A new measure for agreement; an IV approach

A new measure for Agreement; an IV approach

June 29, 2017

Abstract

Using a sample of public firms listed in the U.S. that issued seasoned equity or nonconvertible debt for the years between 1993 and 2016, I examine whether and how agreement between man-agers and investors affects firms’ security issuance behavior. The theory, introduced by Dittmar & Thakor (2007), is only tested with inferior proxies for agreement which also capture information asymmetries. Proxy contests are introduced as a new measure for agreement and it is shown that proxy contests do not have similar problems. The Russell 2000 index inclusion is introduced as a new instrument to induce an exogenous variation in proxy contests. Employing a 2SLS framework shows that agreement has no effect on security issuance.

Master Thesis

Name Arthur van Eeden

Student nr. 10423117

Supervisor R. Almeida da Matta Specialization Corporate Finance

University University of Amsterdam, Amsterdam Business School Faculty Faculty of Economics & Business

Statement of Originality

This document is written by Arthur van Eeden, who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

Contents

2.4 The Agreement theory . . . 10

2.5 The Experiment . . . 11 3 Methodology 12 3.1 Hypothesis . . . 12 3.2 The Model . . . 12 3.3 Description of variables . . . 14 3.3.1 Agreement measures . . . 14 3.3.2 Price variables . . . 15 3.3.3 Control variables . . . 15 4 Empirical evidence 16 4.1 Sample overview . . . 16 4.2 Descriptive statistics . . . 16 4.3 Results . . . 21

5 The Quasi Experiment 27 5.1 The Russell index . . . 27

5.2 Main results . . . 28

6 Conclusion 30

1

Introduction

Many theories try to explain the security issuance behavior of firms, however, research is not able to provide a clear answer. Myers & Majluf (1984) introduce the pecking-order theory which is based on the premise that managers are better informed than investors. At first, internal funds are exhausted before debt is used. Management will only consider equity if both are insufficient. Adverse selection costs are the reason for the reluctance towards equity issuance. Shyam-Sunder & Myers (1999) introduce a broadly used framework and find strong support for the pecking-order theory. Their inference is challenged by Fama & French (2005) who find that more than 50% of their sample do not issue as the pecking-order model expects. Leary & Roberts (2010) question the relevance of the theory and state that interpretation is key. In short, research on the pecking-order theory is mixed. The market-timing theory offers an alternative perspective on firms drive to issue equity. Baker & Wurgler (2002) argue that managers ”time-the-market” to exploit temporary fluctuations in the share price. Managers issue equity if prices are relatively high due to misvaluations by irrational investors. This results in a capital structure that is the result of prior attempts to time the market (Baker & Wurgler, 2002). DeAngelo et al. (2010) post the market-timing theory as the main perspective on firms’ issuance behavior, however, the authors question the completeness of the theory. The market-timing theory is refuted by Butler et al. (2011), who find that it is reasonable for managers to time the market by raising capital when prices are high, however, managers are apparently unable to successfully switch between issuing shares and increasing leverage.

Obviously, the pecking-order and market-timing theory have shortcomings since neither are able to provide a sole explanation for all issuances that occur (DeAngelo & Roll, 2015). This thesis elaborates on the agreement theory, developed by Dittmar & Thakor (2007), which introduces a new perspective on why and when it could be beneficial to issue equity. It is a trade-off between the autonomy that equity provides and the tax advantages of leverage. An equity issue is beneficial if agreement between management and investors, from now on referred to as ”agreement”, is high, since both are more likely to agree on project value (Dittmar & Thakor, 2007). Boot & Thakor (2011) agree and argue that managers seek finance with ”softest” control since autonomy is valued highly. Note that the agreement theory does not rule out market-timing behavior since it supposedly has incremental power aside from the market-timing model. This results in the following hypothesis: Firms prefer to issue equity if agreement between managers and investors is high.

This thesis tests the hypothesis by examining a sample of public firms listed in the US that issued seasoned equity or nonconvertible debt for the years between 1993–2016. At first, similar to Dittmar & Thakor (2007), a logit regression is employed to test the relationship between agreement and secu-rity issuance. This thesis uses, in line with Dittmar & Thakor (2007), Actual-Forecasted earnings per

share (”EPS”) and EPS Dispersion as measures for agreement. My thesis contributes to the literature in numerous ways. First, a new measure for agreement is introduced. The two measures used by Dittmar & Thakor (2007) could be interpreted as measures for asymmetric information and might introduce measurement error and therefore bias in our results. Goldstein & Yang (2015) argue that the size and di-versity of the population of informed traders positively affects price informativeness. If informed traders trade aggressively the quality and quantity of information increases. If so, asymmetric information is lowered by this decrease in uncertainty. It is expected that both Actual-Forecasted EPS and EPS Disper-sion are affected by price informativeness since both measures capture information asymmetries. Proxy contests are introduced as a new measure for agreement since it is an important channel for investors to signal their dissatisfaction (Fos & Tsoutsoura, 2014). Robustness tests show that proxy contests are not subject to the same problems as both other measures. This test strongly substantiates proxy contests as a new measure for agreement since it shows that its effect is even stronger in magnitude and significance if low valuation errors are present. The logit regression first tests the relationship between agreement and issuance for the entire time period. The relationship is separately tested for the years between 1993– 2002. These findings are compared with the results of Dittmar & Thakor (2007), who use a similar time period. The relationship is also tested for the years 2003–2016. Overall, the logit regressions provide evidence that supports the hypothesis which states that firms prefer to issue equity if agreement is high. Although not all measures for agreement can explain issuance in all (adjusted) sample periods the over-all evidence is quite convincing. Dittmar & Thakor (2007) employ a logit analysis to substantiate the agreement theory. Ultimately, the goal of this thesis is to test for causality.

My thesis contributes to the literature in another important matter. The logit analysis, also done by Dittmar & Thakor (2007), cannot offer causal interpretations for several reasons. First, cross-sectional correlations between agreement and unobservable factors that directly affect issuance might bias the results. Second, it could be argued that agreement is affected by issuance. Both concerns are addressed by introducing Russell 2000 index inclusion to induce an exogenous variation in proxy contests. This will be done with an instrumental variable approach using a 2SLS framework. Schmidt & Fahlenbrach (2017), Appel et al. (2016b) and Crane et al. (2016) argue that index assignment for firms around the Russell 2000/1000 cutoff point is approximately random. Apparently, Russell 2000 index inclusion increases passive ownership (Malenko & Shen, 2016; Schmidt & Fahlenbrach, 2017; Appel et al., 2016b; Crane et al., 2016; Appel et al., 2016a). The instrument is relevant since Russell 2000 index inclusion facilitates activism through at least two possible channels. First, the support of passive institutions increases the likelihood of a successful proxy contest (Brav et al., 2008). Second, the presence of passive institutions decreases the coordination cost of activism (Bradley et al., 2010). The exogenous variation in proxy contests is used to test if the effect of agreement on security issuance is causal, rather than a

(partial) correlation. After doing a careful analysis, and a proper identification, I am not able to find any evidence that supports the agreement theory. Firms do not issue more equity after an (exogenous) increase in agreement and therefore this thesis rejects the agreement theory introduced by Dittmar & Thakor (2007) and strengthened Boot & Thakor (2011). I assume that inferior measures for agreement are the reason for the divergence of opinions.

Section 2 examines theories that elaborate on firms’ issuance behavior. Section 3 presents the methodology. Section 4 presents empirical evidence by performing an analysis comparable to Dittmar & Thakor (2007). Section 5 tests for causality. Section 6 concludes.

2

Literature Review

This section examines theories that provide reasons why firms issue equity. Relevant literature for every theory is discussed in the corresponding subsection. Subsection 2.1 elaborates on the trade-off theory. Subsection 2.2 examines the pecking-order. Subsection 2.3 discusses the market-timing theory. Subsection 2.4 introduces the agreement theory. Subsection 2.5 sheds light on literature that uses Russel 2000 index inclusion as exogenous shock to passive ownership, thereby catering proxy contests (Brav et al., 2008; Bradley et al., 2010).

2.1 Trade-off theory

The theory argues that a firm’s current capital structure is the result of prior decisions toward an optimum determined by the benefits and costs of debt. This optimum is a trade-off between the tax advantages of leverage and financial distress and agency costs. Shyam-Sunder & Myers (1999) test a framework which predicts that firms adjust their debt level to arrive at the optimal debt ratio. For example, a raise in the stock price lowers the actual debt ratio and therefore the trade-off theory predicts that debt is issued. The authors do not not find strong evidence that firms balance the marginal benefits of the tax shield against agency and financial distress costs, however, the test lacks power to reject the trade-off specification. Specifically, they argue that the usual tests on the trade-trade-off theory lack statistical power and call for sharper models. Overall, Shyam-Sunder & Myers (1999) question the presence of an optimal debt ratio and therefore cast doubt on the trade-off theory. Fama & French (2002) are one of the first to test predictions of the trade-off theory on the interaction between the dividend payout ratio and debt level. The authors find that profitable firms have more and higher dividend payouts and less leverage. Obviously, the negative relation between leverage and profitability is not consistent with the trade-off theory (Frank & Goyal, 2003). Authors such as Rajan & Zingales (1995) and Shyam-Sunder & Myers (1999) also find a negative correlation between leverage and profitability and reject the theory as

well. Danis et al. (2014) examine the puzzling relationship between leverage and profitability and find a positive correlation for firms with leverage close to the optimal ratio. Note that their dynamic framework allows for a negative correlation between profitability and leverage in the run-up to a debt restructuring. As expected, Danis et al. (2014) only find a negative correlation between profitability and leverage just before a debt restructuring. Strebulaev (2007) find that firms’ capital structure mostly differs from the optimal capital structure as posed by the trade-off theory and stress that evidence on the trade-off theory is very mixed. Similar to Strebulaev (2007) and Danis et al. (2014), Morellec et al. (2012) develop a dynamic trade-off model in which capital structure adjustments are rare. The authors differentiate by focusing on the role of agency conflicts, specifically managers who capture private benefits through misuse of the free cash flow to equity. They show that incorporating a small conflict of interest results in a model that explains why firms are often under-levered according the trade-off theory, and therefore not capturing all the tax advantages of leverage. The theory argues that all firms move toward their optimal capital structure, however, DeAngelo et al. (2011) find that managers intentionally move away from the target ratio to meet capital needs. Additionally, similar to Morellec et al. (2012), DeAngelo et al. (2011) argue that the leverage ratio predicted by the trade-off theory is unrealistic high. DeAngelo et al. (2011) develop and estimate a dynamic model to address both concerns. The introduced model uses debt as a temporary investment vehicle to fund unanticipated investment opportunities. DeAngelo & Roll (2015) state that no single theory can adequately predict firms’ security issuance behavior for all situations, however, the authors stress the importance of dynamic models. Apparently, the dynamic trade-off model has some predictive power, however, the overall evidence on the classic trade-off theory is mixed, at best.

2.2 Pecking-order theory

The pecking-order predicts that firms prioritize the source of their expenditures in a certain order. Firms prefer to finance their expenditures with internal finance, otherwise debt is used. Equity is used only when other funds are not sufficient. It is argued that asymmetric information increases the cost of financing, causing reluctance towards issuing equity. A firm should undertake an investment opportunity if the net present value is positive, however, Myers & Majluf (1984) find indications that managers forgo potential investment opportunities due to their reluctance to issue equity. In an efficient market equity should be a perfect substitute for cash since securities are sold at a fair price (Myers & Majluf, 1984). The authors assume that managers have superior information on investment opportunities and that both managers and investors are aware of this. Consequently, an equity issue could be perceived as a signal of overvaluation resulting in a falling share price. Equity is rarely issued according to Myers & Majluf (1984).

Shyam-Sunder & Myers (1999) test the pecking-order theory against the static trade-off model for public firms in the United States between 1971 and 1989. The authors find that the pecking-order is an excellent theory for issuance behavior if firms are public and mature. However, the authors stress the problems associated with the pecking-order model. Bharath et al. (2009) test whether capital structure decisions are truly driven by information asymmetry and find that the pecking-order model only explains firms issuance behavior if a high degree of information asymmetry is present. Frank & Goyal (2003) find that mostly large firms issue securities as the pecking-order theory predicts. Interestingly, Fama & French (2005) conclude the opposite and find that mostly small firms exhibit issuance behavior in line with the order model. It can also be reasoned that different findings concerning the pecking-order are mainly driven by problems with statistical power (Leary & Roberts, 2010). Leary & Roberts (2010) introduce a new model and testing strategy to address this concern. The authors argue that interpretation of the pecking-order model is crucial for the relevance of the results. They find that only 20% of the firms follow the pecking-order model if a strict interpretation is maintained. On the other hand, a loose interpretation results in significantly more firms that follow the theory. However, even a loose interpretation is not able to accurately explain half of the issuance decisions.

The pecking-order theory has some flaws which should be recognized if equity issues occur on a frequent basis (Shyam-Sunder & Myers, 1999). Fama & French (2005) test the theory between 1973 and 2002. The authors find that more than the half of their sample firms do not issue as the model expects. This closely matches the numbers found by Leary & Roberts (2010). The pervasive occurrence of equity issues implies that the pecking order, as proposed by Myers & Majluf (1984), faces serious problems. Fama & French (2005) argue that transaction costs and asymmetric information problems may not be severe enough to have equity issued only as a last resort. This doesn’t entail that those problem do not exist, however, it indicates that some situations allow for equity issues with modest asymmetric information and transaction costs. Fama & French (2005) argue that we should stop seeing the pecking-order and trade-off as stand-alone theories on capital structure, thereby removing barriers for alternative explanations on issuance.

2.3 Market-timing theory

The failure of the pecking-order as stand-alone theory on capital structure indicates that equity issues are not used only as a last resort (Fama & French, 2005; Leary & Roberts, 2010). This subsections examines the market-timing theory, which is an alternative theory on firms’ issuance behavior. The theory states that managers are more likely to issue equity when prices are high, and repurchase shares when prices are low. Managers ”time-the-market” to exploit temporary fluctuations in the share price (Baker & Wurgler, 2002). This is to the benefit of current shareholders because equity can be sold at

higher prices. Baker & Wurgler (2002) investigate equity issuance behavior between 1968 and 1999 and find that firms issue as the theory predicts. The authors find that the market-timing behavior of managers has a strong and long-lasting effect on firms’ capital structure. Therefore, Baker & Wurgler (2002) argue that the current capital structure is simply the result of prior attempts to time the market.

Alti (2006) researches the effect of market-timing on capital structure between 1971 and 1999. Market-timing effects are identified using initial public offerings (”IPOs”) for hot and cold markets. Alti (2006) give numerous reasons why IPOs can be used to analyze market-timing effects: (i) they are considered as one of the most important financing events, (ii) IPO firms have more asymmetric infor-mation relative to more mature firms; (iii) IPO’s are one of the most volatile financing activities. The three abovementioned reasons cause managers to see a hot-market as an opportunity to lower the cost of issuance. Alti (2006) find that IPO firms offer substantially more equity in hot than in cold markets. Additional robustness tests support the conclusion that the results are truly driven by the market-timing theory; a hot-market has almost no effect on firm and industry characteristics. Moreover, Alti (2006) em-ploy various robustness checks that specifically test for leverage, investment, dividend and cash amounts. All tests support the hypothesis that equity issues are largely driven by the market-timing theory.

Dong et al. (2012) examine whether and how misvaluation affects issuance decisions between 1976 and 2009. The authors find that the effects of misvaluations are stronger for equity than for debt issuers. They define misvaluation with one single measure, V/P, which is a forward-looking benchmark for abnormal returns. Dong et al. (2012) find that more equity issues occur as misvaluation increases, however, this effect is only present among overvalued firms. The authors examine this by creating various subsamples per industry, size, market-to-book ratio and R&D expenditures. The issuance of overvalued equity allows managers to raise cheap capital thereby favoring old shareholders (Dong et al., 2012). Polk & Sapienza (2009) test how mispricing affects corporate financing decisions between 1963 and 2000. The main difference with the paper of Dong et al. (2012) is that Polk & Sapienza (2009) test whether mispricing affects issue decisions through the catering channel. Polk & Sapienza (2009) also find that firms issue more equity if more overvaluation is present. The authors find, in line with their hypothesis, that this relationship is the result of managerial catering to investor overvaluation.

DeAngelo et al. (2010) also examine a statistically and economically significant effect of the market-timing theory on the likelihood that a firm issues equity between 1973 and 2001. The authors present the market-timing theory as the most prominent explanation for equity issues. They find that a large increase in market-timing opportunities increases the likelihood of an equity issue only modestly, however, the marginal impact is large since firms do not sell shares frequently. Although the effects are significant, DeAngelo et al. (2010) find that the market-timing model does not provide a complete explanation. The authors just observe too many firms that forgo the chance to issue equity while having excellent

timing opportunities. Overall, the evidence provided by DeAngelo et al. (2010) supports the market-timing theory but questions its completeness. Butler et al. (2011) find that it is reasonable for managers to time the market by raising capital when prices are high, however, managers are apparently not able to successfully switch between issuing shares or increasing leverage. The authors find no evidence of successful market-timing behavior to fool new investors. Moreover, Dittmar & Thakor (2007) find that a high valuation is one of the key drivers of an equity issue, however, the authors argue that a high valuation is the result of market agreement. This theory is explained in the next subsection.

2.4 The Agreement theory

The previous subsections casts doubt on the completeness of the abovementioned theories in being able to offer a self-contained explanation for the issuance behavior of firms. Fama & French (2005) found that equity is not issued only as a last resort and conclude that the pecking order, as proposed by Myers & Majluf (1984), is dead. The market-timing is better in explaining why firms issue equity, however, DeAngelo et al. (2010) show that this theory fails to provide a complete explanation for their regular occurrence. Dittmar & Thakor (2007) develop and test a new theory that explains why and when it could be beneficial for management to issue equity. The authors introduce the ”managerial investment autonomy” that predicts that an equity issue is beneficial if agreement is high. The choice to issue debt or equity is a trade-off between the autonomy that equity provides and the tax advantages of leverage. This new theory does not rule out market-timing behavior since it supposedly has incremental power aside from the market-timing perspective. Dittmar & Thakor (2007) focus on firms that issued seasoned equity or nonconvertible debt between 1993 and 2002. The main analysis is focused on equity issuers, and debt issuers are used as comparison group. The authors use two variables to measure agreement: (i) Actual-Forecasted EPS, which is increasing in agreement, and (ii) EPS dispersion, which is decreasing in agreement. Managers care about the after-project stock price and long-term equity value (Dittmar & Thakor, 2007). Consequently, the issue decision depends on managers expectations on how the share price will react to the issue. Dittmar & Thakor (2007) substantiate this theory with an analysis through backward induction. The authors argue that a critical cutoff point exists for the agreement parameter. The tax shield causes that firms to prefer to issue debt if p* <p** and equity if p*>p**, given the greater autonomy that it provides. The model predicts that more agreement increases the likelihood that managers and investors will agree on project value. The agreement theory is substantiated with a logit analysis. Furthermore, Dittmar & Thakor (2007) support the market-timing theory, often associated with a high valuation. Note that the authors argue that a high valuation is evidence of agreement.

Boot & Thakor (2011) elaborate on the agreement theory introduced by Dittmar & Thakor (2007). Similar to Dittmar & Thakor (2007), Boot & Thakor (2011) predict that equity is issued when agreement

is sufficiently high, and debt whenever it is low. However, key differences are present. The most important difference is that Boot & Thakor (2011) examine the effect of the value of assets on whether debt or equity is issued. The authors theory revolves around ”managerial autonomy” and states that managers seek finance with ”softest” control since autonomy is valued highly. The authors argue that debt is always issued if the value of assets is sufficiently high; making agreement irrelevant. Boot & Thakor (2011) state that a high value of assets assures that managers have a greater autonomy from bondholders which allows them to engage more proactively in increasing shareholder wealth.

2.5 The Experiment

This subsection discusses various studies that use the Russell 2000 index inclusion as an exogenous shock to passive ownership. Appel et al. (2016b) analyze the effect of passive investors on the campaigns and actions of activists. The authors find that activist shareholders are more likely to enforce their activism when passive institutional investors are present. The idea is that more concentrated ownership helps to overcome the free-ride problem associated with activism. The coordination cost and likelihood of success change favorably once an activist shareholder gains the support of the passive institution. The main concern of Appel et al. (2016b) is that portfolios of passive institutional investors might correlate with the actions and successes of activists, thereby exposing an omitted variable problem. Appel et al. (2016b) use an instrumental variable (IV) regression to address this concern. The authors study the change in passive ownership around the Russell 1000/2000 cutoff between 2008 and 2014. The empirical challenge mentioned is solved if index inclusion correlates sufficiently with passive ownership, and does not affect the outcome of interest directly. Appel et al. (2016b) argue that it is unclear how index inclusion influences the campaigns and actions of activists directly for firms around the cutoff. A sharp difference in passive ownership is shown for firms at the bottom of the Russell 1000 relative to firms at the top of the Russell 2000. Specifically, Appel et al. (2016b) find that firms with more passive ownership, instrumented by Russell 2000 inclusion, have a significantly higher likelihood on being involved in a proxy contest. Schmidt & Fahlenbrach (2017) research the effect of an exogenous variation in passive ownership on corporate governance issues. The authors find that an increase in passive ownership increases CEO power and decreases board independence. Similar to Appel et al. (2016b), Schmidt & Fahlenbrach (2017) use Russel 1000/2000 index inclusion as an exogenous shock to passive ownership. The main difference between both papers is that Schmidt & Fahlenbrach (2017) extract multiple instruments from the index reconstitution event; (i) a dummy that equals one if a firm switches from the Russell 1000 to the Russell 2000 and (ii) an indicator variable that equals one if a firm switches from the Russel 2000 to Russell 1000. Also, Schmidt & Fahlenbrach (2017) do not use the years after 2010 resulting in less exposure to the banding policy introduced by the Russell indexes

in 2007. Moreover, Chang et al. (2015), Appel et al. (2016a) and Crane et al. (2016) also use Russell 1000/2000 index inclusion as an exogenous shock to passive ownership.

3

Methodology

The goal of this thesis is to reveal if agreement between managers and investors has a causal effect on the security issuance behavior of firms. This section focuses on explaining the methodology. Subsection 3.1 presents the hypothesis. Subsection 3.2 elaborates on the model and research method. Subsection 3.3 discusses the variables.

3.1 Hypothesis

Section 2 provided an overview of theories that elaborates on firms’ issuance behavior. Apparently, all theories have shortcomings since none are able to provide a sole explanation for the regular manifestation of equity issues. This thesis examines the agreement theory, introduced by Dittmar & Thakor (2007), and contributes to a long list of researchers who shed light on firms’ issuance behavior. The issue decision is a trade-off between the autonomy that equity provides and benefits of the tax shield (Dittmar & Thakor, 2007). An equity issue is beneficial if agreement is sufficiently high. The idea is that more agreement increases the likelihood that managers and investors will agree on project value and since autonomy is valued highly the hypothesis can be stated as follows:

Hypothesis 1 Firms prefer to issue equity if agreement between managers and investors is high.

Dittmar & Thakor (2007) find strong evidence in favor of the hypothesis between 1993 and 2002. Boot & Thakor (2011) also argue that equity is issued when agreement is high, even after controlling for a high stock price. Note that both Dittmar & Thakor (2007) and Boot & Thakor (2011) argue than a high valuation is evidence of agreement.

3.2 The Model

The goal of this subsection is to elaborate on the model and research method. Ultimately, the goal of this thesis is to test for causality. This will be done with an instrumental variable approach estimated using a 2SLS framework. Prior to this, a maximum likelihood regression is employed to test the relationship between agreement and issuance. The chosen logit model is similar to the model used by Dittmar & Thakor (2007) and is as follows:

Section 4 employs the logit regression and tests the relationship between agreement and firms is-suance behavior. The dependent variable equals one if the choice of finance is equity, and zero if the choice of finance is debt. Three measures for agreement are used: (i) Actual-Forecasted EPS, (ii) EPS dispersion and (iii) proxy contests. Note that the first variable is a measure for agreement and therefore the agreement theory expects a positive effect. The other two variables have an inverted interpretation since they are actually measures for disagreement; a negative effect is expected. If so, a negative corre-lation between Actual-Forecasted EPS and other measures for agreement should be observed. Dittmar & Thakor (2007) use both Actual-Forecasted EPS and EPS dispersion as measures for agreement. It is expected that both measures are significant on the 5% level for the sample period between 1993 and 2002 (Dittmar & Thakor, 2007). The main critique of both variables is that they could be interpreted as measures for asymmetric information. The measurement of a proxy that includes noise is a source of endogeneity and should be avoided. To address this concern a new measure for agreement is introduced. Proxy contests are relevant to measure agreement since it is the most aggressive channel for investors to signal their disagreement to the management (Fos & Tsoutsoura, 2014). A robustness test is done to assure that proxy contests are not subject to the same asymmetric information problems. This test focuses on issuances that have less valuation error according to the Actual-Forecasted EPS measure. Specifically, firms that are in the two quartiles around zero valuation error. Furthermore, several control variables, all winsorized, are added to capture other factors that influence firms’ issuance behavior. At first, the effect of agreement on issuance is tested for the entire sample period, which is between 1993– 2016. The relationship is also tested for the years 1993–2002. These results are compared with Dittmar & Thakor (2007), who test a similar time period. Moreover, the relationship is tested separately for the years 2003–2016. Section 4 tests these different sample periods to check for the robustness of the results and to test which years have a stronger relationship.

The research method, introduced by Dittmar & Thakor (2007), and replicated in Section 4, cannot offer causal interference for several reasons. First, agreement could correlate with unobservable firm characteristics that directly drive the decision to issue equity. Second, reverse causality could be present. For example, investors could be dissatisfied after an issuance. Section 5 addresses these concerns by introducing the Russell 2000 index inclusion as an instrument for proxy contests. Index assignment for firms around the Russell 2000/1000 cutoff point is approximately random and therefore induces an exogenous variation in proxy contests. This exogenous shock is used to test if the relationship between agreement and issuance is causal, rather than just a (partial) correlation. A negative and significant effect of proxy contests on issuance is in line with expectations. An F-test is employed to assure that the results found are not the consequence of a weak instrument.

3.3 Description of variables

3.3.1 Agreement measures

This thesis uses three measures for agreement: (i) Actual-Forecasted EPS, (ii) EPS dispersion and (iii) proxy contests. The first two measures are both used by Dittmar & Thakor (2007). Actual-Forecasted EPS is defined as actual earnings per share of the quarter before issuance minus the most recent analyst forecast, normalized by actual EPS. The idea is that investors have an increased likelihood of supporting managerial decisions if expectations are surpassed. The agreement theory predicts that the likelihood of issuing equity is increasing in Actual-Forecasted EPS (Dittmar & Thakor, 2007). The second measure for agreement used by Dittmar & Thakor (2007) is the average standard deviation of analysts’ forecast, measured in the quarter before issuance, divided by the book value of equity. This measure, referred to as EPS Dispersion, is actually a measure of disagreement among analysts (Diether et al., 2002; Banerjee, 2011; Anderson et al., 2005). Dittmar & Thakor (2007) argue that disagreement among analysts has a high degree of correlation to agreement between management and investors. EPS dispersion has an inverted interpretation as the Actual-Forecasted EPS variable; less EPS dispersion connotes more agreement. The agreement theory predicts that the likelihood that a firm issues equity is decreasing in EPS dispersion.

One of the major disadvantages of the measures used by Dittmar & Thakor (2007) is that both could be interpreted as measures for asymmetric information. Proxy contests are introduced as a new measure for agreement since they are expected to be less sensitive to asymmetric information problems. Fos & Tsoutsoura (2014) argue that the connection between investors and managers is weak if shareholders have no decisive tool to channel their dissatisfaction about firms’ characteristics or goals. The authors state that shareholders should have mechanisms in place that allows them to express their disagreement. Proxy contests are considered as a relevant measure for agreement since they are one of the most aggres-sive channels for investors to signal their dissatisfaction (Fos, 2016). Fos & Tsoutsoura (2014) find that proxy contests have significant effects on the career of managers. Mulherin & Poulsen (1998) agree and state that proxy contests are one of the most important control devices that shareholders possess, which is also used to discipline management. Malenko & Shen (2016) stress the important role of shareholder voting in corporate governance issues. The authors find that proxy contests are facilitated by the rise in institutional ownership. Proxy advisory firms are consulted to deal with those conflicts of interests. Apparently, proxy contests are an important channel for agreement and therefore two measures are in-troduced. The first variable, Proxy1, is a dummy that equals one if a firm has a proxy contest happening one year before the issuance. Furthermore, Proxy2 is a dummy that equals one if a firm has a proxy contest one or two years before the issuance. Proxy1 and Proxy2 are expected to have a high degree of correlation. Both Proxy1 and Proxy2 are measures for disagreement and therefore have an inverted

interpretation; less proxy contests connote more agreement.

3.3.2 Price variables

To distinguish between the agreement and market-timing theory, often associated with a high valuation, this thesis controls for various measures that capture market timing effects. In line with Dittmar & Thakor (2007) three measures are used to capture market timing effects: (i) raw returns for the 3, 6, 9 and 12 months before the issuance, (ii) market adjusted returns, defined as raw minus market returns, for the 3, 6, 9 and 12 months before issuance and (iii) the market-to-book ratio. The abovementioned variables are from now on referred to as price variables. Baker & Wurgler (2002) argue that managers ”time-the-market” to exploit temporary fluctuations in the share price. If so, a greater price run-up increases the likelihood that a firm issues equity and therefore a positive effect of raw and market adjusted returns on issuance is expected (Butler et al., 2011). The market-to-book is another commonly used variable to measure market-timing opportunities (DeAngelo et al., 2010; Baker & Wurgler, 2002; Dong et al., 2012). Baker & Wurgler (2002) argue that the market-to-book ratio captures potential misvaluations by irrational investors and therefore a high market-to-book ratio increases the probability that the choice of finance is equity.

3.3.3 Control variables

Several factors other than agreement measures and market timing factors might affect issuance. Control variables are added to capture those effects. Larger firms are, on average, more diversified and fail less frequently which could lower their cost of debt (Rajan & Zingales, 1995). If so, larger firms would have a preference towards increasing leverage. The natural log of sales is added to control for this effect (Dittmar & Thakor, 2007; Frank & Goyal, 2003; Bharath et al., 2009). A measure for profitability, return on assets, is added as well. This measure is defined as operating income divided by total assets (Dittmar & Thakor, 2007). Jensen (1986) associate profitability with more free cash problems resulting in a preference towards debt. Additionally, Baker & Wurgler (2002) associate profitability with the presence of internal funds resulting in a decreased likelihood that a firm issues equity. Strebulaev (2007) find that more profitable firms decrease costly external finance and therefore reduce their debt level. Dittmar & Thakor (2007) argue that firms with more financial slack, defined as cash and cash equivalents divided by total assets, have less need for costly external finance. Additionally, Frank & Goyal (2003) find that firms have less preference towards increasing leverage if sufficient cash is present. Because of the abovementioned reasons this thesis controls for financial slack. R&D to sales is included since it is a measure for growth opportunities, however, the effect is more controversial. Dittmar & Thakor (2007) state that the agency costs of debt are higher if firms have more growth opportunities resulting

in a preference towards equity. On the contrary, Frank & Goyal (2003) argue that high growth firms are more likely to issue debt. Fama & French (2005) agree and find that firms issue more debt if high growth is present. Furthermore, issuance is allegedly affected by the tangibility of assets. Asset tangibility is defined as the ratio fixed assets over total asset value. It is expected that asset tangibility has a positive effect on the likelihood that firms will issue debt since tangible assets are easier to use as collateral (Rajan & Zingales, 1995; Dittmar & Thakor, 2007; Frank & Goyal, 2003; Leary & Roberts, 2010). It is also expected that issuance is affected by leverage, defined as the book value of leverage divided by assets. Dittmar & Thakor (2007) argue that highly levered firms have a higher probability on an equity issue. All control variables are used by Dittmar & Thakor (2007) and measured at the end of the fiscal year before the issuance.

4

Empirical evidence

This section presents empirical evidence on the relationship between agreement and security issuance by performing an analysis comparable to Dittmar & Thakor (2007). Subsection 4.1 presents an overview of the sample. Subsection 4.2 provides descriptive statistics and Subsection 4.3 analyses the effect of agreement on issuance using a logit regression.

4.1 Sample overview

This thesis uses a sample of public firms listed in the U.S. that issued seasoned equity or nonconvert-ible debt between 1993 and 2016. The Thompson ONE New Issuance database provided information on firms’ issuance behavior. The analysis is focused on equity issuers and debt issuers are used as a comparison group since both are expected to experience similar cash flows (Dittmar & Thakor, 2007). If a firm has multiple debt issues in one year only the first issue is kept. Furthermore, transactions are deleted if debt and equity is issued in the same year. Data on the forecasts of analysts is obtained from IBES, accounting data from Compustat and stock return data from the CRSP database. Additionally, data on proxy contests is obtained from the Thomson One Shareholder Activism database. Moreover, data on Russell 2000/1000 index inclusion is obtained through FTSE Russell indexes. Financial and public administration firms are deleted.

4.2 Descriptive statistics

Figure I presents yearly issuances between 1993 and 2016. Between 1993 and 2007 the quantity of equity issues exhibits a clear up and downward trend with no less than 300 and no more than 450 issues per year. 2008 is the year with the lowest number of equity issues. It is likely that the crisis is the

underlying reason for the occurrence of only 205 issues. After 2008 an upward trend evolved totaling at 607 equity issues in 2016. All years have more equity than debt issues, however, as the only exception, 1998 has slightly more debt. 1998 is the only yearly with over 300 debt issues. Furthermore, the whole sample period exhibits between 70 and 300 yearly debt issues, with the minimum in 2008. The total amount of issuances is 14,409.

Figure I

Security Issuance over time

Table I presents the summary statistics for control variables between 1993 and 2016. Panel A divides the full sample in two subsamples; debt and equity issuers. It shows that equity issuers are smaller, less profitable, have more cash, more R&D expenses, less fixed assets and less leverage relative to debt issuers. All differences are significant on the 1% level. Panel B summarizes control variables for high and low agreement firms. A firm is a high (low) agreement firm if it is in the highest (lowest) quartile of the agreement measure Actual-Forecasted EPS. Again, for high and low agreement firms, it shows that equity issuers are smaller, less profitable, have more cash, more R&D expenses, less fixed assets and less leverage relative to debt issuers. All differences are significant on the 1% level. Panel C summarizes control variables for firms that have or do not have a proxy contest one or two years before the issuance. A firm is a low (high) agreement firm if it does (does not) have a proxy contest one or two years before the issuance. Panel C shows, again, that equity issuers are smaller, less profitable, have more cash, more

R&D expenses, less fixed assets and less leverage relative to debt issuers. All differences are significant on the 1% level for both high and low agreement firms. Dittmar & Thakor (2007) observe comparable firm characteristics as in Table I. The inclusion of these control variables allows us to correctly estimate the effect of agreement on issuance.

Table I

Summary Statistics - Controls Panel A: Full Sample

Debt issuers Equity issuers Difference

lnSales 7.86 4.48 3.39***

Return on assets 0.09 0.00 0.28***

Cash to assets 0.04 0.16 -0.22***

R&D to sales 0.00 0.01 -1.86***

Fixed assets to assets 0.45 0.23 0.16***

Debt to assets 0.32 0.17 0.10***

No. of observations 5680 9729

Panel B: High and Low agreement by EPS forecast High Agreement

lnSales 6.98 3.96 3.00***

Return on assets 0.06 -0.04 0.27***

Cash to assets 0.04 0.16 -0.21***

R&D to sales 0.00 0.02 -1.71***

Fixed assets to assets 0.50 0.23 0.19***

Debt to assets 0.36 0.16 0.18*** No. of observations 1570 5102 Low Agreement lnSales 7.76 4.20 3.49*** Return on assets 0.09 -0.17 0.36*** Cash to assets 0.03 0.30 -0.38*** R&D to sales 0.00 0.10 -3.20***

Fixed assets to assets 0.54 0.22 0.21***

Debt to assets 0.33 0.17 0.11***

No. of observations 1107 1846

Table I–Continued

Debt issuers Equity issuers Difference

Panel C: High and Low agreement by proxy contests High Agreement

lnSales 7.90 4.43 3.57***

Return on assets 0.09 0.00 0.30***

Cash to assets 0.04 0.17 -0.23***

R&D to sales 0.00 0.02 -3.68***

Fixed assets to assets 0.45 0.23 0.16***

Debt to assets 0.32 0.18 0.09*** No. of observations 5583 9671 Low Agreement lnSales 8.54 4.55 4.05*** Return on assets 0.09 -0.14 0.44*** Cash to assets 0.06 0.14 -0.17*** R&D to sales 0.00 0.03 -2.23***

Fixed assets to assets 0.38 0.22 0.12***

Debt to assets 0.31 0.20 0.07***

No. of observations 140 148

Table I presents summary statistics for firms that issue equity or debt between 1993 and 2016. All variables are measured one year before issuance and obtained from the Thomson One, CRSP, Compustat and IBES databases. Return on assets is defined as operating income divided by total assets. Cash to assets is expressed as the sum of cash and cash equivalents divided by total assets. Panel A presents the medians for the complete sample. Panel B summarizes control variables for firms in the highest and lowest agreement quartile according the Actual – Forecasted EPS agreement measure, which is measured one quarter before issuance. Panel C summarizes control variables for firms that have or do not have a proxy contest one or two years before issuance. A firm is a low (high) agreement firm if it does (does not) have a proxy contest one or two years before issuance. Difference is defined as the difference in means between debt and equity issuers. Statistical significance at the 1%, 5%, and 10% level is indicated by ***, **, and *, respectively.

Table II presents summary statistics for the price variables. Raw and adjusted returns are computed for the 3, 6, 9 and 12 months before issuance and the market-to-book ratio is measured one year before. It shows that equity issuers have, on average, a greater price run-up than debt issuers. Additionally, the market-to-book ratio is higher for firms that issue equity. All differences are significant on the 1% level. These results are in line with the market-timing theory, which states that managers are more likely to issue equity if the firm is valued highly. It indicates that managers ”time-the-market” to exploit fluctuations in the share price (Baker & Wurgler, 2002). The issuance of overvalued equity allows managers to raise cheap capital thereby favoring old shareholders (Dong et al., 2012). Obviously, the results are not consistent with the pecking-order model, which predicts that equity is issued only as a last resort (Myers & Majluf, 1984). Firms that issue equity as last resort are not likely to experience high prices (Dittmar & Thakor, 2007). No causal interpretations can be made yet, however, inclusion of

price variables allows us to correctly estimate the effect of agreement on issuance. Note that Dittmar & Thakor (2007) argue that a high valuation is evidence of agreement.

Table II Price Variables

Returns Debt Issuers Equity Issuers Difference Average raw 3-month return 0.03 0.25 -0.22*** Average raw 6-month return 0.08 0.48 -0.39*** Average raw 9-month return 0.09 0.67 -0.58*** Average raw 12-month return 0.13 0.80 -0.68*** Average Market-to-Book ratio 1.76 3.98 -2.22*** Average market-adjusted 3-month return 0.00 0.22 -0.22*** Average market-adjusted 6-month return 0.03 0.42 -0.39*** Average market-adjusted 9-month return 0.02 0.59 -0.58*** Average market-adjusted 12-month return 0.03 0.71 -0.68*** No. of observations 5680 9729

Note: Table II presents the means for several price variables that capture timing effects. Raw and market-adjusted returns are summarized for debt and equity issuers. Raw returns are computed for the 3, 6, 9 and 12 months before issuance. Market adjusted returns are expressed as the difference between raw and market returns. Difference is defined as the difference in means between debt and equity issuers. Statistical significance at the 1%, 5%, and 10% level is indicated by ***, **, and *, respectively.

Table III presents the correlations between different measures for agreement. The correlation be-tween Proxy1 and Proxy2 is high and significant on the 1% level (0.783). This correlation arises me-chanically since both are measures for proxy contests. All other correlations are below 1% and in-significant. The agreement theory expects that agreement is increasing in Actual-Forecasted EPS, and decreasing in all other variables since they have an inverted interpretation. In line with expectations a negative correlation is observed between Actual-Forecasted EPS and all other measures for agreement. Interestingly, EPS Dispersion has a negative correlation with both Proxy1 and Proxy2. Theory predicts a positive correlation since EPS Dispersion and proxy contests have a similar interpretation. Malmendier & Tate (2005) argue that significance and direction of the correlation coefficients are useful indicators to test if variables capture similar effects. Overall, the negative, although insignificant, correlation between EPS Dispersion and proxy fights indicates that EPS Dispersion is not a good measure for agreement. Argued from the price informativeness perspective EPS Dispersion captures information asymmetries and is therefore a bad measures for agreement (Goldstein & Yang, 2015). Moreover, EPS Dispersion is actually a measure for disagreement among analysts. It could be argued that agreement among analyst and agreement between investors and managers is not as highly correlated as Dittmar & Thakor (2007) claim.

Table III

Correlation between Agreement measures

Actual-Forecasted EPS

EPS

Dispersion Proxy1 Proxy2

Actual-Forecasted EPS 1

EPS dispersion -0.002 1

Proxy1 -0.005 -0.002 1

Proxy2 -0.005 -0.003 0.783*** 1

Note: Table III presents the correlations between different measures for agreement. Actual-Forecasted EPS is the difference between actual and forecasted EPS while EPS dispersion is the mean dispersion of analysts forecast. Both are calculated one quarter before issuance. Proxy1 is a dummy that equals one if a proxy contest occurred in the year before the issuance. Proxy2 is a dummy that equals one if a proxy contest occurred one or two years before the issuance. The idea is that agreement is increasing in Actual-Forecasted EPS, and decreases in all other measures. Statistical significance at the 1%, 5%, and 10% level is indicated by ***, **, and *, respectively.

4.3 Results

This subsection tests the relationship between firms’ issuance behavior and agreement between 1993 and 2016. Similar to Dittmar & Thakor (2007) a logit regression is used to test this relationship. The dependent variable equals one if the choice of finance is equity, and zero if the choice of finance is debt. Multiple measures for agreement are used: (i) Actual-Forecasted EPS, (ii) EPS dispersion and (iii) proxy contests. The first variable is a measure for agreement and therefore the agreement theory expects a positive effect. Note that the other measures are expected to have a negative coefficient since both have an inverted interpretation. The variable of interest is presented as both raw coefficient and marginal effects of the means coefficient (”MEMs”). The main analysis is focused on the coefficient that presents marginal effects and it is clearly mentioned if raw coefficients are discussed. Most columns are reported in pairs of three since they differ only by price variable.

Table IV presents the coefficients of the logit analysis in which agreement is measured by Actual-Forecasted EPS. The first three columns show regressions of the entire sample period and differ only by price variable. All three columns show a strong effect of agreement. The results support the agreement theory since Actual-Forecasted EPS is positive and significant on the 1% level. Column 3 presents a coefficient of 0.08 which entails that one unit increase in Actual-Forecasted EPS causes on average a 0.08 increase in the probability that a firm issues equity. The subsequent three columns present the effects of agreement on issuance between 1993 and 2002, which is a similar time period as examined by Dittmar & Thakor (2007). Again, the results support the agreement theory. Regression 4 and 6 present significance on the 1% level, however, Actual-Forecasted EPS is significant on the 5% level if 12-month raw return is included as a price variable. Column 6 presents a coefficient of 0.12 which entails that

one unit increase in Actual-Forecasted EPS causes on average a 0.12 increase in the probability that a firm issues equity. Column 7 presents the effect of agreement on issuance after 2002. Again, the agreement theory is supported by presenting a positive effect of Actual-Forecasted EPS on issuance on the 5% significance level. Column 7 presents a coefficient of 0.05 which entails that one unit increase in Actual-Forecasted EPS causes on average a 0.05 increase in the probability that a firm issues equity. Overall, the results of Table IV support the hypothesis that firms prefer to issue equity if agreement is high. The findings are also consistent with the managerial autonomy theory provided by Boot & Thakor (2011).

Table IV

Logit regressions: Actual-Forecasted EPS

1993–2016 1993–2002 2003–2016 (1) (2) (3) (4) (5) (6) (7) Actual-Forecasted EPS 0.35*** 0.28*** 0.41*** 0.51*** 0.36** 0.58*** 0.29** (0.09) (0.09) (0.09) (0.17) (0.18) (0.17) (0.12) Actual-Forecasted EPS Marginal Effects 0.07*** 0.06*** 0.08*** 0.11*** 0.08** 0.12*** 0.05** (0.02) (0.02) (0.02) (0.04) (0.04) (0.04) (0.02)

3-Month raw Return 0.89*** 0.76***

(0.12) (0.18)

12-Month raw Return 0.47*** 0.48***

(0.06) (0.08)

Market to book ratio 0.25*** 0.27*** 0.21***

(0.04) (0.05) (0.05) lnSales -0.75*** -0.72*** -0.73*** -0.68*** -0.63*** -0.66*** -0.84*** (0.02) (0.02) (0.02) (0.03) (0.04) (0.03) (0.03) Return on assets -2.62*** -3.26*** -3.99*** 0.27 -0.37 -0.83 -4.78*** (0.37) (0.42) (0.57) (0.48) (0.51) (0.69) (0.87) Cash to assets 1.05*** 1.11*** 0.26 2.18*** 3.01*** 1.02* -0.26 (0.26) (0.27) (0.27) (0.60) (0.64) (0.58) (0.35) R&D to sales 0.03 0.02 -0.02 0.14 0.00 -0.01 -0.02 (0.11) (0.09) (0.03) (0.70) (0.17) (0.09) (0.03) Fixed assets to assets -1.14*** -0.99*** -1.10*** -1.69*** -1.57*** -1.64*** -1.00***

(0.11) (0.12) (0.11) (0.19) (0.20) (0.20) (0.15) Debt to assets 0.41** 0.35** 0.48*** -1.23*** -1.19*** -1.18*** 1.51*** (0.17) (0.18) (0.17) (0.28) (0.32) (0.31) (0.22) Intercept 5.60*** 5.27*** 5.22*** 5.34*** 4.87*** 4.87*** 5.96*** (0.17) (0.18) (0.18) (0.28) (0.30) (0.29) (0.26) Pseudo R2 _{0.39 0.39 0.39 0.37 0.37 0.36 0.43} N 9439 8593 9300 3368 2896 3288 5279

Note: Table IV presents coefficients of the logit analysis in which agreement is measured by Actual - Forecasted EPS. All columns have one row that presents raw coefficients and another row that presents marginal effects at the means. Column 1–3 present the results between 1993 and 2016. Column 4–6 present results between 1993 and 2002. Column 7 presents the results after 2002. The dependent variable equals one if the choice of finance is equity, and zero if the choice of finance is debt. Actual EPS is measured one quarter before issuance and is matched with the forecast closest to actual EPS disclosure. The difference is normalized by actual EPS. The price variables are measured one quarter before issuance. All the other variables are computed one year before issuance. Standard errors are reported in parentheses. Statistical significance at the 1%, 5%, and 10% level is indicated by ***, **, and *, respectively.

Table V

Logit regressions: EPS Dispersion

1993–2016 1993–2002 2003–2016

(1) (2) (3) (4) (5) (6) (7)

EPS Dispersion 0.00 0.00 0.00 -0.08*** -0.08** -0.11*** 0.01***

(0.00) (0.00) (0.00) (0.03) (0.04) (0.04) (0.00) EPS Dispersion Marginal

Effects 0.00 0.00 0.00 -0.01** -0.01* -0.02*** 0.001***

(0.00) (0.00) (0.00) (0.01) (0.01) (0.00) (0.00)

3-Month raw Return 0.98*** 1.01***

(0.14) (0.24)

12-Month raw Return 0.58*** 0.72***

(0.06) (0.09)

Market to book ratio 0.24*** 0.23*** 0.22***

(0.04) (0.05) (0.05) lnSales -0.75*** -0.72*** -0.74*** -0.70*** -0.65*** -0.69*** -0.82*** (0.02) (0.02) (0.02) (0.04) (0.04) (0.03) (0.03) Return on assets -2.93*** -3.61*** -4.41*** 0.79 -0.17 -0.69 -5.28*** (0.41) (0.46) (0.60) (0.50) (0.62) (0.81) (0.93) Cash to assets 1.19*** 1.15*** 0.47* 2.21*** 3.19*** 1.39** -0.16 (0.27) (0.28) (0.28) (0.67) (0.78) (0.68) (0.35) R&D to sales 0.18 0.17 0.01 3.09* 2.72 1.85 -0.04** (0.21) (0.19) (0.05) (1.65) (2.03) (1.53) (0.02) Fixed assets to assets -1.09*** -0.93*** -1.06*** -1.65*** -1.51*** -1.64*** -0.93***

(0.11) (0.12) (0.11) (0.20) (0.22) (0.20) (0.15) Debt to assets 0.57*** 0.48*** 0.66*** -0.97*** -0.93*** -0.88*** 1.64*** (0.17) (0.18) (0.17) (0.29) (0.33) (0.31) (0.22) Intercept 5.60*** 5.27*** 5.28*** 5.31*** 4.79*** 5.02*** 5.84*** (0.18) (0.19) (0.18) (0.30) (0.33) (0.31) (0.26) Pseudo R2 0.38 0.38 0.37 0.38 0.39 0.37 0.40 N 8939 8169 8806 3114 2696 3048 5034

Note: Table V presents coefficients of the logit analysis in which agreement is measured by EPS Dispersion. All columns have one row that presents raw coefficients and another row that presents marginal effects at the means. Column 1–3 present the results between 1993 and 2016. Column 4–6 present results between 1993 and 2002. Column 7 present the results after 2002. The dependent variable equals one if the choice of finance is equity, and zero if the choice of finance is debt. The price variables and EPS Dispersion are measured one quarter before issuance. All other variables are computed one year before issuance. Standard errors are reported in parentheses. Statistical significance at the 1%, 5%, and 10% level is indicated by ***, **, and *, respectively.

Table V presents the coefficients of the logit analysis in which agreement is measured by EPS Dis-persion. EPS Dispersion has an inverted interpretation as Actual-Forecasted EPS and therefore the agreement theory expects a negative coefficient. The first three columns show regressions on the entire sample period and differ only by price variable. None of the three columns shows a significant effect of EPS Dispersion on issuance. These findings are not in line with the agreement theory. Interestingly, a negative effect of EPS Dispersion is observed once the same time period as Dittmar & Thakor (2007) is tested. Column 4 and 6 present significance on the 1% level for the raw coefficient, however, EPS Dispersion is significant on the 5% level if a 12-month raw return is used as a price variable. This is as the agreement theory predicts and in line with the research of Dittmar & Thakor (2007) and Boot & Thakor (2011). Interestingly, the effect of EPS Dispersion is weaker if marginal effects are examined.

The significance of column 4 decreases to the 5% level and the significance of column 5 decreases to the 10% level. Column 6 presents a coefficient of -0.02 which entails that one unit decrease in EPS Dispersion causes on average a 0.02 increase in the probability that a firm issues equity. Column 6 is still significant on the 5% level and therefore supports the hypothesis. The last column presents the results after 2002. Inexplicably, EPS Dispersion is positive on the 1% significance level. It presents a coefficient of 0.001 which entails that a one unit increase in EPS Dispersion causing on average a 0.001 increase in the probability that a firm issues equity. In other words, column 7 indicates that more agree-ment decreases the likelihood that a firm will issue equity. This is counter-intuitive and in conflict with our hypothesis. Interestingly, the counter-intuitive findings are only found in the years not examined by Dittmar & Thakor (2007).

Table VI presents the coefficients of the logit analysis in which agreement is measured by proxy contests. Two different measures for proxy contests are introduced: Proxy1 and Proxy2. Proxy1 is used as measure for agreement in columns 1–3 and Proxy2 in columns 4–6. Both are actually measures for disagreement and therefore the agreement theory expects a negative effect of proxy contests on issuance. In line with the agreement theory a negative, and significant on the 5% level, coefficient of proxy contests is found in columns 1–6. Column 3 presents a coefficient of -0.05 with entails that Proxy1 activation causes on average a 0.05 decrease in the probability that a firm issues equity. Column 6 presents a coefficient of -0.04 with entails that Proxy2 activation causes on average a 0.04 decrease in the probability that a firm issues equity. All columns support the hypothesis and indicate that firms have a higher probability of an equity issue if agreement is high. Column 7 focuses on issuances that have low valuation errors. This is done to assure that proxy contests do not encounter the same problems as both measures Actual-Forecasted EPS and EPS Dispersion. Column 7 presents a coefficient of -0.29 with entails that Proxy2 activation causes on average a 0.29 decrease in the probability that a firm issues equity. Again, evidence supports the agreement theory by presenting a negative effect of Proxy2 on the 1% significance level. This test strongly substantiates proxy contest as a new measure for agreement since Table VI shows that its effect is stronger in significance and magnitude if low valuation errors are present.

Table VI

Logit regressions: Proxy contests

(1) (2) (3) (4) (5) (6) (7) Proxy1 -0.60** -0.53** -0.55**

(0.24) (0.25) (0.23) Proxy1 Marginal Effects -0.07** -0.07** -0.05**

(0.03) (0.04) (0.02)

Proxy2 -0.46** -0.39** -0.41** -1.21*** (0.18) (0.18) (0.17) (0.43) Proxy2 Marginal Effects -0.06** -0.05** -0.04** -0.29***

(0.02) (0.03) (0.02) (0.10) 3-Month raw Return 0.52*** 0.52***

(0.17) (0.17)

12-Month raw Return 0.14*** 0.14*** (0.05) (0.05)

Market to book ratio 0.18*** 0.18*** 0.20*** (0.04) (0.04) (0.06) lnSales -0.81*** -0.79*** -0.80*** -0.81*** -0.79*** -0.80*** -0.95*** (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.04) Return on assets -3.31*** -3.84*** -4.34*** -3.34*** -3.86*** -4.37*** -5.74*** (0.56) (0.62) (0.78) (0.56) (0.63) (0.79) (1.31) Cash to assets 1.10*** 0.94*** 0.45 1.11*** 0.94*** 0.46 0.16 (0.27) (0.28) (0.29) (0.27) (0.28) (0.29) (0.44) R&D to sales -0.05*** -0.05*** -0.05*** -0.05*** -0.05*** -0.05*** -0.08*** (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Fixed assets to assets -0.92*** -0.78*** -0.86*** -0.92*** -0.78*** -0.86*** -1.30***

(0.12) (0.12) (0.12) (0.12) (0.12) (0.12) (0.19) Debt to assets 0.92*** 0.77*** 0.83*** 0.92*** 0.77*** 0.83*** 1.90*** (0.17) (0.18) (0.17) (0.17) (0.18) (0.17) (0.28) Intercept 6.08*** 5.83*** 5.80*** 6.07*** 5.82*** 5.79*** 6.82*** (0.18) (0.19) (0.19) (0.18) (0.19) (0.19) (0.33) Pseudo R2 _{0.45 0.46 0.46 0.45 0.46 0.46 0.44} N 9073 8403 9111 9073 8403 9111 3668

Note: Table VI presents coefficients of the logit analysis in which agreement is measured by proxy contests. All columns have one row that presents raw coefficients and another row that presents marginal effects at the means. The sample period is between 2001 and 2016. Two different measures for proxy contests are used: Proxy1 and Proxy2. The dependent variable equals one if the choice of finance is equity, and zero if the choice of finance is debt. Proxy1 is a dummy that equals one if a proxy contest occurs one year before issuance. Proxy2 equals one if a proxy contest occurs one or two year before issuance. Column 7 is an analysis for firms who have less valuation error according the Actual-Forecasted EPS variable. The price variables are measured one quarter before issuance and all other control variables are measured one year before. Standard errors are reported in parentheses. Statistical significance at the 1%, 5%, and 10% level is indicated by ***, **, and *, respectively.

The results in Table IV support predictions made by the hypothesis. It seems that, if agreement is measured by Actual-Forecasted EPS, firms prefer to issue equity if agreement is high, regardless of market-timing factors. The results are strongly significant for the entire time period and for all adjusted periods. Table V presents mixed evidence. Raw coefficients show, similar to Dittmar & Thakor (2007), that firms issue equity as the agreement theory predicts between 1993 and 2002. Note that if marginal effects are examined, which are more important, a weaker effect of EPS Dispersion is observed. Inter-estingly, the years between 2003 and 2016 present results that are the opposite to what the agreement

theory predicts. It indicates that firms are more likely to issue equity if agreement is low. This is in conflict with the hypothesis and the research of Dittmar & Thakor (2007) and Boot & Thakor (2011). Moreover, the entire sample period presents no effect of EPS Dispersion on issuance. As discussed earlier, it is argued that the agreement measure in Table V, EPS Dispersion, is not a good measure for agreement between managers and investors. Literature uses EPS Dispersion as a measure for agreement among analysts (Diether et al., 2002; Banerjee, 2011; Anderson et al., 2005). Perhaps agreement be-tween investors and managers and agreement among analysts is not as highly correlated as claimed by Dittmar & Thakor (2007). Goldstein & Yang (2015) argue that size and diversity of the population of informed traders positively affect price informativeness. The authors assume that informed traders have superior information on varied fundamentals relevant to firm value. The idea is that if an informed trader trades aggressively the quality and quantity of information increases. Consequently, asymmetric infor-mation is lowered by this decrease in uncertainty, plausibly lowering dispersion of analysts’ forecasts. Vives (2014) agree and state that the precision of information about one fundamental affects strategic decisions made by investors who have superior information on other fundamentals. Banerjee (2011) develop a dynamic model which assumes that investors not only trade upon private information but also pay attention to signals of other investors and use this information to condition on the price. The authors find empirical evidence indicating that firms with higher investor disagreement have higher betas, returns and return volatility. In short, the size and diversity of information seem to affect the overall quality and quantity of information available to investors. If so, asymmetric information is lowered by this decrease in uncertainty. It is expected that both Actual-Forecasted EPS and EPS Dispersion are affected by price informativeness since both measures capture information asymmetries. For the abovementioned reasons I cast doubt on the use of EPS Dispersion and Actual-Forecasted EPS as measures for agreement be-tween managers and investors. The mixed results of Table V support the claim that EPS Dispersion is not only a measure for agreement, but for many other things such as asymmetric information. Moreover, Gebhardt et al. (2001) and Kelly & Ljungqvist (2012) use dispersion of analysts forecast as proxy for fundamental cash flow risk and asymmetric information, respectively. Table VI addresses these con-cerns by introducing proxy contests as a new measure for agreement. Table VI supports the hypothesis between 2001 and 2016. It shows that firms prefer to issue equity if agreement is high, regardless of market-timing factors. Additionally, the effect is stronger in magnitude and significance if low valuation errors are present. Overall, Table IV and Table VI support the hypothesis. Evidence backs the claim that agreement plays a role in firms’ issuance behavior. Although EPS Dispersion, used as an agreement measure in Table V, is not able to explain issuance in all (adjusted) time periods the overall evidence indicates that there is at least some (partial) correlation. The next section tests if the relationship between agreement and issuance is causal, rather than just (partial) correlation.

5

The Quasi Experiment

The logit analysis presented in Subsection 4.3, also done by Dittmar & Thakor (2007), cannot offer causal interpretations for several reasons. First, cross-sectional correlations between agreement and fac-tors that affect issuance withhold us from making causal interpretations. For instance, agreement could correlate with unobservable factors that directly influence the decision to issue debt or equity, resulting in a spurious relationship. Second, it could be argued that agreement is affected by issuance. For in-stance, investors could be dissatisfied after an issue. Moreover, besides potentially omitted variables and reverse causality, both measures for agreement employed by Dittmar & Thakor (2007) could be inter-preted as measures for asymmetric information (Kelly & Ljungqvist, 2012). This section describes the identification strategy; an IV approach. Subsection 5.1 elaborates on the Russell 1000/2000 indexes and explains how the yearly index reconstitution event will help to induce an exogenous variation in proxy contests. Subsection 5.2 uses Russell 2000 index inclusion as an instrument for proxy contests, thereby testing if the effect of agreement on issuance is causal.

5.1 The Russell index

The Russel 1000 index consists of the largest 1000 U.S. firms as measured by market cap. The Russell 2000 index includes the 2000 largest U.S. firms after the Russell 1000 firms. Firms on the bottom of the Russell 1000 index are comparable, with respect to market cap, to firms on top of the Russel 2000 index. In 2007, the Russell indexes introduced a banding policy to reduce the quantity of firms that switch indexes. This banding policy implies that firms only switch indexes if the relative change in market cap is sufficiently large. However, before 2007, the cutoff point was simply decided by market cap. Schmidt & Fahlenbrach (2017), Appel et al. (2016b) and Crane et al. (2016) argue that Russell 2000 index inclusion is approximately random for firms around the cutoff. After firms are assigned to a specific index each firm’s weight in the corresponding index is calculated. The index reconstitution is an annual event that occurs every June. Appel et al. (2016b), Chang et al. (2015) and Appel et al. (2016a) show that top firms of the Russell 2000 index have a larger index weight than firms in the bottom of the Russell 1000 index. Specifically, Schmidt & Fahlenbrach (2017) find that a firm’s index weight increases by a factor of 10 if it switches from the Russell 1000 to the Russell 2000 index. Knowingly that market cap decides the cutoff point between both indexes we can finally expose the exogenous link of index inclusion to passive ownership as argued by Appel et al. (2016b) and Schmidt & Fahlenbrach (2017). Passive institutions are required, by mandate, to track an index and are therefore influenced by the composition of that specific index. Consequently, passive institutions increase their ownership if a firm switches from the Russell 1000 to the Russell 2000 index. Russell 2000 index inclusion is a relevant instrument since it facilitates proxy contests through at least two possible channels. First,