Sarbanes-Oxley and Corporate Risk-Taking Leonce Bargerona, Kenneth Lehna,*, and Chad Zuttera a
Katz Graduate School of Business, University of Pittsburgh, Pittsburgh, PA 15260, USA
Draft: March 6, 2008
Abstract This paper empirically examines whether the Sarbanes-Oxley Act of 2002 (“SOX”) discourages risk-taking by publicly traded U.S. companies. Several provisions of SOX are likely to have this effect, including an expanded role for independent directors, an increase in director and officer liability, and rules related to internal controls. We find that several measures of risk-taking decline significantly for U.S. companies as compared with U.K. firms after SOX. The declines are related to several firm characteristics, including pre-SOX board structure, firm size, and R&D expenditures. In addition, the likelihood of initial public offerings (“IPOs”) occurring in the U.S. versus the U.K. declined significantly after SOX, with the decline being significantly higher for R&D intensive industries. Overall, the evidence supports the proposition that SOX discourages risk-taking by public U.S. companies. JEL classification: G1; G18; G30; G32; G38 Keywords: Sarbanes-Oxley; Legislative policy; Corporate risk taking; Investment policy
We thank Charles Calomiris, Allen Ferrell, Kate Litvak, and Peter Wallison, for valuable comments on an earlier version of this paper. The authors gratefully acknowledge funding support from the National Research Institute of the American Enterprise Institute. * Corresponding author. Tel.: + 1-412-648-2034; fax: + 1-412-624-2875. E-mail address:
[email protected] (K. Lehn).
1 Electronic copy available at: http://ssrn.com/abstract=1104063
1. Introduction Prominent scholars and policymakers have criticized the Sarbanes-Oxley Act of 2002 (“SOX”)1 for allegedly discouraging risk-taking by publicly traded U.S. companies. In one of his last interviews, Milton Friedman opined that “Sarbanes-Oxley says to every entrepreneur, ‘for God’s sake don’t innovate. Don’t take chances because down will come the hatchet.’”2 Former Federal Reserve chairman Alan Greenspan notes that “business leaders have been quite circumspect about embarking on major new investment projects”3 because of SOX. William Donaldson, former chairman of the Securities and Exchange Commission, (“SEC”) states that because of SOX “I worry about the loss of risk-taking zeal … [there is] a huge preoccupation with the dangers and risks of making the slightest mistake.”4 This paper empirically examines the relation between SOX and corporate risk-taking. Two provisions of SOX lead us to predict that the legislation has adversely affected risk-taking by publicly traded U.S. companies. First, by increasing the role of independent directors in corporate governance and expanding/criminalizing their liability for corporate misdeeds, SOX5 changed the incentives of directors to approve risky investments. In addition, SOX requires chief executive officers (“CEOs”) and chief financial officers (“CFOs”) to certify their companies’ financial statements and imposes criminal liability for knowing and/or willful violations of this provision. This, too, is predicted to dull the incentives of corporate officers to initiate and pursue risky projects. Second, Section 404 of SOX, which requires companies to evaluate and disclose the adequacy of their internal controls, is expected to discourage companies from investing in risky projects that increase the likelihood that the controls will be compromised and alleged to be 1
Pub. L. 107-204, 116 Stat. 745 (2002). Gerstein (2006). 3 Greenspan (2003). 4 Michaels (2003). 5 In addition to the Sarbanes-Oxley Act itself, SOX, as used in this paper, refers to SEC rules that implement SOX and changes in listing standards on the New York Stock Exchange and Nasdaq associated with SOX. The salient SEC rules and changes in listing standards are discussed in Section 2. 2
2 Electronic copy available at: http://ssrn.com/abstract=1104063
deficient. SEC guidance concerning compliance with Section 404 acknowledges that the risk of financial misstatements is directly related to the complexity of a firm’s operations, the extent to which it relies on specialized knowledge, and the degree to which its organizational structure is decentralized. Under the SEC guidance, as these factors increase, firms are expected to expend more resources and engage in more testing of internal controls to ensure compliance with Section 404. Insofar that these characteristics are related directly to the risk of a firm’s operations, the expected costs of complying with Section 404 are directly related to firm risk, thereby reducing the incentives of firms to invest in risky projects. This paper contributes to a growing academic literature that examines the economic effects of SOX, including the relation between SOX and corporate risk-taking (Cohen, Dey, and Lys (2007), Shadab (2007)). Our analysis consists of two parts. First, we examine whether several measures of corporate risk-taking have changed significantly for publicly traded U.S. companies, as compared with U.K. companies unaffected by SOX, since SOX was signed into law in 2002. Using data for a large sample of U.S. and U.K. firms during 1994-2006, we find that after SOX U.S. companies significantly reduced capital and R&D expenditures, as compared with their U.K. counterparts. In addition, U.S. firms significantly increased their holdings of cash and cash equivalents, which represent non-operating, low-risk investments, as compared with the U.K. firms. Finally, we find that the standard deviation of stock returns, a conventional measure of a company’s equity risk, declined significantly for U.S. firms compared with U.K. firms, after SOX. All of this evidence is consistent with the view that SOX has discouraged corporate risktaking and it complements previous research on this topic (Cohen, Dey, and Lys (2007)). We also find cross-sectional differences consistent with the prediction that SOX has discouraged corporate risk-taking. The changes in the risk-taking variables are significantly greater for large versus small U.S. firms, consistent with the view that the expected costs of complying with Section 404 are greater for firms characterized by more complexity. The changes
3
are also significantly greater for firms with high versus low R&D expenditures before SOX, consistent with the view that the expected costs of complying with Section 404 are directly related to the degree of specialized knowledge in a firm. The changes in cash holdings and stock price volatility are significantly greater for firms that did not have a majority of outside directors before SOX, and hence were most affected by the SOX-related rules governing the independence of boards, than for other firms. The changes in capital and R&D expenditures were not significantly different across these two groups of companies. Second, we examine data on initial public offerings (“IPOs”) in the U.S. and U.K. to test whether the likelihood that an IPO occurs in the U.S. versus the U.K. changed after SOX, and whether this likelihood changed more for IPOs of firms operating in high versus low risk industries as measured by R&D activity. Using a sample of 9,258 IPOs conducted in the U.S. and U.K. during 1990-2006, we find the probability of an IPO being conducted in the U.K. as opposed to the U.S. increased significantly after SOX. Furthermore, we find that the increase in this probability is directly related to R&D expenditures, indicating that high R&D firms were especially less likely to conduct an IPO in the U.S. after SOX. This evidence is consistent with the view that SOX discourages firms with risky operations from accessing U.S. public equity markets. The paper is organized as follows. Section 2 describes relevant provisions of SOX and discusses their relation to corporate risk-taking. Section 3 describes the sample and data. Section 4 contains empirical results on risk-taking by U.S. corporations after SOX. Section 5 presents evidence on the relation between R&D activity and the choice of conducting an IPO in the U.S. versus the U.K. after SOX. Section 6 contains concluding comments. 2. SOX and corporate risk taking We expect SOX to affect corporate risk-taking in two ways. First, by expanding the role played by independent directors and by imposing more liability, including criminal liability, on officers and directors for violations of securities laws, SOX affects the incentives of officers and
4
directors to initiate and approve risky investment projects. Second, Section 404, which requires companies to test and disclose the adequacy of their internal controls is expected to have a discouraging effect on corporate risk-taking. Each of these ways is discussed in turn. 2.1. Board structure and director/officer liability under SOX Independent directors. SOX increased the role played by independent directors in the governance of publicly traded U.S. companies. Section 301 of SOX required the SEC to adopt rules prohibiting national securities exchanges and associations from listing securities of companies that do not have audit committees complying with the requirements of SOX. Under SOX and the associated SEC rules, audit committees of publicly traded companies are required to consist entirely of independent directors.6 Furthermore, Section 407 of SOX requires publicly traded companies to disclose that at least one member of the audit committee is a “financial expert,” defined according to criteria developed by the SEC,7 or, if they do not have a financial expert on the committee, they are required to disclose why not. Shortly after SOX was signed into law, the NYSE and Nasdaq proposed changes to their listing standards that increased the role of independent directors in the governance of firms listed in the two markets. The SEC approved the proposed changes, with amendments, in the fall of 2003.8 The changes included the requirement that a majority of the board of directors of companies listed on the NYSE and Nasdaq be independent, based on new definitions of independent directors. Under the approved changes, the NYSE’s listing standard requires audit, nominating/corporate governance, and compensation committees of companies listed on the NYSE to consist entirely of independent directors. The Nasdaq’s listing standards also require audit committees of companies listed on Nasdaq to consist entirely of independent directors.9
6
SEC Release Nos. 33-8220 and 34-47654, April 9, 2003. SEC Release Nos. 33-8177 and 34-47235, January 23, 2003. 8 SEC Release No. 34-48745, November 4, 2003. 9 Nasdaq’s listing standards do not require companies listed on Nasdaq to have nominating or compensation committees. 7
5
The effect of these changes in regulations and listing standards is to increase the role of independent directors in the governance of publicly traded U.S. companies. Consistent with this view, Linck, Netter, and Yang (2007) find that boards of publicly traded U.S. corporations became larger and consisted of more outside directors after SOX. The increased role for independent directors after SOX is likely to adversely affect corporate risk-taking. Previous research has found that firms with high risk and high growth opportunities have significantly fewer outside directors than other firms. For example, Coles, Daniel, and Naveen find a direct relation between R&D expenditures and the proportion of a board consisting of inside directors. Similarly, Lehn, Patro, and Zhao (2008) find an inverse relation between market to book ratio, a proxy for growth opportunities, and the proportion of a board consisting of inside directors. The extant literature explains these relations in terms of information costs. The costs of monitoring managers’ performance are presumed to be increasing in a firm’s growth opportunities. For example, Smith and Watts (1992) note “It is difficult for shareholders or outside board members who do not have the manager’s specific knowledge to observe all the investments from which the manager chooses (p. 275).” Insofar that it is more costly for outside directors to obtain information about the quality of investment projects in risky, high growth firms than it is for inside directors, then, all else equal, it is optimal for risky, high growth firms to have fewer outside directors than less risky, low growth firms. Hence, firms in high growth industries (e.g., biotechnology and software) historically have had fewer outside directors than firms in low growth industries (e.g., textiles and food processing). By expanding the role played by independent directors in corporate governance, SOX effectively imposed a suboptimal board structure on firms that previously had relied less on independent directors. Based on existing evidence, this effect falls disproportionately on high risk firms with large growth opportunities. As these firms adapt their board structures to comply with SOX, the SEC rules, and the new listing standards, we expect their investments in risky
6
investment projects to decline, as the new independent directors incur the high costs of acquiring information about the projects. For this reason, we expect SOX to result in a decline in risk-taking activity by firms for which the change in board structure was a binding constraint. Expanded liability and increased penalties for officers and directors. SOX significantly increased the liability and penalties faced by officers and directors for violations of U.S. securities laws. For example, SOX imposes criminal liability for knowing violations of U.S. securities laws, with up to 25 years in prison for a single violation.10 In addition, SOX quadruples the maximum prison sentence from five to 20 years for each mail/wire fraud violation related to a securities fraud11 and imposes criminal liability and/or new penalties for other violations of U.S. securities laws.12 SOX also increased the liability of CEOs and CFOs in particular, by requiring, in Section 906, that they annually certify their companies’ financial statements. CEOs and CFOS face up to 10 years in prison for each “knowing” violation of this requirement, and up to 20 years in prison for each “willful” violation of the requirement.13 The creation of criminal liability for securities violations, along with other increases in liability and penalties as specified in SOX, are expected to increase the risk aversion of corporate officers and directors. Previous research has found that the likelihood a firm is charged with accounting fraud by the SEC is directly related to a firm’s growth opportunities, as measured by R&D expenditures (Gerety and Lehn (1997)). This result is consistent with the view that fraud is more likely to occur in industries with substantial intangible assets characterized by high monitoring costs. Insofar that the likelihood of fraud is directly related to the risk of a firm’s 10
The Sarbanes-Oxley Act of 2002, Section 807. Ibid, Section 903. 12 For example, SOX imposes enhanced penalties for conspiracies or attempts to violate the mail and wire fraud law (Section 902), it revises and creates new obstruction of justice crimes related to tampering with evidence and obstructing justice (Section 802), and it imposes a requirement on auditors of public companies to maintain audit and work papers for a minimum of five years, with criminal liability and imprisonment of up to 10 years for anyone who “knowingly and willfully” violates the requirements (Section 802). 13 The Sarbanes-Oxley Act, Section 906. 11
7
assets and its growth opportunities, then the increase in expected penalties for fraud under SOX is expected to reduce the incentives of officers and directors to initiate and approve risky projects that are hard to monitor, such as projects involving R&D expenditures and other investments in intangible assets. 2.2. Internal control provisions of SOX Section 404 of SOX, perhaps the most contentious part of the legislation, directs the SEC to adopt rules requiring companies to evaluate and disclose the adequacy of their internal financial controls in annual reports. In 2003, the SEC adopted rules implementing this requirement of SOX, which took effect in 2004 for “accelerated filers,” defined as companies with market capitalizations of $75 million or more.14 Under the SEC rules, the affected companies are required to disclose the following information about internal controls in their annual reports: a. A statement of management’s responsibility for establishing and maintaining adequate internal controls; b. Identification of the framework used by management to evaluate the adequacy of the internal controls; c. A statement as to whether or not the system of internal controls as of year-end is effective; d. Disclosure of any “material weaknesses” in the system of internal controls; and e. A report by the company’s external auditor attesting to management’s assessment of the firm’s internal controls.
In response to feedback it received on the rules, the SEC issued a release in 2007 that attempted to provide guidance on how issuers should comply with the rules.15 In the introduction to the release, the SEC describes its guidance as entailing a “top-down, risk-based approach:” “The Interpretive Guidance is organized around two broad principles. The first principle is that management should evaluate whether it has implemented controls that adequately address the risk that a material misstatement of the financial statements would not be prevented or detected in a timely manner. The 14 15
SEC Release No. 33-8238, June 11, 2003. SEC Release Nos. 33-8810; 34-55929; June 20, 2007.
8
guidance describes a top-down, risk-based approach to this principle, including the role of entity-level controls in assessing financial reporting risks and the adequacy of controls. The guidance promotes efficiency by allowing management to focus on those controls that are needed to adequately address the risk of a material misstatement of its financial statements … management can focus its evaluation process and the documentation supporting the assessment on those controls that it determines adequately address the risk of a material misstatement of the financial statements (italics added).”16
The SEC then describes the second principle underlying its guidance: “The second principle is that management’s evaluation of evidence about the operation of its controls should be based on its assessment of risk. The guidance provides an approach for making risk-based judgments about the evidence needed for the evaluation. This allows management to align the nature and extent of its evaluation procedures with those areas of financial reporting that pose the highest risks to reliable financial reporting … As a result, management may be able to use more efficient approaches to gathering evidence, such as self-assessments, in low-risk areas and perform more extensive testing in highrisk areas (italics added).”17 In short, the SEC’s guidance emphasizes that management should have the flexibility to design evaluation processes in accordance with the risks of financial misstatements – where the risks are greater, more extensive testing and evaluation is expected. The SEC’s interpretive guidance goes on to describe characteristics of companies that, according to the SEC, give rise to a higher likelihood of financial reporting misstatements: “The methods and procedures for identifying financial reporting risks will vary based on the characteristics of the company. These characteristics include, among others, the size, complexity, and organizational structure of the company and its processes and financial reporting environment … For example, to identify financial reporting risks in a larger business or a complex business process, management’s methods and procedures may involve a variety of company personnel, including those with specialized knowledge. These individuals, collectively, may be necessary to have a sufficient understanding of GAAP, the underlying business transactions and the process activities, including the role of computer technology, that are required to initiate, authorize, record, and process transactions. In contrast, a small company that operates on a centralized basis with less complex business processes and with little change in the risks or processes, management’s daily involvement with the business may provide it with adequate knowledge to appropriately identify financial reporting risks.”18 16
Ibid, pp. 4-5. Ibid, p. 5. 18 Ibid, pp. 13-14. 17
9
Later in the release, the SEC provides additional commentary on factors affecting “whether there is a reasonable possibility that a deficiency, or a combination of deficiencies, will result in a misstatement of a financial statement amount or disclosure,”19 citing, among other factors, “the susceptibility of the related asset or liability to loss or fraud” and “the subjectivity, complexity, or extent of judgment required to determine the amount involved.”20 The SEC’s guidance and identification of firm characteristics associated with a greater risk of financial misstatements indicate that greater evaluation and testing of internal controls is required for firms with activities involving specialized knowledge, decentralized organizational structures, and complex transactions accounted for in ways involving large amounts of subjectivity and judgment. Generally speaking, these are attributes of firms engaged in risky activities involving large growth opportunities. Hence, the incidence of the costs associated with Section 404 is expected to fall disproportionately on these types of firms. Consistent with this view, previous academic research finds that firms disclosing problems with internal controls before the Section 404 rules became effective tend to be younger, more complex, cash constrained, and faster growing than other firms (Ashbaugh-Skaife (2007), Doyle, Ge, and McVay (2007)). A predictable response of firms to the Section 404 rules is to curtail investments in risky investment projects that increase the expected costs of complying with the Section 404 rules adopted by the SEC. Anecdotal evidence of this is found in a letter, commenting on the Section 404 rules, from the Biotechnology Industry Organization to the SEC, stating “Many emerging biotech companies are directing precious resources from core research and development of new therapies for patients due to overly complex controls or unnecessary evaluation of controls.”21 2.3. Summary
19
Ibid, p. 35. Ibid, p. 36. 21 Eisenberg (2007). 20
10
We expect that several provisions of SOX, including those affecting the structure of boards, the liability and penalties associated with securities fraud, and rules related to internal controls, discourage corporate risk-taking. Several academic papers find evidence consistent with this prediction. Graham and Harvey (2003) provide survey evidence that CFOs believe SOX adversely affects corporate risk-taking by “providing an environment in which second-guessing of actions taken by management is more prevalent … and … altering compensation incentives, which might affect risk taking motivations (p. 36).” Consistent with this view, Cohen, Dey and Lys (2007) find evidence of reduced risk-taking by U.S. companies after SOX and link it to a reduction in the use of stock options in executive compensation plans following SOX. Wintoki (2007) finds that residual stock returns around announcements of events related to the passage and adoption of SOX and related regulations vary inversely with firms’ growth opportunities and the uncertainty of their business environments. Litvak (2007) also finds stock price evidence consistent with the view that SOX has adversely affected corporate risk-taking. We add to the literature on the relation between SOX and corporate risk-taking in two ways. First, similar to Cohen, Dey, and Lys (2007), we directly examine whether there was a significant change in investments made by publicly traded U.S. firms and their stock price volatility after SOX. Our analysis, which complements Cohen, Dey, and Lys (2007), differs from their analysis in several ways, including (i) the use of a benchmark sample of U.K. firms to control for intertemporal changes in the key variables, (ii) an analysis of cash holdings, in addition to capital expenditures and R&D, as a proxy for risk-taking, and (iii) a focus on firm characteristics that might explain the change in risk-taking observed after SOX, including board structure, size, and R&D intensity.22 Second, we examine whether the probability that an IPO occurs in the U.S. as opposed to the U.K. changes significantly after SOX, and whether the change in probability is
22
Changes in compensation plans, of course, are likely to be endogenous to the changed incentives for risktaking after SOX, as discussed by Cohen, Dey, and Lys (2007)).
11
different for firms depending on their risk, as measured by the research intensity of the industries in which they operate. The samples and data used in the empirical analysis are described in the next section. 3. Sample and data 3.1. Sample The sample consists of 2,290 U.S. and U.K. publicly traded corporations, including 1,825 U.S. corporations and 465 U.K. corporations, for which sufficient data exists on the Thomson One Banker database. These firms represent all U.S. and U.K firms in the database for which there is consistent time series data, spanning the adoption of SOX, on the key variables used in the analysis. Specifically, to be included in the sample we required that the following variables were available for at least twelve years during the 1994-2006 period: assets, capital expenditures, and cash holdings. Because data on R&D expenditures is considerably sparser, existing for only 885 firms (753 U.S. firms, 132 U.K. firms), we do not require data on R&D expenditures for each period in order for firms to be included in the sample. The sample consists of both large and small companies. For example, 237, or 13%, of the 1,825 U.S. companies in the year 2000 had assets of more than $5 billion and 248, or 13.6%, had assets of less than $50 million. For the UK 53, or 11.4%, of the 465 companies in the year 2000 had assets of more than £2.5 billion and 60, or 12.9%, had assets of less then £25 million. The sample spans a broad array of industries: 71 different 2-digit SIC codes are represented. To determine whether the industry distributions of the two samples are similar, we calculated the correlation coefficient between the number of firms in each industry in the U.S. and U.K. samples. Across the 71 industries, the correlation coefficient is 0.54 and significant at the 0.001 level. Hence, the industry distributions of the two samples appear to be similar. 3.2. Data Financial accounting data for the sample was collected from the Thomson One Banker database. For each firm in each year during the period of 1994 through 2006 we collected the
12
following data: capital expenditures, R&D expenditures, cash holdings, book value of assets, market value of assets, EBIT, and total debt. We express the variables CAPEX, capital expenditures, R&D, research and development expenditures, CASH, cash and short term securities, and EBIT, earnings before interest and taxes, as ratios to the average book value of assets. We express DEBT, total debt, as a ratio to the average market value of assets. Throughout the analysis, because SOX was passed in late July 2002, we define each year as August to July, i.e. the 2002 year begins on August 1, 2001 and ends on July 31, 2002. Data on stock prices, which is used to compute the stock-based risk measure, are collected from the Datastream database for both the U.S. and U.K. samples. Daily stock returns over the period of 1994 through 2006 are calculated from the daily adjusted stock prices. The variable STD is the standard deviation of returns over the August-July year. 4. Empirical results on investment decisions and stock price volatility after SOX To test whether risk-taking by publicly traded U.S. companies declined significantly after SOX, we use both univariate and multivariate regression analysis. Each analysis is discussed in turn. 4.1. Summary statistics and univariate tests The three panels of Table 1 provide summary statistics for the variables of interest based on different divisions of the sample. Panel A tabulates the means and medians over the entire sample period for the U.S sample and the U.K. sample and tests for differences between the two countries. The key variables for our analysis, CAPEX, R&D, CASH, and STD are all significantly higher in the U.S. than the U.K. The control variables, EBIT, M/B, and DEBT are also higher in the U.S. With the exception of the mean values for EBIT these differences are all significant at the 0.01 level. Panel B examines the sample of U.S. firms exclusively and compares the pre-SOX 19942002 period to the post-SOX 2003-2006 period. The results show that the four key variables all change from the pre-SOX to the post-SOX period. Average CAPEX, R&D, and STD decreased
13
while CASH increased. All four changes, significant at the 0.01 level, are consistent with the hypothesis that risk-taking in U.S. firms decreased significantly after SOX. Interestingly, as demonstrated in Panel C, not all of these changes are mirrored in the U.K. sample. Panel C replicates Panel B using U.K. instead of U.S. firms. As in the U.S., average CAPEX decreases significantly in the U.K.; however, the average changes for the U.K sample in R&D, CASH, and STD are in the opposite direction from the respective changes for the U.S. sample. Although the change is not significant, average R&D for the U.K. sample increases in the post-SOX period. Average CASH decreases significantly for the U.K. sample in the post-SOX period. The median increase in average STD for the U.K. sample is also significant. Taken together these univariate results suggest that the decrease in risk-taking after SOX is unique to the U.S. sample. In the next section we investigate this question in a setting that allows for direct tests of the U.S. vs. U.K. sample while controlling for observed and unobserved firm specific characteristics. 4.2. Basic regression results To test more thoroughly whether there was a significant change in the investment decisions and stock price volatility of U.S. firms after SOX, we estimate a series of regression equations. In each set of equations, we estimate separate regressions for the four variables of interest – CAPEX, RD, CASH, and STD. Using a firm fixed effects model, each of these variables is regressed on a set of independent variables for a panel dataset consisting of all 2,290 U.S. and U.K. firms over the period of 1994 through 2006. The regressions routinely include measures of EBIT, M/B, and DEBT (used only in the STD equation) and their interactions with a U.S. dummy. Regressions also include two dummy variables (US Post-SOX, which takes the value of one if the observation is a U.S. firm in the post-SOX period and zero otherwise, and UK PostSOX, which takes the value of one if the observation is a U.K. firm in the post-SOX period and zero otherwise). To test if the variables of interest changed significantly for U.S. firms relative to
14
U.K. firms in the post-SOX period, an F-test is conducted to determine if there is a significant difference in the coefficients on US Post-SOX and UK Post-SOX. Table 2 contains the regression results for two sets of regressions, one in which the variables of interest are regressed only on US Post-SOX and UK Post-SOX and the other in which they are regressed on US Post-SOX, UK Post-SOX, and the control variables. The results show that all four dependent variables changed significantly for U.S. firms in the post-SOX period in the directions consistent with a reduction in risk-taking. For example, column (1) reports the regression results in which CAPEX is regressed only on the two dummy variables for U.S. and U.K. firms in the post-SOX period. The coefficient on both variables is negative and highly significant, indicating that CAPEX was significantly lower for firms in both countries in the post-SOX period. The coefficient on US Post-SOX is more negative (-0.0189 versus -0.0150), indicating that the decline for U.S. firms was greater than the corresponding decline for U.K. firms. The F-statistic of 9.73 indicates the difference in the coefficients is highly significant, consistent with the view that there was a significant reduction in risk-taking, as measured by CAPEX, for U.S. firms after SOX. Table 2 presents similar results for RD, CASH, and STD. In the regressions in which these variables are regressed only on US Post-SOX and UK Post-SOX, there is a significant difference in the coefficients on the two variables, all consistent with the prediction that SOX resulted in a decline in risk-taking by publicly traded U.S. companies. In the RD equation, the coefficients on US Post-SOX and UK Post-SOX are significant and take different signs – the coefficient on US Post-SOX is negative (-0.0019) and highly significant, whereas the coefficient on UK Post-SOX is positive and significant at the 0.05 level. Hence, R&D spending fell significantly for U.S. firms in the post-SOX period, while it increased significantly for U.K. firms. The F-statistic of 16.49 reveals that the difference in these coefficients is highly significant.
15
The coefficients on US Post-SOX and UK Post-SOX also take opposite signs and are highly significant in the CASH equations. The coefficient is positive (0.0149) for U.S. firms and negative (-0.0105) for U.K. firms, indicating that the cash holdings of U.S. firms increased significantly in the post-SOX period and declined significantly for U.K. firms. The F-statistic of 69.94 reveals the difference in coefficients is highly significant. Because cash is a low risk asset, this result also is consistent with the view that SOX has resulted in a reduction in corporate risktaking. In the STD equation, the coefficient on US Post-SOX is negative (-0.0062) and highly significant and the coefficient on UK Post-SOX is positive (0.0002) and not significant. The Fstatistic of 451.88 indicates the difference in the two coefficients is highly significant, indicating that there was a significant decline in the volatility of U.S. stock returns as compared with U.K. stock returns after SOX. This, too, is consistent with the view that SOX has discouraged corporate risk-taking by U.S. firms. Table 2 also contains regression results for the equations that include the three control variables, along with US Post-SOX and UK Post-SOX, as independent variables. With the exception of the CAPEX equation, the F-statistics continue to reveal a significant difference in the coefficients on US Post-SOX and UK Post-SOX, all in the direction consistent with reduced risk-taking by U.S. firms after SOX. Table 3 presents similar results to those shown in Table 2 with one modification. Instead of using the entire pre-SOX period of 1994-2002 as the reference point for the post-SOX period, we decompose the pre-SOX period into two sub periods, 1994-1997, referred to as Period 1, and 1998-2001, referred to as Period 2.23 Data from 2002, the year SOX was adopted, is excluded. We then use the 1994-1997 period as the reference point for both Period 2 and the post-SOX period. Hence, in the regression results reported in Table 3, we include four dummy variables: (i) a
23
For brevity, we report only the results for the regressions including the control variables.
16
dummy variable for U.S. firms in Period 2 (“US Period2”), (ii) a dummy variable for U.K. firms in Period 2 (“UK Period2”), (iii) US Post-SOX, and (iv) UK Post-SOX. As in Table 2, we test whether the coefficients on US Post-SOX and UK Post-SOX are significantly different, which reveals whether there is a significant change in risk-taking by U.S. companies in the post-SOX period as compared with the reference period of 1994-1997. We also test whether the difference in the coefficients on US Post-SOX and US Period2 is significantly different than the corresponding difference in the coefficients on UK Post-SOX and UK Period2. This test reveals if there has been a significant change in risk-taking by U.S. companies in the post-SOX period as compared with the pre-SOX period of 1998-2001. Table 3 shows that the difference in the coefficients on U.S. post-SOX and U.K. post-SOX is highly significant in all four equations, indicating that there was a decline in risk-taking by U.S. firms in the post-SOX period as compared with the pre-SOX period of 1994-1997. We also find that, except for CAPEX, the difference in the coefficients on US Post-SOX and US Period2 is significantly different than the difference in U.K. Post-SOX and U.K. Period2 at the 0.01 level, indicating that there also was a decline in risk-taking by U.S. firms in the post-SOX period as compared with the pre-SOX period of 1998-2001. We next present regression results testing whether the results reported in Tables 2 and 3 vary significantly with three firm characteristics – board structure, size, and R&D intensity. 4.3. Regression results partitioned by board structure To test whether the change in risk-taking by U.S. firms in the post-SOX period is related to the increased role played by independent directors after SOX, we replicate the results in Table 2, with one change. We classify U.S. firms into two groups – (i) firms that had a minority (up to and including 50%) of outsiders on the board as of the end of 2001 (i.e., the year before the adoption of SOX) and (ii) firms that had a majority of outsiders on the board as of the same time. We collect data on the officers and directors from Compact Disclosure CD’s. Any director who is
17
also an officer of the company is defined as an inside director; all others are classified as outsiders.24 We then define two dummy variables: US Minority Outsiders Post-SOX, which takes the value of one for U.S. firms in the first group during the Post-SOX period, and zero otherwise; and US Majority Outsiders Post-SOX, which takes the value of one for U.S. firms in the second group during the Post-SOX period. These two dummy variables are then substituted for US Post-SOX in the regression models reported in Table 2. To test whether the change in risk-taking for U.S. firms differs according to their pre-SOX board structures, we calculate an F-statistic testing whether there is a significant difference in the coefficients on US Minority Outsider Post-SOX and US Majority Outsiders Post-SOX. Table 4 reports the regression results.25 It shows that there is a significant difference in the coefficients on the two variables in the equations in which CASH and STD are the dependent variables. In the CASH equation, the coefficients on firms with outsiders comprising a minority and a majority of the boards before SOX are 0.0404 and 0.0225, respectively, and both are highly significant, indicating that both groups of firms significantly increased cash holdings in the postSOX period. The F-statistic testing the difference between the two coefficients is 7.89, revealing that the increase in cash holdings is significantly greater for firms that had a minority of outsiders on the board before SOX. These are the firms most affected by the SOX-related change in listing standards requiring that boards be comprised of a majority of independent directors. Hence, these results are consistent with the prediction that the increased role played by independent directors after SOX has contributed to a reduction in risk-taking by publicly traded U.S. companies. In the STD equation, the coefficients on the two groups of firms are -0.0081 and -0.0067, respectively, and both are highly significant, indicating that firms with both a minority and a 24
Our definition does not distinguish between outside directors who have contractual relations with the company and other outside directors. 25 For brevity, we report only the results for the regressions including the control variables.
18
majority of outside directors before SOX experienced significant declines in stock price volatility after SOX. The F-statistic testing the difference in the two coefficients is 4.87, revealing that the decline in stock price volatility was greater for firms with a minority of outside directors before SOX, i.e., the firms most affected by the increased role of independent directors after SOX. This result also is consistent with the prediction that the reduction in risk-taking by publicly traded U.S. companies after SOX is, at least in part, due to the increased importance of independent directors after SOX. No significant difference exists in the coefficients on these two variables in the CAPEX and RD equations. Both variables enter with negative and highly significant coefficients in the two equations. Although the coefficient on US Minority Post-SOX is more negative in both equations, the corresponding F-statistics are 0.01 and 2.28 in the CAPEX and RD equations, respectively, revealing no significant difference in the two sets of coefficients. These results do not support the prediction that the expanded role for independent directors after SOX has discouraged risk-taking by publicly traded U.S. companies. 4.4. Regression results partitioned by firm size We also partition U.S. firms by size, as measured by the book value of their assets as of yearend 2001, the year prior to SOX. The median value of assets for U.S. firms in our sample for 2001 is $556 million. We define a U.S. firm as large, for all firm years, if it had more than $556 million in assets in 2001 and small, for all firm years, if it had less than $556 million in assets in 2001. We then create two dummy variables, (i) US Large Post-SOX, which takes the value of one if the firm is a “large” U.S. firm and the year is 2003-2006, and zero otherwise; and (ii) US Small Post-SOX, which takes the value of one if the firm is a “small” U.S. firm and the year is 19942002, and zero otherwise. We substitute these two dummy variables for US Post-SOX in the same regression models reported in Table 2. To test whether the change in risk-taking after SOX
19
differs significantly for large versus small firms, we calculate an F-statistic corresponding to the difference in the two coefficients. The results are reported in Table 5.26 Except for the CASH equation, the difference in the coefficients on the two variables are highly significant. In the CAPEX and RD equations, the direction is consistent with the prediction that large firms, because of their complexity, are more likely to reduce risk-taking so as to reduce the expected costs of complying with Section 404. In the CAPEX equation, the coefficients on US Large Post-SOX and US Small Post-SOX are -0.0178 and -0.0137, respectively, and both are highly significant. The F-statistic corresponding to the difference in the two coefficients is 9.31, indicating that larger U.S. firms reduced their capital expenditures more than smaller U.S. firms in the post-SOX period, a result consistent with the prediction. Similarly, in the RD equation, the coefficients on the two variables are -0.0042 and -0.0014, respectively, and both are highly significant. The F-statistic testing the difference in the two coefficients is 7.36, revealing that larger U.S. firms reduced R&D spending more than smaller U.S. firms. Again, this result is consistent with the prediction. The coefficients on the two variables in the CASH equation are 0.0248 and 0.0193, respectively, and both are highly significant. The F-statistic testing the difference in the two coefficients is 3.03 (with a p-value of 0.082), indicating the difference is not significant at the 0.05 level. In the STD equation, the coefficients on the two variables are -0.0053 and -0.0085, respectively, and both are highly significant. The F-statistic testing the difference in the two coefficients is 78.12, indicating that the stock price volatility of smaller U.S. companies declined more in the post-SOX period than the stock price volatility of larger U.S. companies. These results are not consistent with the prediction that the expected costs of complying with Section 404 are directly related to firm size. 4.5. Regression results partitioned by R&D intensity
26
Again, for brevity, we report only the results corresponding to the regressions that include the control variables.
20
Finally, we partition the sample of U.S. firms by R&D spending during 2001. We create two dummy variables: (i) US R&D Post-SOX, which takes the value of one if the firm is a U.S. firm in the post-SOX period and reported nonzero R&D spending in 2001, and zero otherwise; and (ii) US No R&D Post-SOX, which takes the value of one if the firm is a U.S. firm in the post-SOX period and reported zero R&D spending in 2001, and zero otherwise. These two dummy variables are substituted for US Post-SOX in the regression model reported in Table 2. An F-statistic is then calculated to test whether there is a significant difference in the coefficients on the two variables. Table 6 reports the regression results.27 In all three equations the coefficients on the two variables are significantly different and in the directions consistent with the prediction that firms characterized by a large amount of specialized knowledge, as measured by R&D, are more likely to reduced risk-taking after SOX because of the Section 404 rules. In the CAPEX equation, the coefficients on US R&D Post-SOX and US No R&D Post-SOX are -0.0204 and -0.0128, respectively, and both are highly significant. The F-statistic of 37.73 indicates that U.S. firms with R&D expenditures reduced CAPEX significantly more in the postSOX period than U.S. firms without R&D expenditures. In the CASH equation, the coefficients on the two variables are 0.0282 and 0.0182, respectively, and both are highly significant. The F-statistic is 8.70, revealing that firms reporting nonzero R&D expenditures before SOX increased their cash holdings after SOX by significantly more than other firms. This result is consistent with the prediction that firms characterized by large amounts of specialized knowledge were more likely to reduce their risk after SOX. Finally, in the STD equation, the coefficients on the two variables are -0.0079 and -0.0062, respectively, and both are highly significant. The F-statistic corresponding to the difference in the
27
As in Tables 4 and 5, for brevity, we report results only for the regressions that include the control variables. We exclude the R&D equation because the partition is based on whether or not firms reported R&D expenditures in 2001. The observed decrease in R&D expenditures is driven by firms with prior R&D expenditures, thereby creating a biased comparison.
21
two coefficients is 19.63, indicating that firms with nonzero R&D spending before SOX experienced a significantly larger decline in stock price volatility after SOX. This result also is consistent with the prediction that SOX caused firms with more specialized knowledge to curtail risk-taking by a larger amount than other firms. 5. R&D expenditures, IPOs, and SOX In addition to examining whether measures of risk-taking by U.S. companies changed significantly after SOX, we examine whether firms operating in risky industries were more likely to go public in the U.K. versus the U.S. after SOX. We first describe the sample and data used in the analysis and then present the results. 5.1. Sample and data The sample of U.S. and U.K. IPOs is collected from the Securities Data Company’s (SDC) Global New Issues Database. The sample consists of all completed common stock IPOs for the period 1990 through 2006 where the listing exchange is located in either the U.S. or U.K. The sample includes 9,262 IPOs, consisting of 1,882 U.K. IPOs (20% of the sample) and 7,380 U.S. IPOs (80% of the sample). Table 7 contains summary information on characteristics of the IPOs before and after SOX. Of the 7,380 U.S. IPOs, 6,417, or 87%, occurred before SOX. Of the 1,882 U.K. IPOs, 1,284, or 68%, occurred before SOX. Hence, a substantially higher percentage of U.K. IPOs occurred after SOX than in the U.S. Of the 6,417 U.S. IPOs that occurred before SOX, 2,472, or 39%, were done by firms in high R&D industries.28 Of the 963 U.S. IPOs after SOX, only 283, or 29%, were done by firms in high
28
We define high, moderate, and low R&D industries based on total U.S. industry R&D expenditure over the benchmark period of 1994-1997. Seven 2-digit SIC code industries account for 92.1% of R&D expenditure in the U.S. during this period. We disaggregate the seven 2-digit SIC code industries into their thirty-seven 3-digit SIC code industry components. We classify the fifteen of these 3-digit SIC code industries that account for at least 1% of the total U.S. R&D expenditure during the period as high R&D industries. We define the remaining twenty-two of these 3-digit SIC code industries as moderate R&D industries. We also classify the seventeen remaining 2-digit SIC code industries with at least $100 million in total R&D expenditure during the period as medium R&D industries. Finally, we classify the fifty-four
22
R&D industries. Hence, the percentage of U.S. IPOs accounted for by firms in high R&D industries declined substantially after SOX. In contrast, there was little change in this percentage for the U.K. sample, from 24% before SOX (i.e., 314 of 1,284 IPOs) to 22% after SOX (i.e., 154 of 598 IPOs). Similarly, the percentage of U.S. IPOs accounted for by firms in low R&D industries increased substantially after SOX, from 39% (i.e., 2,501 of 6,417 IPOs) to 52% (i.e., 501 of 963 IPOs). The percentage of U.K. IPOs accounted for by firms in low R&D industries actually declined from 53% before to 49% after SOX. Hence, the difference in these percentages across the U.S. and U.K. changed substantially from before to after SOX. Total proceeds raised in IPOs differ substantially across the two samples. Over the entire period of 1990-2006, total proceeds raised in U.S. IPOs were $911 billion, as compared with $186 billion in U.K. IPOs. Of the total proceeds, 75% were raised before SOX for the U.S. sample, versus 79% for the U.K. sample. The percentage of proceeds raised by firms in high R&D industries declined substantially for the U.S. sample after SOX, from 34% (i.e., $229,088 of $680,840) to 17% (i.e., $39,259 of $230,286). In contrast, this percentage increased slightly in the U.K., from 10% (i.e., $14,690 of $147,639) to 12% (i.e., $4,484 of $38,687). A dramatic difference exists in the percentage of proceeds raised by firms in low risk industries in the U.S. and U.K. before and after SOX. For the U.S. sample of IPOs, this percentage increased substantially from 39% before SOX (i.e., $263,519 of $680,840) to 65% after SOX (i.e., $148,609 of $230,286). For the U.K. sample, this percentage decreased substantially, from 59% before SOX (i.e., $87,203 of $147,639) to 40% after SOX (i.e., $15,512 of $38,687).
2-digit SIC code industries with less than $100 million in total R&D expenditure during the period as low R&D industries.
23
To the extent that SOX has dampened risk-taking by U.S. corporations, it is possible that the U.S. IPO market would soften as a result. In particular, private firms in the U.S. may opt more often to remain private vis-à-vis go public in the post-SOX era. Figure 1 shows the relative percentages of IPOs going public in the U.S. versus the U.K. each year during the period of 1990 through 2006. It can be seen that the relative percentage of U.K. IPOs has increased dramatically, suggesting that more U.S. firms are opting to remain private in response to the SOX legislation. It is possible, however, that the change in relative percentages shown in Figure 1 is not due to a reduction in U.S. IPOs, but rather an increase in the number of U.K. IPOs vis-à-vis the number of U.S. IPOs. Figure 2 reveals that the number of IPOs in the U.S. has decreased substantially post SOX, whereas the number of IPOs in the U.K. is not remarkably different from historic levels. Thus, Figure 2 strengthens the notion that fewer U.S. firms are going public after SOX. Another potential explanation for the rise in the relative percentage of U.K. IPOs shown in Figure 1 and the decrease in U.S. IPOs shown in Figure 2 is that the reduction in U.S. IPO activity occurred because of a downturn in the U.S. stock market, not SOX. Although there is considerable empirical evidence relating IPO volume to stock market performance, for the U.S. stock market to explain the relative changes shown in Figures 1 and 2 between U.S. and U.K. IPO markets the U.K. stock market cannot be highly correlated with the U.S. stock market. On the contrary, Figure 3 shows that the U.K. stock market is highly correlated with the U.S. stock market, represented by the FTSE All Share and the CRSP Value Weighted indices respectively. The figure shows that the two stock markets run in parallel through time. The correlation in returns over the 1990 through 2006 time period is 0.73. Moreover to the extent that a difference in stock market performances explains the difference in IPO markets, Figure 3 suggests that the U.S. IPO market, and not the U.K. IPO market, should be gaining in relative strength since the U.S. stock market has been outpacing the U.K. stock market during the post-SOX period.
24
Overall, the summary data discussed above suggests that IPO activity has decreased in the U.S. vis-à-vis the U.K. since SOX, especially among firms operating in high R&D industries. We next turn to empirical tests of this proposition. 5.2. Empirical results Table 8 presents results from a logit model in which we estimate the likelihood that a firm goes public in the U.K. versus the U.S. over the period of 1990-2006. The dependent variable takes the value of 1 if the IPO occurs in the U.K. and 0 if the IPO occurs in the U.S. The independent variables include the log of proceeds raised in the IPO, dummy variables for whether or not the firm is in a high or moderate R&D industry, a dummy variable if the IPO occurred after SOX, and two interaction variables consisting of the product of the post-SOX dummy variable and the dummy variables for firms in high and moderate R&D industries. All of the independent variables enter with estimated coefficients that are significant at the 0.01 level. Log of proceeds enters with a negative coefficient, indicating that larger firms are more likely to go public in the U.S. Both dummy variables for firms operating in high and moderate R&D industries enter with negative and significant coefficients (-0.865 and -0.423, respectively), indicating that compared with low R&D firms these firms were more likely to do IPOs in the U.S. The dummy variable for post-SOX IPOs enters with a positive coefficient, indicating that in the years after SOX there was an increase in the likelihood that firms would do IPOs in the U.K. The two variables that interact the post-SOX dummy variable with high and moderate R&D enter with positive coefficients (0.542 and 0.707, respectively), indicating that after SOX, firms operating in high and moderate R&D industries were more likely to do IPOs in the U.K. This result is consistent with the proposition that SOX has dampening corporate risk-taking by publicly traded U.S. companies and that firms operating in risky industries are less likely to conduct an IPO (in the U.S.) than their U.K. counterparts (in the U.K.). 6. Concluding comments
25
This paper finds empirical support for the proposition that the Sarbanes-Oxley Act of 2002 has had a negative effect on risk-taking by publicly traded U.S. companies. We find evidence of this in both the investment decisions of U.S. firms and their stock price volatility, as compared with U.K. firms, after SOX. The results complement results found in Cohen, Dey, and Lys (2007), who find that the reduction in risk-taking is associated with changes in executive compensation contracts after SOX. We find that the changes are related to various firm characteristics, including pre-SOX board structure, firm size, and the degree of specialized knowledge, as measured by R&D expenditures. We make no normative judgment as to whether the reduction in risk-taking after SOX is socially efficient. However, if on average firms had adopted value-maximizing governance structures before SOX, and we presume they did, then the changes in risk-taking induced by SOX are inefficient. Furthermore, insofar that a major source of agency problems is that managers are more risk averse than stockholders want them to be, then SOX may be aggravating this problem, leading to further value destruction. These issues await further investigation.
26
References Ashbaugh-Skaife, Hollis, Collins, Daniel W., and Kinney, William R., “The discovery and reporting of internal control deficiencies prior to SOX-mandated audits,” Journal of Accounting and Economics 44 (2007), 166-192. Cohen, Daniel A., Dey, Aiyesha and Lys, Thomas Z., “The Sarbanes-Oxley Act of 2002: Implications for compensation contracts and managerial risk-taking” (November 9, 2007). Available at SSRN: http://ssrn.com/abstract=568483 or DOI: 10.2139/ssrn.568483. Coles, Jeffrey L., Daniel, Naveen D., and Naveen, Lalitha, “Boards: does one size fit all?,” Journal of Financial Economics 87 (2008), 329-356. Doyle, Jeffrey, Ge, Weili, and McVay, Sarah, “Determinants of weaknesses in internal control over financial reporting,” Journal of Accounting and Economics 44 (2007), 193-223. Eisenberg, Alan F., comment letter to the SEC, July 12, 2007. Gerety, Mason and Lehn, Kenneth, “The causes and consequences of accounting fraud,” Managerial and Decision Economics, (November-December 1997), 587-599. Gerstein, Josh, “Friedman, 93, set to unleash power of choice,” New York Sun, March 22, 2006. Graham, John R., Harvey, Campbell R., and Rajgopal, Shiva, “The economic value versus reported earnings trade-off and voluntary disclosure,” working paper, Duke University, December 22, 2003. Greenspan, Alan, Testimony before the Committee on Financial Services, U.S. House of Representatives, July 15, 2003. Lehn, Kenneth, Patro, Sukesh, and Zhao, Mengxin, “Determinants of the size and structure of corporate boards: 1935-2000,” working paper, January 2008. Linck, James S., Netter, Jeffry M., and Yang, Tina, “The determinants of board structure,” Journal of Financial Economics 87 (2008), 308-328. Litvak, Kate, “Did the Sarbanes-Oxley Act affect risk-taking by cross-listed companies?,” working paper, University of Texas, 2007. Michaels, Adrian, “After a year of U.S. corporate clean-up, William Donaldson calls for a return to risk-taking,” FinancialTimes.com, July 24, 2003. Pub. L. 107-204, 116 Stat. 745 (2002). Shadab, Houman B., “Innovation and corporate governance: The impact of Sarbanes-Oxley,” University of Pennsylvania Journal of Business and Employment Law, 10 (2008), forthcoming. Smith, Jr. Clifford W. and Watts, Ross L., “The investment opportunity set and corporate financing, dividend, and compensation policies,” Journal of Financial Economics 32 (1992), 263-292.
27
SEC Release Nos. 33-8177 and 34-47235, Disclosure required by Sections 406 and 407 of the Sarbanes-Oxley Act of 2002, January 23, 2003. SEC Release Nos. 33-8220 and 34-47654, Standards relating to listed company audit committees, April 9, 2003. SEC Release No. 33-8238, Management’s reports on internal control over financial reporting and certification of disclosure in Exchange Act periodic reports, June 11, 2003. SEC Release No. 34-48745, NASD and NYSE rulemaking: Relating to corporate governance, November 4, 2003. SEC Release Nos. 33-8810 and 34-55929, Commission guidance regarding management’s report on internal control over financial reporting under Section 13(a) or 15(d) of the Securities Exchange Act of 1934, June 20, 2007. Testimony of Alan Greenspan before the Committee on Financial Services, U.S. House of Representatives, July 15, 2003. Wintoki, M. Babajide, “Corporate boards and regulation: The effect of the Sarbanes-Oxley Act and the exchange listing requirements on firm value,” Journal of Corporate Finance 13 (2007), 229-250.
28
Percentage of IPOs by Year by Country 100% 90% 80%
Percentage of IPOs
70% 60% 50% 40% 30% 20% 10% 0% 1990
1991
1992
1993
1994
1995
1996
1997
1998
US
1999
2000
2001
2002
2003
2004
2005
2006
UK
Fig. 1 Percentage of IPOs by Year by Country
29
Number of IPOs by Year by Country 1000 900 800 700
Number
600 500 400 300 200 100 0 1990
1991
1992
1993
1994
1995
1996
1997
1998
US
1999
2000
2001
2002
2003
2004
2005
2006
UK
Fig. 2 Number of IPOs by Year by Country
30
Monthly CRSP Value Weighted vs FTSE All Share
700 600
Index Level
500 400 300 200 100 0
CRSP Value Weighted
FTSE All Share
Fig. 3 Monthly CRSP Value Weighted and the FTSE All Share Indices Returns
31
Table 1: Summary Statistics by Country and Time Period The sample includes all US and UK firms with at least twelve years of data available from Thomson ONE Banker database during the 1994 through 2006 period. Panel A reports mean and median values for US and UK firms over the entire date range. The panel lists p-values resulting from difference in means test between the US and UK samples as well as p-values resulting from Wilcoxon rank-sum test between the US and UK samples. Panel B reports mean and median values for US firms in the pre-SOX period (19942002) and the post-SOX period (2003-2006). The panel lists p-values from difference in means tests and Wilcoxon rank-sum tests between the pre-SOX and post-SOX periods. Panel C replicates panel B for the UK sample. The variable CAPEX is the capital expenditures for the year divided by the average assets for the year. The variable R&D is the R&D expenditures for the year divided by the average assets. The R&D statistics in this table include only firm-years with non-zero R&D expense. The variable CASH is the year end level of cash and short-term investments divided by average assets. The variable STD is the standard deviation of the returns for the year. The variable EBIT is the earnings before interest and taxes divided by average assets. The variable M/B is the year end market value of the assets divided by the year end book value of the assets. The variable DEBT is the average debt divided by the average market value of assets. Panel A: Summary Statistics for US and UK Firms Means US UK p-value CAPEX 0.0642 0.0545 0.000 R&D 0.0831 0.0367 0.000 CASH 0.1545 0.1111 0.000 STD 0.0307 0.0170 0.000 EBIT 0.0800 0.0752 0.764 M/B 1.9202 1.4034 0.000 DEBT 0.2246 0.1808 0.000
Medians UK 0.0348 0.0168 0.0664 0.0144 0.0732 0.9358 0.1405
p-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Panel B: Summary Statistics for US Firms Pre-SOX & Post-SOX Means Pre-SOX Post-SOX p-value Pre-SOX CAPEX 0.0700 0.0512 0.000 0.0495 R&D 0.0852 0.0784 0.005 0.0494 CASH 0.1498 0.1651 0.000 0.0539 STD 0.0326 0.0265 0.000 0.0284 EBIT 0.0791 0.0822 0.865 0.0950 M/B 1.9739 1.8002 0.017 1.2313 DEBT 0.2234 0.2273 0.210 0.1600
Medians Post-SOX 0.0341 0.0448 0.0824 0.0224 0.0731 1.2226 0.1685
p-value 0.000 0.000 0.000 0.000 0.000 0.001 0.346
Panel C: Summary Statistics for UK Firms Pre-SOX & Post-SOX Means Pre-SOX Post-SOX p-value Pre-SOX CAPEX 0.0609 0.0398 0.039 0.0388 R&D 0.0359 0.0384 0.430 0.0163 CASH 0.1144 0.1037 0.006 0.0678 STD 0.0169 0.0171 0.537 0.0142 EBIT 0.0846 0.0534 0.000 0.0820 M/B 1.5303 1.1148 0.043 0.9575 DEBT 0.1680 0.2097 0.000 0.1258
Medians Post-SOX 0.0270 0.0181 0.0634 0.0148 0.0548 0.8965 0.1847
p-value 0.000 0.992 0.094 0.019 0.000 0.000 0.000
US 0.0442 0.0477 0.0625 0.0262 0.0885 1.2278 0.1623
32
Table 2: Multivariate Regression Analysis The sample includes all US and UK firms with at least twelve years of data available from Thomson ONE Banker database during the 1994 through 2006 period. The dependent variable in models (1) and (5) is CAPEX, in models (2) and (6) is R&D, in models (3) and (7) is CASH, and in models (4) and (8) is STD. US Post-SOX (UK Post-SOX) is an indicator variable equal to one for US firms (UK firms) in the years 2003 through 2006. US*EBIT is equal to EBIT for US firms and zero otherwise. US*M/B is equal to M/B for US firms and zero otherwise. US*DEBT is equal to DEBT for US firms and zero otherwise. R&D is set equal to zero if R&D is missing. All remaining variables are defined in the header to Table 1. The control variables other than the Post-SOX indicators are lagged one year. All variables are winsorized at the 1% and 99% levels. The estimates are from firm fixed effects regressions. p-values are in brackets. The F-tests test if the coefficient on US Post-SOX equals the coefficient on UK Post-SOX. The symbols ***, **, and * denote significance at the 1%, 5%, and 10% level respectively.
US Post-SOX UK Post-SOX
1 CAPEX -0.0189*** [0.000] -0.0150*** [0.000]
2 R&D -0.0019*** [0.000] 0.0013** [0.029]
3 CASH 0.0149*** [0.000] -0.0105*** [0.000]
4 STD -0.0062*** [0.000] 0.0002 [0.497]
5 CAPEX -0.0159*** [0.000] -0.0137*** [0.000] 0.0346*** [0.001] 0.0114 [0.324] 0.0027*** [0.000] 0.0008 [0.370]
6 R&D -0.0029*** [0.000] 0.0008 [0.213] -0.0161** [0.011] -0.0168* [0.061] 0.0009 [0.368] -0.0001 [0.955]
7 CASH 0.0223*** [0.000] -0.0037 [0.147] 0.0336 [0.149] -0.0203 [0.430] 0.0127*** [0.000] 0.0039 [0.151]
0.0664*** [0.000] Yes 29,110 2,290 0.035 9.73*** 0.002
0.0293*** [0.000] Yes 29,110 2,290 0.001 16.49*** 0.000
0.1423*** [0.000] Yes 29,110 2,290 0.004 69.94*** 0.000
0.0294*** [0.000] Yes 28,980 2,284 0.049 451.88*** 0.000
0.0624*** [0.000] Yes 25,747 2,282 0.066 2.56 0.109
0.0328*** [0.000] Yes 25,747 2,282 0.017 19.76*** 0.000
0.1418*** [0.000] Yes 25,747 2,282 0.059 74.23*** 0.000
EBIT US*EBIT M/B US*M/B DEBT US*DEBT Constant Firm Fixed Effects Observations Number of Firms Adjusted R-squared F-test: US = UK p-value
8 STD -0.0069*** [0.000] -0.0004 [0.179] -0.0031 [0.122] -0.0057** [0.018] 0.0005** [0.017] -0.0003 [0.125] 0.0041** [0.045] 0.0105*** [0.000] 0.0301*** [0.000] Yes 25,544 2,276 0.079 389.84*** 0.000
33
Table 3: Multivariate Regressions with Four Year Sub-periods The sample includes all US and UK firms with at least twelve years of data available from Thomson ONE Banker database during the 1994 through 2006 period. All 2002 firm-years are excluded. The dependent variable in model (1) is CAPEX, in model (2) is R&D, in model (3) is CASH, and in model (4) is STD. The table compares three distinct four year sub-periods. US Period2 (UK Period 2) is an indicator variable equal to one for US firms (UK firms) in the firm-years 1998 through 2001. US Post-SOX (UK Post-SOX) is an indicator variable equal to one for US firms (UK firms) in the firm-years 2003 through 2006. The excluded period is 1994 through 1997. All remaining variables are defined in Table 1 and Table 2. The control variables other than the period indicators are lagged one year. All variables are winsorized at the 1% and 99% levels. The estimates are from firm fixed effects regressions. p-values are in brackets. The first set of F-tests test if the coefficient on US Period2 equals the coefficient on Period2. The second set of F-tests test if the coefficient on US Post-SOX equals the coefficient on UK Post-SOX. . The final set of F-tests test if the difference in the coefficients on US Post-SOX and US Period2 equals the difference in the coefficients on UK Post-SOX and UK Period2. The symbols ***, **, and * denote significance at the 1%, 5%, and 10% level respectively. 1 2 3 4 CAPEX R&D CASH STD US Period2 -0.0058*** -0.0023*** -0.0146*** 0.0073*** [0.000] [0.000] [0.000] [0.000] UK Period2 -0.0013 0.001 -0.0099*** 0.0074*** [0.348] [0.131] [0.002] [0.000] US Post-SOX -0.0207*** -0.0046*** 0.0135*** -0.0025*** [0.000] [0.000] [0.000] [0.000] UK Post-SOX -0.0159*** 0.0014* -0.0112*** 0.0045*** [0.000] [0.067] [0.001] [0.000] EBIT 0.0359*** -0.0144** 0.0334 -0.0042** [0.003] [0.027] [0.193] [0.035] US*EBIT 0.0144 -0.0220** -0.0272 -0.0061** [0.254] [0.018] [0.340] [0.014] M/B 0.0026*** 0.0008 0.0129*** 0.0003* [0.001] [0.447] [0.000] [0.066] US*M/B 0.0009 0.0002 0.0044 -0.0002 [0.308] [0.888] [0.133] [0.283] DEBT 0.0035* [0.088] US*DEBT 0.0101*** [0.000] Constant 0.0663*** 0.0343*** 0.1514*** 0.0257*** [0.000] [0.000] [0.000] [0.000] Firm Fixed Effects Yes Yes Yes Yes Observations 23,553 23,553 23,553 23,356 Number of Firms 2,281 2,281 2,281 2,275 Adjusted R-squared 0.077 0.019 0.063 0.142 F-Tests: US Period2 = UK Period2 7.30*** 12.44*** 1.55 0.20 p-value 0.007 0.000 0.213 0.657 US Post SOX = UK Post SOX 8.14*** 31.66*** 37.23*** 366.71*** p-value 0.004 0.000 0.000 0.000 US[Post SOX-Period2] = 0.05 8.45*** 73.87*** 314.85*** UK[Post SOX-Period2] p-value 0.832 0.004 0.000 0.000
34
Table 4: Multivariate Regressions with Minority Outsider Boards and Majority Outsider Boards The sample includes all US and UK firms with at least twelve years of data available from Thomson ONE Banker database during the 1994 through 2006 period. The dependent variable in model (1) is CAPEX, in model (2) is R&D, in model (3) is CASH, and in model (4) is STD. The table compares US firms with minority outsider boards (including 50%) in the year 2001 to US firms with majority outsider boards in the year 2001. [The year prior to SOX, 2001, is chosen as the reference year for all firms.] US Minority Outsiders Post-SOX (US Majority Outsiders Post-SOX) is an indicator variable equal to one for US firms that had a minority (majority) of outsiders on the board for the firm-years 2003 through 2006. All remaining variables are defined in Table 1 and Table 2. The control variables other than the Post-SOX indicators are lagged one year. All variables are winsorized at the 1% and 99% levels. The estimates are from firm fixed effects regressions. p-values are in brackets. The F-tests test if the coefficient on US Minority Outsiders Post-SOX equals the coefficient on US Majority Outsiders Post-SOX. The symbols ***, **, and * denote significance at the 1%, 5%, and 10% level respectively.
US Minority Outsiders Post-SOX US Majority Outsiders Post-SOX UK Post-SOX EBIT US*EBIT M/B US*M/B
1 CAPEX -0.0196*** [0.000] -0.0193*** [0.000] -0.0137*** [0.000] 0.0346*** [0.001] 0.0183 [0.129] 0.0027*** [0.000] 0.0009 [0.305]
2 R&D -0.0065*** [0.000] -0.0041*** [0.000] 0.0008 [0.213] -0.0161** [0.011] -0.0221** [0.014] 0.0009 [0.368] -0.0001 [0.937]
3 CASH 0.0404*** [0.000] 0.0225*** [0.000] -0.0037 [0.147] 0.0336 [0.149] -0.0202 [0.464] 0.0127*** [0.000] 0.0050* [0.082]
0.0619*** [0.000] Yes 19,117 1,691 0.096
0.0350*** [0.000] Yes 19,117 1,691 0.02
0.1487*** [0.000] Yes 19,117 1,691 0.061
4 STD -0.0081*** [0.000] -0.0067*** [0.000] -0.0004 [0.179] -0.0031 [0.122] -0.0060** [0.017] 0.0005** [0.017] -0.0003 [0.227] 0.0041** [0.045] 0.0099*** [0.000] 0.0295*** [0.000] Yes 19,041 1,688 0.084
0.01
2.28
7.89***
4.87**
0.908
0.131
0.005
0.027
DEBT US*DEBT Constant Firm Fixed Effects Observations Number of Firms Adjusted R-squared F-Test: US Minority Post-SOX = US Majority Post-SOX p-value
35
Table 5: Multivariate Regressions with Large and Small Firms The sample includes all US and UK firms with at least twelve years of data available from Thomson ONE Banker database during the 1994 through 2006 period. The dependent variable in model (1) is CAPEX, in model (2) is R&D, in model (3) is CASH, and in model (4) is STD. The table compares large US firms (above the US median assets in the year 2001) to small US firms (below the US median assets in the year 2001). [The year prior to SOX, 2001, is chosen as the reference year for all firms.] US Large Post-SOX (US Small Post-SOX) is an indicator variable equal to one for large (small) US firms for the firm-years 2003 through 2006. All remaining variables are defined in Table 1 and Table 2. The control variables other than the Post-SOX indicators are lagged one year. All variables are winsorized at the 1% and 99% levels. The estimates are from firm fixed effects regressions. p-values are in brackets. The F-tests test if the coefficient on US Large Post-SOX equals the coefficient on US Small Post-SOX. The symbols ***, **, and * denote significance at the 1%, 5%, and 10% level respectively. 1 2 3 4 CAPEX R&D CASH STD US Large Post-SOX -0.0178*** -0.0042*** 0.0248*** -0.0053*** [0.000] [0.000] [0.000] [0.000] US Small Post-SOX -0.0137*** -0.0014 0.0193*** -0.0085*** [0.000] [0.139] [0.000] [0.000] UK Post-SOX -0.0137*** 0.0008 -0.0037 -0.0004 [0.000] [0.213] [0.147] [0.179] EBIT 0.0346*** -0.0161** 0.0336 -0.0031 [0.001] [0.011] [0.149] [0.122] US*EBIT 0.0114 -0.0169* -0.0205 -0.0058** [0.326] [0.060] [0.427] [0.017] M/B 0.0027*** 0.0009 0.0127*** 0.0005** [0.000] [0.368] [0.000] [0.017] US*M/B 0.0008 -0.0001 0.0039 -0.0003 [0.372] [0.955] [0.153] [0.116] DEBT 0.0041** [0.045] US*DEBT 0.0096*** [0.000] Constant 0.0623*** 0.0328*** 0.1419*** 0.0301*** [0.000] [0.000] [0.000] [0.000] Firm Fixed Effects Yes Yes Yes Yes Observations 25,725 25,725 25,725 25,522 Number of Firms 2,280 2,280 2,280 2,274 Adjusted R-squared 0.066 0.017 0.059 0.083 F-Test: US Large = US Small (Post-SOX) 9.31*** 7.36*** 3.03* 78.12*** p-value 0.002 0.007 0.082 0.000
36
Table 6: Multivariate Regressions with R&D and Non-R&D Firms The sample includes all US and UK firms with at least twelve years of data available from Thomson ONE Banker database during the 1994 through 2006 period. The dependent variable in model (1) is CAPEX, in model (2) is CASH, and in model (3) is STD. The table compares US firms with R&D expenses in the year 2001 to US firms with no R&D expenses in the year 2001. [The year prior to SOX, 2001, is chosen as the reference year for all firms.] US R&D Post-SOX (US No R&D Post-SOX) is an indicator variable equal to one for R&D (no R&D) US firms for the firm-years 2003 through 2006. All remaining variables are defined in Table 1 and Table 2. The control variables other than the Post-SOX indicators are lagged one year. All variables are winsorized at the 1% and 99% levels. The estimates are from firm fixed effects regressions. p-values are in brackets. The F-tests test if the coefficient on US Large Post-SOX equals the coefficient on US Small Post-SOX. The symbols ***, **, and * denote significance at the 1%, 5%, and 10% level respectively.
US R&D Post-SOX US No R&D Post-SOX UK Post-SOX EBIT US*EBIT M/B US*M/B
1 CAPEX -0.0204*** [0.000] -0.0128*** [0.000] -0.0137*** [0.000] 0.0346*** [0.001] 0.0108 [0.351] 0.0027*** [0.000] 0.0007 [0.420]
2 CASH 0.0282*** [0.000] 0.0182*** [0.000] -0.0037 [0.147] 0.0336 [0.149] -0.0197 [0.444] 0.0127*** [0.000] 0.004 [0.144]
0.0623*** [0.000] Yes 25,725 2,280 0.067
0.1419*** [0.000] Yes 25,725 2,280 0.059
3 STD -0.0079*** [0.000] -0.0062*** [0.000] -0.0004 [0.179] -0.0031 [0.122] -0.0059** [0.016] 0.0005** [0.017] -0.0003 [0.108] 0.0041** [0.045] 0.0103*** [0.000] 0.0301*** [0.000] Yes 25,522 2,274 0.08
37.73*** 0.000
8.70*** 0.003
19.63*** 0.000
DEBT US*DEBT Constant Firm Fixed Effects Observations Number of Firms Adjusted R-squared F-Test: US R&D = US No R&D (Post-SOX) p-value
37
Table 7: Proceeds for Firms Going Public in the U.K. vs. the U.S. This table presents total proceeds in millions of end-of-year 2005 CPI-adjusted U.S. dollars. The sample includes 9,262 IPOs from 1990 through 2006. Panel A: U.S. IPOs Number of IPOs Sum of Proceeds in Million of U.S. Dollars Mean Proceeds in Million of U.S. Dollars Median Proceeds in Million of U.S. Dollars Panel B: U.K. IPOs Number of IPOs Sum of Proceeds in Million of U.S. Dollars Mean Proceeds in Million of U.S. Dollars Median Proceeds in Million of U.S. Dollars
1990-2006 All Deals 7,380 911,126 123 41 1990-2006 All Deals 1,882 186,326 99 14
All Deals 6,417 680,840 106 35 All Deals 1,284 147,639 115 17
Pre SOX (1990-2002) High R&D Mod. R&D Low R&D 2,472 1,444 2,501 229,088 188,233 263,519 93 130 105 32 33 43 Pre SOX (1990-2002) High R&D Mod. R&D Low R&D 314 287 683 14,690 45,746 87,203 47 159 128 13 16 20
All Deals 963 230,286 239 122 All Deals 598 38,687 65 10
Post SOX (2003-2006) High R&D Mod. R&D Low R&D 283 179 501 39,259 42,418 148,609 139 237 297 72 135 173 Post SOX (2003-2006) High R&D Mod. R&D Low R&D 154 153 291 4,484 18,691 15,512 29 122 53 11 11 9
38
Table 8: Logistic Regression of Going Public in the U.K. vs. the U.S. This table presents a logistic regression that models the occurrence of an IPO in the U.K. vs. the U.S. The sample includes 9,258 IPOs from 1990 through 2006. There are 1,878 U.K. IPOs and 7,380 U.S. IPOs. The dependent variable is 1 for U.K. IPOs and 0 for U.S. IPOs. The logit model estimates the probability of a U.K. IPO. Log of Proceeds is the natural log of the total proceeds in millions of end-of-year 2005 CPIadjusted U.S. dollars. Top R&D Group is equal to 1 for IPOs in the Top R&D group. Mid R&D Group is equal to 1 for IPOs in the Mid R&D group. Post SOX is equal to 1 for IPOs taking place after 2002 or from 2003 through 2006. P-values refer to t-tests of parameter estimates equal to zero. ***, **, * Significant at the one, five, and ten percent levels, respectively. Logistic regression of U.K. vs. U.S. IPOs Parameter Standard Estimate Error P-value Intercept 0.460 0.076 0.0001*** Log of Proceeds in Million of U.S. Dollars -0.543 0.020 0.0001*** High R&D Group -0.865 0.077 0.0001*** Mod. R&D Group -0.423 0.082 0.0001*** Post SOX 1.138 0.096 0.0001*** High R&D Group * Post SOX 0.542 0.156 0.0005*** Mod. R&D Group * Post SOX 0.707 0.169 0.0001*** Correct predictions Number of observations
72.5% 9,258
39
