Does Corporate Governance Matter More for Firms with High Financial Slack? ∗ Kose John

†

Yuanzhi Li

‡

Jiaren Pang

§

This version: June 30, 2015

Abstract We examine whether and how the effect of corporate governance depends on a firm’s financial slack, financial resources not committed to any specific use. On one hand, financial slack may be spent by self-interested managers for their private benefits, so its level is positively associated with the degree of agency conflicts. This implies that corporate governance matters more for high financial slack firms (wasteful spending hypothesis). On the other hand, financial slack provides insurance against future uncertainties; a low level of financial slack may signal that managers are imprudent and engage in excess risk taking. Then corporate governance is more effective for low financial slack firms (precautionary needs hypothesis). We differentiate the two hypotheses using the passage of anti-takeover laws to identify exogenous variation in governance. Consistent with the wasteful spending hypothesis, we find that the laws’ passage has a larger negative impact on the operating and stock market performance of high financial slack firms. Further analysis of the source of wasteful spending shows that these firms do not invest more but are less efficient at cost management than low financial slack firms after the passage of BC laws. Our findings suggest that shareholder activism and government regulations aiming to improve corporate governance can be more efficient by focusing on firms with high financial slack. Keywords: corporate governance, financial slack, business combination laws JEL Codes: G34 G38

∗

We are grateful to Xavier Giroud, Kun Huang, David Reebs, David Yermack, Feng Zhang, seminar participants at Temple University, University of New Orleans, and Tulane University, and conference participants at the 2011 Chinese International Conference in Finance for their helpful comments. We especially thank Martijn Cremers for sharing G-index data of the 1977-89 period, and David Yermack for sharing various governance data of Fortune 500 firms from 1984 to 1991. All errors are our own. † Kose John is from the Department of Finance, Stern School of Business, New York University. E-mail: [email protected]. Phone: 212-998-0337. ‡ Yuanzhi Li is from the Department of Finance, Fox School of Business, Temple University. E-mail: [email protected]. Phone: 215-204-8108. § Jiaren Pang is from the Department of Finance, School of Economics and Management, Tsinghua University. E-mail: [email protected]. Phone: (+86)10-6279-4800.

Does Corporate Governance Matter More for Firms with High Financial Slack?

Abstract We examine whether and how the effect of corporate governance depends on a firm’s financial slack, financial resources not committed to any specific use. On one hand, financial slack may be spent by self-interested managers for their private benefits, so its level is positively associated with the degree of agency conflicts. This implies that corporate governance matters more for high financial slack firms (wasteful spending hypothesis). On the other hand, financial slack provides insurance against future uncertainties; a low level of financial slack may signal that managers are imprudent and engage in excess risk taking. Then corporate governance is more effective for low financial slack firms (precautionary needs hypothesis). We differentiate the two hypotheses using the passage of anti-takeover laws to identify exogenous variation in governance. Consistent with the wasteful spending hypothesis, we find that the laws’ passage has a larger negative impact on the operating and stock market performance of high financial slack firms. Further analysis of the source of wasteful spending shows that these firms do not invest more but are less efficient at cost management than low financial slack firms after the passage of BC laws. Our findings suggest that shareholder activism and government regulations aiming to improve corporate governance can be more efficient by focusing on firms with high financial slack. Keywords: corporate governance, financial slack, business combination laws JEL Codes: G34 G38

1

Introduction Is corporate governance equally important for the performance of all firms? If not, for which

type of firms does it matter more? These are not trivial questions because they can improve our understanding of corporate governance and help us design policies to promote better governance practices and protect shareholder wealth. To answer these questions, this paper attempts to examine whether corporate governance interacts with certain firm characteristics and exerts differential impacts on firm performance. While firms are different in many dimensions, we specifically focus on financial slack, financial resources not committed to any specific use, for two reasons. First, liquid assets account for a significant fraction of total corporate wealth. Bates, Kahle, and Stulz (2009) report that the average cash-to-assets ratio of U.S. firms has steadily increased and reached 23.2% in 2006. Second, and more importantly, the use of financial slack is largely at the discretion of managers, and is considered a central issue among the conflicts between managers and shareholders (Jensen, 1986). Harford, Mansi, and Maxwell (2008) also suggest that “any discussion of the efficacy of corporate governance mechanisms to control managers must address this issue.” Theoretically, corporate governance should matter more for firms with more severe agency conflicts as its main goal is to mitigate agency problems. However, it is unclear whether and how the effect of corporate governance depends on financial slack. On one hand, Jensen’s free cash flow theory points out that financial slack is not subject to the same scrutiny and monitoring by the capital markets as externally raised funds, and self-interested managers are likely to spend these excess funds for their private benefits at the expense of shareholders. Because of the potential wasteful spending, the level of financial slack is positively related to the degree of agency conflicts. This implies that, everything else being equal, corporate governance has a greater impact on firms with high financial slack. We call this the wasteful spending hypothesis. On the other hand, a low level of financial slack may also signal severe agency problems. Due to financial market imperfections, external financing is generally more costly than internal financing (Myers and Majluf, 1984). Hence financial slack is beneficial to firms as it provides insurance against future uncertainties. This is consistent with the precautionary motive for holding cash

1

proposed by Keynes (1936) and elaborated by Opler, Pinkowitz, Stulz, and Williamson (1999), Lins, Servaes, and Tufano (2010), and Disatnik, Duchin, and Schmidt (2014), among others. In this view, low financial slack may be an indication of insufficient financial reserves for future uncertainties and may imply that managers are imprudent and engage in excess risk taking. If that is the case, the level of financial slack would be negatively associated with the degree of agency conflicts, and corporate governance should have a greater impact on firms with low financial slack. This prediction is in contrast to the wasteful spending hypothesis, and we name it the precautionary needs hypothesis. To test the two hypotheses, we follow the literature and use the passage of business combination laws (henceforth, BC laws) in different states from 1985 to 1991 to identify exogenous changes in governance.1 These laws impose stringent restrictions on hostile takeovers of firms in the legislating state, thus reduce the disciplinary role of capital markets on managers and weaken corporate governance. They provide a natural experiment to study the effect of corporate governance as they were exogenous to most firms and were passed in a staggered manner in different states. Our main proxy for financial slack is excess cash, which is the difference between actual cash holdings and the predicted amount of cash for future liquidity and investment needs calculated from a regression as in Dittmar and Duchin (2011), Bates et al. (2009), and Opler et al. (1999). Using a triple-difference approach, we find differential impacts of BC laws on firms with different levels of financial slack. On average, firms’ return on assets (ROA) drops 0.5% after the passage of BC laws, and a one standard deviation increase in excess cash reduces ROA further by 0.53%. We also examine the impact of the laws’ passage on firms’ stock market performance and find similar results. Firms with more financial slack experience a larger decline in stock prices after the laws’ passage. This also implies that at the time of the laws’ passage, the stock 1

The passage of anti-takeover laws is well studied in the literature. Early papers conduct event studies surrounding the news release date of the laws’ passage (Pound, 1987; Schumann, 1988; Ryngaert and Netter, 1988; Romano, 1987; Margotta, McWilliams, and McWilliams, 1990; Karpoff and Malatesta, 1989). Bertrand and Mullainathan (2003) investigate plant-level data before and after the laws’ passage, and find increases in worker wages, decreases in destruction of old plants and creation of new plants, and decreases in productivity and profitability. They conclude that managers choose to enjoy a quiet life rather than engage in empire building after the passage of anti-takeover laws.

2

market does not completely realize its full impact. Both results suggest that the weakening of corporate governance leads to a larger decline in performance of firms with higher financial slack, which is consistent with the wasteful spending hypothesis, but not with the precautionary needs hypothesis. Karpoff and Wittry (2014) point out several issues that may lead to biased inference when using BC laws to identify exogenous changes in governance. Specifically, the coverage of BC laws is likely endogenous to lobbying firms and firms that used the opt-out or opt-in provisions of some BC laws to adjust their status. Excluding these firms from the sample does not change our results. The effect of BC laws may also depend on important court decisions that validated the constitutionality of these laws, other anti-takeover laws, and firm-level anti-takeover defenses. We control for these factors and continue to find similar results. One might be concerned that a firm’s financial slack is endogenous. We address this endogeneity concern in three ways. First, we directly control for a number of firm-level governance measures and other characteristics that may cause omitted variables bias. Second, we use a sticky measure of financial slack that is not affected by the laws’ passage. Third, we perform 2SLS estimation using two instruments for financial slack. One is the average of other firms’ financial slack in the same industry incorporated in states that had not passed BC laws; the other is the average of sticky financial slack of other firms in the same state but in different industries. We continue to find supporting evidence for the wasteful spending hypothesis. Our results also hold with various subsample analysis as robustness checks. (1) Our results could be driven by reverse causality. Some firms might expect a decline in profitability, and they lobbied to have the state pass BC laws to protect themselves. Thus the decline in performance is the cause rather than the result of the laws’ passage. Even though our results hold when we drop the list of lobbying firms, it is possible that some lobbying activities are not reported and such firms are not on the publicized list. As managers in larger firms have stronger incentives and more resources to engage in lobbying activities, we repeat the analysis excluding large firms. (2) To ensure that the results are not specific to the sample period, we repeat the analysis for

3

different sample periods. (3) Since one half of the sample firms are incorporated in Delaware, our results could be driven by a Delaware effect. We exclude firms incorporated in Delaware from the treatment group to address this concern. (4) Some states never passed BC laws in the sample period, so firms incorporated in these states could be fundamentally different from the rest. We repeat the analysis excluding those firms from the control group. (5) The entry of new firms and exit of old firms in the sample period may bias our results, so we analyze a sub-sample of firms with data available in all sample years. In all robustness checks, we continue to find that the passage of BC laws exerts a larger negative impact on high financial slack firms. Finally, we investigate the source of the larger decline in performance for high financial slack firms. Motivated by Bertrand and Mullainathan (2003), we focus on their investment and cost management. Due to agency conflicts, both types of activities of high financial slack firms may be associated with wasteful spending of financial slack that brings private benefits to managers. The problem can become relatively more severe for these firms after the laws’ passage, and explains their larger decline in performance. We find that these firms do not increase their capital expenditure, over-investment, asset growth, PPE growth, or acquisition ratio more than firms with low financial slack after the laws’ passage, but have higher overhead costs, operating expenses, costs of goods sold, and more employees relative to sales. These findings suggest that the larger decline in performance for high financial slack firms is likely due to their managers not maintaining cost efficiency. Our study echoes the burgeoning literature on shareholder activism. Brav, Jiang, Partnoy, and Thomas (2008) show that hedge fund activism tends to target firms with lower growth and higher cash flows, and the activism leads to increases in payout, operating performance, and higher CEO turnover of the target firms. Our results reinforce their findings and suggest that policies aiming to improve corporate governance will be more effective by focusing on firms with high financial slack. We also contribute to the literature that examines the conditional effect of corporate governance on firm performance. In a similar setting, Giroud and Mueller (2010) show that corporate governance matters more for firms in noncompetitive industries, because the agency problem of firms in competitive industries is already mitigated by competition. Duchin,

4

Matsusaka, and Ozbas (2010) find that the effectiveness of outside directors is conditional on the cost of acquiring information about the firm: when the cost of acquiring information is low (high), performance improves (worsens) when outsiders are added to the board. We complement this line of research by suggesting that the importance of corporate governance varies with the level of financial slack. Our paper is closely related to several recent papers that study the differential value of cash conditional on governance. Dittmar and Mahrt-Smith (2007) find that cash is more valuable for well-governed firms. Fr´esard and Salva (2010) find that the value that investors attach to excess cash is substantially larger for foreign firms listed on US exchanges than for their domestic peers. Kalcheva and Lins (2007) study international data and find that when external country-level shareholder protection is weak, firm values are lower when controlling managers hold more cash. We distinguish our paper from these studies in two ways. First, we focus on the conditional nature of corporate governance. Second, most governance measures used in these studies suffer from endogeneity, while we identify exogenous variation of corporate governance by using the BC laws’ passage. The rest of the paper is organized as follows. Section 2 provides background knowledge regarding the passage of state anti-takeover laws. Section 3 describes the empirical methodology and the data. Section 4 discusses the main findings and various robustness checks. Section 5 studies the impact of the laws’ passage on stock market performance for firms with different levels of financial slack. Section 6 investigates the source of the larger impact of BC laws on high financial slack firms. Section 7 concludes and discusses the policy implications of our findings.

2

State anti-takeover laws There are two generations of state anti-takeover laws. In the 1970s, the first generation

state anti-takeover laws were passed by extending the Williams Act, a federal statute enacted in 1968 that regulates tender offers. These laws based their jurisdiction over tender offers on the

5

relationship between the target and the legislating state, which is determined by a number of factors, such as the target’s state of incorporation, its principal place of business, and where it holds substantial assets. However, the early laws were invalidated by a 1982 U.S. Supreme Court decision (Edgar v. Mite Corp.) on the grounds of excessive jurisdictional reach.2 In response to this decision, states began to pass a second wave of anti-takeover legislation, which was less aggressive and restricted the jurisdiction to only firms incorporated in the legislating state. The Supreme Court upheld an Indiana state control share acquisition law in 1987 (CTS Corp. v. Dynamics Corp. of America), and the U.S. court of Appeals, Seventh Circuit upheld a Wisconsin BC law in 1989 (Amanda Acquisition Corp. v. Universal Foods Corp.).3 The rulings generated the presumption that other anti-takeover laws are also valid and stimulated further enactments of such legislations across the country.4 Most of the second generation anti-takeover statutes can be classified into business combination (BC laws), fair price, control share acquisition, poison pill, and constituency laws. BC laws impose a moratorium on certain kinds of transactions (e.g., mergers and asset sales) between a bidder and the target for a period of three to five years after the stake of the bidder has reached a threshold level. These statutes make it more costly for successful bidders to realize gains from a takeover, hence discouraging potential buyers from bidding. Fair price laws require a bidder, when acquiring shares beyond a pre-specified threshold, to pay a “fair price”, which is usually determined by share prices prior to the takeover announcement. Control share acquisition laws require a bidder intending to make a “control share acquisition”, defined by several threshold levels, to present its offer to the target’s shareholders. If the bidder fails to comply and purchases a large block of shares, it may be disqualified from voting with these shares and will not be able to gain control until its voting rights are reinstated. Poison pill laws allow a firm to grant current shareholders the rights to buy stocks at a low price when a bidder acquires a significant amount of shares without the approval of the board. This can 2

See 457 U.S. 624 (1982). See 481 U.S. 69 (1987) and 877 F.2d 496 (1989), respectively. 4 Some scholars name the laws enacted after the Supreme Court’s decision in (CTS Corp. v. Dynamics Corp. of America) as the third generation laws, but this paper refers to all state anti-takeover laws passed after 1982 as the second generation laws. 3

6

substantially increase takeover costs and deter potential bidders. Constituency laws allow the board of directors to take into account the interests of constituencies besides shareholders, such as employees, customers, and suppliers, when the board decides how to respond to a takeover bid. They are considered to be anti-takeover statutes because they provide managers with additional resources to oppose the takeover. In order to be comparable to previous studies, we follow the literature and mainly focus on BC laws in our analysis. In robustness checks, we also control for the influence of other anti-takeover laws. The passage of BC laws weakens corporate governance by reducing the disciplinary role of capital markets. It provides an ideal setting to examine the conditional nature of corporate governance. First, the laws’ passage can be viewed as an exogenous event to most treated firms; it is endogenous to only a small fraction of the firms, and the endogeneity concern is addressed in later sections. Second, the laws were passed at different points in time for different states. It reduces the clustering of observations at the time of the laws’ passage. Lastly, these laws were passed on a state basis, so they induced a common change of corporate governance in all affected firms. Holding constant the change of corporate governance in treated firms enables us to take firms with different levels of financial slack and compare their changes in performance before and after the laws’ passage. While the BC laws’ passage has been used extensively to identify exogenous variation in corporate governance, Karpoff and Whittry (2014) point out several important issues that may bias the estimation. First, the laws’ coverage may be endogenous to two small groups of firms. One group is those firms that faced takeover threats and lobbied legislators to pass the laws. The other group consists of those firms that opted out of or into the coverage of the laws. Specifically, many anti-takeover laws, including some BC laws, have opt-out provisions that allow affected firms to opt out of the laws. Meanwhile, the BC law of Georgia requests firms to opt into the law. Second, the BC laws were not officially declared constitutional until the court ruling on Amanda Acquisition Corp. v. Universal Foods Corp. in 1989, so they may have differential effects before and after the ruling. Third, the effect of the BC laws may depend on the coverage by first-generation and other second-generation anti-takeover laws. Finally, the results may be

7

confounded by existing firm-level anti-takeover provisions. We address all these issues in our empirical analysis.

3

Empirical methodology and Data

3.1

Empirical methodology

Our research question is how the passage of BC laws affects the performance of firms with varying levels of financial slack. The main regression equation is as follows: ROAijkl,t = αi + αt + β1 BC lawijkl,t + β2 (BC lawijkl,t × F Sijkl,t−1 ) + β3 F Sijkl,t−1 + γ 0 Xijkl,t + εijkl,t ,

(1)

where i, j, k, l, and t index firms, industries, states of incorporation, states of location, and years, respectively. ROAijkl,t is the return on assets. αi and αt are firm and year fixed effects, respectively. BC lawijkl,t is equal to one if firm i is subject to the BC law in year t, and zero otherwise. F Sijkl,t−1 is the financial slack of firm i measured in year t − 1. Xijkl,t is a vector of control variables, and εijkl,t is the error term. In the baseline specification, Xijkl,t includes firm size, firm age, squared terms of size and age, and two proxies for time-varying local and industry shocks, state year and industry year. We follow Bertrand and Mullainathan (2003) and define state year and industry year as the annual mean of the dependent variable in the firm’s state of location and three-digit SIC industry, respectively, excluding the firm itself. Including the two controls enables us to separate the effect of the laws’ passage from other contemporaneous shocks in the state of location and the industry. It also helps address the concern that a coalition of firms located in the same state or operating in the same industry lobbied for an anti-takeover law to gain better protection against hostile takeovers when they expect a decline in profitability. The marginal effect of BC laws on performance is given by β1 + β2 × F S. We are most

8

interested in whether and how it varies with the level of financial slack, which is captured by the coefficient of the interaction term, β2 . The wasteful spending hypothesis predicts β2 < 0 and the precautionary needs hypothesis predicts β2 > 0. In all regressions, standard errors are adjusted for heteroskedasticity and clustered at the state of incorporation level. Clustering the standard errors at the state of incorporation level accounts for three types of correlations among the error terms: across different firms in the same state of incorporation and year (cross-sectional correlation), across different firms in the same state of incorporation over time (across-firm serial correlation), and within the same firm over time (within-firm serial correlation). Cross-sectional correlation is likely since firms in the same state of incorporation are subject to the same shock in corporate governance due to the passage of BC laws. Serial correlation is a concern since the dummy variable of the laws’ passage is persistent over time. The regression specification is essentially a difference-in-differences-in-differences approach. The first level of difference is the performance difference of firms before and after the laws’ passage. The second level is the difference of the first-level difference among firms across incorporating states with and without BC laws. The first two differences can identify the average treatment effect of the laws’ passage on firm performance. The interaction term BC law × F S captures the third-level difference. It is the difference in the second-level differences among treated firms with different levels of financial slack. For example, suppose we want to estimate the differential impact of the New York BC law passed in 1985 on the performance of firms incorporated in New York with different levels of financial slack. First, we would compare the performance before and after 1985 for high financial slack firms incorporated in New York. The performance difference could reflect the effect of the law, but could also be related to other shocks in the economy, such as an unexpected increase in oil price. To control for the impact of other contemporaneous shocks, we select a control state that had not passed the law by 1985, such as California. We compare firm performance before and after 1985 for high financial slack firms incorporated in California. Since firms incorporated in California are subject to the same economic shocks, but not to the passage of the law in New York, the difference of the two identifies the average effect of the law on high

9

financial slack firms. Then we repeat the same process for low financial slack firms incorporated in New York and California. The difference of the final two differences would reflect the differential impact of the law’s passage on high and low financial slack firms in New York.

3.2

Data and Variables

The exact years of the BC laws’ passage are obtained from Karpoff and Wittry (2014).5 They also point out that not all firms incorporated in a state that passed an anti-takeover law were affected by the law. Specifically, many anti-takeover laws, including some BC laws, have opt-out provisions. For those firms that opted out of the laws’ coverage, the corresponding observations are coded as not subject to the laws, i.e., the dummy variable BC law is equal to zero. Meanwhile, the Georgia BC law requires firms to opt into coverage. Therefore, firms incorporated in Georgia are considered as not subject to the BC law unless they opted into coverage. The data on firms’ opt-out and opt-in decisions are taken from RiskMetrics. We collect accounting data of all publicly listed US firms from Compustat. We drop observations with missing values on total assets, sales, or operating income before depreciation. We also drop observations that have no data on any of the financial slack proxies discussed below. All financial firms are excluded because their cash reserves and cash flows may have different interpretations, and utility firms are dropped because they are highly regulated. To make our analysis comparable to the literature (Bertrand and Mullainathan, 2003; Giroud and Mueller, 2010), we choose the sample period of 1976 to 1995.6 The final sample contains 8,025 firms and 68,008 firm-year observations. Our main variable is financial slack, which is the extra financial resource not committed to 5

A total of 31 states passed BC laws during our sample period of 1976 to 1995. Specifically, New York passed the law in 1985; Indiana, Kentucky, Missouri, and New Jersey passed the law in 1986; Arizona, Minnesota, Washington, and Wisconsin passed the law in 1987; Connecticut, Delaware, Georgia, Idaho, Maine, Nebraska, Pennsylvania, South Carolina, Tennessee, and Virginia passed the law in 1988; Illinois, Kansas, Maryland, Massachusetts, Michigan, and Wyoming passed the law in 1989; Ohio, Rhode Island, and South Dakota passed the law in 1990; Nevada and Oklahoma passed the law in 1991. 6 In robustness checks, we use different sample periods to ensure that our results are not driven by the specific time period of the sample.

10

future specific use such as liquidity and investment needs. Since the exact amount needed for future specific use is not directly observable, we start by using two gross proxies of a firm’s overall financial resource. The first proxy is the current ratio, also known as the liquidity ratio. It is defined as the ratio of current assets to current liabilities, and is a popular measure for financial slack in management literature.7 The second proxy is cash ratio, defined as the ratio of cash and short-term investments to total assets. Practitioners in accounting and finance normally use this ratio as one of the measures for financial slack. Both variables are noisy proxies for financial slack as they do not take into account the portion reserved for future specific needs. Our main proxy for financial slack throughout the analysis is excess cash, which is the difference between actual cash holdings and the amount of cash committed to future specific needs, normalized by total assets. It is also referred to as “unexpected cash” (Dittmar and Duchin, 2011) and “cash residual” (Harford et al., 2008). Specifically, it is calculated as follows,

Excess cashi,t = Cash ratioi,t − Cash ratio∗i,t ,

(2)

where Cash ratio∗i,t represents firm i’s expected needs for cash in year t, which is unobservable and must be estimated. Following Dittmar and Duchin (2011), we first estimate an empirical cash model similar to the one in Opler et al. (1999) and Bates et al. (2009) over a rolling five-year window [t − 5, t − 1]. The dependent variable is cash ratio, and the explanatory variables include lagged cash flow, cash flow volatility, Tobin’s Q, firm size, net working capital, leverage, capital expenditure, R&D expenditure, and the dividend payout dummy variable (variable definitions are given below). We then use the estimated model to obtain the predicted value of cash holdings of year t, Cash ratio∗i,t . As the regression explicitly controls for future specific financial needs, it is the cleanest among all three financial slack proxies. We construct the rest of the variables as follows. ROA is operating income before depreciation divided by total assets. Firm size is the log of total assets. Firm age is the log of one plus the number of years between the first year the firm is covered in Compustat and the current year. 7

See Daniel, Lohrke, Fornaciari, and Turner Jr (2004) for a review.

11

Cash flow is earnings after interests, dividends, and taxes but before depreciation divided by total assets. Cash flow volatility is the standard deviation of cash flow in the previous five years. Tobin’s Q is the ratio of the market value of assets to the book value of assets, where the market value of assets is equal to the book value of assets, plus the market value of common equity, minus the sum of the book value of common equity and deferred taxes. Net working capital is net working capital excluding cash divided by total assets. Leverage is the sum of long-term and short-term debt divided by total assets. Capital expenditure is the ratio of capital expenditure to total assets. R&D expenditure is R&D divided by sales, and is set equal to zero if R&D is missing. The dividend payout dummy equals one in years when a firm pays a common dividend, and zero otherwise. All continuous variables are winsorzied at 1% and 99% levels to reduce the influence of outliers. Panel A of Table 1 provides summary statistics of the main variables used in Equation (1). The average ROA of our sample is 8.9%. The averages of current ratio, cash ratio, and excess cash are 1.176, 0.131, and 0.001, respectively. It is not surprising that the average of excess cash is close to zero as it is similar to a regression residual by construction. Panel B presents the correlations among main variables. The three financial slack proxies are highly correlated with each other, and all three have a slightly negative correlation with ROA. They are also negatively correlated with firm size and age, suggesting that larger and older firms hold less liquid assets.

4

Results

4.1

Baseline regression results

Table 2 presents our baseline regression results. Column (1) does not include financial slack or its interaction with BC laws’ passage, while other columns include both. Without the interaction term, the regression imposes the restriction that the laws’ passage affects all firms equally and the coefficient β1 shows the average effect of the laws’ passage on the performance of all treated firms. The coefficient on the law dummy is -0.005, suggesting that ROA drops 0.5% after the

12

laws’ passage. This is consistent with the finding of Bertrand and Mullainathan (2003) that on average the passage of BC laws hurts firm performance. In columns (2) to (4), a proxy for financial slack and its interaction term with the laws’ passage are included, and the proxies are current ratio, cash ratio, and excess cash, respectively. The interaction term captures the differential impact of the laws’ passage conditional on the level of financial slack. From column (2) to column (4), we consistently find that the coefficient of the interaction term is significantly negative. This implies that the negative impact of the laws’ passage on performance is more pronounced for those firms with more financial slack, which is consistent with the wasteful spending hypothesis, but not with the precautionary needs hypothesis. The differential effect is also economically meaningful. Take the regression of Column (4) that uses excess cash as the proxy for financial slack as an example. According to the coefficient estimates, the marginal effect of BC laws is −0.002 − 0.056 × Excess cash. It implies that an increase in excess cash by one standard deviation (0.106) is associated with a relative drop of 0.56 percentage points in ROA after the laws’ passage, which is about 6% of the average ROA of our sample. In columns (2) and (3), the coefficient estimates on the stand-alone current ratio and cash ratio are significantly negative. In column (3), the coefficient on the excess cash is negative but statistically insignificant. Combined with the negative coefficients on the interaction terms, these results imply that the average effect of financial slack on firm performance is negative. This is also consistent with the wasteful spending hypothesis, but not with the precautionary needs hypothesis. The signs of coefficient estimates on firm size, firm age, and their squared terms are as expected. As a firm becomes larger or older, its ROA increases first, and then decreases, which is an inverted U-shaped relationship. We also obtain positive and highly significant coefficient estimates on the two proxies for local and industry shocks, which confirms the necessity to control for them.

13

4.2

Using BC laws’ passage for identification

As mentioned in Section 2, there are several important issues that may generate biased inference when using BC laws to identify exogenous variation in corporate governance (Karpoff and Wittry, 2014). First, while it is exogenous to most firms, the laws’ coverage is likely endogenous to two small groups of firms. One group includes those firms lobbying for specific state anti-takeover laws. In most cases, the lobbying firm was the target of an actual or rumored takeover bid, and the state passed anti-takeover legislation quickly upon such news. It is obvious that there exists the possibility of reverse causality for this group of firms. Karpoff and Wittry (2014) identify nearly 30 lobbying firms that motivated BC laws. The small quantity of firms involved in lobbying activities is consistent with Romano (1987), who concludes that BC laws are unlikely caused by broad-based lobbying. We exclude the lobbying firms from the analysis to directly address the endogeneity concern. The results remain qualitatively unchanged and are presented in column (1) of Panel A, Table 3. Endogeneity is also a concern for the group of firms that exercised the opt-out or opt-in options offered by some BC laws because whether they are covered by the laws reflects their endogenous choice. In our sample, 26 firms opted out of the coverage of their states’ BC laws, and three firms incorporated in Georgia opted into coverage. The small quantity is probably due to the significant adjustment costs. For example, opting out of the Ohio BC law requires the approval of at least two thirds of the outstanding shares and two thirds of the outstanding shares not owned by a 10% stockholder. Moreover, the opt-out would be ineffective for 12 months and would not apply to a control transaction of a shareholder with more than 10% shares before the approval of the opt-out amendment. Therefore, the laws’ coverage is still exogenous to most firms because the high transaction costs prevent them from making adjustment (Karpoff and Wittry, 2014). Nevertheless, we drop firms that opted out of or opted into coverage as well as lobbying firms, and continue to find similar results, as shown in column (2) of Panel A, Table 3. Second, the impact of anti-takeover laws may depend on the legal environment. The consti-

14

tutionality of BC laws was not established until a ruling by the U.S. Court of Appeals, Seventh Circuit in Amanda Acquisition Corp. v. Universal Foods Corp. on May 24, 1989. Before the court ruling, it was uncertain how much protection against takeover bids that BC laws could offer. Thus it is possible that the impact of BC laws became meaningful only after the ruling. To differentiate the effects before and after the ruling, we replace the law dummy and the interaction term in the baseline regression with two dummy variables, BC law before ruling and BC law after ruling, and their interactions with excess cash in column (3) of Panel A, Table 3. BC law before ruling is the same as the dummy BC law for any year t < 1989, and is equal to zero for t ≥ 1989. Similarly, BC law after ruling is the same as BC law for any year t ≥ 1989, and is equal to zero for t < 1989. By construction, the two dummies capture the effect of BC laws before and after the ruling, respectively. The coefficients of the two interaction terms BC law bef ore ruling × Excess cash and BC law af ter ruling × Excess cash are -0.045 and -0.077, respectively, and both are significant. The results have three implications. First, both before and after the ruling, the negative impact of BC laws is stronger for firms with more excess cash, which is consistent with our previous finding. Second, the size of the two coefficients suggests that the conditional effect of BC laws is indeed larger in magnitude after the ruling. Finally, in the baseline regression the coefficient of the interaction term BC law × Excess cash is -0.053 in Table 2, somewhat between these two coefficients. Thus the baseline regression can be considered as an estimation of the average effect of the interaction term over the whole sample period, since it does not take into account the court ruling. In column (4), we repeat the analysis excluding lobbying firms and firms that opted out of or opted into the laws’ coverage, and the results are similar. Third, the effect of BC laws may be influenced by four other types of state anti-takeover laws passed during the sample period, which are fair price, control share acquisition, poison pill, and constituency laws. To control for their influence, we construct four dummy variables: F air price, Control share, P oison pill, and Constituency. They are equal to one if the corresponding laws are effective and zero otherwise. In columns (1) to (4) of Panel B, Table 3, we separately add one of the four dummies and its interaction with excess cash into the baseline specification. Column

15

(5) includes all of them. When added separately, each new interaction term enters the regression negatively, and both F air price × Excess cash and P oison pill × Excess cash are statistically significant. When added together, only P oison pill × Excess cash is significant among new interaction terms. This suggests that some of the other anti-takeover laws also exert a negative impact on firm performance conditional on financial slack. Admittedly, including these other laws reduces the magnitude of the coefficient estimate on the interaction term BC law × Excess cash, but it remains significant in all specifications.8 A related concern is that some of the first generation anti-takeover laws were also effective in the early years of the sample period, and part of the BC laws’ effect we observe could come from these laws. To address this concern, we re-run the baseline regression excluding observations during the 1976-1982 period. Because the first generation anti-takeover laws were deemed unconstitutional in 1982, the sample period after 1982 is free from their impact.9 With the shorter sample period of 1983-1995, column (6) presents the results of the baseline regression, and column (7) controls for all other second generation anti-takeover laws and their interactions with excess cash. The coefficients on BC law × Excess cash remain significant in both regressions. We conclude that our main findings are not driven by the first generation or other second generation anti-takeover laws.

4.3

Endogeneity of financial slack

Financial slack, as a part of a firm’s liquidity management, is likely to be endogenous. In particular, both firm performance and financial slack can be jointly determined at equilibrium, and are both driven by other firm characteristics. There is omitted variables bias if these characteristics are not accounted for. In the earlier baseline specification, we address this concern with two 8

Like the BC law dummy, the dummy variables for these anti-takeover laws are appropriately coded to account for the opt-out and opt-in decisions of firms. We also control for the court rulings that upheld these laws and find similar results. 9 An alternative solution is to control for these first generation laws explicitly. However, it is difficult to find out whether a firm is under the jurisdiction of these laws because it is determined by a number of factors, such as the firm’s state of incorporation, its principal place of business, or where it holds substantial assets.

16

considerations. First, we use excess cash as the main proxy for financial slack, which is the difference between actual and expected cash holdings. When estimating expected cash holdings, we control for a number of firm characteristics that capture a firm’s specific needs for cash, so excess cash is unlikely a response to changes in these firm characteristics. Perhaps this is why it is called “unexpected cash” in Dittmar and Duchin (2011). Second, we include firm fixed effects in all regressions to control for unobserved time-invariant firm characteristics. Nevertheless, excess cash can still be correlated with some other time-varying firm characteristics that are not included in the regression of estimating expected cash holdings. Our results are biased if they are correlated with firm performance and are not taken into account. In this section, we tackle the problem in three ways. First, we control for various firm-level governance measures and other characteristics that may cause the omitted variables bias. Second, we construct a sticky measure of financial slack that is not affected by the laws’ passage. This sticky measure can address the concern that the passage of BC laws induces a change in financial slack that is correlated with time-varying omitted variables. Third, we construct two instrumental variables to identify exogenous variation in financial slack and conduct 2SLS estimation.

4.3.1

Firm-level characteristics

Firm-level governance measures Some firms adopt anti-takeover provisions to protect themselves from takeover threats. If these provisions affect firm performance and are correlated with financial slack, the regression results will be biased. To address this concern, we control for a firm’s takeover defenses using the G-index proposed by Gompers, Ishii, and Metrick (2003). Martijn Cremers graciously shared with us their hand-collected G-index data for the period of 1977-1989, as described in Cremers and Ferrell (2014). We then complement their data with the G-index data of 1990-1995 provided by RiskMetrics.10 We control for G-index as a stand-alone variable and interact it with the law 10

The data in Cremers and Ferrell (2014) covers 12,366 firm-year observations of 1,297 firms, and 8465 observations have data on excess cash. With the additional G-index from RiskMetrics, the final sample contains 11650 observations.

17

dummy in the regression, and present the results in column (1) of Panel A, Table 4. We find that firms with higher G-index on average indeed perform worse than their peers, but there is no differential impact of BC laws conditional on G-index. Meanwhile, the interaction term of BC law and financial slack continues to be negative and significant. A related concern is that other governance measures may also be related to financial slack and affect firm performance. We use data on CEO ownership, CEO compensation structure, CEO duality, board size, and board independence to control for these effects. CEO ownership is the percentage of a firm’s equity owned by its CEO. CEO compensation structure is measured by CEO equity pay, which is one minus the percentage of cash pay including salary and bonus. CEO duality is a dummy variable that is one if the CEO is also the chairperson of the board, and zero otherwise. Board size is the log of the number of board directors. Board independence is the percentage of directors that are classified as outsiders. David Yermack kindly provided us with his hand-collected data on these governance variables for the largest 500 companies ranked by Forbes during 1984-1991, as described in Yermack (1996). For the period of 1992-1995, We obtain CEO compensation data from ExecuComp and ownership and board characteristics from Compact Disclosure. To be consistent, we only include past and current Forbes 500 companies. After deleting observations with missing values on excess cash, the final sample contains less than 5,000 firm-year observations. To avoid multicollinearity, we separately add these governance variables and their interactions with the law dummy into the regression, and present the results in columns (2) to (6) of Panel A, Table 4.While CEO duality and board size are negatively associated with firm performance, we find no evidence that the effect of BC laws on firms performance is conditional on these governance measures. Meanwhile, the interaction term BC law×Excess cash is significant at the 10% level in all regressions except the one with board size (t statistics = 1.63 and p value = 0.103). The decrease in statistical significance of these regressions is likely due to the much smaller sample size, which is about one tenth of that of the baseline regression.11 Other firm characteristics 11

We also perform additional robustness checks. Our results continue to hold when we control for ownership nonlinearity by including the squared term of ownership, replace CEO ownership with ownership of all officers and directors, and control for other CEO characteristics such as age and tenure.

18

Giroud and Mueller (2010) show that the passage of BC laws has a larger negative impact on the operating performance of firms in less competitive industries. Given that firms in less competitive industries are more likely to enjoy higher profits and have more financial slack, it is possible that our results just manifest the findings in Giroud and Mueller (2010). To control for the effect of competition on the importance of corporate governance, we include the industry Herfindahl-Hirschman Index (HHI) in the regression analysis. The industry HHI is defined as the sum of squared market shares of firms with the same three-digit SIC code, where the market shares are calculated based on sales. Column (1) of Panel B, Table 4 adds the HHI and its interaction with the law dummy variable to the baseline specification. Consistent with the findings of Giroud and Mueller (2010), the coefficient estimate on the interaction term BC law × HHI is negative and significant. Meanwhile, we continue to find a significant coefficient estimate on BC law × Excess cash. A firm’s financial slack may be related to its investment opportunities. If a firm’s investment opportunities also affect the importance of corporate governance, our findings could be driven by the difference in investment opportunities rather than in agency costs. To address this concern, we control for lagged Tobin’s Q and its interaction with the law dummy variable. The results are shown in column (2) of Panel B, Table 4. While Tobin’s Q is significant, its interaction term with the law dummy variable is not. More importantly, controlling for Tobin’s Q does not change our main findings. Another issue is that the accounting performance measure ROA is not adjusted by risk. Thus a decline in ROA may result from lower asset risk. Before the passage of BC laws, firms may invest excessively in highly risky projects to appear more profitable to deter hostile bidders. After the laws’ passage, this incentive is weakened, and firms may choose to switch to less risky projects which command lower returns. Meanwhile firms with abundant financial slack are more likely to be the targets of hostile takeovers, which implies that they have stronger incentives to take risky projects than low financial slack firms before the laws’ passage. Our findings may just reflect changes in the asset risk of high financial slack firms. To test this alternative hypothesis, we follow Zhang (2006) and use cash flow volatility as a measure for asset risk. As defined in Section

19

3.2, cash flow volatility is the standard deviation of cash flows in the past five years. Column (3) of Panel B, Table 4 presents the results of the regression that includes cash flow volatility and its interaction with the law dummy. While the two additional controls are both significant, the positive coefficient on the interaction of cash flow volatility and the law dummy is inconsistent with the alternative story. More importantly, our main results remain qualitatively unchanged. Firms with more financial slack tend to use less debt. Given that external debt is a device of monitoring managers, it could be a substitute governance mechanism for the market of corporate control. When the takeover market is weakened by the laws’ passage, overall governance should be hurt more for low leverage firms. Our results might just reflect the substitution between debt monitoring and the takeover market. Column (4) of Panel B, Table 4 tests this hypothesis by including lagged leverage and its interaction with the law dummy variable in the baseline specification. The coefficient on the interaction term of the laws’ passage and leverage is significantly positive, suggesting that the laws’ passage hurts firms with lower leverage more. This means that debt monitoring and the takeover market are indeed substitutes. However, our main results remain robust. Suppose firm performance is related to financial slack in a nonlinear way and the laws’ passage affects all firms equally. A regression that ignores the nonlinear relationship between firm performance and financial slack could produce a dubious relationship between performance and the interaction term of the laws’ passage and financial slack. To check this possibility, we add the squared term of excess cash to the baseline specification. The results in column (5) of Panel B, Table 4 show that our findings are robust to this alternative specification.

4.3.2

A sticky measure of financial slack

One might be concerned that the laws’ passage drives variation in financial slack and that the variation is correlated with omitted determinants of firm performance, leading to biased estimation. To address this concern, we construct a sticky measure of financial slack so that it is measured by the levels before the coverage of BC laws. Specifically, if a firm is subject to the

20

BC law starting in year τ , then for any t < τ , financial slack is measured in year t; for any t ≥ τ , financial slack is measured in year τ − 1. By construction, this new measure is sticky and not affected by the passage of BC laws. Furthermore, its sticky nature suggests that it is less likely correlated with omitted variables that cause the bias unless there are anticipation effects or the omitted variables are persistent. Column (1) of Table 5 presents the results of the baseline regression using the sticky measure of excess cash.12 The interaction term between the BC law dummy and excess cash is negative and highly significant, consistent with our previous findings. Since the sticky financial slack variable remains constant when the firm is subject to the BC law, one concern is that it conveys little information about the firm’s actual financial slack many years after the law’s coverage when using the whole sample period. For example, the BC law was passed in New York in 1985, and the sample period ends in 1995. This means that the sticky measure of financial slack does not change from 1985 to 1995 for firms covered by New York’s BC law. To address this issue, we repeat the analysis and use the sample period that begins n years before the law change and ends n years after, n = 1, 2, 3, 4. Columns (2) to (5) of Table 5 present the results. In all regressions, the coefficient estimates of the interaction term remain consistently negative and highly significant.13

4.3.3

2SLS estimation

A more general solution to the endogeneity problem is to find an instrumental variable for financial slack, which is correlated with a firm’s financial slack but not its performance. We construct two instruments for financial slack. For the financial slack of firm i in year t, the first instrument is peer financial slack, defined as 12

The sample size in the regression is smaller because firms are dropped when there is no data available to construct the sticky measure of excess cash. This is true for firms incorporated in a state that passed the BC law and it enters the sample after the law’s passage. 13 As a robustness check, to construct excess cash in year t ≥ τ , instead of using its level in year τ − 1, we also use the average levels from years τ − 2 to τ − 1 and from τ − 3 to τ − 1. The results continue to hold.

21

the average financial slack of other firms in year t in the same industry as firm i but incorporated in states that have not passed the BC law by year t.14 Accordingly, the instrument for the interaction term is the interaction of peer financial slack with the law dummy. Columns (1), (2), and (3) of Table 6 present the 2SLS regression results using excess cash as the measure for financial slack. Columns (1) and (2) show the first stage results, where the dependent variables are excess cash and its interaction with the law dummy, respectively. Column (3) presents the second stage results. Consistent with the OLS regression results, the coefficient estimate of the interaction term is negative and highly significant. The first stage regressions show that the two endogenous variables are highly correlated with the two instruments. This is not surprising since the peer financial slack is constructed from a firm’s industry peers and the financial slack of firms in the same industry is likely affected by common industry shocks. Regarding the exclusion restriction, one concern is that some of those shocks may also impact firm performance, but the control variable industry year already takes this into account. Furthermore, this instrument is unlikely to be correlated with performance changes caused by the laws’ passage, because it is not affected by BC laws by construction. However, peer financial slack may affect a firm’s performance directly. In particular, it may be a proxy for rivals’ financial strength and could exert a negative impact on the firm (Bolton and Scharfstein, 1990). If this is true, then the instrument violates the exclusion restriction and the inference is biased. Given that peer financial slack may not satisfy the exclusion restriction, we construct an alternative instrument, local sticky financial slack, which is the average sticky financial slack of other firms in the same state but not in the same industry. The sticky measure of financial slack is defined as in Section 4.3.2. This instrument is likely to be correlated with financial slack because the financial slack of firms in the same state is probably affected by some common local shocks at the state level. It is possible that these shocks also affect firm performance, but their impacts are already accounted for by the control variable state year. Furthermore, this instrument is not 14

When a firm is incorporated in a state that has not passed a BC law, other firms in the same state are also included to calculate the instrument. When a firm is incorporated in a state that has passed a BC law, other firms in the same state are not included to calculate the instrument.

22

affected by the laws’ passage since it is constructed from the sticky measure of financial slack. Most importantly, its construction does not include firms in the same industry, so it does not proxy for rival firms’ financial strength. Therefore, it is likely to satisfy the exclusion restriction. Accordingly, its interaction with the law dummy serves as the instrument for the interaction of financial slack and the law dummy. The 2SLS regression results are presented in columns (4), (5), and (6) of Table 6. The first stage regressions are in columns (4) and (5), where the dependent variables are excess cash and its interaction with the law dummy, respectively. The results show that the two endogenous variables are indeed significantly correlated with the two instruments. Column (6) presents the second stage regression results and shows that the coefficient estimate of the interaction term is significantly negative, which is consistent with our previous findings. In summary, we address the endogeneity concern of financial slack in three ways. While they can address the endogeneity of financial slack to some extent, they also have their own shortcomings. First, directly controlling for more firm characteristics can reduce omitted variables bias, but it is always possible that some unobservable time-varying variables cause the bias. Second, the sticky measure of financial slack could be correlated with omitted variables that affect firm performance if there are anticipation effects or the omitted variables are persistent. In that case, the estimation would be biased. Finally, as the exclusion restriction is untestable, it is uncertain whether the instruments are truly exogenous to firm performance. Therefore, the three methods might not be perfect individually. However, the fact that the results are robust in all these specifications provides support for our findings.

4.4

Robustness checks

4.4.1

Subsample Analysis

We repeat the analysis using different subsamples to address various concerns. Reverse Causality

23

First, an alternative interpretation of our regression results is reverse causality: the passage of BC laws could be the result of expected decline in profitability rather than the cause of it. Some firms might foresee that they will experience a decline in profitability due to common negative economic shocks, especially the ones with financial slack. As a result these firms lobby the state to pass BC laws to protect themselves. We addressed this problem in Section 4.2 directly by excluding lobbying firms. But some lobbying activities might not be publicized and our list of lobbying firms could be incomplete. Here we address the possibility of reverse causality directly by excluding large firms from our sample. From a cost and benefit perspective, managers in larger firms have stronger incentives and more resources to engage in lobbying activities. When the passage of the laws is indeed driven by lobbying activities of a group of large firms incorporated in the same state expecting profit decline, the event is still exogenous to smaller firms incorporated in the same state. If we find the same results with the sample of smaller firms, reverse causality is unlikely to be the cause of our findings. For each state and each year, we rank firms based on their total assets, and perform the regression analysis excluding the top 50% of the firms. The results are presented in column (1) of Table 7.15 We continue to find that the passage of BC laws hurts the performance of high financial slack firms more. Different sample periods Second, in the baseline regression, we choose the sample period from 1976 to 1995 to be consistent with Bertrand and Mullainathan (2003) and Giroud and Mueller (2010). To check if our results depend on a specific sample period, we repeat the analysis using different time intervals. Since the first BC law was passed in 1985 and the last one was passed in 1991, we choose alternative sample periods that are symmetric around the period of 1985-1991 by expanding the period by one, two, three, and four years on each end. Our main results still hold for the four different sample periods. For brevity, we only report the results based on the period of 1984-1992 in column (2) of Table 7. Other sample periods produce similar results. Non-Delaware firms 15

Dropping the top 10%, 20%, 30% or 40% of firms ranked by size produces similar results.

24

Third, since half of the sample firms are incorporated in Delaware, one might suspect that our results are merely a Delaware effect. To address this concern, we exclude all Delaware firms from the treatment group and repeat the regression analysis. Column (3) of Table 7 shows that our main result does not change. Firms incorporated in states that eventually passed the BC laws Fourth, some firms are incorporated in states that never passed BC laws during the sample period. These firms are used as part of the control group in the regression. One might suggest that firms incorporated in states that never passed the laws are fundamentally different from those incorporated in other states. This questions the validity of using these firms as controls. Column (4) of Table 7 conducts the analysis using only firms incorporated in states that passed BC laws at some point between 1976 and 1995. The control group in any year t only includes firms incorporated in states that have not passed the law by year t but later did. Our finding is robust to this specification. Firms with data available for the whole sample period Fifth, one might be concerned with the entry of new firms and the exit of old firms during the sample period. If a firm’s decision regarding where to incorporate is endogenous and affected by whether a particular state has passed BC laws, including firms that enter the sample during the sample period could induce a selection bias. Another possibility is that our finding is caused by survivorship bias in the data. Ex ante the laws’ passage affects all firms equally, but firms with low financial slack that experience decline in performance enter bankruptcy and drop out of the sample. As a result, we observe all high financial slack firms, and only a fraction of low financial slack firms that perform relatively better. On average, it appears that after the laws’ passage, firms with more financial slack perform worse. To address the issue of firm entry and exit, we repeat the analysis with the subsample of firms that have performance data available in the whole sample period and present the results in column (5) of Table 7. Our main finding continues to hold.

25

4.4.2

More robustness checks

We conduct more analysis to check the robustness of our main finding.16 So far our measure for accounting performance is ROA before depreciation. We also perform the analysis using four alternative performance measures: ROA after depreciation, return on equity (ROE), net profit margin, and sales growth. ROA after depreciation is the ratio of operating income after depreciation to total assets. ROE is the ratio of net income to common equity. Net profit margin is defined as operating income before depreciation divided by sales. Sales growth is the annual growth rate of sales. The results are qualitatively the same. When estimating excess cash, we normalize cash by total assets, following Bates et al. (2009) and Dittmar and Duchin (2011). We also find similar results if we normalize cash by net assets (total assets excluding cash) or sales. Furthermore, there are also studies that use the natural logarithm of the cash-to-net-assets ratio or cash-to-sales ratio (e.g., Opler et al., 1999; Harford et al., 2008). Using these measures to estimate excess cash does not change our main finding. In all regressions, standard errors are clustered at the state of incorporation level because the BC law dummy is a possible source of both cross-sectional and serial correlations as discussed in Section 3.1. As a robustness check, we follow Bertrand and Mullainathan (2003) and Bertrand, Duflo, and Mullainathan (2004) and consider a number of alternative methods to correct correlations in the error term. We find similar results if we cluster the standard errors at the state of location level, if we use an AR(1) correction method, or if we block bootstrap the standard errors using 51 blocks with 200 bootstrap samples.

5

Stock market performance In previous analysis we use the firm accounting performance as the main dependent variable

when studying the effect of the laws’ passage on performance. If the stock market is unable to perfectly predict the effect of law at the time of the laws’ passage, stock performance should also 16

For the sake of brevity, these results are not reported, but available upon request.

26

be affected. We construct characteristics-adjusted annual returns to measure the stock performance as in Daniel, Grinblatt, Titman, and Wermers (1997). Specifically, at the end of each calendar year, we classify stocks into size quintiles based on their market capitalization using the breakpoints determined by NYSE stocks. Stocks in each quintile are further split into quintiles based on their book-to-market ratios. We then assign stocks in each of the 25 groups into quintiles based on their past six-month returns. This generates 125 benchmark portfolios based on size, book-to-market ratios, and past returns. We calculate the equally-weighted return for each group and use it as the benchmark portfolio return. A stock’s adjusted annual return is defined as the difference between its raw return and the return of its benchmark portfolio. We use the adjusted annual stock return as the dependent variable and estimate a specification similar to Equation (1). The purpose is to examine how the laws’ passage affects stock market performance differently across firms with different levels of financial slack. Because size has been accounted for when calculating the adjusted return, we do not include size or square of size in the regression. The results are presented in Table 8. For all four measures of financial slack, the coefficient estimates on the interaction between BC laws and financial slack are negative and highly significant. This is consistent with our earlier finding using firm accounting performance.

6

Channels of wasteful spending We have shown that the passage of BC laws leads to a larger decline in the performance of firms

with high financial slack. This is consistent with the argument that the laws’ passage weakens corporate governance and exacerbates agency conflicts associated with Jensen’s free cash flow problem, i.e., managers with excess financial resources tend to engage in more wasteful spending. It would be interesting to find out what kind of wasteful spending is behind the performance drop of high financial slack firms after the laws’ passage. In particular, managers can engage in “empire building” to entrench themselves with more resources under control. Alternatively, because greater financial slack provides a comfort zone to managers, they may just enjoy a “quiet life” and do not work as hard to minimize costs. They may spend resources beyond the optimal

27

level in order to avoid “cognitively difficult activities” (Bertrand and Mullainathan, 2003), such as negotiating with suppliers, labor unions, and business units within the firm demanding bigger overhead budgets. Motivated by these two conjectures, we investigate firms’ investment activities and cost management to identify the channels of wasteful spending of high financial slack firms after the passage of BC laws. Table 9 presents the empirical results. In Panel A, we estimate the baseline specification using five proxies for investment activities as the dependent variable. Column (1) uses the ratio of capital expenditure to total assets. Column (2) uses a measure for over-investment, which is the difference between the actual investment and expected investment calculated from a regression model based on Richardson (2006), as described in the Appendix. Column (3) uses the annual growth rate of total assets. Column (4) uses the annual growth rate of fixed assets. Column (5) uses the acquisition ratio, defined as the sum of the value of all acquisitions made by the firm in a given year divided by the firm’s market capitalization in that year. The coefficients on the stand-alone excess cash are consistently positive across all regressions, implying that indeed firms with more financial slack on average engage in more investment activities. However, the average effect of the laws’ passage is insignificant. Most importantly, the coefficients on BC law × Excess cash are mostly insignificant. This means that, although on average firms with more financial slack tend to engage in more investment and expansion, such tendency is not exacerbated by the passage of the BC laws. In column (3) with asset growth as the dependent variable, the coefficient on the interaction term is significantly negative, which suggests that the positive relationship between asset growth and financial slack is actually moderated by the laws’ passage. Overall the results do not seem to suggest that the larger performance drop of high financial slack firms after the laws’ passage is due to relatively more investment and expansion. Panel B turns to the cost management of firms before and after the passage of BC laws. We estimate the same regression using six proxies for cost management as the dependent variable. The proxies include overhead costs (selling, general, and administrative expenses), advertising

28

expenses, operating expenses, costs of goods sold (COGS), the number of employees, and wages.17 All variables are normalized by sales except that wages are normalized by the number of employees. Managers in high financial slack firms might not work as hard to maintain cost efficiency after the passage of BC laws. Indeed, the positive and significant interaction terms in columns (1), (3), (4), and (5) suggest that firms with high financial slack experience an increase in overhead costs, operating expenses, costs of goods sold, and the number of employees after the laws’ passage. This provides direct evidence that inefficient cost management, a type of wasteful spending, is behind the larger performance drop of high financial slack firms after the laws’ passage. We want to emphasize that our main goal in this section is not to test the “empire building” hypothesis vs. the “quiet life” hypothesis. Instead, the analysis aims to examine the channels of wasteful spending derived from the two hypotheses. Actually, the two sets of proxies are imperfect and may be related to both empire building and enjoying the quiet life.18 In particular, the dependent variables of panel A could also represent substantial cash outlays, so an increase in these proxies might suggest that managers do not prudently monitor the use of financial slack and thus enjoy a quiet life. On the other hand, all the dependent variables of panel B could also be positively related to empire building activities. Increasing overhead costs, advertising, and the number of employees, is often an indication of a greater scale of operations and suggests possible empire building. Therefore, the empirical evidence might actually provide some support to both hypotheses. However, whether it is empire building or enjoying the quiet life is not crucial for this study; what really matters is that the analysis provides direct evidence and identifies the channel of wasteful spending by managers of high financial slack firms after the laws’ passage.

7

Conclusion In this paper, we examine whether and how the effect of corporate governance depends on

the level of financial slack. Theoretically, corporate governance matters more for firms with 17

Because many Compustat firms do not report wages, the number of observations is much smaller in column (6) compared to other columns. 18 We thank one anonymous referee for pointing this out.

29

more severe agency problems as its main goal is to mitigates agency conflicts. However, the relationship between financial slack and agency conflicts is ambiguous. On one hand, the use of financial slack is largely at the discretion of managers and lacks disciplines from the capital market, so managers may spend financial slack for their private benefits at the expense of shareholders (Jensen, 1986). This implies that agency conflicts are potentially more severe for firms with high financial slack, and corporate governance should be more effective for this type of firms. On the other hand, financial slack can serve as precautionary savings and protect firms again future economic and political uncertainties, but self-interested managers can be imprudent and hold insufficient financial slack. In this case, a low level of financial slack signals more severe agency problems and corporate governance should matter more for firms with low financial slack. Using the passage of BC laws to identify exogenous variation in corporate governance, we find that the weakening of corporate governance causes a larger decline in performance of firms with high financial slack. This finding has important policy implications. It suggests that governance mechanisms do not matter equally for all firms; instead, shareholder activism and government regulations aiming to improve corporate governance can be more effective by targeting firms with high financial slack. In a nutshell, our paper investigates a particular aspect of the conditional nature of corporate governance, i.e., whether and how financial slack affects the effectiveness of corporate governance. It would be interesting for future research to examine wether and how the functioning of corporate governance depends on other firm traits, industry characteristics, macroeconomic environment, and other factors. This would further deepen our understanding of corporate governance and help implement governance mechanisms to better address the conflicts between shareholders and managers.

30

Appendix Following Richardson (2006), we construct the expected investment in new positive NPV projects ∗  Inew and over-investment Inew in two steps.

Step one, total investment Itotal is defined as the sum of all outlays on capital expenditure, acquisitions, and research and development, less receipts from the sale of property, plant, and equipment. Total investment Itotal can be decomposed into two components: (i) required expenditure to maintain assets in place Imaintenance , and (ii) investment in new projects Inew . A proper proxy for Imaintenance is amortization and depreciation. Thus investment in new projects Inew can be computed as: Inew,t = Itotal,t − Imaintenance,t = CAP EXt + Acquisitionst + R&Dt − SaleP P Et − Depreciationt .

Step two, observed Inew,t is used as the dependent variable to fit the following regression model: Inew,t = α + βV Pt−1 + ϕZt−1 + εt . ∗ The predicted value is the expected investment expenditure in new positive NPV projects Inew  . The main explanatory variable in , and the residual is the measure for over-investment Inew

estimating the expected investment expenditure is a firm’s growth opportunities denoted as V P . It is calculated as the ratio of firm value (Vaip ) to the market value of equity. Vaip is estimated as Vaip = (1 − αr)BV + α(1 + r)X − αrd, where, α = (ω/(1 + r − ω)), r = 12%, and ω = 0.62. ω is the abnormal earnings persistence parameter from the Ohlson (1995) framework, BV is the book value of common equity, d is annual dividends, and X is operating income after depreciation. Z is a vector of control variables including leverage, size, age, stock of cash, past stock returns, prior firm level investment, year fixed effects, and industry fixed effects.

31

References Bates, T. W., K. M. Kahle, and R. M. Stulz (2009). Why do us firms hold so much more cash than they used to? Journal of Finance 64 (5), 1985–2021. Bertrand, M., E. Duflo, and S. Mullainathan (2004). How much should we trust differences-indifferences estimates? Quarterly Journal of Economics 119 (1), 249–275. Bertrand, M. and S. Mullainathan (2003). Enjoying the quiet life? Corporate governance and managerial preferences. Journal of Political Economy 111, 1043–1075. Bolton, P. and D. S. Scharfstein (1990). A theory of predation based on agency problems in financial contracting. The American Economic Review , 93–106. Brav, A., W. Jiang, F. Partnoy, and R. Thomas (2008). Hedge fund activism, corporate governance, and firm performance. Journal of Finance 63 (4), 1729–1775. Cremers, M. and A. Ferrell (2014). Thirty years of shareholder rights and firm valuation. Journal of Finance 69 (3), 1167–1196. Daniel, F., F. T. Lohrke, C. J. Fornaciari, and R. A. Turner Jr (2004). Slack resources and firm performance: a meta-analysis. Journal of Business Research 57 (6), 565–574. Daniel, K., M. Grinblatt, S. Titman, and R. Wermers (1997). Measuring mutual fund performance with characteristic-based benchmarks. Journal of Finance 52 (3), 1035–1058. Disatnik, D., R. Duchin, and B. Schmidt (2014). Cash flow hedging and liquidity choices. Review of Finance 18 (2), 715–748. Dittmar, A. and R. Duchin (2011). The dynamics of cash. Working paper . Dittmar, A. and J. Mahrt-Smith (2007). Corporate governance and the value of cash holdings. Journal of Financial Economics 83 (3), 599–634. Duchin, R., J. Matsusaka, and O. Ozbas (2010). When are outside directors effective? Journal of Financial Economics 96 (2), 195–214. Fr´esard, L. and C. Salva (2010). The value of excess cash and corporate governance: Evidence from US cross-listings. Journal of Financial Economics 98 (2), 359–384. Giroud, X. and H. Mueller (2010). Does corporate governance matter in competitive industries? Journal of Financial Economics 95 (3), 312–331. Gompers, P., J. Ishii, and A. Metrick (2003). Corporate governance and equity prices. Quarterly Journal of Economics 118 (1), 107–155. Harford, J., S. A. Mansi, and W. F. Maxwell (2008). Corporate governance and firm cash holdings in the US. Journal of Financial Economics 87 (3), 535–555. Jensen, M. (1986). Agency costs of free cash flow, corporate finance, and takeovers. The American Economic Review 76 (2), 323–329.

32

Kalcheva, I. and K. V. Lins (2007). International evidence on cash holdings and expected managerial agency problems. Review of Financial Studies 20 (4), 1087–1112. Karpoff, J. and P. Malatesta (1989). The wealth effects of second-generation state takeover legislation. Journal of Financial Economics 25, 291–322. Karpoff, J. M. and M. D. Wittry (2014). Test identification with legal changes: The case of state antitakeover laws. Working paper . Keynes, J. M. (1936). The General Theory of Employment, Interest and Money. Harcourt Brace, London. Lins, K. V., H. Servaes, and P. Tufano (2010). What drives corporate liquidity? An international survey of cash holdings and lines of credit. Journal of Financial Economics 98 (1), 160–176. Margotta, D., T. McWilliams, and V. McWilliams (1990). An analysis of the stock price effect of the 1986 Ohio takeover legislation. Journal of Law, Economics, and Organization 6 (1), 235–251. Myers, S. C. and N. S. Majluf (1984). Corporate financing and investment decisions when firms have information that investors do not have. Journal of Financial Economics 13 (2), 187–221. Ohlson, J. (1995). Earnings, book values, and dividends in equity valuation. Contemporary Accounting Research 11 (2), 661–687. Opler, T., L. Pinkowitz, R. Stulz, and R. Williamson (1999). The determinants and implications of corporate cash holdings. Journal of Financial Economics 52 (1), 3–46. Pound, J. (1987). The effects of antitakeover amendments on takeover activity: Some direct evidence. Journal of Law and Economics 30 (2), 353–367. Richardson, S. (2006). Over-investment of free cash flow. Review of Accounting Studies 11 (2), 159–189. Romano, R. (1987). The political economy of takeover statutes. Virginia Law Review 73 (1), 111–199. Ryngaert, M. and J. Netter (1988). Shareholder wealth effects of the Ohio antitakeover law. Journal of Law, Economics, and Organization 4 (2), 373–383. Schumann, L. (1988). State regulation of takeovers and shareholder wealth: the case of New York’s 1985 takeover statutes. Rand Journal of Economics 19 (4), 557–567. Yermack, D. (1996). Higher market valuation of companies with a small board of directors. Journal of Financial economics 40 (2), 185–211. Zhang, X. (2006). Information uncertainty and stock returns. Journal of Finance 61 (1), 105–137.

33

Table 1: Regression Variables Panel A reports the summary statistics. Panel B reports the correlation matrix. ROA is operating income before depreciation divided by total assets. Size is the log of total assets. Age is the log of one plus the number of years between the first year when the firm is covered by Compustat and the current year. Current ratio is current assets divided by total assets. Cash ratio is cash and shortterm investments divided by total assets. Excess cash is the cash ratio minus a predicted cash ratio estimated by a five-year rolling window; the explanatory variables of the estimation regression include lagged measures of cash flow, cash flow volatility, Tobin’s Q, firm size, net working capital, leverage, capital expenditure, R&D expenditure, and the dividend payout dummy variable. Panel A: Summary Statistics N

MEAN

STDEV

MIN

MAX

ROA

68008

0.089

0.185

−0.883

0.407

Size

68008

4.357

2.116

−2.847

12.008

Age

68008

2.485

0.773

0.693

3.829

Current ratio

68008

1.176

0.515

0

7.45

0

0.808

Cash ratio

68000

0.131

0.169

Excess cash

44045

0.001

0.106

−0.214

0.47

Panel B: Correlation Matrix ROA

Size

Age

Current ratio

Cash Ratio

ROA

1

Size

0.305

1

Age

0.148

0.582

Current ratio

−0.022

−0.257

−0.139

1

Cash ratio

−0.129

−0.231

−0.180

0.571

1

Excess cash

−0.047

−0.11

−0.110

0.596

0.862

34

Excess cash

1

1

Table 2: Impact of BC Laws’ Passage on Firm Performance for Different Levels of Financial Slack Coefficient estimates and their t-statistics (in parentheses) are presented for the following regression model: ROAijkl,t = αi + αt + β1 BC lawijkl,t + β2 (BC lawijkl,t × F Sijkl,t−1 ) + β3 F Sijkl,t−1 + γ 0 Xijkl,t + εijkl,t , where i, j, k, l, and t index firms, industries, states of incorporation, states of location, and time, respectively. BC lawijkl,t is equal to one if firm i incorporated in state k is under the coverage of BC laws in year t and zero otherwise. F S is financial slack. X is a vector of control variables, which includes firm size, firm age, squared terms of size and age, the mean ROA of other firms in the same industry-year group (“industry year”), and the mean ROA of other firms in the same state-year group (“state year”). Column (1) estimates the equation without financial slack. In columns (2) to (4), financial slack is measured by current ratio, cash ratio, and excess cash, respectively, which are defined as in Table 1. Firm and year fixed effects are included. Standard errors are adjusted for heteroskedasticity and clustered at the state of incorporation level. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. (1) BC law

(2)

(3)

(4)

-0.005∗

0.014∗∗

(-1.70)

(2.35) -0.030∗∗∗ (-8.83) -0.015∗∗∗ (-3.80)

-0.000 (-0.12)

-0.002 (-0.88)

Current ratio BC law × Current ratio

-0.043∗∗∗ (-3.93) -0.028∗∗ (-2.31)

Cash ratio BC law × Cash ratio Excess cash BC law × Excess Cash Size Age Size squared Age squared Industry year State year R2 N

0.117∗∗∗ (21.49) 0.155∗∗∗ (8.50) -0.009∗∗∗ (-22.83) -0.071∗∗∗ (-10.16) 0.224∗∗∗ (9.91) 0.242∗∗∗ (8.12) 0.692 68008

0.121∗∗∗ (23.02) 0.130∗∗∗ (8.06) -0.010∗∗∗ (-25.06) -0.064∗∗∗ (-10.24) 0.222∗∗∗ (9.82) 0.238∗∗∗ (7.46) 0.696 68008

35

0.117∗∗∗ (22.34) 0.140∗∗∗ (8.52) -0.009∗∗∗ (-23.99) -0.067∗∗∗ (-10.36) 0.222∗∗∗ (9.87) 0.242∗∗∗ (7.88) 0.693 68000

-0.009 (-0.57) -0.053∗∗∗ (-2.92) 0.085∗∗∗ (9.80) 0.262∗∗∗ (4.63) -0.006∗∗∗ (-8.59) -0.095∗∗∗ (-6.29) 0.208∗∗∗ (9.01) 0.166∗∗∗ (6.56) 0.653 44045

Table 3: Using BC Laws’ Passage for Identification This table reports the regression analyses that deal with various issues of using the passage for identification. We consider the influence of lobbying firms, firms that chose to opt in or out of state anti-takeover laws, the legal regime as reflected in important court decisions, and first-generation and other second-generation state anti-takeover laws. BC law before ruling equals one if the law was effective in year t < 1989, and zero otherwise. BC law after ruling equals one if the law was effective in year t ≥ 1989, and zero otherwise. All other variables are defined as in Tables 1 and 2. Firm and year fixed effects are included. Standard errors are adjusted for heteroskedasticity and clustered at the state of incorporation level. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Panel A: Issues of BC Laws In Panel A, column (1) drops all lobbying firms; column (2) drops all firms that opted in or opted out of the law as well as lobbying firms; column (3) studies the differential effect of BC laws before and after the court ruling on Amanda Acquisition Corp. v. Universal Foods Corp. in 1989 that established the constitutional status of BC laws; column (4) is the same as column (3) except that it drops all lobbying firms and firms that opted in or opted out of the law. (1) BC law Excess cash BC law × Excess cash

-0.002 (-0.78) -0.009 (-0.53) -0.053∗∗∗ (-2.81)

(2)

(3)

(4)

-0.003 (-0.93) -0.008 (-0.50) -0.053∗∗∗ (-2.74)

-0.009 (-0.57)

-0.009 (-0.53)

-0.007 (-1.19) -0.000 (-0.11) -0.045∗∗∗ (-2.71) -0.077∗∗ (-2.48) 0.085∗∗∗ (9.22) 0.263∗∗∗ (4.46) -0.006∗∗∗ (-8.04) -0.096∗∗∗ (-6.10) 0.207∗∗∗ (8.55) 0.165∗∗∗ (6.20) 0.653 44045

-0.007 (-1.13) -0.000 (-0.03) -0.046∗∗∗ (-2.74) -0.078∗∗ (-2.45) 0.086∗∗∗ (9.44) 0.253∗∗∗ (4.33) -0.006∗∗∗ (-8.38) -0.093∗∗∗ (-5.94) 0.213∗∗∗ (7.98) 0.168∗∗∗ (6.17) 0.653 43319

BC law before ruling BC law after ruling BC law before ruling × Excess cash BC law after ruling × Excess cash Size Age Size squared Age squared Industry year State year R2 N

0.086∗∗∗ (9.45) 0.253∗∗∗ (4.19) -0.006∗∗∗ (-8.44) -0.093∗∗∗ (-5.73) 0.214∗∗∗ (7.90) 0.167∗∗∗ (6.12) 0.653 43507

0.087∗∗∗ (9.48) 0.252∗∗∗ (4.23) -0.006∗∗∗ (-8.47) -0.092∗∗∗ (-5.75) 0.214∗∗∗ (7.91) 0.168∗∗∗ (6.14) 0.653 43319

36

Panel B: Other State Anti-takeover Laws Panel B controls for the effect of other state anti-takeover laws. In columns (1) to (4), Fair price, Control share, Poison pill, and Constituency are dummy variables that stand for the passage of fair price laws, control share laws, poison pill laws, and constituency laws, respectively. Column (5) controls for all four types of second generation anti-takeover laws simultaneously. In column (6), observations from 1976 to 1982 are dropped to exclude the effect of first-generation anti-takeover laws that were deemed unconstitutional in 1982. Column (7) controls for all four second generation anti-takeover laws and excludes data from 1978 to 1982. (1) BC law Excess cash BC law × Excess Cash Fair price Fair price × Excess cash

-0.003 (-0.81) -0.007 (-0.44) -0.043∗∗∗ (-2.99) 0.001 (0.18) -0.043∗ (-1.70)

Control share

(2)

(3)

(4)

(5)

(6)

(7)

-0.002 (-0.84) -0.008 (-0.47) -0.049∗∗ (-2.52)

-0.002 (-0.76) -0.006 (-0.39) -0.042∗∗∗ (-2.88)

-0.002 (-0.86) -0.008 (-0.46) -0.044∗∗ (-2.51)

-0.004 (-1.17) -0.006 (-0.41) -0.039∗∗ (-2.67) 0.005 (1.47) -0.017 (-0.62) 0.002 (0.57) -0.008 (-0.28) -0.005 (-1.66) -0.053∗ (-1.86) -0.005 (-1.49) 0.013 (0.41) 0.085∗∗∗ (9.23) 0.260∗∗∗ (4.28) -0.006∗∗∗ (-8.14) -0.095∗∗∗ (-5.81) 0.208∗∗∗ (8.50) 0.163∗∗∗ (6.19) 0.653 44045

0.000 (0.12) -0.000 (-0.02) -0.056∗∗∗ (-3.01)

0.000 (0.01) 0.002 (0.09) -0.047∗∗∗ (-2.87) -0.001 (-0.49) -0.027 (-0.96) 0.003 (0.68) 0.021 (0.72) -0.005 (-1.59) -0.045∗ (-1.72) -0.005 (-1.32) 0.008 (0.30) 0.123∗∗∗ (11.09) 0.374∗∗∗ (6.31) -0.010∗∗∗ (-11.68) -0.122∗∗∗ (-7.69) 0.182∗∗∗ (12.07) 0.154∗∗∗ (4.42) 0.695 31657

0.001 (0.24) -0.022 (-0.89)

Control share × Excess cash

-0.006∗∗∗ (-3.10) -0.054∗∗ (-2.59)

Poison pill Poison pill × Excess cash Constituency Constituency × Excess cash Size Age Size squared Age squared Industry year State year R2 N

0.085∗∗∗ (9.22) 0.259∗∗∗ (4.29) -0.006∗∗∗ (-8.08) -0.094∗∗∗ (-5.82) 0.208∗∗∗ (8.50) 0.166∗∗∗ (6.17) 0.653 44045

0.085∗∗∗ (9.24) 0.261∗∗∗ (4.37) -0.006∗∗∗ (-8.11) -0.095∗∗∗ (-5.94) 0.208∗∗∗ (8.48) 0.166∗∗∗ (6.21) 0.653 44045

0.085∗∗∗ (9.24) 0.262∗∗∗ (4.35) -0.006∗∗∗ (-8.14) -0.095∗∗∗ (-5.90) 0.208∗∗∗ (8.56) 0.165∗∗∗ (6.20) 0.653 44045

37

-0.006∗∗ (-2.34) -0.033 (-1.55) 0.085∗∗∗ (9.29) 0.263∗∗∗ (4.38) -0.006∗∗∗ (-8.18) -0.096∗∗∗ (-5.95) 0.209∗∗∗ (8.46) 0.164∗∗∗ (6.17) 0.653 44045

0.122∗∗∗ (11.20) 0.373∗∗∗ (6.32) -0.010∗∗∗ (-11.78) -0.121∗∗∗ (-7.73) 0.182∗∗∗ (11.86) 0.163∗∗∗ (4.34) 0.694 31657

Table 4: Controlling for More Firm Characteristics This table reports the regressions that control for various governance measures and other firm characteristics. Coefficient estimates and their t-statistics (in parentheses) are presented for the following regression model: ROAijkl,t = αi + αt + β1 BC lawijkl,t + β2 (BC lawijkl,t × F Sijkl,t−1 ) + β3 F Sijkl,t−1 + β4 Zijkl,t−1 + β5 (BC lawijkl,t × Zijkl,t−1 ) + γ 0 Xijkl,t + εijkl,t . Zijkl,t−1 is the additional firm characteristic to be controlled for. All other variables are defined as in Tables 1 and 2. Coefficients on the control variables are not reported for the sake of brevity. Firm and year fixed effects are included. Standard errors are adjusted for heteroskedasticity and clustered at the state of incorporation level. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Panel A: Firm level Governance Measures In Panel A, Zijkl,t−1 is G-index, CEO ownership, CEO equity pay, CEO duality, board size and board independence in columns (1) to (6), respectively. CEO ownership is the percentage of a firm’s equity owned by the CEO. CEO equity pay is one minus the percentage of cash pay including salary and bonus. CEO duality is a dummy variable that is one if the CEO is also the chairperson of the board, and zero otherwise. Board size is the log of the number of board directors. Board independence is the percentage of directors that are qualified as outsiders. (1) BC law Excess cash BC law × Excess cash G-index BC law × G-index

-0.001 (-0.22) 0.003 (0.14) -0.058∗∗ (-1.96) -0.002∗∗∗ (-2.91) 0.000 (0.44)

CEO ownership

(2)

(3)

(4)

(5)

(6)

-0.007∗∗

-0.012∗∗∗

(-2.17) 0.018 (0.69) -0.061∗ (-1.79)

(-3.22) 0.017 (0.62) -0.059∗ (-1.67)

-0.004 (-1.10) 0.012 (0.45) -0.062∗ (-1.82)

-0.011 (-1.37) 0.003 (0.12) -0.055 (-1.63)

-0.003 (-0.38) 0.012 (0.44) -0.062∗ (-1.81)

-0.015 (-0.70) -0.012 (-0.34)

BC law × CEO ownership CEO equity pay

-0.007 (-1.07) 0.012 (1.46)

BC law × CEO equity pay

-0.006∗ (-1.95) 0.004 (1.22)

CEO duality BC law × CEO duality

-0.002∗∗∗ (-3.35) 0.000 (0.64)

Board size BC law × Board size Board independence BC law × Board independence R2 N

0.613 11650

0.424 4108

0.418 4623

38

0.411 4477

0.414 4491

0.001 (0.10) -0.005 (-0.38) 0.412 4491

Panel B: Other Firm Characteristics In Panel B, Zijkl,t−1 is the lagged value of industry Herfindahl-Hirschman Index (HHI), Tobin’s Q, cash flow volatility, and leverage in columns (1) to (4), respectively. Column (5) includes the squared term of excess cash. (1) Additional Controls BC law Excess cash BC law × Excess cash HHI BC law × HHI

(2)

(3)

(4)

(5)

HHI

Tobin’s Q

Cash Flow Volatility

Leverage

Nonlinearity of Excess Cash

-0.003 (-0.95) -0.008 (-0.48) -0.052∗∗ (-2.57) 0.016 (1.26) -0.020∗∗ (-2.35)

0.000 (0.12) -0.003 (-0.18) -0.046∗∗ (-2.62)

-0.005∗∗ (-2.14) -0.010 (-0.58) -0.055∗∗∗ (-2.77)

-0.009∗ (-1.89) -0.010 (-0.59) -0.048∗∗ (-2.58)

-0.002 (-0.82) 0.004 (0.25) -0.052∗∗ (-2.57)

0.017∗∗∗ (8.25) -0.003 (-1.12)

Tobin’s Q BC law × Tobin’s Q

0.096∗∗∗ (2.74) 0.060∗∗ (2.33)

CF volatility BC law × CF volatility

-0.042∗∗∗ (-5.96) 0.027∗∗ (2.04)

Leverage BC law × Leverage Excess cash squared R2 N

0.658 42927

0.658 44045

0.654 44045

39

0.654 44045

-0.084 (-1.18) 0.653 44045

Table 5: Sticky Measure of Excess Cash This table reports the results of the baseline regression using the sticky measure of excess cash, which is measured by the level of excess cash before the coverage of BC laws. Specifically, if a firm is subject to the BC law starting in year τ , then for any t < τ , excess cash is measured in year t; for any t ≥ τ excess cash is measured in year τ − 1. Column (1) estimates the whole sample from 1976 to 1995. Columns (2) to (5) estimate the sub-periods starting n years before and ending n years after the laws’ coverage, with n equal to 1, 2, 3, and 4, respectively. All other variables are defined as in Tables 1 and 2. Firm and year fixed effects are included. Standard errors are adjusted for heteroskedasticity and clustered at the state of incorporation level. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. (1) BC law Sticky excess cash BC law × Sticky excess cash Size Age Size squared Age squared Industry year State year R2 N

0.000 (0.17) 0.003 (0.24) -0.069∗∗ (-2.64) 0.074∗∗∗ (13.44) 0.302∗∗∗ (6.81) -0.006∗∗∗ (-12.50) -0.103∗∗∗ (-9.03) 0.189∗∗∗ (7.08) 0.192∗∗∗ (10.71) 0.569 34239

(2)

(3)

(4)

(5)

0.003 (0.27) -0.015 (-0.43) -0.064∗∗ (-2.28) 0.146∗∗∗ (3.90) 0.824∗ (1.92) -0.013∗∗∗ (-4.35) -0.266∗∗ (-2.28) 0.107∗∗∗ (3.40) -0.004 (-0.03) 0.869 3951

-0.003 (-0.64) 0.010 (0.17) -0.085∗∗∗ (-3.55) 0.133∗∗∗ (9.12) 0.689∗∗∗ (3.37) -0.012∗∗∗ (-9.31) -0.210∗∗∗ (-3.52) 0.208∗∗∗ (3.31) 0.132∗∗ (2.32) 0.739 7766

-0.004 (-1.00) -0.004 (-0.09) -0.093∗∗∗ (-4.17) 0.108∗∗∗ (7.26) 0.662∗∗∗ (4.94) -0.010∗∗∗ (-8.06) -0.193∗∗∗ (-4.91) 0.213∗∗∗ (7.09) 0.159∗∗∗ (4.26) 0.685 11432

0.001 (0.33) -0.010 (-0.35) -0.083∗∗∗ (-4.46) 0.102∗∗∗ (7.76) 0.640∗∗∗ (6.30) -0.009∗∗∗ (-8.16) -0.187∗∗∗ (-6.33) 0.197∗∗∗ (10.69) 0.164∗∗∗ (3.98) 0.648 14959

40

Table 6: 2SLS Estimation This table reports the results of 2SLS regressions. In columns (1) to (3), the instrumental variable for excess cash is peer excess cash, defined as the average excess cash of firms in the same industry but incorporated in states that have not passed BC laws; its interaction with the law dummy serves as the instrument for the interaction of excess cash with the law dummy. Columns (1) and (2) show the first stage regression results, and column (3) shows the second stage regression results. In columns (4) to (6), the instrumental variable for excess cash is local sticky excess cash, defined as the average sticky pre-law level of excess cash of firms located in the same state but in different industries; its interaction with the law dummy serves as the instrument for the interaction of excess cash with the law dummy. Columns (4) and (5) show the first stage regression results, and column (6) shows the second stage regression results. All other variables are defined as in Tables 1 and 2. Firm and year fixed effects are included. Standard errors are adjusted for heteroskedasticity and clustered at the state of incorporation level. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Instrument based on industry peer firms

Dependent Variable BC law

First stage

First stage

Second stage

First stage

First stage

Second stage

(1)

(2)

(3)

(4)

(5)

(6)

Excess cash

Excess cash × BC law

ROA

Excess cash

Excess cash × BC law

ROA

0.001 (0.19)

-0.001 (-0.43) -0.057 (-1.51) -0.189∗∗∗ (-2.63) 0.093∗∗∗ (7.95) 0.228∗∗∗ (5.13) -0.007∗∗∗ (-6.66) -0.085∗∗∗ (-7.07) 0.229∗∗∗ (8.68) 0.150∗∗∗ (5.39)

0.003 (1.35)

-0.001 (-0.13)

0.021∗∗∗ (5.07) 0.049 (0.71) -0.003∗∗∗ (-10.21) -0.020 (-1.10) -0.049∗∗∗ (-3.73) -0.014 (-0.53)

0.035∗∗∗ (9.65) -0.418∗∗∗ (-8.48) -0.004∗∗∗ (-13.26) 0.132∗∗∗ (10.01) -0.005 (-0.41) 0.027 (1.05)

-0.003 (-0.93) 0.330 (1.57) -0.120∗∗ (-2.29) 0.081∗∗∗ (6.33) 0.204∗∗ (2.55) -0.006∗∗∗ (-4.41) -0.075∗∗∗ (-3.34) 0.224∗∗∗ (7.10) 0.176∗∗∗ (6.43)

40745

0.188∗∗∗ (4.13) -0.037 (-0.82) 0.711 43577

0.001 (0.03) 0.501∗∗∗ (12.81) 0.629 43577

43577

0.003 (0.89)

Excess cash BC law × Excess cash 0.024∗∗∗ (5.63) Age -0.120 (-1.61) Size squared -0.003∗∗∗ (-10.54) Age squared 0.032 (1.62) Industry year -0.007 (-0.48) State year -0.049∗ (-1.82) Peer excess cash 0.615∗∗∗ (10.83) BC law × Peer excess cash -0.479∗∗∗ (-3.94) Local sticky excess cash Size

BC law × Local sticky excess cash R2 N

Instrument based on local firms

0.737 40745

0.030∗∗∗ (7.76) -0.319∗∗∗ (-4.07) -0.004∗∗∗ (-9.48) 0.102∗∗∗ (4.84) -0.018∗ (-1.99) -0.028 (-1.11) -0.243∗∗∗ (-3.01) 0.398∗∗∗ (9.08)

0.648 40745

41

Table 7: Subsample Analysis This table reports the results of the baseline regression model as in Table 2 using various subsamples. All variables are defined as in Tables 1 and 2. Column (1) excludes large firms whose assets are above sample median. Column (2) examines a symmetric time period of 1984-1992. Column (3) excludes firms incorporated in Delaware from the treatment group. Column (4) excludes firms incorporated in states that never passed BC laws in the sample period. Column (5) includes only firms with observations available in the whole sample period from 1976 to 1995. Firm and year fixed effects are included. Standard errors are adjusted for heteroskedasticity and clustered at the state of incorporation level. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. (1) BC law Excess cash BC Law × Excess Cash Size Age Size squared Age squared Industry year State year R2 N

-0.006 (-0.98) -0.013 (-0.49) -0.056∗∗∗ (-2.72) 0.156∗∗∗ (5.35) 0.084 (0.58) -0.017∗∗∗ (-4.16) -0.039 (-0.89) 0.205∗∗∗ (6.76) 0.267∗∗∗ (4.81) 0.658 22021

(2)

(3)

(4)

(5)

0.002 (0.68) 0.009 (0.47) -0.083∗∗∗ (-5.10) 0.131∗∗∗ (12.49) 0.577∗∗∗ (6.39) -0.011∗∗∗ (-12.11) -0.171∗∗∗ (-7.52) 0.185∗∗∗ (10.04) 0.164∗∗∗ (3.80) 0.708 21605

-0.004 (-1.33) -0.025 (-1.59) -0.056∗∗ (-2.22) 0.072∗∗∗ (9.65) 0.217∗∗∗ (3.30) -0.005∗∗∗ (-7.89) -0.084∗∗∗ (-4.62) 0.233∗∗∗ (11.53) 0.159∗∗∗ (5.07) 0.637 35210

0.001 (0.43) 0.013 (1.04) -0.071∗∗∗ (-3.21) 0.091∗∗∗ (12.45) 0.281∗∗∗ (5.02) -0.007∗∗∗ (-11.36) -0.101∗∗∗ (-7.00) 0.199∗∗∗ (8.61) 0.191∗∗∗ (9.20) 0.656 38650

-0.001 (-0.33) -0.004 (-0.18) -0.047∗∗ (-2.19) 0.073∗∗∗ (7.13) 0.339∗∗∗ (4.64) -0.005∗∗∗ (-6.29) -0.109∗∗∗ (-5.50) 0.190∗∗∗ (8.60) 0.167∗∗∗ (5.78) 0.649 35542

42

Table 8: Impact of the Laws’ Passage on Stock Performance Coefficient estimates and their t-statistics (in parentheses) are presented for the following regression model: Rijkl,t = αi + αt + β1 BC lawijkl,t + β2 (BC lawijkl,t × F Sijkl,t−1 ) + β3 F Sijkl,t−1 + γ 0 Xijkl,t + εijkl,t , where Rijkl,t is the characteristics-adjusted annual stock return of firm i, constructed using benchmark portfolios based on size, book-to-market ratios, and past returns. “Industry year” is the mean adjusted return of other firms in the same industry-year group. “State year” is the mean adjusted return of other firms in the same state-year group. All other variables are defined as in Tables 1 and 2. Firm and year fixed effects are included. Standard errors are adjusted for heteroskedasticity and clustered at the state of incorporation level. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. (1) BC law Current ratio BC law × Current ratio

0.019 (1.39) -0.008∗∗∗ (-3.92) -0.009∗∗∗ (-3.34)

Cash ratio

(2)

(3)

0.016 (1.19)

-0.007 (-0.62)

-0.001 (-0.03) -0.172∗∗∗ (-2.98)

BC law × Cash ratio Excess cash BC law × Excess cash Age Age squared Industry year State year R2 N

-0.131∗∗ (-2.30) 0.021 (1.12) 0.432∗∗∗ (24.55) 0.310∗∗∗ (10.23) 0.190 51609

-0.082 (-1.54) 0.005 (0.26) 0.429∗∗∗ (24.72) 0.308∗∗∗ (10.94) 0.188 53026

43

0.013 (0.18) -0.263∗∗∗ (-3.61) 0.039 (0.13) -0.025 (-0.32) 0.396∗∗∗ (20.84) 0.283∗∗∗ (9.21) 0.179 35921

Table 9: Channels of Wasteful Spending Coefficient estimates and their t-statistics (in parentheses) are presented for the following regression model: Yijkl,t = αi + αt + β1 BC lawijkl,t + β2 (BC lawijkl,t × F Sijkl,t−1 ) + β3 F Sijkl,t−1 + γ 0 Xijkl,t + εijkl,t . In Panel A, column (1), Y is capital expenditure divided by total assets. In column (2), Y is overinvestment divided by total assets, where over-investment is constructed with the regression model of Richardson (2006) as described in the Appendix. In column (3), Y is asset growth, the percentage increase in total assets from one year to the next. In column (4), Y is PPE growth, the percentage increase in fixed assets. In column (5), Y is acquisition ratio, defined as the sum of the value of all acquisitions made by the firm in a given year divided by the firm’s market capitalization in that year. The acquisition data are collected from the Securities Data Corporation’s (SDC) database. In Panel B, column (1), Y is the overhead costs (selling, general, and administrative expenses) divided by sales. In column (2), Y is the advertising expenses divided by sales. In column (3), Y is the operating expenses divided by sales. In column (4), Y is the costs of goods sold (COGS) divided by sales. In column (5), Y is the number of employees divided by sales. In column (6), Y is wages, measured by the natural logarithm of labor and related expenses divided by the number of employees. All other variables are defined as in Tables 1 and 2. For the sake of brevity, coefficients on the control variables are not reported. Firm and year fixed effects are included. Standard errors are adjusted for heteroskedasticity and clustered at the state of incorporation level. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively.

Panel A: Investment Activities

Dependent Variable BC law Excess cash BC law × Excess cash R2 N

(1)

(2)

(3)

(4)

(5)

Capital Expenditure

Over-investment

Asset Growth

PPE Growth

Acquisition Ratio

-0.006 (-0.60) 0.052∗ (1.76) -0.135∗∗∗ (-5.35) 0.474 44071

-0.006 (-0.52) 0.572∗∗∗ (17.69) 0.004 (0.10) 0.364 44016

0.003∗ (1.79) 0.021∗∗∗ (4.33) 0.011 (1.36) 0.570 43931

0.001 (0.48) 0.089∗∗∗ (8.07) 0.000 (0.02) 0.305 24086

-0.002 (-0.95) 0.046∗∗∗ (5.62) 0.012 (1.30) 0.245 40355

Panel B: Cost Management (1)

(2)

(3)

(4)

(5)

(6)

Dependent Variable

Overhead Costs

Advertising Costs

Operating Expenses

COGS

Number of Employees

Wages

BC law

-0.002 (-0.93) 0.041∗∗∗ (3.78) 0.052∗∗∗ (4.13) 0.840 36940

0.000 (0.30) 0.010∗∗∗ (3.51) 0.001 (0.17) 0.853 16470

0.003 (1.14) -0.001 (-0.16) 0.059∗∗∗ (4.17) 0.765 38937

0.004 (1.39) -0.012 (-1.63) 0.034∗∗ (2.47) 0.708 43138

-0.000 (-0.51) -0.003 (-0.87) 0.008∗ (1.75) 0.606 43117

-0.001 (-0.22) 0.000 (0.00) -0.014 (-0.74) 0.896 5583

Excess cash BC law × Excess cash R2 N

44