The Sarbanes-Oxley Act and Corporate Investment: New Evidence from a Natural Experiment *

Transcription

1 The Sarbanes-Oxley Act and Corporate Investment: New Evidence from a Natural Experiment * Ana Albuquerque Boston University Católica-Lisbon School of Business and Economics albuquea@bu.edu Julie L. Zhu Boston University juliezhu@bu.edu First Draft: August 2011 This draft: January 2012 Abstract Prior studies conclude that one of the economic costs of complying with the Sarbanes-Oxley Act (SOX) is lower corporate investment. U.S. firms with a public float above $75 million during had to comply with Section 404 of SOX, whereas firms with a smaller public float in each of those three years could delay compliance until at least Using this setting as a natural quasiexperiment to isolate the effects that were uniquely due to SOX, we compare investment activities for the two groups of firms around the $75 million threshold. In contrast to prior studies, we do not find a reduction in investment for firms that had to comply with SOX relative to those that could delay compliance. Our results challenge the conventional wisdom that SOX caused firms to decrease corporate investment. JEL Classifications: K0, G2, O5. Keywords: SOX, investment, regulation, natural experiment. * We gratefully acknowledge the financial support of Boston University. We thank Rui Albuquerque, Miguel Ferreira, Francesca Franco, Krish Menon, George Papadakis, Jun Qian, Brian Quinn, Scott Richardson, Eddie Riedl, Irem Tuna, Rodrigo Verdi and Paul Zarowin for comments and suggestions. This work has also benefited from the comments of workshop participants at Boston University and London Business School. The research leading to these results has received funding from the European Union Seventh Framework Programme (FP7/ ) under grant agreement n PCOFUND-GA

2 The Sarbanes-Oxley Act says to every entrepreneur, For God s sake, don t innovate. Don t take chances because down will come the hatchet. Milton Friedman 1. Introduction The passage of the Sarbanes-Oxley Act (SOX) in 2002 by the U.S. Congress resulted in the most significant change in securities regulation since the Securities Act of In response to a string of high-profile corporate scandals, the main goal of SOX was to strengthen corporate governance while improving disclosure quality and transparency. An unintended consequence, as some argue, is a reduction in investment activities that generate profits for shareholders and produce long-term economic growth (Brady, 2007). A year after enactment, William Donaldson, former chairman of the Securities Exchange Commission (SEC), stated that SOX would lead to a loss of risk-taking zeal due to a huge preoccupation with the dangers and risks of making the slightest mistake. Several recent studies suggest that compliance with SOX caused companies to decrease investment (see Litvak, 2008; Shadab, 2008; Bargeron, Lehn, and Zutter, 2010; and Kang, Liu, and Qi, 2010). However, Leuz (2007) cautions about attributing these findings to SOX, pointing out that changes in market conditions and other concurrent events may have been at work and that it is crucial to find a control group comparable U.S. firms that are not affected by SOX to isolate the effects of SOX (see also Coates, 2007; Ball, 2009; and Hochberg, Sapienza, and Vissing-Jorgensen, 2009). In this paper, we use a quasi-natural experiment to isolate the impact of SOX on corporate investment by comparing two groups of similar U.S. firms that are affected differently by the legislation. Arguably the most demanding requirement of SOX is Section 404 (hereafter SOX404), which mandates that a firm s annual report filing include a management report, in which management evaluates the firm s internal control system on financial reporting and discloses any material weakness found. The management report must be personally certified by the chief executive officer (CEO) and chief financial officer (CFO) and the company s auditor needs to attest to management s 1

3 assessment of the firm s internal controls. 1,2 Prior studies suggest compliance with Section 404 as a reason for companies to avoid risky investments that could compromise the internal control systems (e.g., Bargeron et al., 2010). 3 However, the extent to which SOX404 has dampened corporations investment activities is debatable. First, a stream of literature suggests that SOX may have a positive impact on corporate investment by lowering firms cost of capital. For example, Coffee (2007) points out that, as a result of SOX404, investors benefit from more reliable financial statements, greater transparency, and greater accountability, all of which can lead to a lower cost of capital. 4 Ashbaugh-Skaife, Collins, Kinney, and LaFond (2009) show that increases in the effectiveness of internal controls yield decreases in the firm s cost of equity, providing evidence that SOX404 reduces the information asymmetry between a firm and its investors. Hochberg et al. (2009) also find evidence suggesting that the improved disclosure due to SOX was perceived as beneficial to investors. Using a sample of European firms that cross-list in the U.S., Arping and Sautner (2011) show that firms level of transparency increased after having to comply with SOX. Second, it has long been argued that security laws in general and mandatory disclosure requirements in particular are inconsequential (Stigler, 1964), because firms can evade the requirements by disclosing boiler-plate statements and by taking advantage of loopholes and deficiencies in the regulations and in the enforcement process. Specifically, even though the 1 The SEC s final rule on management s report on internal control over financial reporting and certification of disclosures following Section 404 can be found at 2 Note that some certifications pursuant to Section 302 of SOX were amended to conform to the requirement under Section 404. See as examples, Section 302(a)(4)(A), Section 302(a)(4)(B), and Section 302(a)(4)(C). 3 Bargeron et al. (2010) mention two other measures of SOX that could have led to a decrease in investment: the increased role of independent directors and increased liabilities and penalties faced by officers and directors. However, the former had already been imposed by the New York Stock Exchange and NASDAQ corporate rules and became effective for fiscal year 2003, before SOX. As for the latter, Ball (2009) questions its effectiveness in deterring accounting fraud because these penalties also existed before SOX. We thus focus on the most demanding requirement of SOX, Section Leuz and Verrecchia (2000) find evidence of decreased cost of capital for firms that increase their level of (voluntary) disclosure. 2

4 certification requirements of SOX404 can, in theory, generate significant civil and criminal liabilities for CEOs and CFOs, in practice, only officers who certify financial statements knowing them to be materially misleading are in a position to face liability and may escape any punishment, depending on other factors. 5 In addition, officers can protect themselves through directors and officers liability insurance to cover damages or defense costs resulting from a lawsuit for alleged wrongful acts while acting in their capacity as directors and officers for the organization. 6 Third, Ball (2009, page 314) questions the effectiveness of the new legislation regarding the expanded liability of officers and directors because, prior to SOX, CEOs and CFOs were already required to sign and attest to the veracity of financial statements and faced heavy penalties for knowingly attesting to false certification (see also Coffee, 2007; Coates, 2007). Given the importance of investment to the growth of firms and of the economy in general, it is of great significance to ascertain whether or not SOX has negatively affected firms investment. Unfortunately, uniquely identifying the impact of SOX404 on corporate investment during a period characterized by other significant events including the burst of the tech bubble in 2000/2001, the recession following 9/11, new NYSE and NASDAQ rules, and the Enron and WorldCom scandals is a challenge for researchers. These events alone could have caused firms to become more cautious in their investment decisions. The key is to find a control group of firms that were not affected by SOX404, but were subject to the same concurrent events. 5 Corporate officers who rely in good faith on reports from subordinates or from outside auditors and accountants are not likely to be subject to civil or criminal liability themselves. As a result of this reliance defense, very few cases in which plaintiffs bring actions against defendant corporate officers proceed beyond a motion to dismiss in the very early stages of litigation. In the vast majority of cases where plaintiffs allege violations pursuant to 302, the motions are dismissed. Only two motions appear to have survived the motion-to-dismiss stage as of May 2011: (1) Bear Stearns Securities Derivative Litigation, 2011 U.S. Dist. LEXIS 6026, dated Jan. 19, 2011, and (2) LDK Solar Securities Derivative Litigation, 584 F. Supp. 2d 1230; 2008 U.S. Dist. LEXIS 42425, dated May 29, Although neither of these cases has been adjudicated to a final judgment, they did survive early dismissal. We thank Brian Quinn, Assistant Professor of Law at Boston College, for providing us with this information. 6 Cases in which directors of a company are required to personally pay for their misconduct are also rare (e.g., Davidoff, 2011). 3

5 In this paper, we use a quasi-natural experiment to isolate the impact of SOX on corporate investment by comparing two groups of similar U.S. firms that were affected differently by the legislation. SOX required firms with a public float above $75 million in 2002 to comply with Section 404 in 2004, while firms with a public float below $75 million in 2002, 2003, and 2004 could delay compliance until the end of Specifically, we compare a sample of small firms with public float just above $75 million (the filers ) to firms with public float just below this threshold (the control group ) to benchmark the changes in investment made by similar firms forced to comply with SOX404. In addition, the specific threshold of $75 million used to define Section 404 in 2002 was not known prior to 2002 and is not related to firm characteristics, which lowers the risk associated with manipulation of or endogeneity in the public float as of 2002 (see Ilieve, 2010). This allows for a difference-in-difference research design, which should mitigate potential biases from unobservable factors that might be correlated with investment. If SOX404 caused firms to engage in less risk-taking behavior, as suggested by prior studies, we expect the level of investment for filers to drop in the post-sox period relative to that of the control group. But if the change in risk-taking behavior is due to the general economic conditions or if filers are able to enjoy greater transparency and lower cost of capital due to SOX, then the level of investment for filers would not drop and might even increase in the post-sox period compared to that of the control group. Using a sample of approximately 440 unique firms for the sample period of 1994 through 2006, we first find that post-sox investment by filers falls from the pre-sox levels; the magnitude of the reduction is similar to that documented in earlier studies of firms larger than those in our study. However, post-sox investment by the control group also dropped over the same period. We show 7 Securities and Exchange Act Rule 12b-2 defines public float as the aggregate market value of the issuer's outstanding voting and non-voting common equity held by non-affiliates. An affiliate is a person who, directly or indirectly through one or more intermediaries, controls, or is controlled by, or is under common control with the person specified. The term control means the possession, direct or indirect, of the power to direct or cause the direction of the management and policies of a person, whether through the ownership of voting securities, by contract, or otherwise. 4

6 that the decrease in investment for both groups starts in 1999, not in 2003 when SOX became effective. The result is consistent with both filer firms and control firms reducing investment to adjust to a business and legal environment that had gone through substantial changes prior to the introduction of SOX. Moreover, we do not find that filers reduced their investment significantly more than the control group did. In addition, we use an instrumental variables approach to examine investment activities in 2004, the first year of SOX compliance, when the difference between the two groups should be most pronounced. Specifically, we use firms public float in 2002, along with stock returns and stock return volatilities, to instrument firms compliance status in 2004 (i.e., whether a firm becomes a filer or not). 8 We find that the filers invest more after SOX than firms in the control group do, which suggests that filers may actually have benefited from the increased disclosures mandated by SOX404. Our results thus are inconsistent with the hypothesis that Section 404 of SOX reduces investment for firms that had to comply. We also test whether our findings are limited to small firms; we do not find that the impact of SOX404 on large firms investment is statistically different from that on small firms in our sample. We then explore whether the impact of SOX404 varies cross-sectionally for our sample firms. First, we examine large-scale acquisition, an activity considered to have adverse effects on a firm s internal control system, and do not find that filers make fewer acquisitions than control group firms do after SOX. Second, we analyze whether investment decreased significantly more for firms with a higher percentage of insider holdings, as these firms would presumably be more concerned about the potential liability costs associated with risky investments that could compromise internal controls. We find that filers with a higher percentage of insider ownership (measured before SOX) decrease investment less, not more, in the post-sox period than filers with lower levels of insider holdings. 8 Gao, Wu, and Zimmerman (2009) show evidence that firms take actions to avoid complying with SOX. 5

7 Overall, these results cast doubt that SOX increased the level of risk-aversion among executives and directors. We only find evidence of a negative impact of Section 404 on investment for filers with material internal-control weaknesses (58 firms or 15.9 percent of our filers), possibly due to the fact that these firms had to reallocate investment capital to fix their internal control systems. This paper contributes to the debate over the effects of SOX on firms investment decisions and challenges the extant evidence that SOX had an adverse effect on investment. Bargeron et al. (2010) and Kang et al. (2010) use UK firms as a benchmark and find that U.S. firms decreased their investment more than UK firms. Litvak (2008) compares foreign cross-listed firms in the U.S. (thus bound by U.S. law) to similar non-cross-listed firms and finds that the risk of those firms subject to SOX declined after SOX. In these papers, however, a causal link is subject to differences in legal rules, economic conditions, and contemporaneous events across countries. Our paper also adds to the literature that focuses on the impact of SOX on small firms although we do not find any differential investment behavior across small and large firms. Our sample of small firms is particularly relevant as Gao et al. (2009) show that small firms have an incentive to manipulate their public float to avoid complying with SOX404 and several studies show that the compliance costs of SOX404 are a significant burden to small firms (e.g., Engel, Hayes, and Wang, 2007; Chhaochharia and Grinstein, 2007; Kamar et al., 2007; Kang et al., 2010; and Alexander et al., 2010). The remainder of the paper proceeds as follows. In Section 2, we review the institutional background and timeline of Section 404 of SOX. Section 3 describes the sample selection and data. Section 4 presents the empirical tests and results. Section 5 offers additional robustness tests and Section 6 concludes. 6

8 2. Institutional Background and Sequence of Events In this section, we provide a brief description of the institutional background and sequence of key events regarding SOX and, in particular, Section 404 (see Figure 1). We then explain why the implementation of Section 404 provides an empirical setting that allows us to isolate its effects from those of other confounding events. Shortly after SOX was signed into law on July 30, 2002, the SEC required publicly listed companies to declare accelerated filer status in their 2002 annual reports, based on the size of their public float in the second (fiscal) quarter of Public float is the fraction of the common stock not held by insiders such as managers, employees, and board members and is reported on the first page of the company s 10-K. Firms with a float over the $75 million threshold became accelerated filers and had to keep that status going forward (see Exchange Act Rule 12b-2). 9 Accelerated filers had to complete their 10-K filing within 75 days of their fiscal year-end; the prior deadline had been 90 days. In May of 2003, under the SOX404 guidelines, the SEC announced that all accelerated filers must file a new management report beginning in However, firms with a public float less than the $75 million threshold in 2002, 2003, and 2004 did not need to comply with SOX404 and file the management report until a later date. These firms were also exempted from an auditor s attestation of the management report until June We develop our empirical tests based on this sequence of events. First, we use companies public floats, as reported in their Ks, to define the treatment and control groups. The key for our empirical strategy is that firms declared their accelerated filer status at the end of fiscal year Moreover, firms did not know that the declaration of accelerated filer status would subsequently require them to file a management report, so our classification into treatment and 9 A company with accelerated filer status can become a non-accelerated filer only if its revenues and public float are smaller than $25 million for two consecutive years (SEC release ). Firms who do not have public equity but have public debt are, by definition, non-accelerated filers as they do not meet the criteria of an accelerated filer. 7

9 control groups based on the 2002 public float avoids concerns regarding manipulation or endogeneity of public float. Specifically, firms with a public float over the $75 million threshold are mandatory filers, since they must comply with SOX404 starting in 2004 (regardless of the size of float in 2003 or 2004) and hence are directly affected by it (treatment group). Firms with a public float below the threshold in 2002 are the control group; their changes from the pre- to post-sox periods are used as a benchmark to measure changes in the treatment group. We define the pre-sox and post-sox periods to include the fiscal years of 1994 to 2002 and 2003 to 2006, respectively; these periods are also used in Bargeron et al. (2010). Since firms were required to file their first management report in 2004 (for the fiscal year ending on or after June 15, 2004 and later postponed for fiscal years ending on or after November 15, 2004), we conduct a robustness test that excludes the transition year of 2003 and defines the post-sox period to include only the fiscal years of 2004 to We obtain qualitatively similar results. 3. Sample Selection and Descriptive Statistics 3.1. Sample Selection Our sample consists of approximately 440 unique firms that appear on Compustat with a public float between $50 million and $150 million in 2002, excluding financial firms (with SIC codes between 6000 and 6900) and regulated firms (with SIC codes between 4900 and 4939). The public float equals the value of the common stock owned by outside investors; we hand-collect a firm s public float from the cover page of its 10-K report. Accounting and financial information is obtained from Compustat. Several studies show that SOX404 compliance costs lead many small firms to deregister their common stock ( go dark ) or to go private so as to avoid complying with SOX (see Leuz, Triantis, and Wang, 2008; Engel et al., 2007). These decisions can induce a survivorship bias 8

10 that could lead us to find no effect across the filers and the control group if the filers most affected by Section 404 delisted. To alleviate such concerns, we use a constant sample of filer and control firms with available data for the full sample period of 1994 to Our final sample has approximately 3,700 firm-year observations covering that period Descriptive Statistics Table 1 reports descriptive statistics for the sample firms, which include filers and the control group. Panel A presents summary statistics for firm characteristics across both groups for the fiscal year of 2002, just before SOX took effect, as well as the p-values for the difference in means and medians across the two groups. All variables are defined in Appendix A. We winsorize all the variables used in the tests at the 1st and 99th percentiles to minimize the effect of outliers. As shown in Panel A, the average 2002 public floats for the filers and the control group are $105 million and $62 million, respectively. Sixty-three percent (277 / [ ]) of the firms are mandatory filers; the remaining thirty seven percent are the control group. 10 Panel A also shows that the total pre-sox investment scaled by total assets (INVEST) is statistically significantly higher for the control group than for filers; the mean investment values are and 0.129, respectively. The mean control firm is relatively smaller than the mean filer firm, when measured by market capitalization, and less profitable; mean earnings before interest and taxes scaled by assets (EBIT) are and for control firms and filers, respectively. When comparing the median filer to the median control firm, we do not find that they differ in their investment levels or profitability measures. Filers and control firms do not seem to exhibit different levels of growth opportunities before SOX; for example, neither the stock returns (STKRET) nor the market-to-book ratios (MTB) of the two groups are statistically different from each other. Because prior studies argue that SOX404 leads firms to engage 10 In the robustness section, we replicate the tests using a sample with the same number of filers and non-filers and obtain qualitatively similar results. 9

11 in less risk-taking behavior and thus to invest less, we also analyze firms level of riskless assets that is, cash and stock-return volatility. Bargeron et al. (2010) argue that if a firm engages in less risk-taking behavior due to SOX that would be reflected in the amount of cash it holds and the volatility of its stock returns. Table 1 shows that short-term investments scaled by total assets (CASH) and the standard deviation of returns (STD) are also not statistically different across the two groups. In summary, filers exhibit relatively lower levels of investment and higher accounting performance and are larger than control firms prior to the 2002 enactment of SOX. However, the growth prospects do not seem to be different across the two groups. Panel B provides descriptive statistics for the variables of interest across the pre-sox ( ) and post-sox ( ) periods for control firms and filers. Filers exhibit a decrease in investment (CAPEX, INVEST) and volatility (STD) and an increase in cash holdings from the pre- SOX to the post-sox period. During the post-sox period, the mean level of growth options (MTB) declined, but the median (unreported) level of growth options actually increased. Control firms also show a decrease in capital expenditures (CAPEX) and volatility (STD) and an increase in cash holdings in the post-sox period. Post-SOX INVEST also decreases for the control group, but the difference is not statistically significant at conventional levels. In the last column of Panel B, we test whether the differences between the filers and control firms in the changes observed from the pre- SOX to the post-sox periods are statistically significant. None of the univariate changes is statistically different between filers and control firms. For example, although, in the post-sox period, both filers and control firms decreased their investment in capital expenditures in the amount of and , respectively the decrease in CAPEX for filers is not statistically different from that for the control group (p-value of 0.211). Figures 2 through 6 show how each of the variables capturing investment and risk-taking 10

12 activities CAPEX, R&D, INVEST, CASH, and STD change through the sample period for filers, control firms, and large firms, the latter defined as all Compustat firms with market capitalization above $150 million. The figures show how these variables change through time unconditionally. Figure 2 shows that CAPEX moves in tandem for all three sets of firms, decreasing between 1996 and 2003, with a more accentuated decline immediately after 2000, and then increasing between 2003 and Figure 3 shows R&D for the three groups. It increases from 1995 to 1998 and declines in 1999 and 2000 for both the filers and control firms, followed by a slight increase from 2000 to 2006 for the filers and an increase from 2000 to 2002 and then a slight decline for the control group. These results suggest that during the post-sox period, it is the control firms not the filers that exhibit a slight decline in R&D expenditures. The mean R&D for the large firms does not change significantly throughout the sample period. Figure 4 shows the results for total investment. Consistent with Figure 2, the decline in investment for all three groups starts in 1999, rather than in 2003 when SOX became effective. During the post-sox period, investment increased rather than decreased, which raises questions about prior findings that U.S. corporations became more risk-averse after SOX. Figure 5 reports the trend in CASH holdings. For all three groups, the level of cash holding increases throughout the entire period, while the rate of increase seems to decline slightly in the post-sox period. The argument that firms have become more risk-averse suggests that they prefer to hold more of the risk-free asset cash. However, the evidence suggests that the rate of cash holdings actually declined after SOX, particularly for filers. Finally, Figure 6 shows how the average firm s stock volatility evolved throughout the sample period. Volatility increased up to 2000, when the tech bubble burst, then steadily declined. To the extent that the decrease in volatility is found across all the groups, including the control group, it calls into question whether it was caused by any risk-aversion attributable to SOX

13 4. Empirical Tests and Results 4.1 The Impact of Section 404 on Filers versus Control Firms Using a Difference-in-Differences Approach To test whether compliance to SOX s Section 404 impacted firm s willingness to take risks, we estimate the following regression models: - -, (1) (2) We follow Bargeron et al. (2010) and measure Y i,t, a proxy for a firm s willingness to take risks, using the following five variables: CAPEX, R&D, INVEST, CASH, and STD. The control variables used are EBIT, MTB, and DEBT, as in Bargeron et al. (2010). EBIT and MTB are included to account for any variation in investment or cash holdings that is related to a firm s profitability or growth opportunities. The STD regressions include EBIT, MTB, and DEBT as controls to account for the effect of a firm s profitability, growth prospects, and debt on the volatility of its stock. All the control variables are lagged one year. The variable Post-SOX is an indicator variable that takes the value of one for the years 2003 through The variable Filer is an indicator variable that takes the value of one if a firm s public float is above the $75 million threshold in The regression models are estimated using year and firm fixed effects. The year-fixed effects,, control for any unobservable time-aggregate effect or time trend in the dependent variables. 11 The firm fixed effects, 11 Bargeron et al. (2010) also include GDP growth and an index return to control for the impact of the UK and U.S. economies growth on the variables of interest. We do not include these variables because our sample consists only of 12

14 , control for any omitted firm characteristics that are time-invariant. We use two empirical models to test the hypothesis that SOX404 caused firms to become more risk-averse. The empirical model specified in Equation 1 tests whether there is a change in the variable of interest, Y it, for filers in the post-sox period. However, showing such a change is not sufficient to show causality, because the same change in Y it could have been caused by other concurrent events. In order to uniquely identify the impact of SOX on the variables of interest, we use a sample of non-filer U.S. firms as a control group and estimate Equation 2. This regression model also adds interaction terms between the control variables and a dummy variable for the filers in order to capture any differential marginal impact of the control variables on the variable of interest for these firms. The empirical model specified in Equation 2 allows for a difference-in-differences test, where the coefficient β 1 captures the change in the outcome variable between the pre-sox and post- SOX periods for the filers when benchmarked against the change for the non-filers during the same period. If SOX404 caused firms to engage in less risk-taking behavior, as suggested by prior studies, we expect β 1 <0. If, instead, the level of risk-taking behavior is due to the general economic conditions or if filers are able to enjoy greater transparency and lower cost of capital thanks to SOX404, then β 1 0. Table 2 presents the results of estimating Equations 1 and 2 for each of the five dependent variables. Columns 1, 3, 5, 7, and 9 report the results of estimating Equation 1 using the sample of filers only. Columns 2, 4, 6, 8, and 10 report the results of estimating Equation 2 using the sample of filers and the control group. The results in Columns 1, 3, 5, and 9 show that CAPEX, INVEST, and STD are significantly lower for the filers during the post-sox period when compared to the reference period of 1994 through 2002 (coefficients of , , and respectively). These results U.S. firms; GDP growth and the index return are subsumed by the inclusion of the year fixed effects. Inclusion of the S&P500 index and GDP growth yields qualitatively similar results. 13

15 are consistent with those obtained in Bargeron et al. (2010, Table 2) and Kang et al. (2010, Table 2) for U.S. firms. 12 However, Columns 2, 4, and 6 show that the coefficients on Post-SOX*Filer are not statistically different from zero, suggesting that the change in investment during the post-sox period for filers is not statistically different from that for the control group, which casts doubt on the effect of Section 404 on a firm s level of investment. Moreover, the coefficients on the Post-SOX indicator remain negative and significant suggesting that both filers and control group decrease investment in the post-sox period for reasons other than the regulation. The results for CASH are reported in Columns 7 and 8. Contrary to the findings in Bargeron et al. (2010), we do not find that CASH increased significantly post-sox for the filers, nor when compared to the control group. Our finding that small firms do not tend to hold more of this risk-free asset in the post-sox period differs from the finding in Bargeron et al. (2010). The difference is due to the fact that we include year dummies to account for the time-trend identified in Figure 5. If we exclude the year dummies, the coefficient on Post-SOX becomes positive and statistically significant, while the remaining coefficients remain qualitatively the same. When analyzing the impact of SOX404 on stock volatility for both filers and control firms in Column 10, we find that the coefficient on Post-SOX is still negative (-0.012) and is statistically significant, providing evidence that both filers and control firms stock volatility declined following SOX. However, the results in Column 10 show that the post-sox decline in volatility is marginally smaller for filers than for control firms; the coefficient on Post-SOX*Filer is (t-statistic=1.72). This casts doubt on the view that SOX404 discouraged risk-taking behavior on the part of firms that were subject to compliance (the filers), causing a more significant decline in stock volatility for these firms. 12 In contrast with these papers, we do not find that R&D decreased during the post-sox period for our sample of constant firms. If, however, we do not restrict the sample to reflect the same firms before and after SOX, we do find that R&D decreased after SOX. 14

16 We further investigate whether the lack of a statistically significant difference in risk-taking and investment behavior between filers and control firms is due to a lack of statistical power to detect such a difference. First, we check whether the lack of statistical significance in Table 2 for the coefficient Post-SOX*Filer is due to a high correlation between this variable and Post-SOX. Although the correlation between these two variables is 0.73, excluding Post-SOX from the regressions in Columns 2, 4, 6, 9, and 10 yields very similar results. 13 Second, the economic significance of the coefficient associated with Post-SOX*Filer is about one-fifth of the economic significance of the Post-SOX coefficient. During the post-sox period, CAPEX decreased by 2.4 percent for both filers and control firms. However, the incremental economic impact on CAPEX for the filers relative to the control firms is a decrease of only 0.5 percent. Because the size of the coefficient estimate is not affected by the power of the test, we conclude that it is the lack of economic significance and not of statistical power that primarily explains our results. 4.2 The Impact of Section 404 on Filers versus Control Firms Using Instrumental Variables To avoid endogeneity concerns, we define firms as the treatment group (filers) and the control group (non-filers) based on their float in However, it is possible that some of the firms in the control group end up having to comply with SOX404 in 2004 if their float crosses the threshold of $75 million in 2003 or Moreover, we want to examine whether filers and the control group behaved differently in terms of corporate investment in the first year after SOX404 became effective, when the effect of SOX is arguably the strongest. We thus follow Iliev (2010) and, in order to account for potential misclassification of filers as non-filers, use a two-stage least square instrumental variables (IV) approach by first estimating whether a firm will become a filer (Wooldrige, 2009; Angrist and Pischkle, 2009). In the second stage, we use the estimated filer status obtained from the 13 The inclusion of Post-SOX leads to the statistical insignificance of the year dummies for the post-sox period (2003 to 2006). 15

17 first stage to examine the difference in corporate investment between filers and non-filers. Specifically, we estimate the following IV model:. (3) First, we need to find instruments to estimate whether the firm will become a filer; these instruments must be highly correlated with the probability of becoming a filer but not with the variables of interest in the second stage. Therefore, we use the exogenous variable Float the firm s public float in Following the suggestion of Hayes (2009, page 513), we also use the expected returns (STKRET) and the stock return volatility (STD), measured in 2002, as instruments to estimate whether a firm will become a filer in Firms with a larger public float, higher expected returns, or greater return volatilities are assumed to have a higher ex-ante probability of crossing the $75- million threshold and having to comply with SOX404 by Second, we use the estimated filer value from the first-stage regression as an explanatory variable along with EBIT, MTB, DEBT, and industry dummies to predict CAPEX, R&D, INVEST, CASH, and STD in Table 3 presents the results using the IV approach. As can be seen from Column 1, becoming a filer in 2004 is highly correlated with the size of the firm s public float in 2002, its stock return volatility, and its expected returns (the latter proxied by the actual returns) in The second-stage regression results show that the estimated filers invest more in 2004, the first year of SOX implementation, than the control group, which contrasts with the findings of prior studies. Column 5 shows that the estimated filers tend to hold more of the riskless asset (cash). Finally, Column 6 shows that estimated filers have lower volatility of returns. These results are inconsistent with firms 16

18 avoiding risky investment due to SOX, but reinforce our results using the difference-in-difference approach. We also replicate the second-stage results using the CAPEX, R&D, INVEST, CASH, and STD in 2005 and 2006 and obtain similar (untabulated) results for the investment variables. However, we no longer find that the estimated filers cash and stock return volatility are statistically different from that of the control group. 4.3 The Impact of SOX s Section 404 on Small versus Large Filers To investigate whether the above findings are driven by the fact that small firms investment levels and risk-taking behaviors are inherently different from those of larger firms, we replicate Table 2 using a sample that includes both small accelerated filers and large accelerated filers. Table 4 shows that investment and volatility decreased, while the level of cash increased, during the post-sox period. However, the interaction between Post-SOX and the indicator variable Large is not statistically significant in any of the regressions, except for the STD equation. These results are inconsistent with the notion that small firms differ from large firms in post-sox investment level and risk-taking behavior. Column 10 shows that larger firms exhibit a smaller decrease in volatility in the post-sox period than smaller firms do (coefficient of 0.002, t-statistic of 2.73). To the extent that changes in STD reflect changes in a firm s level of risk-taking behavior, this result is consistent with smaller firms decreasing their level of risk taking more than larger firms do in the post-sox period Other Cross-sectional Evidence In this subsection, we examine whether the impact of SOX s Section 404 on filers varies for those with a higher proportion of insider holdings and for those with internal control problems (or material weaknesses) Filers with a Large Proportion of Insider Holdings 14 Several studies find that SOX compliance imposed higher relative costs on smaller firms (Chhaochharia and Grinstein, 2007 and Kang et al., 2010). 17

19 We investigate whether filers with a higher proportion of insider holdings and presumably a greater concern about potential liability costs associated with Section 404 compliance invest less than filers with a lower proportion of insider holdings. Specifically, if SOX404 reduces incentives to engage in risk-taking activities, this negative effect would be more pronounced in firms with higher insider ownership since the risk-adverse executives would have more personal wealth at stake. We test this hypothesis by adding the interaction terms Post-SOX*InsiderHold and Post- SOX*InsiderHold*Filer in Equation 2, where InsiderHold is an indicator variable equal to one if the level of insider holdings of the firm s equity in 2001 (before SOX) is above the sample median. The insider ownership data is hand-collected from the firms proxy statements (Form DEF 14A) and it includes shares owned by all the insiders (such as executives and board members). Note that we do not include InsiderHold as a control because Equation 2 is estimated with firm fixed effects and thus any time-invariant firm characteristic drops out, as is the case for the 2001 level of insider ownership. Table 5 shows the results of this test. We find some evidence that filers with a high level of insider holdings invest more, not less, during the post-sox period than do filers with a lower level of insider holdings (coefficients of (t-value of 1.73) for CAPEX and (t-value of 1.81) for INVEST). To the extent that a high level of insider holdings reflects the higher potential costs to which insiders are exposed, these results cast doubt that SOX404 has discouraged risk-taking among executives and directors Filers with Material Weaknesses Eldridge and Kealey (2005) find that firms that reported ineffective internal controls after SOX404 experienced higher SOX audit costs than firms that reported effective internal controls. Ge and McVay (2005) analyze firms with material weaknesses and show that poor internal controls are 18

20 usually associated with insufficient resources committed for accounting controls. We thus predict that filers that report having material weaknesses in their internal control systems need to allocate more resources to their accounting controls in order to comply with Section 404 and are thus more likely to cut investment. We test this hypothesis by adding the interaction term Post-SOX*Filer*Weakness in Equation 2, where Weakness is an indicator variable equal to one if the filer discloses at least one internal control (material) weakness in its SEC filings during the post-sox period. The information regarding material weaknesses related to SOX404 is obtained from Audit Analytics. The coefficient on this interaction term captures the extent to which, during the post-sox period, filers with material weaknesses exhibit lower CAPEX, R&D, and INVEST than do filers without material weaknesses. Table 6 presents the results. We do not include firm fixed effects in Table 6 because about 60 percent of the firms reporting a material weakness in their internal controls do so for two or more years during the post-sox period. The inclusion of firm fixed effects would prevent us from detecting any statistically significant difference because disclosure of a material weakness is a firm characteristic that is weakly time-invariant. Consistent with our prediction, we find that filers with material weaknesses do decrease their investment subsequent to Section 404 compliance, but that filers without material weaknesses do not invest less than the control firms. Economically, the total investment scaled by assets of a filer with a material weakness decreases by 3.8 percent more than those of a filer without a material weakness. This result needs to be interpreted cautiously because the classification of filers with material weaknesses suffers from potential endogeneity biases, as these firms are classified after SOX404 is implemented. It is possible that firms that decrease their investment because of lack of growth opportunities or profitability are also more likely to report a material weakness, in which case it maybe the lack of future growth prospects rather than reporting a material weakness that leads these firms to cut investment. 19

21 5. Robustness Tests Section 404 is more likely to cause internal control problems for filers that grow through acquisition, since acquired firms may have poor accounting systems. To investigate whether filers reduced their level of acquisition more than the control firms did, we replicate the results in Table 2 using acquisitions as our measure of investment level. The (untabulated) results are consistent with those in Table 2; we do not find that filers exhibit significantly lower levels of acquisition than control firms do. Recent studies show that U.S. firms are pursuing growth strategies by outsourcing and by taking large equity holdings in other companies as a form of investment, approaches which are not necessarily reflected in larger capital expenditures or R&D investments (e.g., Bhagwati, Panagariya, and Srinivasan, 2004; McCarthy, 2002). This change in investment strategy raises the question of whether this paper s findings, using capital expenditures and R&D expenditures as measures of investment, accurately portray the current risk-taking characteristics of U.S. firms. To investigate the possibility that they do not, we replicate Table 2 and Equation 2, using three variables to measure investment that are independent of investment strategy: employment growth, asset growth, and sales growth. The (untabulated) results show that the employment and sales growth for both filers and nonfilers decline during the post-sox period (while asset growth remain stable for both groups), but the decline is not statistically different across filer and control firms. The results are thus consistent with those presented above. The sample in this study covers firms with a float between $50 million and $150 million in order to have a large enough sample of similar firms. However, one can question whether our results are due to a lack of power to detect statistically significant differences between the filers and the 20

22 control group due to the lower proportion of control firms in the main sample. To address this concern, we redefine the sample to be composed of firms with a public float between $50 million and $100 million and between $25 million and $150 million and find that the (untabulated) results remain qualitatively the same. 6. Conclusion The impact of the Sarbanes-Oxley act on corporate investment has attracted significant attention from both practitioners and academics since its enactment in Prior studies conclude that the decrease in corporate investment in the U.S. during the post-sox period results from managers unwillingness to take risks due to the increase in litigation and compliance costs associated with SOX. However, a stream of literature suggests that SOX could actually have a positive impact on corporate investment as investors benefit from greater transparency conferred by improved disclosure, which can lead to lower cost of capital. Moreover, because the post-sox period is also a period of major changes in corporate governance and of other significant events, it is a challenge to uniquely identify the impact of SOX on corporate investment. We use a natural experiment to identify a control group to isolate the effects of Section 404, arguably SOX s most demanding requirement, on corporate investment. We compare the impact of SOX404 on a sample of small firms with public float just above the $75 million threshold to the impact on firms with a public float just below that threshold. Filers are required to apply Section 404, while non-filers are not, leading to natural treatment (filers) and control (non-filer) groupings. Even though both groups significantly decrease their investment in the post-sox period ( ) relative to that of the pre-sox period ( ), we do not find that firms complying with SOX404 changed their level of investment to a greater or lesser degree than firms in the control 21

23 group did. We also find that our results are not a characteristic of small firms only, as we do not find that the impact of SOX404 on large firms investment is statistically different from the impact on small firms. We further show that for small filers, for the control group, and for large filers, the decrease in investment starts in 1999, not in 2003 when SOX became effective. In fact, when we use an instrumental variables approach to estimate which firms will comply with SOX404 in 2004, we find that the filers actually invest more than the control group do after SOX404 is implemented, suggesting that filers may actually have benefited from the newly mandated disclosures. Our cross-sectional tests further corroborate the above results. We examine large-scale acquisition and do not find that filers make fewer acquisitions after SOX than the control group does. We also find that filers with a higher percentage of insider ownership and thus greater concern about litigation costs invest more, not less, in the post-sox period than filers with lower levels of insider holdings do. Only for filers with material internal-control weaknesses do we find some evidence of a decrease in investment, probably because these firms had to reallocate investment capital to fix their internal control systems. Thus, our results challenge the conventional wisdom that SOX had a chilling effect on risk-taking by publicly traded firms (Bargeron et al., 2010). 22

24 References: 1. ALEXANDER, C., S. BAUGUESS, G. BERNILE, Y. LEE, and J. MARIETTA-WESTBERG. The Economic Effects of SOX Section 404 Compliance: A Corporate Insider Perspective, Unpublished Paper, Securities and Exchange Commission, Available at 2. ANGRIST, J., and J. PISCHKLE. Mostly Harmless Econometrics: An Empiricist s Companion. Princeton: Princeton University Press, ARPING, S., and Z. SAUTNER. Did the Sarbanes-Oxley Act of 2002 Make Firms Less Opaque? Evidence from Analyst Earnings Forecasts, Unpublished Paper, University of Amsterdam, Available at SSRN: 4. ASHBAUGH-SKAIFE, H., D. COLLINS, W. KINNEY, and R. LAFOND,. The Effect of SOX Internal Control Deficiencies on Firm Risk and Cost of Equity, Journal of Accounting Research 47 (2009): BALL, R. Market and Political/Regulatory Perspectives on the Recent Accounting Scandals, Journal of Accounting Research 47 (2009): BARGERON, L., K. LEHN, and C. ZUTTER. Sarbanes-Oxley and Corporate Risk-Taking, Journal of Accounting and Economics 29 (2010): BHAGWATI, J., A. PANAGARIYA, and T. SRINIVASAN. The Muddles over Outsourcing, Journal of Economic Perspectives 18 (2004): BRADY, D. Sarbanes-Oxley = A Downturn in Corporate Risk-taking, Business Week, September 26, Chhaochharia, V., and Y. Grinstein. Corporate Governance and Firm Value: The Impact of the 2002 Governance Rules, Journal of Finance 62 (2007): COATES, J. The Goals and Promise of the Sarbanes-Oxley Act, Journal of Economic Perspectives 21 (2007): COFFEE, J. Law and the Market: The Impact of Enforcement, University of Pennsylvania Law Review 156 (2007): DAVIDOFF, S., Despite Worries, Serving at the Top Carries Little Risk, The New York Times, June 7th. 13. ELDRIDGE, S., and B. KEALEY. SOX Costs: Auditor Attestation under Section 404, Unpublished paper, University of Nebraska at Omaha, Available at SSRN: ENGEL, E., R. HAYES, and X. WANG. The Sarbanes-Oxley Act and Firms Going-private Decision, Journal of Accounting and Economics 44 (2007), GAO, F., J. WU, and J. ZIMMERMAN. Unintended Consequences of Small Firms Exemptions from Securities Regulation: Evidence from SOX, Journal of Accounting Research 47 (2009): GE, W., and S. MCVAY. The Disclosure of Material Weaknesses in Internal Control after the Sarbanes-Oxley Act Accounting Horizons 19 (2005): HAYES, R. Discussion of Unintended Consequences of Small Firms Exemptions from Securities Regulation: Evidence from SOX, Journal of Accounting Research 47 (2009): HOCHBERG, Y., P. SAPIENZA, and A. VISSING-JORGENSEN. A Lobbying Approach to Evaluating the Sarbanes-Oxley Act of 2002, Journal of Accounting Research 47 (2009):