The Determinants of Corporate PAC Activity

Transcription

1 The Determinants of Corporate PAC Activity Brett W. Myers August 28, 2007 Abstract Corporate political action committee (PAC) data has been used to identify politically favored/connected firms in recent finance applications and has been found to predict such things as stock returns and capital structure. However, the interpretation of corporate PAC data is not clear and the endogeneity problems it raises muddy the results of studies that use it; PAC activity may have little to do with the political process. This study describes the characteristics of firms that form PACs. Firm characteristics (such as size, market-to-book ratio, profitability, stock price volatility, etc.) explain the majority of corporate PAC activity. There is limited evidence that managerial hubris/distractedness also increases a firms level of political activity. Importantly, measures of the intrinsic political value of a firm also predict higher levels of PAC activity. PAC data, therefore, does contain useful information as to the political connectedness of firms, but the information is noisy. Myers is a PhD student at the UCLA Anderson School of Management. comments at brett.myers@anderson.ucla.edu. All errors are my responsibility. I welcome any and all

2 1 Introduction There is a growing body of research into the effect of the political process on finance topics, particularly in the area of asset pricing. The popular idea that politically connected firms exert a pervasive influence of over the political process, and promote private enrichment at public expense, provides at least part of the motivation for these studies. A key challenge facing researchers is determining which firms are politically favored or connected. This is typically not an obvious or easy exercise. To this end, campaign finance data has been widely used with the idea that firms that donate to politicians through an affiliated corporate political action committee (PAC) are better connected than those that do not. While the reasoning behind this idea is straightforward, there are endogeneity issues associated with campaign finance data that are not easy to resolve. This study examines the financial determinants of corporate political activity in order to better understand the endogeneity issues associated with using this type of data in financial topics. I find that firm characteristics explains the majority of PAC activity, as do measures of the intrinsic political value of the firm. I present limited evidence that managerial distractedness/hubris explains PAC activity. Specifically, I find that firms that form PACs tend to be older, larger firms with thin margins. They tend to be stable: their stock price volatility is lower, as is their debt levels and interest coverage ratios. Firms that experience a recent drop in sales also are more likely to form PACs, as are firms with high capital expenditures. I find limited evidence that firms whose CEOs win prestigious awards tend to form PACs, though CEO awards do not predict levels of PAC funding. Finally, I find evidence that measures of a firms intrinsic political value also help explain corporate PAC activity; firms in smaller states (and therefore have greater representation in Congress) and firms whose Senators chair important committees that are relevant to the firm are more likely to form PACs. This study does not seek to resolve all of the endogeneity issues associated with PAC data. For example, firm characteristics (such as profitability or sales) are likely related to political importance. Nevertheless, this study does contribute to the understanding of the kinds of firms that are active politically and contributes to the literature on corporate PAC activity in at least two important ways. First, the primary focus is on the financial 1

3 determinants of corporate PAC activity, particularly on variables that are of interest in finance applications. Second, it is the largest study on corporate PACs conducted to date, and examines all firms in the COMPUSTAT universe over a 24 year period (of which 900 or so have been found to have PACs). Prior studies have tended to focus on one or two election cycles and for a limited universe of firms. 1 Popular and academic interest has motivated a large amount of recent work into the value of political connections to firms. Studies have found that political activity and firm connectedness can influence expected returns, operating performance, capital structure, and other traditional finance topics. The challenge facing researchers is determining whether or not a firm is politically connected or favored, particularly on a large scale. In the vast majority of instances, making this determination is not straightforward. For studies in the United States, researchers often rely on the corporate PAC contribution data in order to infer a level of corporate connectedness. Corporate PAC data includes donations from firms to the political campaigns of politicians. Either explicitly or implicitly, the assumption is made that the more firms spend on politics through an affiliated PAC, the better connected they are. 2 This fits well with the conventional wisdom that corporations can buy politicians and influence the political process through campaign donations. For example, Jayachandran (2006) studies the effect on asset prices due to an unanticipated change in control of the Sentate in 2001 to the Democrats and finds that the market capitalization of firms that donated primarily to Democrats increased while those donated primarily to Republicans decreased. In a series of studies surrounding the close presidential elections of 2000 and 2004, Cheng (2005), Goldman, Rocholl, and So (2006), Knight (2006), and Shon (2007) find that firms and industries that donated primarily to Republicans through an affiliated PAC appear to benefit from the outcomes of the election outcomes, while firms that donated primarily to Democrats did not. Roberts 1 For instance, Hart (2001) examines approximately 100 firms over a 20 year period (of which 30 operated PACs), while Mitchell, Hansen, and Jepsen (1997) examine 524 firms over a two year period (of which 282 had affiliated PACs). 2 The level of corporate political activity can be taken to be total PAC disbursements, total donations to candidates, number of candidates to which a firm donates, etc. All of these measures are highly correlated (with each other and firm size) and for most analytical purposes are interchangeable. This study defines the level of PAC activity as being the total disbursements of the PAC. 2

4 (1990) uses PAC donations to study the effects of asset prices following the unexpected death of an influential senator. Finally, Cooper, Gulen, and Ovtchinnikov (2007) and Myers (2006) find that campaign contributions predict stock returns, though the results of these studies are not consistent; Cooper, Gulen, and Ovtchinnikov (2007) finds that the number of politicians to whom a firm donates through an affiliated PAC predicts higher future returns, while Myers (2006) finds that the level of PAC activity, adjusted for firm size and industry, is associated with lower future returns. There is no doubt valuable information contained in corporate political spending as the results in the aforementioned studies imply. However, the conclusion that it is the political process that drives the results is muddied by the endogeneity issues inherent in campaign finance data. Put simply, the useful information contained in campaign finance data may have nothing to do with politics! Consider a hypothetical study on the ability of corporate political donations to predict financial distress. It may be the case that politically connected firms enjoy bailout protection from sympathetic politicians (see, for example, Faccio, Masulis, and McConnell (2006)) and thus we might expect politically favored firms are less likely to experience financial distress. Suppose campaign finance data was used to infer which firms are politically connected and which are not. Such a study may indeed find that firms that donate to politicians are less likely to experience financial distress, and one might be tempted to conclude that it is, in fact, the political connections that drive the result. However, it may be the case that firms with good management (and an excess of managerial capital) are the ones that become involved with and donate to politicians. Under such a scenario, it may not be the value of political connections that matter in predicting financial distress, but rather the effective management (which drives both financial distress probabilities and political involvement). Alternatively, political involvement may be partially driven by desperation; management foresees economic or financial trouble on the horizon and attempts to mitigate it through the political process and campaign donations. If unsuccessful, campaign contributions (but not the relationship between politicians and firms) may be seen as predicting distress. Campaign finance data is not the only means available to a researcher when measur- 3

5 ing the level of connectedness between a firm and a politician. For example, among the aforementioned studies, Goldman, Rocholl, and So (2006) focuses on the political affiliation of board members to make determinations as to political connectedness. Mitton and Johnson (2003) identify politically connected firms in Malaysia as having officers or controlling shareholders with close relationships with key government officials. Faccio (2006) and Faccio, Masulis, and McConnell (2006) identify politically connected firms as those whose large shareholders or top directors are members of parliament, government ministers, chiefs of state, or closely related to a top politician, while Myers (2007) examines firms whose U.S. Senators chair powerful Senate committees that are valuable to that firms industry. However, campaign finance data is by far the most detailed available to researchers, particularly in the United States. Unlike data involving the political affiliations of corporate board members and officers, or senate committee chair data, campaign finance data is available for thousands of firms and politicians spanning nearly three decades. It is highly detailed and can be parsed across different industries and other firm categories and allows detailed study into the dynamics of political involvement. For these reasons, corporate PAC data is the only dataset available to researchers who hope to conduct large-scale studies on the effects of the political process on finance topics. This study examines the determinants of corporate political activity through an affiliated PAC. A small but active body of research has been devoted to this topic. Economists have generally preferred a profit-maximization motivation to the selection of explanatory variables towards PAC formation and levels of activity while behaviorists prefer other explanations such as industry peer pressure, community norms, or internal politics. The current study is more closely aligned with the latter category and is primarily concerned with three questions. First, to what extent do firm characteristics explain corporate PAC activity? A large number of financial variables have been linked to various financial phenomena such as value and size effects on asset pricing and the effect of collateral on a firm s optimal capital structure. To what extent do these variables predict PAC activity and, by extension, to what extend does PAC activity reflect these underlying firm characteristics? Second, is political activity explained by managerial distractedness/hubris? 4

6 Politically active managers have been criticized as being distracted and set on empire building. 3 Alternatively, it may be the case that good management teams may be more likely to become politically active. Most of the endogeneity concerns regarding PAC data can fall loosely into one of these first two categories. Finally, is corporate PAC activity related to how intrinsically important firms are to politicians? That is, are politically valuable firms the ones that contribute? This question, though central to the purposes for which PAC data is typically employed, is not trivial. If politically important firms tend to contribute more, or more generally, if political spending contains information regarding the quality of the relationship between firms and politicians, it would help justify the use of PAC data to infer relationships between firms and politicians. In fact, the assumption that this is the case is implicit in most of the studies that use campaign finance data. I find that firm characteristics explain a large portion of corporate PAC activity with firm size, age, and stock volatility explaining the majority of PAC formation and level of disbursements. I also find evidence that measures of the intrinsic value of the firm explain PAC activity; my proxy for intrinsic political importance includes small state and large state dummy variables (smaller states have proportionally greater representation in Washington and their needs, presumably, garner greater attention) as well as firms whose U.S. Senators chair powerful committees that are relevant to that firms industry. 4 I find limited evidence that managerial distractedness predicts PAC formation. In this study, I use CEO awards, such as CEO of the Year, as a proxy for managerial distractedness. Malmendier and Tate (2007) finds that firms whose CEOs achieve superstar status through prestigious national awards from the business community subsequently spend more time on and effort on activities outside their company (such as writing memoirs and other books and sitting on outside boards), appear to increase their manipulation of company earnings, and underperform beyond mean reversion in subsequent years. 5 The remainder of the paper is organized as follows: section 2 discusses corporate 3 The Center for Political Accountability, in a report titled The Green Canary (2005), argues that corporate political activity...can function as an effective means of alerting shareholders and financial analysts to possible management, reputational, and other problems that may negatively affect shareholder value. Also,...shareholders and analysts can be made aware of potential company problems and questionable management behavior by scrutinizing a company s political contributions. 4 See Myers (2007) 5 My thanks to Geoffrey Tate and Ulrike Malmendier for the relevant data. 5

7 political activity and the role of corporate PACs, as well as outlines some of the problems with using PAC data and why researchers use it. Section 3 outlines the determinanges of corporate PAC activity and contains the results of empirical tests. Section 4 concludes. 2 Corporate Political Activity and Corporate PACs The incentives for corporate political involvement are enormous. Governmental policy has the potential to influence almost every variable in a firms profit function. 6 In no particular order, taxes, regulation (labor, environmental, safety, health, pension, etc.), education, social welfare programs, monetary policy, the government-as-customer, etc., can all impact the price a corporation can charge for its products, product demand, worker wages and hours, and the cost of capital. As a consequence, firms have incentive to attempt to affect political outcomes (legislative and electoral) and one of the most visible ways in which they do this is through an affiliated PAC. 2.1 A Brief History of Corporate PACs A corporate PAC is established by a firm for the express purpose of funnelling money into the political process. It is the only corporate affiliated entity that can legally make direct contributions to candidates (so-called hard money ), though it can make other donations as well. 7 Corporations are not allowed to fund their PACs from company revenues, exempting only administrative and overhead costs. Any PAC funding used for political purposes is restricted to voluntary donations from management, stockholders, and non-managerial employees (from whom the firm can solicit donations twice per year). PAC disbursements are controlled by management. Corporate financial involvement in the political process via PACs increased in 1971 with the enactment of the Federal Election Campaign Act (FECA) and its subsequent amendments later in the decade. Prior to this time the Taft-Hartley Act of 1947 placed most official corporate campaign finance activity on uncertain legal ground, which limited direct corporate financial involvement in politics, especially relative to unions. FECA 6 As a conceptual starting point in their study on the motivations for corporate PAC formation, Grier, Munger, and Roberts (1994) note that the profit function of firm i is π i = P i Q i r i K i w i L i, with P, Q, r, and w being partially politically determined. 7 Hard money contributions are donations directly to a campaign and are heavily regulated. 6

8 allowed for the formation of corporate PACs which could, among other things, make hardmoney donations to candidates. After FECA, the number of corporate PACs increased rapidly, reaching a high of 2,008 in the 1988 election cycle before dropping slightly in subsequent years. Despite FECA, the majority of corporations have not formed a PAC, though many contribute to industry or trade PACs (which are not considered in this study). The Federal Election Commission (FEC) requires that PACs report their financial activity in great detail and at regular intervals throughout an election cycle. This information is then made available to the public. PAC categories include those affiliated with a political campaign or political party and interest groups, such as labor unions, trade associations, and corporations with publicly traded stock (which are the category considered in this study). 2.2 Previous Work on Corporate PAC Activity There is a small but active literature on the dynamics of corporate PACs, and has centered around two questions. First, what are the determinants of PAC formation and level of funding? And second, what are PAC disbursement strategies and what is their impact on the political process, including the outcome of elections and their influence in shaping the legislative agenda. As for the first question, the majority of studies use Heckman (1976, 1979) corrected OLS to estimate the determinants of PAC formation and level of activity. PAC activity has been defined as the total number of disbursements, donations to candidates, receipts, number of candidates to whom the PAC donates, etc. All of these definitions of PAC activity are highly correlated. Empirically, the most important factors predicting corporate PAC formation and activity are firm size, the degree of government regulation affecting the firm, and the extent to which the government is a customer of the firm. Other factors have been studied as well, such as whether or not the firm has a Washington office (Humphries (1991)), the percent of unionization at the firm, industry classification, industry and geographic concentration, etc. (see Masters and Keim (1985), Grier, Munger, and Roberts (1994), and Hart (2001) for examples). 7

9 2.3 The Trouble with PAC Data Politicians need money in order to be elected and are often under heavy pressure from their party to raise it. With this in mind, the common wisdom is that politicians are more than willing to offer political influence in order to raise campaign funds and that, essentially, campaign donations function as nothing more than legalized bribes. At first glance, the assumption that firms that spend more on politics should consequently enjoy greater levels of political influence appears straightforward. However, there are serious problems with this view. Campaign contributions from PACs to candidates are restricted to $10,000 per candidate per two-year election cycle. This is small in the face of multi-million dollar campaign budgets. Further, this constraint rarely binds! Most contributions fall well below this limit; Kroszner and Stratmann (2000) note that the average contribution from a corporate PAC to a political campaign during the was $700. Similarly, in a more recent study, Ansolabehere, de Figueiredo, and Snyder (2003) find that only 4% of PAC contributions are at or near the $10,000 limit, and that the average total corporate PAC contribution to a candidate is only $1,400 per election cycle. Also, corporate donations make up a small part of most political campaign budgets. It can be seen in Table I that the category of politicians that receive the largest portion of their campaign budget are incumbents in the House. But even in that category, corporate donations total less than 18% of the overall campaign budget (and comes from a large number of firms). In fact, the average political donation appears to come from individual voters in small amounts. The typical U.S. Senate election in California in the 1990s cost around $24 million whereas the typical U.S. Senate election in small states cost around $2 million, with the difference almost entirely accounted for by individual contributions in small amounts. The same is true in close elections (see Ansolabehere, de Figueiredo, and Snyder (2003)). It appears that both the average dollar in politics, as well as the marginal dollar, come from individuals in small amounts. In short, it is difficult to see how an individual firm could influence the political process on the basis of campaign donations alone, given the small notional amount of their donations with respect to the size of the typical campaign budget, particularly when firms do not donate the marginal dollar. Evidence that corporate campaign donations from firms affect electoral and legislative 8

10 Table I Candidate Campaign Receipts ( ) This table contains the PAC donations received by politicians from corporate and labor PACs as a percentage of their total campaign budget. Corporate PAC donations are from firms which self-identify as having publicly traded stocks; labor PAC donations are included for comparison. Total PAC donations also include donations from trade organizations, private firms, cooperatives, party PACs, ideological PACs, etc. Body I/C/O Party Corporate PAC Labor PAC Total PAC Senate Challenger Democrat 1.3% 5.4% 10.5% Republican 5.0% 0.0% 9.1% Incumbent Democrat 8.3% 5.5% 23.6% Republican 14.5% 0.7% 27.2% Open-seat Democrat 2.5% 3.9% 10.5% Republican 7.9% 0.1% 14.3% House Challenger Democrat 1.0% 13.0% 20.3% Republican 4.1% 0.1% 12.2% Incumbent Democrat 13.3% 15.9% 47.1% Republican 17.9% 1.8% 38.5% Open-seat Democrat 2.4% 10.9% 21.1% Republican 7.0% 0.3% 18.7% outcomes is limited. Ansolabehere, de Figueiredo, and Snyder (2003) surveys nearly 40 articles in economics and political science that examines the effects of campaign contributions on roll call votes by members of Congress and finds that in three out of four instances, campaign contributions had no statistically significant effect on votes or had the wrong sign (which would suggest that more contributions would lead to less support). Further, many studies fail to resolve the endogeneity questions surrounding votes and campaign donations: do politicians respond to the preferences of their donors, or to interest groups donate to politicians that already agree with them and would vote their way whether or not a donation was made? These results, as well as the fact that firms donate little to politicians (relative to their total campaign budget) calls into question the effectiveness of campaign contributions from firms as legalized bribes and with it the 9

11 assumption that campaign contributions reliably proxy for the extent to which a firm is politically favored. Of course, a PAC is not the only vehicle at a firms disposal with which to affect the political process. Other activities are available as well, including lobbying and, for most of the sample period used in this study, virtually unlimited so called soft money donations from the corporate treasury directly to political parties. 8 Other firm activity can be interpreted as having political objectives as well, such as corporate philanthropy, corporate social responsibility or corporate citizenship initiatives and programs, etc., inasmuch as these efforts affect the image of the corporation among opinion-makers or affects the level of access to policy makers. 9 As noted by Ansolabehere, de Figueiredo, and Snyder (2003) and Milyo, Primo, and Groseclose (2000), corporate PAC activity is not the largest category of corporate political activity in dollar terms; lobbying expenditures are typically an order of magnitude higher, and charitable giving is higher still, by multiple orders of magnitude. 10 In summary, corporate PAC donations are small relative to politicians campaign budgets, and corporate PACs do not donate the marginal political dollar in close or expensive elections. There is little empirical evidence that campaign donations affect the behavior of political roll call votes. And finally, corporate PAC activity is a small portion of overall corporate political activity. These considerations call into question the appropriateness of using PAC data to make inferences about the relationships that exist between politicians and firms. 2.4 The Value of PAC Data: Why PAC Data is Here to Stay The issues raised in the previous section highlights some of the difficulties with PAC data. While the popular press is awash with reports on the extent to which corporations buy 8 Soft money contributions are donations directly to political parties ostensibly for non-partisan partybuilding activities, such as get-out-the-vote efforts (however, they have often been used in effectively partisan ways, such as funding so-called issue-ads which do not specifically name candidates, but nevertheless effectively promote one candidate over another). There were typically no limits on soft money donations for most of the sample period. With the passage of the Bipartisan Campaign Reform Act of 2002 (BCRA), soft money donations have since been severely restricted. 9 See Himmelstein (1997), Smith (1994), and Navarro (1988) for studies related to these issues. 10 For example, Philip Morris spent $120 million on philanthropic activities between 1997 and 1998, which is close to the $140 million spent by all corporate PACs for the same period. 10

12 politicians, academic studies simply do not support this. So why do many finance and economic studies of corporate involvement in politics focus on PACs when attempting to gauge the level of corporate political connectedness? If corporate PAC activity is so comparatively small, and there are other means by which a corporation involves itself in the political process, why are they excluded, particularly when the financial commitment to PACs is comparatively smaller? There are several reasons for the focus on PACs. As a practical matter, PAC data is simply the most readily available; data involving corporate lobbying and soft money donations from corporate treasuries are more widely disbursed, if they are available at all. These types of difficulties with the available data are even more acute in the case of corporate charitable giving and similar activity. By way of contrast, corporate PAC activity is reported in great detail and made public by the Federal Election Commission (FEC). As a consequence, the disproportionate attention given to PACs is essentially a case of looking under the lampost. 11 More importantly, though it is unlikely that firms are able to buy political support through campaign donations, PAC donation may nevertheless contain useful information as to the political connectedness of the firm. While corporate political spending may not reliably buy influence in the form of desired political outcomes (electoral or legislative), it may buy access to politicians, a subtle but important distinction. More to the point, it may signal a level of access a firm enjoys with a politician. A preeminent concern among politicians is reelection, which involves satisfying their constituents. While politicians have prior beliefs regarding the preferences of their constituents, they do not know these preferences perfectly, nor are they ideologically dogmatic when it comes to satisfying them. 12 Politicians therefore entertain lobbyists and interest groups in an attempt to discern the preferences of their constituents. However, politicians time is limited. 13 Therefore, those interest groups that best represent the constituent preferences tend to be the ones that are granted access to politicians (who do not want to waste their time 11 Milyo, Primo, and Groseclose (2000), page Thus we see liberal Democratic U.S. Senators from Michigan adamantly opposed to raising fuel economy standards and conservative Republican U.S. Senators from South Carolina vigorously supporting protectionist legislation in favor of the domestic textile industry. 13 By one estimate, a legislator s time is worth around $10,000 per hour (see Langbein (1986)). 11

13 by entertaining lobbyists whose interests are opposed to their own). 14 PAC donations are highly correlated with lobbying and access, and thus while PAC money may not buy political support per se, it may nevertheless proxy for the intrinsic political value of the sponsoring organization, as evidenced by the amount of political access that attends it. For example, the auto manufacturers tend to have large PACs. However, their influence does not derive from their donations, but rather the fact that the auto manufacturers employ millions of voters. As Humphries (1991) observes: it is not PACs per se that corporations value, but the ability of PAC money to facilitate the job of Washington representatives and that their results suggest that there is a strong link between what might be termed the amount of lobbying and the amount of money contributed to corporate PACs. 15 Attempts have been made to disaggregate and explain different forms of corporate political activity (notably, Hansen and Mitchell (2000)), but there is a consensus among many researchers that differing measures of political involvement (such as lobbying, charitable giving, etc.) are highly correlated with each other and that using PAC activity as a proxy for overall levels of political involvement and influence is therefore justified The Determinants of Corporate PAC Activity Many studies that use corporate political data do so in order to make a determination as to whether or not politicians favor a given firm. For example, in studies of stock returns surrounding elections, campaign finance data has been employed to determine whether firms prefer a Democratic or Republican outcome. 17 The idea that politically favored firms are the ones that donate to politicians (whether or not the relationship is causal) is typically why PAC data is used in most finance studies. However, this need not be the 14 For a theoretical treatment of political activity along these lines, see Austen-Smith (1995). 15 In related studies, Grenzky (1989) finds that in cases where a relationship between contributions and votes are established, the donations serve as a proxy for a more important and larger package of support. Similarly, Hojnacki and Kimball (2001) conclude that, while the presence of a PAC does not necessarily increase the level of access to the sponsoring organization per se, groups that form PACs tend to have a base of support that does result in increased access (which could include lobbying and other indirect political expenditures in the case of corporate sponsors). 16 According to Berelson, Lazersfeld, and McPhee (quoted in Verba and Nie (1972) page 44): Almost all measures of political involvement and participation are highly correlated with one another and for analytical purposes, interchangeable. 17 See Goldman, Rocholl, and So (2006), Knight (2006), and Cheng (2005) for examples. 12

14 case; firms may give to politicians for reasons unrelated to the political process! For the purposes of this study, I divide the determinants of corporate political activity into three general categories: firm characteristics, managerial distractedness, and measures of a firms intrinsic political value. Firm characteristics that have already been shown to explain various financial phenomena may also determine levels of political activity in a way that is completely unrelated to the political process and the relationship between the firm and the politician. Thus it may be the underlying firm characteristics that drive some of the phenomena explained by PAC activity, not the PAC activity itself or the level of political connectedness of the firm at all. For example, large, mature firms may donate to politicians simply because they have the economies of scale to do so. Certain industries may donate to politicians because government happens to be a significant customer (i.e., the aerospace industry) and political donations proxy for communications costs. Stable, profitable firms may donate more because they have the excess capital necessary to lobby. Or, conversely, volatile, unprofitable firms may donate more in an attempt to change unfavorable business conditions. Value firms may donate more as they tend to be older and better established; on the other hand, growth firms may be more likely to be politically active in order to help break through regulatory barriers that are unfriendly towards new technologies or because they need help moving into the global economy and need help navigating or changing the international regulatory environment. Additionally, when a firm becomes successful (i.e., is profitable of has a high return on equity), it may feel a need to give back or otherwise become involved in social issues, and becomes involved in politics (and other social and charitable activities) as a consequence. 18 All of these financial characteristics have been linked to various financial phenomena and drive political spending or involvement in the political process; however, it is not necessarily the case that these types of firm characteristics have anything to do with the level of political connectedness or importance of a particular firm. Managerial agency problems has been linked to a large number of financial phenomena, such as the optimal capital structure of the firm, corporate investment decisions, 18 For example, Hansen and Mitchell (2000) find that firms that operate an affiliated PAC are almost three times as likely to make charitable donations. 13

15 asset prices, etc. Agency problems arise when managers have an agenda that is different from maximizing the value of the firm for shareholders. Problems include empire building, managerial distractedness, hubris, etc. The quality of management may determine whether or not a firm becomes politically active. For example, it may be the case the firms with good management have the excess managerial capital to spend on the political process; well run firms may be the ones that are politically active. On the other hand, bad management may be overly focused on empire building and other outside pursuits and not as focused on running their firms as they should. It may be this category of management that becomes involved in politics. Finally, firms that are active politically may be intrinsically more important to politicians than firms that do not. The political science literature has largely discredited the idea that U.S. firms buy political influence that they otherwise would not have had. 19 However, political spending may still contain useful information regarding the intrinsic political importance of a firm. For example, models of strategic information transmission where campaign contributions act as statements of policy preferences and of informational quality predict that campaign contributions and the intrinsic political importance of the donors are positively correlated. 20 These types of models suggest that the greater the intrinsic political importance of a firm (i.e., a firm that employs a large amount of voters), the greater the political donations. This is not because such firms can buy influence; they already have it. Rather, there exists a separating equilibria wherein important firms donate, unimportant firms less so. In most finance applications, the idea that politically important firms are the ones that donate is the primary motivation for using PAC data in the first place. Of course, firm characteristics, managerial behavior, and political importance can all simultaneously impact corporate political activity; these drivers are not mutually exclusive. Further, certain firm characteristics may also contain information as to how politically important a firm is. Firm size, for example, has been shown to have important asset pricing implications. It also has important political implications; large firms typ- 19 The majority of studies that attempt to link campaign contributions to legislators with roll call votes have found little or no impact. 20 See Austen-Smith (1995). 14

16 ically have a large number of employees (or, in the eyes of politicians, voters) and may enjoy favorable political treatment as a consequence. Nevertheless, it is useful to examine these categories separately in order to develop a clearer picture of what drives corporate political activity and, consequently, what PAC spending implies in a finance application. 3.1 Methodology I employ the standard analytic technique used in the literature on corporate PACs, the Heckman selection model. This model ostensibly solves some of the problems that arise in alternate techniques, OLS in particular, but also a truncated regression and Tobit approach. The Heckman selection model considers the PAC formation and level of funding questions separately, while correcting for bias in the model of PAC size that arises from the censoring of that dependent variable. It firsts estimates a Probit model for PAC formation and then estimates a modified OLS model for levels of PAC size, correcting for any bias caused by censoring. This is accomplished by including the Inverse Mills ratio (IMR), or Hazard rate, as estimated from the Probit model. A Tobit approach requires that the determinants of both PAC formation and PAC size have exactly the same effects and Grier, Munger, and Roberts (1994) strongly rejects the Tobit assumption when applied to PAC data. 21 However, I find in the current application that both a Tobit approach and truncated regression yield coefficient estimates and levels of significance that are very similar to those of the Heckman selection model in determining levels of PAC activity. While the decision to form a PAC and the level of PAC activity are conceptually related, there are good reasons to consider them separately. Forming a PAC is entirely a management decision; it requires some paperwork, but not much else. Any management team that wishes to form a PAC and register it with the FEC faces minimal hurdles in doing so. However, funding a PAC is a different matter. Firms cannot fund their PACs from corporate revenue (and hence cannot be solely determined by management), the funds must come from voluntary donations from employees, whose contribution levels to the PAC are limited by law. Successfully funding a PAC, then, requires senior manage- 21 See The Trouble with Tobit in Grier, Munger, and Roberts (1994). 15

17 ment to successfully make the case to employees that it is in their best interest to donate. Thus the level of PAC activity is not entirely under the control of management, nor is it a one time decision; keeping a PAC well funded requires constant effort by management to successfully convince employees to contribute. The determinants of PAC formation are not entirely the same as the determinants of PAC activity. Many previous PAC studies that use the Heckman selection model include the same variables in the probit step as they do in the corrected OLS step. This creates identification problems. It is better to include variables that affect PAC formation (but not PAC size) in the probit step, and omit them in the corrected OLS step. I argue that firm age affects PAC formation, but not firm size, and therefore exclude it from the OLS step. The older a firm is, the more likely management has made the decision to, at some point, form a PAC. Once a PAC is formed it tends to persist, even as funding fluctuates. PAC funding, on the other hand, requires management to continually a case to current employees to donate, and there is little reason that a 10 year old firm should be better or worse at this than a 20 or 50 year old firm (or that firm age should be related to capital expenditures/assets, or similar). Other studies have used sales in the probit step and log sales in the OLS step, though this solution is imperfect (see Hart (2001)). In the current application, such an approach (and others like it) yield very similar results. The data used in this study (both PAC information and firm characteristics) are highly autocorrelated and consequently all standard errors are estimated by clustering at the firm level. Federal elections occur on a 2-year cycle. 22 While both campaign finance data and the relevant accounting data are available on an annual basis, campaign donations are irregular within an election cycle; they tend to be lower during the first year, and higher in the second. It is therefore most appropriate to examine an election cycle in its entirety and consequently the time periods considered in this study each span two years. I examine the 12 political cycles from Finally, year dummies and 3-digit SIC code dummies are included in order to mitigate any omitted variable bias. Similar to findings in previous studies, firm-fixed effects reduce 22 For example, the 1992 election cycle spans the calender years 1991 and

18 the significance of all but the strongest results for the independent variables, particularly for PAC formation. This is not surprising as the time-period is short (12 periods) and the dependent variables highly persistent and in many cases the firm fixed effects completely determine the dependent variable (particularly PAC formation). Consequently, they are not included. 3.2 Data PAC Data The FEC requires that PACs report their financial activity in great detail and at regular intervals throughout a given election cycle. This information is then made public through the FEC s website ( In this study, I restrict my attention to those PACs that have self-identified as being affiliated with a publicly-traded corporation. 23 data are available from 1979 to the present and is segmented by election cycles. The data are highly detailed and includes transaction information between PACs and political campaigns, as well as records of each donation (above a minimum threshold) to a given PAC. There exists in the data approximately 40,000 corporate PAC-years but no direct way to match corporate PAC records to a firm in CRSP. For this study, matching was performed by hand based on the name of the PAC, the name of the PACs sponsoring organization and the name of the firm in CRSP. A significant effort was also spent attempting to match PACs of subsidiary organizations to a parent firm in CRSP, matching PACs to the appropriate firms after mergers and acquisitions, etc. 24 In general, matching was made conservatively; if a PAC was not unambiguously associated with a firm in CRSP based on careful investigation, a match was not specified. Once the matching of PAC data to CRSP was complete, I was able to generate data files for firms that included PAC data (in particular, whether or not a firm had an active PAC) and perform the analysis. Approximately 60% of PACs that have self-identified as being affiliated with a pub- 23 Corporate PACs are classified by the FEC as special interest group (SIG) PACs. Other classifications for SIG PACs include labor unions, trade groups, cooperatives, and private companies. 24 Delisting codes in CRSP (such as for mergers and acquisitions) and various web resources, such as corporate affiliation websites as well as the official websites for individual firms, were also invaluable in this effort. The 17

19 Table II Corporate PAC Summary Statistics ( ) This table contains summary statistics for corporate PACs by election cycle from Standard Election Cycle Mean Median Deviation Min Max ,039 10,061 41, , ,264 22,973 63, , ,259 27,033 85, , ,374 29, , ,852, ,645 33, , ,974, ,684 32, , ,142, ,068 36, , ,604, ,403 37, , ,730, ,356 39, , ,846, ,119 47, , ,408, ,477 50, , ,317, ,197 57, , ,651,116 licly traded corporation were successfully matched to a firm in CRSP for each election cycle. Further, approximately 80% of PAC disbursements have been traced to a firm in CRSP. 25 There are a number of reasons more corporate PACs were not matched to CRSP: the matching scheme was simply not exact, many PACs have incorrectly self-identified as being sponsored by a corporation with capital stock (the sample includes numerous activist groups, law firms, etc.), and not all capital stock trades on the NYSE, NASDAQ, or AMEX (the markets from which the CRSP sample is drawn). Nevertheless, the majority of corporate PAC money is accounted for. Summary statistics for PAC data by election cycle can be seen in Table II Financial Data Financial data comes from COMPUSTAT and CRSP databases. Financial variables considered include sales (DATA12), profitability, return on equity (ROE), interest coverage, stock return volatility (the annualized standard deviation of monthly stock returns in the 24-month election cycle), market-to-book, capital expenditures divided by assets, re- 25 The matching success rates reported here are similar to those in Grier, Munger, and Roberts (1994) which matched 50-60% of corporate PACs and approximately 80% of donations to firms in COMPUSTAT for the time period. 18

20 Table III Summary Statistics ( ) This table contains summary statistics for independent and dependent variables from Standard Variable Mean Median Deviation Min Max ln(sales) Sales Profit/Assets Profit ROE ROE INTCOV Stock Volatility MKTBK CAPEX/Assets R&D/Sales Firm Age CEO Award CEO Award t Senate Committee Chair Dummy ln(electoral Votes) Committee Dummy ln(sales) Committee Dummy ln(votes) ln(pac Disbursements) search and development divided by sales, and the age of the firm (the number of years the firm has been listed in COMPUSTAT). These variables were calculated as follows: Profitability = ROE = INTCOV = Market-to-Book Ratio = CAPEX = R&D = Income Before Depreciation (DATA13) Book Value of Assets (DATA6) Net Income (DATA172) Stockholders Equity (DATA216) Interest Expense (DATA15) Income Before Depreciation (DATA13) Market Value of Assets (MVA) Book Value of Assets (DATA6) Capital Expenditures (DATA128) Total Assets (DATA6) Research and Development (DATA46) Sales (DATA12) 19

21 The market value of assets used above is calculated as M V A = [Shares (DATA199) Share Price (DATA54)] + Debt in Current Liabilities (DATA34) +Long Term Debt (DATA9) + Pref Stock Liquidating Value (DATA10) Deferred Taxes (DATA35) In addition, I also examine the effects of changes in in sales, profitability, and ROE, calculated in percentage terms from one election cycle to the next. Summary statistics for these financial variables is contained in Table III. Financial variables (except sales) have been Winsorized at the 1% tails. 3.3 Empirical Tests Results of empirical tests can be seen in Tables V and VI. A correlation matrix for all independent variables is contained in Table IV. Independent variables are divided to address three general questions that surround the explanation of corporate PAC activity: firm characteristics, managerial distcactedness/hubris, and variables that contain information on the intrinsic political value of the firm. First, do firm characteristics determine PAC activity? Second, does managerial distractedness/hubris explain PAC activity? And finally, does PAC activity have anything to do with the intrinsic political value of the firm? Most PAC data is used with the implicit assumption that PAC spending either buys political influence or is a signal that the firm is politically important for other reasons (e.g., has intrinsic political importance). Each of these categories are outlined in greater detail, and the empirical results discussed, below. Importantly, I do not attempt to resolve certain simultaneity problems in this study. For example, a firm characteristic such as size also predicts the intrinsic political importance of a firm; large firms, employing large numbers of voters, are more important to politicians on aggregate than are smaller firms. Rather, my intent is to describe those variables that explain PAC data so they can be accounted for, or resolved, in future work. 20

22 Table IV Correlation Matrix ( ) This table contains a correlation matrix for the independent variables considered. Profit/ Stock CAPEX/ ln(sales) Sales Assets Profit ROE ROE INTCOV Vol MKTBK Assets ln(sales) Sales Profit/Assets Profit ROE ROE INTCOV Stock Volatility MKTBK CAPEX/Assets R&D/Sales Firm Age CEO Award CEO Awardt Senate Committee Chair Dummy ln(electoral Votes) Committee Dummy ln(sales) Committee Dummy ln(votes) R&D/ Firm Comm Dummy Dummy Sales Age Award Awardt 1 Dummy ln(votes) ln(sales) ln(votes) R&D/Sales Firm Age CEO Award CEO Awardt Senate Committee Chair Dummy ln(electoral Votes) Committee Dummy ln(sales) Committee Dummy ln(votes)