition.

26

References

Aghion, Philippe, John Van Reenen, and Luigi Zingales. "Innovation and institutional ownership." The American Economic Review 103.1 (2013): 277-304. Appel, Ian R., Todd A. Gormley, and Donald B. Keim. "Passive investors, not passive owners." Journal of Financial Economics 121.1 (2016): 111-141. Appel, Ian R., Todd A. Gormley, and Donald B. Keim. "Standing on the shoulders of giants: The effect of passive investors on activism". No. w22707. National Bureau of Economic Research, 2016. Biktimirov, Ernest N., Arnold R. Cowan, and Bradford D. Jordan. "Do demand curves for small stocks slope down?." Journal of Financial Research 27, no. 2 (2004): 161-178. Boos, Roger J., and Michele Ruotolo. "General criteria for S&P US index membership." Standard and Poor’s Corporation (2000). Cai, Jie, Jacqueline L. Garner, and Ralph A. Walkling. "Electing directors." The Journal of Finance 64, no. 5 (2009): 2389-2421. Chou, Julia, Lilian Ng, and Qinghai Wang. "Are better governed funds better monitors?" Journal of Corporate Finance 17.5 (2011): 1254-1271. Chen, Honghui, Gregory Noronha, and Vijay Singal. "The price response to S&P 500 index additions and deletions: Evidence of asymmetry and a new explanation." The Journal of Finance 59.4 (2004): 1901-1930. Crane, Alan D., Sébastien Michenaud, and James P. Weston. "The effect of institutional ownership on payout policy: Evidence from index thresholds." The Review of Financial Studies 29.6 (2016): 1377-1408.

27

Davis, Gerald F., and E. Han Kim. "Business ties and proxy voting by mutual funds." Journal of Financial Economics 85.2 (2007): 552-570. Demsetz, Harold, and Kenneth Lehn. "The structure of corporate ownership: Causes and consequences." Journal of Political Economy 93.6 (1985): 1155-1177. Demsetz, Harold, and Belen Villalonga. "Ownership structure and corporate performance." Journal of Corporate Finance 7.3 (2001): 209-233. Denis, Diane K., John J. McConnell, Alexei V. Ovtchinnikov, and Yun Yu. "S&P 500 index additions and earnings expectations." The Journal of Finance 58, no. 5 (2003): 1821-1840. Ertimur, Yonca, Fabrizio Ferri, and Volkan Muslu. "Shareholder activism and CEO pay." Review of Financial Studies (2010): hhq113. Ferri, Fabrizio, and David A. Maber. "Say on pay votes and CEO compensation: Evidence from the UK." Review of Finance 17, no. 2 (2013): 527-563. Gillan, Stuart and Laura Starks, 2000. Corporate Governance Proposals and Shareholder Activism: The Role of Institutional Investors, Journal of Financial Economics 57: 275-305. Gowri Shankar, S., and James M. Miller. "Market reaction to changes in the S&P SmallCap 600 index." Financial Review 41, no. 3 (2006): 339-360. Iliev, Peter, and Michelle Lowry. "Are Mutual Funds Active Voters?" Review of Financial Studies 28, no. 2 (2014): 446-485. Lynch, Anthony W., and Richard R. Mendenhall. "New evidence on stock price effects associated with changes in the S&P 500 index." The Journal of Business 70, no. 3 (1997): 351383. Matvos, Gregor, and Michael Ostrovsky. "Heterogeneity and peer effects in mutual fund proxy voting." Journal of Financial Economics 98, no. 1 (2010): 90-112. Morgan, Angela, Annette Poulsen, Jack Wolf, and Tina Yang. "Mutual funds as monitors: Evidence from mutual fund voting." Journal of Corporate Finance 17, no. 4 (2011): 914-928. Mullins, William. "The governance impact of index funds: Evidence from regression discontinuity." Work. Pap., Sloan Sch. Manag., Mass. Inst. Technol (2014). Ng, Lilian, Qinghai Wang, and Nataliya Zaiats. "Firm performance and mutual fund voting." Journal of Banking & Finance 33, no. 12 (2009): 2207-2217. Pruitt, Stephen W., and KC John Wei. "Institutional ownership and changes in the S&P 500." The Journal of Finance 44, no. 2 (1989): 509-514.

28

Schmidt, Cornelius and Rudiger Fahlenbrach “Do exogenous changes in passive institutional ownership affect corporate governance and firm value?” Journal of Financial Economics 124.2 (2017): 285-306.

29

1

Tables

Table 1: S&P500 Index Additions by Year of Addition.This table presents the number of S&P 500 Index addition firms by year for the sample of 256 additions to the S&P 500 Index from January 2004 through December 2014. Year 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Total

Additions 20 16 31 38 35 29 16 19 18 19 15 256

Matched Firms 18 14 25 34 34 29 16 19 18 18 14 239

30

Additions % of Sample 7.87 6.30 12.20 14.57 13.78 11.42 6.30 7.48 7.09 7.09 5.91 100.00

Table 2: S&P500 Additions and Matched Firms Summary Statistics. This table reports firm characteristics for S&P 500 Additions and Matched Firms for the year of the addition. All variable definitions can be found in Appendix A.1 Additions Median

Variable

N

Mean

Total Assets ($ Million) Number of Common Shareholders Asset Turnover CapEx/Assets Cash Flow/Assets Firm Age Leverage (LT/AT) Market Cap ($ Million) Market to Book ROA Size - log(Total Assets ($ Million)) Tobin’s Q

245 245 245 245 245 245 245 245 245 245 245 245

13403.82 22.5 0.83 0.05 0.12 20.26 0.55 11500.92 2.49 0.08 8.83 2.49

6622 2.79 0.62 0.03 0.11 15 0.55 8190.76 1.81 0.06 8.8 1.81

Std

N

26317.15 47.48 0.71 0.07 0.08 16.58 0.23 14908.1 1.82 0.07 1.06 1.82

214 214 214 214 214 214 214 214 214 214 214 214

Matched Firms Mean Median 12685.5 29.41 0.7 0.06 0.09 18.46 0.57 10506.53 2.3 0.05 8.73 2.31

5894.04 1.64 0.5 0.04 0.09 15 0.56 6584.75 1.78 0.05 8.68 1.79

Std

T Statistic

22743.61 113.44 0.58 0.08 0.11 14.19 0.22 18752.25 1.84 0.11 1.15 1.83

0.31 -0.77 2.17 -0.9 2.4 1.25 -0.68 0.62 1.09 2.6 1 1.05

31

Table 3: Ownership Characteristics. This table reports ownership characteristics for S&P 500 Additions and Matched Firms before (PRE) and after (POST) the addition. All ownership values are annualized and averaged over 4 to 1 year before the addition (for PRE) and 0 to 4 years after the addition (for POST). All variable definitions can be found in Appendix A.1

Variable

PRE

No Funds No Index Funds No Non Index Funds Own (%) Index Fund Own (%) Non Index Fund Own (%) Index Fund Own / Own (%) Non Index Fund Own / Own (%)

402.67 59.67 317.87 31.7 3.99 27.89 14.25 85.75

Additions POST T Statistic 682.07 143.36 515.04 33.7 7.23 26.28 22.46 77.54

11.09 32.55 9.57 1.94 13.9 -1.47 8.61 -8.61

PRE 344.59 53.78 299.54 27.58 3.04 24.45 11.43 88.57

Matched Firms POST T STatistic 443.39 80.81 360.38 29.83 4.56 25.04 15.1 84.9

4.76 7.21 3.05 1.8 5.76 0.48 4 -4

32

Table 4: S&P500 Proposals.Panel A contains the number of proposals by sponsor and split into proposals before the addition (PRE) and after the addition (POST). Panel B shows only governance proposals. Panel A-Total Number of Agenda Items Additions Matched Firms PRE POST PRE POST Management 6,198 11,487 6,358 9,624 Shareholder 114 467 99 276 Total 6,312 11,954 6,457 9,900

Panel B - Governance Proposals Distribution Management Shareholder PRE POST PRE POST Addition 99 285 47 187 Matched Firm 303 247 25 89 Total 402 532 72 276

Total 618 664 1,282

Panel C - Executive Compensation Proposals Distribution Management Shareholder Total PRE POST PRE POST Addition 543 1,249 12 62 1,866 Matched Firm 530 995 16 31 1,572 Total 1,073 2,244 28 93 3,438

33

Table 5: Proposal Support Regressions (Goverance Proposals). This table shows coefficients and standard errors (in parenthesis) for panel regressions of proposal voting support (fraction of FOR votes) for our sample of SP500 additions and matched firms from 4 years before to 4 years after the addition. ADD is an indicator variable taking the value 1 if the firm was added to the SP500 during our sample period, and 0 otherwise. POST is an indicator variable taking the value 1 for observations occurring after the addition and 0 otherwise. Industry fixed effects are based on 2-digit SIC codes. All variable definitions can be found in Appendix A.1 Dependent variable: Votes FOR the Proposal (%) (1)

(2) ∗∗

(3)

(4)





ADD

−0.067 (0.029)

−0.049 (0.028)

−0.057 (0.029)

−0.041 (0.028)

POST

−0.026 (0.029)

−0.020 (0.028)

0.013 (0.028)

0.012 (0.027)

POST * ADD

0.068∗∗ (0.033)

0.049 (0.031)

0.077∗∗ (0.033)

0.058∗ (0.032)

0.163∗∗∗ (0.024)

ISS Recommendation

Idx Fd Own (%)

0.014∗∗∗ (0.004)

0.166∗∗∗ (0.024)

0.012∗∗∗ (0.004)

Non Idx Fd Own (%)

−0.001 (0.001)

−0.0003 (0.001)

Firm Size

−0.006 (0.010)

0.001 (0.009)

−0.014 (0.010)

−0.005 (0.009)

Leverage

−0.037 (0.041)

−0.065 (0.040)

−0.019 (0.043)

−0.055 (0.041)

CapEx/Assets

0.098 (0.125)

0.030 (0.120)

0.045 (0.127)

−0.013 (0.121)

Cash-Flow/Assets

−0.511 (0.428)

−0.242 (0.409)

−0.654 (0.438)

−0.326 (0.419)

ROA

0.577 (0.412)

0.302 (0.394)

0.699∗ (0.422)

0.371 (0.404)

Firm Age

−0.0004 (0.001)

−0.0004 (0.001)

0.0001 (0.001)

0.0001 (0.001)

Constant

0.709∗∗∗ (0.131)

0.530∗∗∗ (0.128)

0.850∗∗∗ (0.132)

0.636∗∗∗ (0.129)

YES YES 546 0.799 0.752

YES YES 546 0.819 0.775

YES YES 546 0.794 0.746

YES YES 546 0.814 0.770

Industry FE Proposal Type FE Observations R2 Adjusted R2



Note:

34

p