CORPORATE POLITICAL SPENDING AND THE SIZE EFFECT

* Juris Doctor 2017, Columbia Law School.

Introduction

In the years since Citizens United v. FEC, 1 558 U.S. 310 (2010). corporate-political-spending disclosure has become an increasingly heated public policy issue. The portion of the Court’s opinion that championed shareholder rights to make decisions about corporate political speech generated a substantial, interdisciplinary literature, 2 See, e.g., Lucian A. Bebchuk & Robert J. Jackson, Jr., Corporate Political Speech: Who Decides?, 124 Harv. L. Rev. 83, 83 (2010); Rajesh K. Aggarwal et al., Corporate Political Donations: Investment or Agency?, 14 Bus. & Pol., no. 1, 2012, Article. 3, at 1, 6; For several critical views of efforts to force disclosure, see Symposium on Corporate Political Spending, 3 Harv. Bus. L. Rev. 363 (2013). and shareholders responded by demanding political-spending disclosure through a bevy of shareholder proposals. 3 See Lucian A. Bebchuk & Robert J. Jackson, Jr., Shining Light on Corporate Political Spending, 101 Geo. L.J. 923, 938–39 (2013) [hereinafter Bebchuk & Jackson, Shining Light]; infra Appendix at Table A.1. However, many commentators have argued that shareholder activists’ efforts to bring about disclosure on their own are bound to be inadequate, and in 2011, a group of law professors petitioned the Securities and Exchange Commission (SEC) for rules that would mandate disclosure by public companies—a petition that generated over a million comment letters (and considerable controversy). 4 Lucian Bebchuk & Robert J. Jackson, Jr., The Million-Comment-Letter Petition: The Rulemaking Petition on Disclosure of Political Spending Attracts More than 1,000,000 SEC Comment Letters, Harvard Law Sch. Forum on Corp. Governance and Fin. Regulation (Sept. 4, 2014), http://corpgov.law.harvard.edu/2014/09/04/the-million-comment-letter-petition-the-rulemaking-petition-on-disclosure-of-political-spending-attracts-more-than-1000000-sec-comment-letters [http://perma.cc/J64F-54PA]. Critics of the petition have charged that mandatory rules are unnecessary because private ordering 5 Private ordering refers to the mechanism by which private actors—here, shareholders and boards of directors—set the rules that govern their conduct. 3 James D. Cox & Thomas Lee Hazen, Treatise on the Law of Corporations § 15:23, Westlaw (3d ed., database updated Dec. 2016). should be sufficient to induce the optimal level of disclosure, 6 See, e.g., Larry Ribstein, Should the SEC Regulate Corporate Political Speech?, Truth on the Mkt. (Aug. 4, 2011), http://truthonthemarket.com/2011/08/04/
should-the-sec-regulate-corporate-political-speech [http://perma.cc/TW5L-7FC3] (favoring “experimentation and evolution” instead of “a one-size-fits-all rule”); infra note 30 and accompanying text. while supporters have questioned the assumption that private ordering is enough in light of shareholders’ collective-action problems. 7 See Bebchuk & Jackson, Shining Light, supra note 3, at 947–49; infra Part I. To be sure, there are other facets to this debate. Compare Bebchuk & Jackson, Shining Light, supra note 3, at 937–49 (offering several arguments in favor of mandatory rules), with Symposium on Corporate Political Spending, supra note 2 (presenting several pieces opposed to mandatory rules on various grounds). However, this Comment focuses on collective-action problems. This debate has continued unabated, with Senator Elizabeth Warren calling on President Obama in October 2016 to remove then–SEC Chair Mary Jo White for failing to act on the 2011 petition. 8 See Andrew Ackerman, Elizabeth Warren to Obama: Fire SEC Chief Mary Jo White, Wall St. J. (Oct. 14, 2016), http://www.wsj.com/articles/elizabeth-warren-to-obama-fire-sec-chief-mary-jo-white-1476439200 (on file with the Columbia Law Review). Although the 2011 petition may, given the current political environment, languish for the time being, the push for political-spending disclosure seems unlikely to disappear for long.

Despite the persistence of this debate, there has been virtually no empirical investigation into vote outcomes for shareholder proposals related to political spending. 9 Empirical research related to corporate political spending has focused on different issues, such as the drivers of political-spending proposals themselves and the effects of corporate political spending on shareholder value. See generally John C. Coates IV, Corporate Politics, Governance, and Value Before and After Citizens United, 9 J. Empirical Legal Stud. 657 (2012); Michael J. Cooper et al., Corporate Political Contributions and Stock Returns, 65 J. Fin. 687 (2010); Michael Hadani & Douglas A. Schuler, In Search of El Dorado: The Elusive Financial Returns on Corporate Political Investments, 34 Strategic Mgmt. J. 165 (2012); Aggarwal et al., supra note 2; Geeyoung Min & Hye Young You, Political Origins of Shareholder Activism: Corporate Political Spending and Shareholder Proposals (Va. Law & Econ. Research Paper No. 15, 2015), http://ssrn.com/
abstract=2601181 (on file with the Columbia Law Review). Although one study has examined the determinants of proposal implementation, the study did not consider the determinants of vote outcomes. See Vishal P. Baloria et al., Shareholder Activism and Voluntary Disclosure Initiation: The Case of Political Spending 56 tbl.5, panel B (Oct. 15, 2015) (unpublished manuscript), http://ssrn.com/abstract=2079131 (on file with the Columbia Law Review) (analyzing the probability of implementation but not including the percentage of votes cast in favor of a proposal as a dependent variable). This Comment begins to fill this gap in the literature through an empirical study of approximately six years of vote-outcome data for political-spending proposals, focusing on the potential for a “size effect.” If shareholders’ collective-action problems are indeed a barrier to private implementation of effective political-spending disclosure, then it should be more difficult for shareholder proposals to achieve high levels of support at larger companies—where one would expect these collective-action problems to be the most severe. 10 See infra Part I. Consistent with this theory, I find that even controlling for firm-specific characteristics, doubling a company’s market capitalization is associated with a two-percentage-point decline in shareholder support. 11 Infra Table 1. Although this research design does not permit a conclusive causal inference, the evidence nevertheless provides suggestive empirical support for one of the most compelling justifications for mandating disclosure of companies’ political spending. 12 See infra Part III.

The rest of this Comment proceeds in three Parts. Part I provides background information and explains the Comment’s theory, while Part II undertakes the Comment’s empirical analysis and considers objections and limitations. Part III discusses implications for lawmakers and policymakers. A brief conclusion follows.

I. Corporate Political Spending, Shareholder Proposals, and Collective-Action Problems

In the landmark case Citizens United v. FEC, the Supreme Court invalidated certain federal restrictions on corporate political spending, 13 See 558 U.S. 310, 365 (2010). spurring considerable interest among investors in how corporations would use their newfound freedom. 14 See Bebchuk & Jackson, Shining Light, supra note 3, at 937–41. However, the law does not currently require corporations to disclose their political spending to shareholders. 15 Id. at 925. Notwithstanding the extensive mandatory disclosure requirements imposed by federal securities law, 16 See, e.g., SEC, Form 10-K, Annual Report Pursuant to Section 13 or 15(d) of the Securities Exchange Act of 1934, http://www.sec.gov/about/forms/form10-k.pdf [http://perma.cc/ES6K-9JKV] (last visited Sept. 13, 2017) (setting forth the requirements as to public corporations’ annual disclosures). investors must instead resort to shareholder proposals 17 Exchange Act Rule 14a-8 requires public companies to include qualifying shareholder proposals in their annual proxy statements. See 17 C.F.R. § 240.14a-8 (2016). and other forms of private pressure to convince companies to disclose their political spending—a task investors have eagerly taken up since Citizens United, albeit with mixed results. As Figure 1 shows, political-spending shareholder proposals have been very common since 2012: An average of approximately eighty proposals per year have gone to a vote, 18 Infra Figure 1; infra Appendix at Table A.2. making corporate political spending one of the most popular topics for shareholder proposals over the past several proxy seasons. 19 See Bebchuk & Jackson, Shining Light, supra note 3, at 938–40. But as Figure 2 demonstrates, many corporations still fail to disclose their political spending, or they do so only piecemeal. 20 See Ctr. for Political Accountability, The 2016 CPA Zicklin Index of Corporate Political Disclosure and Accountability 10–11 (2016), http://files.politicalaccountability.net/
index/2016CPAZicklinIndex.pdf [http://perma.cc/NW4T-E674] (noting over half of companies fail to disclose “dark money” payments and there continue to be “persistent basement-dwellers” that “lag[] behind” in their political-spending disclosure and accountability); infra Figure 2.

Figure 1: Number of Political-Spending Shareholder Proposals

Voted on by Year 21 For an explanation of the procedures used to collect these data, see infra notes 34–35 and accompanying text. I define a “proxy year” as beginning on November 1 of the preceding calendar year and ending on October 31 of the current calendar year.

Figure 2: Political-Spending Accountability and Disclosure Levels in the S&P 500 22 The x-axis variable is the full score as a percentage of total points from the 2016 CPA–Zicklin Index. For an explanation of this index, see infra notes 39–41 and accompanying text.

Shareholders face well-known collective-action problems when lobbying individual companies for corporate change, which may explain why they often fail to obtain effective political-spending disclosure. First, the process is expensive for shareholders, yet they receive only a small portion of any benefits that their actions produce. Although companies bear the cost of including shareholder proposals in their proxy statements, 23 See 17 C.F.R. § 240.14a-8 (2016). corporate management can also devote considerable resources—and company funds—to oppose these proposals. Shareholder-proponents, on the other hand, must bear the costs of convincing their fellow shareholders themselves, 24 Rule 14a-8 does not provide for reimbursement of shareholders’ costs. See id. creating a classic public-goods problem. Moreover, the act of drafting and submitting a proposal, which includes navigating the shareholder proposal rules’ procedural and eligibility limitations, can also be costly. 25 Cf. id. (describing the requirements that a shareholder must meet to use the shareholder proposal rule). These costs are likely to reduce the number of proposals brought in the first place. Second, the modern-day dominance of institutional investors likely exacerbates this problem. The political preferences of the individuals who make voting decisions on political-spending proposals—such as managers of large mutual funds—may diverge from those of the beneficial owners for whom they are trustees. Specifically, one might suspect that relative to ordinary investors, these asset managers have a stronger interest in blocking political-spending disclosure. 26 This is plausible for numerous reasons. For example, asset managers may tend to be wealthier than fund beneficiaries, and corporations may tend to spend money on candidates who advance tax policies that favor wealthier individuals. Accordingly, to the extent that these asset managers vote against political-spending disclosure that beneficiaries want, vote totals would understate ordinary investors’ desire for this disclosure. 27 See infra note 87 and accompanying text (describing theoretical reasons for why investment-fund managers’ votes may diverge from beneficiaries’ preferences and empirical evidence from political-spending proposals that is consistent with this theory). Although investors could theoretically lobby these funds to vote in accordance with their political preferences, investment-fund beneficiaries are likely to face the same collective-action problems vis-à-vis investment funds as shareholders are likely to face vis-à-vis traditional firms.

Resting their case partially on the understanding that shareholders’ collective-action problems are likely to impede their efforts to obtain effective political-spending disclosure, 28 See Bebchuk & Jackson, Shining Light, supra note 3, at 948 (supporting the case for mandatory political-spending disclosure on the grounds that “shareholders face collective-action problems that make it costly for them to take action at individual firms”). There may be other reasons for mandatory disclosure as well. See generally id. at 941–49. This Comment, however, focuses on the collective-action-problem rationale. a group of law professors petitioned the SEC in 2011 to create rules requiring public companies to disclose their political spending. 29 Letter from Comm. on Disclosure of Corp. Political Spending to Elizabeth M. Murphy, Sec’y, SEC (Aug. 3, 2011) [hereinafter Petition], http://www.sec.gov/rules/
petitions/2011/petn4-637.pdf [http://perma.cc/UD6K-M2Q8]. In response, critics disputed the necessity of mandatory disclosure, claiming that private ordering should be enough to engender whatever disclosure regime shareholders may desire 30 See, e.g., Ribstein, supra note 6. These critics argue that because shareholders can privately pressure companies to disclose their political spending, mandatory disclosure imposed by the government is unnecessary. The critics maintain that a “privately ordered” disclosure regime—that is, one set by management and shareholders—is likely to better reflect investor preferences. See id. —and apparently doubting the severity of shareholders’ collective-action problems. To date, however, there has been no systematic empirical examination of the magnitude (or existence) of these collective-action problems. Thus, the extent to which shareholders who desire political-spending disclosure face collective-action problems remains an unresolved question with important policy implications.

How can one empirically test this theory? One possibility is to look at the relationship between firm size and outcomes related to shareholder support for political-spending proposals. 31 It is important to emphasize that this type of test speaks to collective-action problems only at the voting stage. As previously noted, shareholders may also face collective-action problems at the proposal design and submission stages. This Comment leaves examination of the severity of shareholders’ collective-action problems at those stages to future research. Standard economic intuition suggests that these collective-action problems will be more severe at larger companies: A group’s ability to overcome collective-action problems decreases as group size increases. 32 See Mancur Olson, The Logic of Collective Action: Public Goods and the Theory of Groups 53–65 (1965) (providing the classic theoretical treatment of the relationship between group size and collective-action problems). There are at least two reasons to suppose that this would hold in the shareholder-voting context. First, the average shareholder activist is likely to own a smaller proportion of shares in a larger company, 33 This is a function of the higher cost of obtaining the same proportion of a company’s shares as market capitalization increases. which reduces the activist’s incentive and ability to “get out the vote.” Second, it may be costlier for activists to convince other shareholders to support their proposals because larger firms have more dispersed shareholder bases. These two problems may manifest in both lower shareholder support among those who cast votes and lower voter “turnout” in general. Thus, an empirical test might search for evidence that either shareholder support or shareholder turnout is negatively associated with firm size. Although such a finding would not provide a conclusive basis for a causal inference, it could nevertheless provide a helpful starting point. Part II turns to this analysis.

II. Empirical Analysis of Shareholder Proposal Vote Outcomes

This Part reports the results of my empirical analysis of vote outcomes for shareholder proposals related to corporate political spending. Section II.A describes the data set. Section II.B reports a simple preliminary test of the relationship between firm size and voter turnout. Section II.C, the heart of the Comment, reports the results of a cross-sectional regression analysis with the percentage of votes cast in favor as the dependent variable. Section II.D considers issues related to political-spending disclosure among a larger population of corporations, including those at which a shareholder proposal did not reach a vote.

A. Data

This section describes the data and measures used in the remainder of this Part’s empirical analysis. To investigate the determinants of vote outcomes and the possibility of a size effect, I gather voting results and proposal data from FactSet’s Shark Repellent database, limiting the search to “political issues” shareholder proposals made under SEC Rule 14a-8 34 17 C.F.R. § 240.14a-8 (2016). Rule 14a-8 requires companies to include qualifying shareholder proposals in their annual proxy statements. See id. For helpful background on shareholder proposals under Rule 14a-8, see generally Patrick J. Ryan, Rule 14a-8, Institutional Shareholder Proposals, and Corporate Democracy, 23 Ga. L. Rev. 97, 104–23 (1988). for annual meetings from November 1, 2011, to October 31, 2016. 35 SharkRepellent, http://sharkrepellent.net [http://perma.cc/22VA-4CZA] (last visited Sept. 13, 2017). FactSet defines “political issues” proposals as those that “request that the board provide a report detailing the company’s policies regarding political contributions.” Id. I excluded proposals for which voting-result data were not available and proposals that management recommended. I collect financial, ownership, and industry data from FactSet Fundamentals, 36 Data: Fundamentals, FactSet, http://www.factset.com/data/company_data/fundamentals [http://perma.cc/FXC4-VVPX] (last visited Sept. 13, 2017). FactSet Ownership, 37 Data: Ownership, FactSet, http://www.factset.com/data/company_data/ownership [http://perma.cc/B5L3-JPXU] (last visited Sept. 13, 2017). and the CPA–Zicklin Index, 38 CPA–Zicklin Index, Ctr. for Political Accountability, http://politicalaccountability.net/index [http://perma.cc/6DA3-KTSM] (last visited Sept. 13, 2017) [hereainafter Current Report]; Past CPA–Zicklin Index Reports, Ctr. for Political Accountability, http://politicalaccountability.net/reports/past-cpa-zicklin-index-reports [http://perma.cc/
W6ET-TB8S] (last visited Sept. 13, 2017) [hereinafter Past Reports]. respectively

I measure vote outcomes using the percentage of votes cast in favor of each proposal (excluding abstentions); company size using each company’s market capitalization, which equals the number of shares outstanding multiplied by the market price per share at the end of the fiscal year that precedes the annual meeting, in billions of dollars; and preexisting political accountability and transparency using each company’s overall CPA–Zicklin Index score as a percentage of the total available points. A joint project of the Center for Political Accountability and the Zicklin Center for Business Ethics Research at the University of Pennsylvania’s Wharton School, the CPA–Zicklin Index quantifies corporations’ political accountability along three dimensions: disclosure, oversight, and policy. 39 See Current Report, supra note 38. To calculate each company’s final score, CPA researchers reviewed the company’s website to assign ratings along a number of subdimensions, 40 CPA used twenty-nine different subdimensions in the 2011 index, twenty-five dimensions in the 2012 index, and twenty-four dimensions thereafter. See Current Report, supra note 38; Past Reports, supra note 38. These year-to-year methodological changes may slightly change companies’ scores over time, but because this Comment uses each company’s raw score as a percentage of the total available points, the year-to-year changes are unlikely to materially affect the results. summed the numeric scores from each category to calculate a total “raw score,” and computed the score as a percentage of the total available points. 41 Additional information on CPA’s methodology, including on the various subdimensions along which it scores companies, is available in each individual report. See Current Report, supra note 38; Past Reports, supra note 38. I match each proposal with the CPA–Zicklin report published most recently prior to the annual meeting at which the proposal was presented. 42 Unfortunately, the index tracks a limited number of companies—just under 500 in 2016 and only 100 in 2011—which resulted in the loss of a number of observations when the CPA–Zicklin score was introduced into the regression models. To account for any potential time-based biases this may introduce, I also run multivariate regressions for a subset of the data that includes only proposals voted on after the release of the 2015 CPA–Zicklin Index. See infra note 69.

To measure financial performance, I use several accounting measures, each reported as of the end of the fiscal year preceding the annual meeting: return on assets (ROA), which equals operating income divided by total assets, multiplied by 100 to report a percentage; operating margin, which equals operating income divided by sales, again multiplied by 100; price-to-book ratio (PTB), which equals market capitalization divided by the accounting value of common equity, again multiplied by 100; and earnings-per-share (EPS) growth, a dummy variable with 0 equal to no growth. I measure company ownership using two separate variables: first, the percentage of shares owned by institutional investors, and second, the percentage of shares owned by corporate executives and directors. I measure proponent identity using a dummy variable that indicates whether the shareholder-proponent was an institutional investor or an individual investor (with 0 marking an individual). Finally, I measure industry using Global Industry Classification Standard (GICS) sector (a nominal variable converted into dummies for fixed effects). The Appendix presents summary statistics, information on modifications made to these data, and selected descriptive statistics by year.

B. Firm Size and Voter Turnout

To test the plausibility of a link between firm size and voter turnout, I run a bivariate ordinary least squares (OLS) regression. 43 This simple statistical technique allows the researcher to assess the relationship between a “dependent” or “outcome” variable and an “independent” or “predictor” variable, producing a “best-fit” line that can be laid on a scatterplot. The dependent variable is a proxy for voter turnout, measured as votes cast either “for” or “against” the proposal as a percentage of the total votes outstanding. 44 Algebraically, this is equivalent to: (expressed as a percentage). The independent variable, a proxy for firm size, is the natural log of market capitalization. Figure 3 plots the results of this regression.

Figure 3: Voter Turnout by Market Capitalization 45 The red line is a best-fit line from an OLS regression. The light-gray area represents a 95% confidence interval.

Figure 3 provides initial support for the theory that there is a connection between firm size and shareholders’ collective-action problems. Figure 3 shows a negative association between firm size and voter turnout. The estimated slope coefficient for this regression is −1.662, which is statistically significant at conventional thresholds (p < 0.01). This result implies that doubling a company’s market capitalization is associated with an approximately one-percentage-point decrease in voter turnout, and multiplying a company’s market capitalization by ten is associated with an approximately four-percentage-point decrease in turnout. 46 For an explanation of how to interpret logarithmically transformed coefficients, see infra note 61. This result is consistent with the theory that shareholders’ rational apathy is greater at larger firms, which in turn leads to reduced turnout. This evidence alone does not, however, directly speak to the relationship between firm size and shareholders’ ability to gain support for their proposals. The next section turns to this issue more directly.

C. Firm Size and Shareholder Support

To further investigate the possibility of a size effect, I run several different regressions with the percentage of votes cast “for” the proposal as the dependent variable. 47 Algebraically, the percentage of “for” votes is equivalent to: (expressed as a percentage). These regressions speak more directly to the heart of the issue: the possibility that firm size may impede shareholders’ ability to obtain support for their proposals. This section reports the results of these regressions.

1. Bivariate Regression Analysis. — To begin, I run a bivariate OLS regression with the percentage of votes cast in favor as the dependent variable and market capitalization as the independent variable. The estimated slope coefficient for the line produced by this bivariate model is −0.047, which is statistically significant at conventional levels (p < 0.01). 48 See infra Figure 4. Figure 4 plots vote outcomes against market capitalization, with the OLS best-fit line in red, a LOESS best-fit curve in blue, 49 The LOESS best-fit curve attempts to capture nonlinear aspects of the relationship between voting results and market capitalization, generating a “smoothed” curve rather than a straight line. and a 95% confidence interval for the OLS line in light gray.

Figure 4: Vote Outcome by Market Capitalization 50 The blue curve is a best-fit curve from a LOESS regression, while the red line is a best-fit line from an OLS regression. The light-gray area represents a 95% confidence interval for the OLS line.

Based on this bivariate analysis alone, the results are striking. Under the OLS model, a $100 billion increase in market capitalization is associated with a five-percentage-point decrease in “for” votes, 51 See supra Figure 4. while the LOESS regression shows that the relationship is strongest between $0 and $100 billion in market capitalization. 52 See supra Figure 4. Indeed, among the eleven instances in which a proposal received majority support, each company had a market capitalization of under $50 billion, and among those fifty-eight instances in which a proposal garnered over 40% support, only one firm’s market capitalization exceeded $100 billion (with no firm’s market capitalization exceeding $150 billion). 53 See supra Figure 4. Overall, these results suggest that it is considerably more difficult for a proposal to achieve support at a large company than at a small one.

However, it would be inappropriate to draw a causal inference from this simple bivariate analysis. Although the negative association between voting results and market capitalization is clear, correlation does not imply causation—there is always a concern that “lurking” variables may account for the observed relationship. 54 See, e.g., Brian L. Joiner, Lurking Variables: Some Examples, 35 Am. Statistician 227, 227 (1981) (defining a lurking variable as “a variable that has an important effect and yet is not included among the predictor variables under consideration”). For example, large companies might tend to already comprehensively disclose their political spending, causing shareholders to view additional disclosure as unnecessary at such companies. Alternatively, large companies may generally exhibit stronger financial performance, and shareholders may tend to vote against proposals by other shareholders when a company is doing well. Or it may be that ownership characteristics systematically differ according to company size, that the most effective types of shareholder-proponents target smaller companies, or that shareholders perceive companies in certain industries as particularly prone to misspend corporate funds on political contributions (with industry being correlated with size).

2. Multivariate Regression Analysis. — To overcome these difficulties, I run several multivariate OLS regressions with additional variables added as controls for preexisting political accountability and transparency, financial performance, ownership characteristics, proponent identity, and industry. Table 1 reports the results of these regressions. 55 Not reported are models that test interaction effects between operating margin and CPA–Zicklin score. The interaction-term coefficients in these models were neither statistically nor substantively significant. In each case, I logarithmically transform market capitalization to improve the fit of the linear model to the data, consistent with the dominant approach in the literature. 56 See, e.g., Randall S. Thomas & James F. Cotter, Shareholder Proposals in the New Millennium: Shareholder Support, Board Response, and Market Reaction, 13 J. Corp. Fin. 368, 378 (2007) (adopting a similar logarithmic transformation of market capitalization).

Table 1: Regression Analysis of Vote Outcomes 57 The figures reported without parentheses in the main part of the table are slope coefficients, while the figures reported in parentheses are t-statistics calculated using robust standard errors.

Dependent Variable:

Percentage of Votes Cast in Favor of Proposal

(1)

(2)

(3)

(4)

(5)

(6)

Log(Market Capitalization)

−3.464***

−3.141***

−3.368***

−3.205***

−3.038***

−2.327*

(−6.083)

(−3.970)

(−4.333)

(−3.429)

(−2.791)

(−1.827)

CPA–Zicklin Score

−0.102***

−0.108***

−0.123***

−0.130***

−0.121***

(−3.851)

(−3.913)

(−4.142)

(−3.802)

(−3.315)

ROA

−0.059

−0.054

−0.010

0.053

(−0.440)

(−0.414)

(−0.062)

(0.279)

Operating Margin

0.038

0.041

0.002

(0.627)

(0.689)

(0.654)

(0.024)

PTB

0.0001

0.0003

0.001

0.003

(0.034)

(0.174)

(0.236)

(1.106)

EPS Growth

−0.177

−0.380

−1.338

−0.507

(−0.116)

(−0.245)

(−0.799)

(−0.295)

Proponent Identity

5.167**

3.664

(2.131)

(1.443)

Constant

37.351***

41.520***

42.798***

43.413***

39.044***

32.128***

(16.265)

(13.591)

(13.477)

(5.197)

(4.218)

(2.952)

Ownership Controls 58 Ownership controls include the percentage of shares owned by institutional investors and the percentage of shares owned by corporate insiders. Industry FE

Yes

Industry FE

Yes

Observations

391

291

278

246

R²

0.104

0.151

0.167

0.173

0.198

0.260

Adjusted R²

0.101

0.145

0.149

0.167

0.201

F Statistic

36.997***

27.771***

11.462***

8.994***

8.284***

5.670***

Note:

*p < 0.1; **p < 0.05; ***p < 0.01

Table 1 shows that even controlling for preexisting political accountability, financial performance, ownership characteristics, proponent identity, and industry, the percentage of “for” votes continues to be negatively associated with market capitalization. In each model, this association is statistically significant (p < 0.01 in all models without industry fixed effects; p < 0.1 in the full model with industry fixed effects). 59 Supra Table 1. Because reverse causality is practically impossible 60 See infra note 67 and accompanying text. and the multivariate specifications include covariates tracking the more obvious potential alternate causes, these results suggest—but cannot conclusively demonstrate—a causal relationship.

The results are also substantively significant. According to the full model without industry fixed effects, doubling a company’s market capitalization is associated with a two-percentage-point decrease in “for” votes, 61 Supra Table 1. This result is achieved by multiplying the estimated slope coefficient for the natural log of market capitalization in model five by the natural log of two. For an explanation of how to interpret slope coefficients for logarithmically transformed variables, see FAQ How Do I Interpret a Regression Model When Some Variables Are Log Transformed?, UCLA Inst. for Dig. Research & Educ., http://stats.idre.ucla.edu/other/mult-pkg/faq/general/faqhow-do-i-interpret-a-regression-model-when-some-variables-are-log-transformed [http://perma.cc/3F2Q-FC2L] (last visited Sept. 13, 2017). and multiplying a company’s market capitalization by ten—approximately the factor by which Apple is larger than PayPal 62 Compare PayPal Holdings, Inc. (PYPL), Yahoo! Fin., https://
finance.yahoo.com/quote/PYPL [http://perma.cc/YV3R-6WJP] (last visited Oct. 9, 2017) (reporting a market capitalization of $79 billion after closing on October 6, 2017), with Apple Inc. (AAPL), Yahoo! Fin., http://finance.yahoo.com/quote/AAPL [http://
perma.cc/L5FU-U4HT] (last visited Oct. 9, 2017) (reporting a market capitalization of $802 billion after closing on October 6, 2017). —is associated with a seven-percentage-point decrease in “for” votes. 63 Supra Table 1. Meanwhile, multiplying a company’s market capitalization by 100—approximately the factor by which ExxonMobil is larger than Bed Bath & Beyond 64 Compare Bed Bath & Beyond Inc. (BBBY), Yahoo! Fin., http://
finance.yahoo.com/quote/BBBY [http://perma.cc/D84V-C24S] (last visited Oct. 9, 2017) (reporting a market capitalization of $3.28 billion after closing on October 6, 2017), with Exxon Mobil Corporation (XOM), Yahoo! Fin., http://finance.yahoo.com/
quote/XOM [http://perma.cc/35U6-NUCE] (last visited Oct. 9, 2017) (reporting a market capitalization of $346 billion after closing on October 6, 2017). —is associated with a fourteen-percentage-point decline. 65 Supra Table 1. Such a drop could easily spell the difference between a proposal’s success and its failure. And even controlling for industry, the apparent size effect remains. According to the full model with industry fixed effects, doubling a company’s market capitalization is still associated with a two-percentage-point decrease in support, multiplying a company’s market capitalization by ten is associated with a five-point decline, and multiplying market capitalization by 100 is associated with an eleven-point decline. 66 Supra Table 1.

3. Objections and Limitations. — As with many observational empirical studies, it is difficult to draw causal inferences based on the evidence presented in the previous subsections. Although reverse causality is not a concern here, 67 Because size is measured as of the end of the preceding fiscal year, it is necessarily prior to vote outcomes. there is still a possibility that omitted variables may account for the observed associations. Additionally, there may be other concerns related to the chosen measurements and the possibility of a selection effect. This subsection considers some of these objections. On balance, while it is impossible to definitively rule out alternative causal theories, the evidence is still suggestive as to a size effect.

First, one might question whether the CPA–Zicklin Index accurately measures each company’s preexisting political accountability and transparency. After all, the CPA researchers who ranked each company are only human, and humans make mistakes. 68 Indeed, the CPA–Zicklin Index has been subject to some criticism. See Joe Trotter, New CPA–Zicklin Index Flawed, Inst. for Free Speech (Sept. 24, 2015), http://www.ifs.org/
2014/09/24/new-cpa-zicklin-index-flawed-2/ [http://perma.cc/CW9X-ZNZ3]. However, as at least one other person has noted, this criticism seems to stem from the CPA’s goals and overall function, not its methodologies. Jacquelyn E. Ryberg, Note, The Train Has Left the Station, Folks: The Inevitability of Widespread Adoption of Voluntary Political Spending and Lobbying Disclosure, 10 Va. L. & Bus. Rev. 1, 33 (2015). However, even if a company’s CPA–Zicklin score is not a perfect metric, there is no indication that any measurement errors are occurring in the sort of systematic way that would be necessary to bias the results of this Comment’s analysis. 69 A related critique of the CPA–Zicklin Index might question its utility because of the potential that year-to-year changes in its coverage—the index began by covering only the S&P 100, with coverage extending to the entire S&P 500 only in 2015—may bias the results. To investigate whether this bias was present, I run another set of regressions using a subset of the data for proposals with annual meetings occurring after the publication of the 2015 CPA–Zicklin Index. Table A.3 in the Appendix reports the results of these regressions. In all models without industry fixed effects, the slope coefficient for the log of market capitalization remains substantively and statistically significant. See infra Appendix at Table A.3. When industry fixed effects are introduced (in the sixth model), the coefficient remains negative but falls short of statistical significance at conventional levels. See infra Appendix at Table A.3.

Second, as with all research into the determinants of shareholder proposal vote outcomes, there is a possibility that nuanced differences in proposal content are driving the outcomes—a lurking-variable problem. For example, some proposals may be more ambitious than others in the level of disclosure that they appear to request, and this ambition may be correlated with company size. However, to the extent that one can expect material differences in proposal content to be associated with certain types of proponents (e.g., individual versus institutional investors), 70 Supposing that differences in proposal content relate to the proposals’ effectiveness, we might expect certain proponents—namely, institutional investors—to be better at designing effective proposal content than others. Cf. Reilly S. Steel, Proxy Access and Optimal Standardization in Corporate Governance: An Empirical Analysis, 23 Fordham J. Corp. & Fin. L. (forthcoming 2017) (manuscript at 35–37) (on file with the Columbia Law Review) (showing that the substantive terms of proxy-access shareholder proposals are associated with the type of shareholder-proponent, and that nearly all “oddball” proposals being brought by individual retail investors). the proponent-identity controls should also effectively control for proposal content.

Perhaps more significantly, there is a possibility of a selection effect. As other recent research has shown, shareholders have frequently withdrawn their proposals before any vote can take place as part of private “settlements” in which companies promise to disclose their political spending. 71 Sarah C. Haan, Shareholder Proposal Settlements and the Private Ordering of Public Elections, 126 Yale L.J. 262, 277–79 (2016). If these settlements occur more frequently at larger companies (and are attributable to managers’ fear that those particular proposals, if not “settled” before a vote, will receive significant shareholder support), then the large-company proposals that actually make it to a vote may be more likely to fail for reasons that have nothing to do with the companies’ size per se. Nevertheless, there does not currently appear to be any evidence that settlements are positively associated with company size. 72 A lack of data creates considerable difficulties for those who wish to study shareholder proposal settlements. See id. at 277–78. If anything, one would expect the settlements that lead to withdrawals to occur more frequently at smaller companies that lack the resources to fight contentious shareholder proposals. 73 In fact, one study suggests that the likelihood of withdrawal is negatively associated with market capitalization—consistent with this Comment’s findings regarding vote outcomes. See Baloria et al., supra note 9, at 55 tbl.5, panel A (reporting negative coefficients for market capitalization in three probit models with withdrawal as the dependent variable). Thus, the possibility of a selection effect seems remote enough here so as not to cast significant doubt on the previous analysis.

Finally, one might object that even if this Comment’s evidence is consistent with the theory that shareholders’ collective-action problems are impeding proposal success, there are other potential causal mechanisms that would also be consistent with a size effect. Indeed, it is difficult for a cross-sectional regression analysis to truly rule out alternative causal theories. However, this Comment’s empirical analysis casts doubt on perhaps the most notable alternative theory. Critics might contend that it is possible that larger companies have legitimately earned shareholders’ trust, which could explain the reduced shareholder support for political-spending proposals at such companies. However, to the extent that trust is a function of a company’s strong financial performance, one would expect the appearance of a size effect to diminish once controls for financial performance are included in the regression model. Here, though, the appearance of a size effect increases once financial-performance controls are introduced. 74 See supra Table 1 (showing that the slope coefficient for the log of market capitalization decreases from −3.141 to −3.368 once financial-performance controls are introduced). This suggests that the apparent size effect is not the result of shareholders’ greater trust in large companies.

D. Firm Size and Actual Political-Spending Accountability and Disclosure Levels

Consistent with the theory that collective-action problems are impeding shareholders’ efforts to obtain political-spending disclosure, the previous section demonstrated a negative relationship between shareholder support for political-spending proposals and firm size. But one might also be curious as to the relationship between size and firms’ actual practices regarding political-spending accountability and disclosure. This section assesses this relationship, showing that among S&P 500 firms, size is positively associated with actual levels of political-spending accountability and disclosure. 75 See infra Table 2. However, this relationship appears to be attributable to particularly large (or “big-cap”) companies’ practices: When the sample is restricted to only small- and medium-sized companies (“small cap” and “mid cap”), the relationship becomes much weaker and is no longer statistically significant. 76 See infra Table 2. Overall, this evidence would be consistent with a theory that other channels, such as the product markets, can supplement shareholder initiatives to effect political-spending disclosure, but these channels work for only the largest, most well-known companies.

To evaluate the relationship between company size and companies’ actual practices regarding political-spending accountability and disclosure, I run bivariate OLS regressions for those S&P 500 firms tracked by the 2016 CPA–Zicklin Index. In each case, the company’s 2016 overall score as a percentage of the total points available was the dependent variable, and the independent variable was either the company’s market capitalization or the natural logarithm of the company’s market capitalization (in each case as of the end of the fiscal year prior to the publication of the index on September 29, 2016). Figure 5 plots the results of the first of these regressions.

Figure 5: Accountability and Disclosure by Market Capitalization 77 The solid red line is an OLS regression line, and the blue dotted curve is a LOESS regression curve. The gray area represents a 95% confidence interval for the OLS line.

Figure 5 shows a positive association between market capitalization and firms’ CPA–Zicklin scores. The estimated slope coefficient is 0.19, which is statistically significant at conventional levels (p < 0.01). This implies that a $1 billion increase in market capitalization is associated with a 0.19-point increase in CPA–Zicklin score. The direction and statistical significance of this result hold with logarithmic transformations of market capitalization and CPA–Zicklin score. 78 A regression using the natural log of market capitalization as the independent variable produces an estimated slope coefficient of 16.41, which is again statistically significant (p < 0.01). Taking the natural log of the CPA–Zicklin score in addition to market capitalization produces a slope coefficient of 0.59, which remains statistically significant (p < 0.01). Overall, these results imply that larger companies tend to score higher on political-spending accountability and disclosure.

However, the relationship is clearly not linear, and casual visual investigation of Figure 5 suggests that the high scores of particularly large companies can explain a great deal of the observed relationship. To more rigorously investigate the nature of this relationship, I run another set of regressions with the sample in one specification restricted to small- and mid-cap companies (i.e., companies with a market capitalization of less than $10 billion). 79 These classifications are consistent with standard industry practice. See Mid Cap, Investopedia, http://www.investopedia.com/terms/m/midcapstock.asp [http://perma.cc/
Y5RY-3CEK] (last visited Sept. 13, 2017). Table 2 displays the results of these regressions, in each case using a logarithmic transformation of market capitalization. As Table 2 shows, while market capitalization appears to explain a healthy amount of the variation in CPA–Zicklin score in the full sample (with an adjusted R² of 0.240), it explains virtually none of the variation in CPA–Zicklin score when the sample is restricted to small- and mid-cap companies (with an adjusted R² of −0.003), and the slope coefficient loses its statistical significance in the restricted sample.

Table 2: Regression Analysis of Accountability and Disclosure 80 The figures reported without parentheses in the main part of the table are slope coefficients, while the figures reported in parentheses are t-statistics calculated using robust standard errors.

	Dependent Variable:
	CPA–Zicklin Score (2016)
	Full Sample	Small- and Mid-Cap Only
	(1)	(2)
Log(Market Capitalization)	16.413***	5.794
	(14.634)	(0.848)
Constant	−6.983*	16.669
	(−1.864)	(1.289)
Observations	480	123
R²	0.242	0.006
Adjusted R²	0.240	−0.003
F Statistic	214.156***	0.720
Note:	p < 0.1; p < 0.05; **p < 0.01

Overall, these results suggest that while company size has a positive relationship with political-spending accountability and disclosure, the influence of particularly large companies—big-cap companies with a market capitalization of at least $10 billion—explains the lion’s share of this relationship. Thus, while size appears to be related to actual political accountability and disclosure, the influence of particularly large companies can likely explain this relationship.

Taking a broader view, it is difficult to come to any definite conclusion as to the causal mechanism that underlies the relationship between company size and actual accountability and disclosure levels. One possibility, however, relates to pressure from the product markets. Given that the public generally supports political-spending disclosure, 81 See Scott Hirst, Social Responsibility Resolutions, J. Corp. L. (forthcoming 2017) (manuscript app. tbls.A.2 & A.3), http://ssrn.com/abstract=2773367 (on file with the Columbia Law Review). a lack of political-spending accountability and disclosure may quite plausibly hurt a company with strong brand recognition and public visibility. If larger companies tend to have stronger brand recognition and public visibility, these companies’ boards and management may voluntarily disclose their political spending to avoid harm to the companies’ reputations among customers. This theory is consistent with the data if one supposes that there is a nonlinear relationship between size and scrutiny from the product markets, whereby small- and mid-cap companies tend to fly under the radar but big-cap companies quickly attract greater public attention. Overall, this causal channel would still be consistent with the theory that shareholders’ collective-action problems are impeding their efforts to obtain disclosure at the governance level—even if at particularly large companies, pressure from the product markets supplements shareholders’ efforts.

III. Implications for Lawmakers and Policymakers

Part II undertook an empirical examination of the determinants of vote outcomes for shareholder proposals related to corporate political spending and found a significant “size effect.” This Part discusses the implications of that empirical analysis for lawmakers and policymakers, arguing that it supports the case for SEC rulemaking that would require corporations to disclose their political spending.

According to one common justification for mandatory disclosure of corporate political spending, mandatory disclosure is warranted to the extent that shareholders would benefit from disclosure but cannot obtain it due to market failure. 82 See, e.g., Bebchuk & Jackson, Shining Light, supra note 3, at 937–49 (advocating mandatory rulemaking based on shareholders’ informational interest). See generally John C. Coffee, Market Failure and the Economic Case for a Mandatory Disclosure System, 70 Va. L. Rev. 717 (1984) (presenting a market-failure-based justification for mandatory disclosure in the broader securities law context); Ciara Torres-Spelliscy et al., Running the D.C. Circuit Gauntlet on Cost-Benefit Analysis After Citizens United: Empirical Evidence from Sarbanes-Oxley and the JOBS Act, 9 Duke J. Const. L. & Pub. Pol’y 135 (2014) (presenting empirical evidence that investors value transparency). One might also justify mandatory political-spending disclosure on the basis of voters’ informational interests, see Sarah C. Haan, Voter Primacy, 83 Fordham L. Rev. 2655, 2680–81 (2015), but this Comment leaves consideration of those interests to other work. Indeed, SEC staff guidance has identified market failure resulting from collective-action problems as a justification for SEC regulation in other contexts. 83 See Memorandum from the SEC, Div. of Risk, Strategy & Fin. and SEC, Office of the Gen. Counsel, on Current Guidance on Economic Analysis in SEC Rulemakings, to the Staff of the Rulewriting Divs. and Offices Regarding Current Guidance on Economic Analysis in SEC Rulemakings 1 (Mar. 16, 2012), https://www.sec.gov/divisions/riskfin/rsfi_
guidance_econ_analy_secrulemaking.pdf [http://perma.cc/HW3Z-E2AL]. Applied to this context, the justification would be that private ordering is inadequate because shareholders’ collective-action problems significantly impede their ability to cause corporations to adopt effective disclosure regimes. 84 See Bebchuk & Jackson, Shining Light, supra note 3, at 947–49. This rationale would supplement other justifications for mandatory disclosure, such as the public-goods rationale, 85 See Coffee, supra note 82, at 723–37. the network-benefits rationale, 86 See Michael Klausner, Corporations, Corporate Law, and Networks of Contracts, 81 Va. L. Rev. 757, 836 n.240 (1995). This view is related—and perhaps essentially equivalent—to the public-goods rationale. Compare Coffee, supra note 82, at 723–37 (describing the public-goods rationale), with Klausner, supra (characterizing the facilitation-of-comparisons rationale as “essentially a network externality argument”). and an agency-costs rationale in which asset managers do not vote corporate proxies consistently with the preferences of the funds’ ultimate beneficiaries. 87 See Hirst, supra note 81 (manuscript at 18–22) (presenting evidence that mutual funds do not vote in accordance with their beneficiaries’ preferences and focusing on political-spending proposals in particular); see also Ronald J. Gilson & Jeffrey N. Gordon, The Agency Costs of Agency Capitalism: Activist Investors and the Revaluation of Governance Rights, 113 Colum. L. Rev. 863, 889–90 (2013) (characterizing the “gap between the beneficiaries’ and the fund’s interests” as “a particular kind of agency cost that is of special concern because it interacts with . . . managerial slack at the portfolio companies”). In the political-spending context, one would expect this problem to be particularly acute to the extent that large asset managers tend to have different wealth and income profiles—and thus are taxed differently—from their clients. Further research is necessary to investigate this possibility.

This Comment’s findings support the market-failure case for mandatory political-spending disclosure. The negative association between proposal success and company size suggests that collective-action problems are impeding shareholders’ efforts to bring about effective political-spending disclosure. As noted in Part II, this association persists even when controlling for various firm-specific characteristics. 88 See supra Table 1. Moreover, the fact that achieving high levels of shareholder support becomes more difficult as company size increases should be particularly concerning to the extent that such companies weigh most heavily in investors’ portfolios and, as a function of their size, have the greatest capacity to engage in political spending.

The apparently positive relationship between company size and the actual level of political-spending accountability and disclosure is still consistent with this theory. As discussed in section II.D, the influence of larger companies, which may be subject to greater pressure from the product markets, appears to explain most of the positive relationship between size and actual accountability and disclosure. To the extent that the normatively salient question is whether collective-action problems are impeding investors’ efforts to obtain the level of disclosure that they truly desire, 89 The 2011 petition calls for rulemaking under section 14 of the Securities Exchange Act of 1934, see Petition, supra note 29, at 1, which permits the SEC to promulgate such rules as are “necessary or appropriate in the public interest or for the protection of investors,” 15 U.S.C. § 78n(a)(1) (2012); see also J.I. Case Co. v. Borak, 377 U.S. 426, 431–32 (1964) (noting “the protection of investors” as among section 14(a)’s “chief purposes” (internal quotation marks omitted)). Thus, the extent to which mandatory disclosure would protect investors is—at least as a legal matter—a relevant question. Scholars may debate the deeper normative question—relating to whom securities law should serve—for a long time, but this Comment assumes that investor protection is a relevant consideration. the evidence from shareholder proposal vote outcomes is much more relevant.

Granted, there may still be reasons to question the utility of political-spending disclosure, at least if one believes that shareholders’ financial interests are the only interests that such disclosure should serve. 90 See James R. Copland, Against an SEC-Mandated Rule on Political Spending Disclosure: A Reply to Bebchuk and Jackson, 3 Harv. Bus. L. Rev. 381, 385–89 (2013) (arguing that political spending is immaterial to shareholders’ financial interests); Thomas Stratmann & J.W. Verret, How Does Corporate Political Activity Allowed by Citizens United v. Federal Election Commission Affect Shareholder Wealth?, 58 J.L. & Econ. 545, 551–57 (2015) (finding “public companies experienced positive abnormal returns” following Citizens United); J.W. Verret, The Securities Exchange Act Is a Material Girl, Living in a Material World: A Response to Bebchuk and Jackson’s “Shining Light on Corporate Political Spending,” 3 Harv. Bus. L. Rev. 453, 461–68 (2013). If shareholders’ financial interests are the only interests that matter and corporations’ political expenditures are immaterial to these interests, then mandating political-spending disclosure serves no purpose—regardless of any market failure. 91 See Copland, supra note 90, at 385–89.

However, even assuming that corporate political spending is immaterial to shareholders’ financial interests—a highly contestable claim, as others have noted 92 See Bebchuk & Jackson, Shining Light, supra note 3, at 941–42, 943 n.63, 956–57. First, managers may use corporate funds to lobby for policies that benefit them as individuals, such as tax policies that are favorable to corporate executives, but do not benefit shareholders as a whole. Second, managers may use corporate funds in ways that benefit individual firms but produce externalities that negatively impact diversified shareholders’ investments in other firms. See Leo E. Strine, Jr. & Nicholas Walter, Conservative Collision Course?: The Tension Between Conservative Corporate Law Theory and Citizens United, 100 Cornell L. Rev. 335, 379–87 (2015). For example, while coal-industry lobbying might increase the value of one’s investments in the coal industry, it might reduce the value of one’s investments in natural gas, solar energy, and other noncoal industries. —corporate and securities law should not disregard investors’ nonfinancial interests. Even if shareholders invest primarily with a view to earning economic returns, both state and federal law have long allowed shareholders to pursue nonpecuniary objectives—including political goals. For example, in one Delaware case, Food & Allied Service Trades Department v. Wal-Mart Stores, Inc., Chancellor William Allen held that political ends could constitute a “proper purpose” for the exercise of shareholder inspection rights. 93 See No. 12551, 1992 WL 111285, at *1, *4 (Del. Ch. May 20, 1992). In one notable exception, the Minnesota Supreme Court, applying Delaware law, held that political purposes constituted an improper exercise of shareholder inspection rights. See State ex rel. Pillsbury v. Honeywell, Inc., 191 N.W.2d 406, 411–13 (Minn. 1971). However, the Delaware Court of Chancery subsequently recognized that case as bad law. See Wal-Mart, 1992 WL 111285, at *1, *4. For more background on state law’s treatment of shareholders’ nonfinancial interests, see generally Reilly S. Steel, Note, The Underground Rulification of the Ordinary Business Operations Exclusion, 116 Colum. L. Rev. 1547, 1583 n.256 (2016) [hereinafter Steel, Underground Rulification]. Meanwhile, both the SEC and federal courts have affirmed shareholders’ rights to present proposals that focus on “significant social policy issues,” 94 See, e.g., Roosevelt v. E.I. Du Pont de Nemours & Co., 958 F.2d 416, 427–29 (D.C. Cir. 1992) (R.B. Ginsburg, J.) (observing that the exception to the shareholder proposal rule for proposals relating to the company’s “ordinary business operations” does not apply to shareholder proposals that implicate “significant policy issues”); Amendments to Rules on Shareholder Proposals, Exchange Act Release No. 34–40018, 63 Fed. Reg. 29,106, 29,108 (May 28, 1998) (providing that proposals that focus on “significant social policy issues” are not excludable under the ordinary business operations exclusion); see also Med. Comm. for Human Rights v. SEC, 432 F.2d 659, 681 (D.C. Cir. 1970) (suggesting that section 14 of the Securities Exchange Act allows shareholders to “present to their co-owners . . . the question of whether they wish to have their assets used in a manner which they believe to be more socially responsible but possibly less profitable”), vacated, 303 U.S. 403 (1972). See generally Steel, Underground Rulification, supra note 93, at 1559–62 (explaining the history of the so-called “social policy exception” to the ordinary business operations exclusion). and the SEC staff has applied this approach to political-spending proposals. 95 See, e.g., Home Depot, Inc., SEC No-Action Letter, 2011 WL 291324, at *1 (Mar. 25, 2011) (declining to permit exclusion of a proposal that focused on the company’s “general political activities”). More recently, in Burwell v. Hobby Lobby Stores, the Supreme Court observed that corporate law permits for-profit companies to pursue nonpecuniary objectives “with ownership approval.” 96 134 S. Ct. 2751, 2770–71 (2014). Indeed, the Court recognized shareholders’ interests in corporate political spending in Citizens United itself. See Bebchuk & Jackson, Shining Light, supra note 3, at 944–45.

Moreover, this Comment’s findings suggest—as a positive matter—that to the extent shareholders care about corporate political spending, it is primarily for reasons unrelated to political expenditures’ capacity to cut into corporate profits. First, while there was a statistically and substantively significant negative association between proposal support and preexisting political accountability, there was no such association between vote outcomes and financial performance. 97 See supra Table 1. Yet if profitability were shareholders’ primary concern, one would expect proposal support to rise as profits fall. Second, a test for interaction effects between operating margin and political accountability—the theory being that as operating margins become thinner, the negative effect of a lack of political accountability should amplify—did not uncover a statistically or economically significant interaction term. 98 See supra note 55. An interaction term measures whether the effect of one of two variables depends on the other variable. Karen Grace-Martin, Interpreting Interactions in Regression, The Analysis Factor, http://www.theanalysisfactor.com/
interpreting-interactions-in-regression [http://perma.cc/A25A-LZDA] (last visited Sept. 13, 2017). Thus, financial performance does not appear to matter to shareholders when they decide whether to support political-spending proposals. Indeed, this should not be particularly surprising. Political scientists have long believed that voters consider their actions’ capacity to produce benefits for society generally, 99 See, e.g., James H. Fowler, Altruism and Turnout, 68 J. Pol. 674, 674–75 (2006) (discussing the role of altruism in causing voter turnout). and it seems plausible that shareholders may behave similarly. Furthermore, shareholders may have legitimate interests (financial and otherwise) at stake beyond their investments, and corporate political spending may implicate those interests. 100 See Bebchuk & Jackson, Shining Light, supra note 3, at 942–44 (arguing that donations may frequently fail to reflect shareholders’ political preferences and that shareholders may have expressive interests in even small amounts of political spending); supra note 92 (explaining that political spending may produce externalities that harm investors with diversified portfolios).

Conclusion

As lawmakers and policymakers continue to debate the necessity of mandatory disclosure of corporate political spending, it is important to develop a complete understanding of how the current ad hoc regime of private ordering works—and fails. Undertaking the first empirical investigation into the determinants of vote outcomes for shareholder proposals related to corporate political spending, this Comment has uncovered a significant “size effect” in which proposal support is negatively associated with company size. Consistent with the theory that shareholders’ collective-action problems are impeding their ability to bring about effective political-spending disclosure, these results provide suggestive empirical support for one of the strongest arguments for mandatory disclosure rules.

Appendix

Table A.1: Summary Statistics

Statistic	N	Mean	St. Dev.	Min	Max
Votes Cast in Favor of Proposal	397	24.79	13.86	0.60	66.00
Proponent Identity 101 I manually recoded FactSet’s “Other Stake Holders” values based on whether the main proponent was an individual investor or institutional investor.	359	0.88	0.32	0	1
CPA–Zicklin Score	295	48.46	26.06	0.00	98.57
Market Capitalization (Millions)	391	71,167	86,056	371	521,615
Book Value of Equity (Millions)	393	32,459	49,768	−12,629	219,333
Total Debt (Millions)	393	44,204	120,286	0	634,321
Total Assets (Millions)	393	176,447	446,306	382	2,593,211
PTB 102 To reduce the influence of outliers and correct measurement error, I excluded observations for which PTB was above 2000% or less than or equal to 0%. PTB above 2000% is likely due to an accounting anomaly. I excluded negative and zero PTB because each instance was attributable to a negative or zero book value. A negative (or zero) book value, combined with a high market capitalization, would imply that investors value the company much more than its book value suggests, but a negative (or zero) PTB would not reflect this.	373	345.60	338.02	27.67	1995.00
Sales (Millions)	393	50,097	68,345	627	476,294
Operating Income (Millions)	393	6,796	9,011	−5,067	53,973
Operating Margin	393	17.11	12.11	−29.22	75.00
ROA 103 To reduce the influence of outliers, I excluded observations with ROA above 50%. ROA above 50% would imply a company’s earnings before interest and taxes exceeded half the company’s assets, which is implausible and likely reflects an accounting anomaly. Observations from the bottom end of ROA were not removed because there were no extreme observations at the low end of ROA.	390	8.82	6.25	−9.07	39.16
EPS Growth	397	0.673	0.47	0	1
Institutional Ownership 104 I changed all percentages above 100% to 100% to correct what is likely measurement error. By definition, a maximum of 100% of shares can be owned.	393	76.35	14.36	2.93	100.00
Insider Ownership	393	3.21	8.84	0.002	80.17

Table A.2: Descriptive Statistics by Proxy Year 105 I define a “proxy year” as beginning on November 1 of the preceding calendar year and ending on October 31 of the current calendar year. For example, proxy year 2012 ran from November 1, 2011 through October 31, 2012.

Proxy Year	Number of Proposals	Mean Votes in Favor (%)	Mean Market Capitalization (Millions) ($)	Mean CPA–Zicklin Score (%)
2012	74	20.88	65,957	43.53
2013	86	24.16	53,179	42.36
2014	94	25.10	63,999	51.57
2015	72	27.92	80,600	51.90
2016	71	26.01	98,272	50.97

Table A.3: OLS Regression Analysis of Vote Outcomes (Post–2015 CPA–Zicklin Index) 106 The figures reported without parentheses in the main part of the table are slope coefficients, while the figures reported in parentheses are t-statistics calculated using robust standard errors.

	Dependent Variable:
	Percentage of Votes Cast in Favor of Proposal
	(1)	(2)	(3)	(4)	(5)	(6)
Log(Market Capitalization)	−3.468***	−2.988**	−3.271**	−5.898***	−7.552***	−4.523
	(−2.860)	(−2.025)	(−2.264)	(−2.906)	(−3.503)	(−1.626)
CPA–Zicklin Score		−0.042	−0.069	−0.047	−0.012	−0.030
		(−0.776)	(−1.046)	(−0.639)	(−0.157)	(−0.415)
ROA			−0.019	0.042	−0.073	0.706
			(−0.056)	(0.127)	(−0.170)	(1.066)
Operating Margin			−0.020	−0.077	−0.017	−0.224
			(−0.178)	(−0.712)	(−0.146)	(−1.450)
PTB			−0.002	−0.001	0.00002	−0.007
			(−0.355)	(−0.138)	(0.002)	(−0.718)
EPS Growth			0.221	2.590	3.737	2.740
			(0.071)	(0.736)	(0.954)	(0.461)
Proponent Identity					−2.462	1.080
					(−0.627)	(0.233)
Constant	39.443***	39.805***	44.081***	77.164***	91.471***	61.991**
	(7.553)	(7.701)	(8.479)	(4.943)	(4.546)	(2.320)
Ownership Controls 107 Ownership controls include the percentage of shares owned by institutional investors and the percentage of shares owned by corporate insiders.	No	No	No	Yes	Yes	Yes
Industry FE	No	No	No	No	No	Yes
Observations	70	70	63	63	52	52
R²	0.149	0.154	0.235	0.284	0.318	0.529
Adjusted R²	0.136	0.129	0.153	0.178	0.172	0.272
F Statistic	8.181***	4.927**	2.991**	3.231***	2.637**	7.227***
Note:	p < 0.1; p < 0.05; **p < 0.01

CLR Forum

CORPORATE POLITICAL SPENDING AND THE SIZE EFFECT

Reilly S. Steel*

Introduction

I. Corporate Political Spending, Shareholder Proposals, and Collective-Action Problems

Figure 1: Number of Political-Spending Shareholder Proposals

Figure 2: Political-Spending Accountability and Disclosure Levels in the S&P 500 22 The x-axis variable is the full score as a percentage of total points from the 2016 CPA–Zicklin Index. For an explanation of this index, see infra notes 39–41 and accompanying text.