The influence of the Sarbanes-Oxley Act on excessive
risk taking in the pharmaceutical- and
non-pharmaceutical market in the United States of America
Nienke Sterk
10787526
Bsc. Economie en Bedrijfskunde
Specialization: Organisation and Finance
Supervisor: A.R.S. Woerner
Date: January 31st 2018
Statement of own work
This document is written by Nienke Sterk who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
This paper studies the influence of the Sarbanes-Oxley act on risk taking by pharmaceutical and non-pharmaceutical companies in the USA. This research is important to figure out whether the instalment of the SOX Act successfully decreased excessive risk-taking in the pharmaceutical market. The way this is researched is by doing a literature review and a panel data regression on different proxies for risk taking and interpreting the results. The main results found are that for pharmaceutical companies the SOX Act has not diminished excessive risk-taking but had no real effect. However, for the non-pharmaceutical companies it had a diminishing effect. To take a look at future research there is the possibility of doing a regression with a larger data sample with all kinds of sizes of firms and there could be more industry specific research.
Table of contents
1. Introduction 5
2. Background & Literature 7
3. Data & Methodology 10
4. Results 14
5. Conclusion 17
List of tables and figures 19
Bibliography 20
1. Introduction
Fraudulent accounting activities are unfortunately nowadays still a problem. Excessive risk taking might be a consequence of these kinds of activities, can potentially be very harmful for a firm and lead to great losses or bankruptcy. Further, if the financial numbers (i.e. financial statement, balance sheet, etc.) of companies are not correct, it can steer investors into taking bad investment decisions. If the financial numbers overstate a firm’s value, investors might eventually lose some of their money. Other stakeholders are also affected by inaccurate reporting, most of all when these inaccuracies are found out. In recent years there were several accounting scandals involving all kinds of big US companies. To prevent these kinds of fraudulent accounting activities, the government enacted the Sarbanes-Oxley Act in 2002 (henceforth, SOX).
The SOX was installed to protect investors and other parties against fraudulent accounting activities by corporations. It contains all kinds of new or expanded requirements to improve financial reporting practice of corporations and prevent accounting fraud. It also instated very harsh punishments for corporations or top management teams who are not compliant with the new rules of the Act.
As conclusively identifying fraud is often not possible, this paper looks at an identifiable proxy, excessive risk taking. To put it in a little more perspective this paper will look at the excessive risk taking done by firms in the pharmaceutical and non-pharmaceutical sector in the US in particular.
The SOX Act was made for overseeing the financial landscape of finance professionals to make sure that there would be less fraudulent accounting activities. Previous research has found that the presence of fraud has a very strong correlation with a firm’s growth opportunities which is (most of the time) measured by R&D expenditures (Gerety and Lehn, 1997) and also capital expenditures. Since pharmaceutical companies are running their daily business on R&D expenditures I believe it is interesting to see if the SOX Act has made an impact in the industry which has to do a lot with one of the main proxies for fraud compared to other industries.
The way I research this is by doing a literature review, collecting data and running several regressions with capital expenditures and R&D expenditures as proxies for risk taking.
The main results are that the instalment of the SOX Act had a significant impact of around zero for the pharmaceutical firms and a strong, not significant effect for the non-pharmaceutical firms. When the connection between SOX and non-pharmaceutical firms was let go, I found a negative impact with a 10% significance level.
This paper is organized as follows. Section 2 provides background information about the SOX Act and reviews the literature. Section 3 describes the data and empirical approach. Section 4 presents and explains the estimation results. Finally, section 5 concludes, highlights limitations of this research and discuss possible future research.
2. Background & Literature
This section defines and explains the important concepts of the Sarbanes-Oxley Act. I then discuss the literature on the relationship between the installment of SOX and risk-taking, particularly for the pharmaceutical market. The Sarbanes-Oxley Act1 was designed as a reaction to the major scandals of Enron and WorldCom, among others. The scandal around Enron included accounting fraud, poor financial reporting and creation of shady special purpose entities by their CFO and other executives who together were responsible for the embezzlement of billions of dollars. The reason the SEC (Securities and Exchange Commission) got notice of the fraud was that the stock of Enron went from US $90.75 per share in June 2000 to less than US $1 per share by the end of 2001. That was a big indicator that something was not right, so when the investors filed a lawsuit the SEC started an investigation and discovered the fraud (New York Times, 2013). After Enron had not accepted a hostile takeover bid by their biggest competitor Dynegy, Enron filed for bankruptcy with still $63.4 billion in assets which made it the bankruptcy with the highest amount of assets in the U.S. until then (Benston, 2003). The next year even a bigger bankruptcy by WorldCom happened and together (amongst other accounting scandals) they were part of the events that triggered the installment of SOX. The act contains eleven sections ranging from additional corporate board responsibilities to criminal penalties. In the appendix you can find a table which provides an overview of the main elements of SOX.
The overall goal of SOX is to make it harder to commit fraud by committing top management to sign off on the correctness and completeness of the financial information of their firm. It also increases the penalties for top management if they involve in fraudulent activities. This is described by section 906 of the Act which states that top management needs to annually certify their company’s financial statement. In theory, one would expect that this is a simple solution for a much more complicated situation. By holding top management responsible for the completeness and correctness of the financial information the probability that they will commit fraud or take excessive risk will likely decline, since they will now bear the direct consequences of their own actions and the actions of their employees. These
1
The bill is named after U.S. Senator Paul Sarbanes and U.S. Representative Michael Oxley who together co-sponsored the Act.
repercussions can go up to a prison sentence. Note that it is not enough for top management to not engage in these activities themselves, they also need to make sure that nobody else in their firm is committing to fraud or taking excessive risk since they will be the ones who will pay for the consequences. Because of this new internal controls might be instated. Section 404, which requires companies to test and disclose the adequacy of their internal controls, is therefore also expected to have a discouraging effect on corporate risk taking according to Bargeron et al. (2009). Also, according to research about the effect of the installment of the Act on corporate risk taking, the increased litigation risk may encourage boards to reduce the level of risk taken by their corporations and change the reward structure to one that induces CEOs to take less risk which is in line with the expectations of this paper (Cohen et al., 2009). By imposing criminal liability on officers and directors from the board SOX discourages officers and directors from initiating and approving risky investment projects. These increased penalties consist out of a higher prison sentence for a single violation which can go up to 25 years. The overall view is that SOX supposedly discourages companies to take risk with respect to investments because of all the new or expanded requirements for U.S. public company board of directors which is similar to my own research (Wood, 2012).However, in the U.S. there is a lot of ongoing debate of the perceived pros and cons of SOX. Benefits of the Act are that the given financial information becomes more reliable. This was also the original purpose behind SOX, namely to restore public confidence in the financial statements made by the public companies. Also, corporate governance is supposed to be strengthened and there should be a reduction of financial statement fraud (Jahmani and Dowling, 2008). One big negative aspect, as perceived from the opponents point of view, is that the Act has made large American corporates perform worse compared to their competitors since it is a very complex and time-consuming job to comply with all the regulations of the Act. Executives have made complaints that because of the compliance costs related to SOX they are distracted from running their companies (Solomon and Low, 2004).
The main finding of this stream of literature is that the installment of the SOX Act has diminished corporate risk taking which is in line with the expected results of this paper. However, to take a more detailed look at risk taking this paper looks at the difference between
pharmaceutical companies and non-pharmaceutical companies. The reason to look at the difference between the industries is to see if the SOX Act had a different effect on firms in a sector which in general has a lot of risky projects, compared to other sectors with not as much as risky projects. Since the stream of literature says that when the SOX Act was installed the risk taking by firms has gone down it is interesting to see if this also accounts for an industry where risk taking is the daily main event and part of their core operations. Previous research has found that the presence of fraud has a very strong correlation with a firm’s growth opportunities which is (most of the time) measured by R&D expenditures (Gerety and Lehn, 1997) and also capital expenditures. Assuming this is the case, then the increase in punishments of SOX for fraud will probably reduce the likelihood of undertaking risky projects which require large expenditures on R&D, capital or intangible assets. I believe it is relevant to research this, because if the findings are the opposite of the other literature main findings then there is an ambiguous outcome which would imply that new research needs to be done.
3. Data & Methodology
This section describes the data collection and the empirical methodology.
3.1 Data
The sample consists out of 40 pharmaceutical companies and 40 non-pharmaceutical companies in North-Amerika who all existed before the instalment of the SOX act in the year 2002. For these firms I used yearly data between 1997 to 2007, thus 5 years before and after the Act was installed. I collected data for firm’s capital expenditures, R&D expenditures, assets, ticker sector classifications and national GDP data. This data was obtained from the Compustat North America database which was collected via WRDS which stands for the Wharton Research Data Services research platform. The nominal GDP data per year was collected from the OECD (Organisation for Economic Co-operation and Development) database in US $. In table 3.1.1 you can find a table with the summary statistics of all dependent and independent variables.
All firms Pharmaceutical companies Non-pharmaceutical companies CAPEX 395.2743 (635.328) 663.370 (788.532) 221.297 (431.564) R&D 1051.084 (1953.718) 1892.678 (2312.773) 249.703 (1026.417) Ln Assets 7.405 (2.077) 8.292 (2.152) 6.829 (1.809) Ln GDP 16.228 (0.163) 16.228 (0.163) 16.228 (0.163) SOX 0.545 (0.499) 0.545 (0.499) 0.545 (0.499) Pharma 0.389 (0.488) 1 (0) 0 (0) SOX*Pharma 0.213 (0.409) 0.499 (0) 0 (0) Min 0.039 (0.181) 0 (0) 0.056 (0.229) Constr 0.136 (0.343) 0 (0) 0.222 (0.416) TransCom 0.017 (0.129) 0 (0) 0.0278 (0.165) Whole 0.068 (0.252) 0 (0) 0.111 (0.315) Ret 0.136 (0.343) 0 (0) 0.222 (0.416) Serv 0.119 (.324) 0 (0) 0.194 (0.396) Manu 0.102 (0.302) 0 (0) 0.167 (0.373) 3.2 Methodology
Table 3.1.1: Summary statistics of the variables
I test whether the instalment of SOX has made an impact on risk taking. In particular, I check whether the impact differs between pharmaceutical and non-pharmaceutical companies. The way this is done is by looking at the capital expenditures and research and development expenditures from a sample of firms, where the capital- and R&D expenditures are used as a proxy for risk taking, and then do a regression on them. The reason why I am going to run two regressions on the different proxies is because it is unlikely that one variable can be hold responsible for the whole risk taking actions a firm does. The reason why capital expenditures and R&D expenditures are chosen as proxies is because these proxies are commonly used for risk in other literature with research methods similar to my own research.
The independent variables are ln Assets, ln GDP, a dummy SOX and several industry dummies. The reason to include the logarithm of assets is to take into account the firm size. A big firm has probably more capital to invest in R&D or capital expenditures when compared to a smaller firm. Ln GDP was included to make the results of the regression less noisy. When the gross domestic product is higher there is likely to be more spending and investing in risky projects compared to when times are hard. The industry dummies were added to look at the difference between the pharmaceutical and non-pharmaceutical industry and to account voor industry specific effects.
I used a panel data regression. The regression uses data for several firms before and after the instalment of the Act. This means that it is a combination of cross-sectional data (behaviour of multiple firms) and time-series data (one firm throughout a certain period of time) which implies that the best option for a regression is using panel data. The advantage of using panel data is control of heterogeneity and reducing the collinearity between the explaining variables which results in a better model.
There are a few models regarding panel data and I decided to use the random effects model with firm-fixed effects. The reason is because the differences across entities have some influence on my dependent variable so in the model I allow for individual effects, which indicates the random effects model. An advantage of random effects is that you can include time invariant variables such as gender, or in this case industries. Also panel data gives more informative data, more variability, less collinearity among the variables, more degrees of
freedom and higher efficiency. Panel data is also better able to study the dynamics of adjustment and it is better able to identify and measure effects that are simply not detectable in pure cross-section or pure time-series data. Drawbacks are that there could be design and data collection problems, measurement errors may arise because of faulty responses due to unclear questions or there could be the case of cross-section dependence (Baltagi, 2005).
3.2 Methodology
This is the tested model:
𝐶𝐴𝑃𝐸𝑋𝑖𝑡 = 𝛽0 + 𝛽 1 ∗ ln 𝐴𝑠𝑠𝑒𝑡𝑠𝑖𝑡 + 𝛽2 ∗ ln 𝐺𝐷𝑃𝑡 + 𝛽3 ∗ 𝑆𝑂𝑋𝑡 + 𝛽4 ∗ 𝑃ℎ𝑎𝑟𝑚𝑎𝑖 + 𝛽5 ∗ ( 𝑆𝑂𝑋𝑡 ∗ 𝑃ℎ𝑎𝑟𝑚𝑎𝑖 ) + 𝛽6 ∗ 𝑀𝑖𝑛𝑖 + 𝛽7 ∗ 𝐶𝑜𝑛𝑠𝑡𝑟𝑖 + 𝛽8 ∗ 𝑇𝑟𝑎𝑛𝑠𝐶𝑜𝑚𝑖 + 𝛽9 ∗ 𝑊ℎ𝑜𝑙𝑒𝑖 + 𝛽10 ∗ 𝑅𝑒𝑡𝑖 + 𝛽11 ∗ 𝑆𝑒𝑟𝑣𝑖 + 𝛽12 ∗ 𝑀𝑎𝑛𝑢𝑖 + 𝜀𝑖𝑡 𝑅𝐷𝐸𝑋𝑃𝑖𝑡 = 𝛽0+ 𝛽1∗ ln 𝐴𝑠𝑠𝑒𝑡𝑠𝑖𝑡 + 𝛽2 ∗ ln 𝐺𝐷𝑃𝑡 + 𝛽3 ∗ 𝑆𝑂𝑋𝑡 + 𝛽4 ∗ 𝑃ℎ𝑎𝑟𝑚𝑎𝑖 + 𝛽5 ∗ (𝑆𝑂𝑋𝑡 ∗ 𝑃ℎ𝑎𝑟𝑚𝑎𝑖) + 𝛽6 ∗ 𝑀𝑖𝑛𝑖 + 𝛽7 ∗ 𝐶𝑜𝑛𝑠𝑡𝑟𝑖 + 𝛽8 ∗ 𝑇𝑟𝑎𝑛𝑠𝐶𝑜𝑚𝑖 + 𝛽9 ∗ 𝑊ℎ𝑜𝑙𝑒𝑖 + 𝛽10 ∗ 𝑅𝑒𝑡𝑖 + 𝛽11 ∗ 𝑆𝑒𝑟𝑣𝑖 + 𝛽12 ∗ 𝑀𝑎𝑛𝑢𝑖 + 𝜀𝑖𝑡
CAPEX stands for the capital expenditures, RD_EXP stands for research and development
expenses, ln Assets stands for the natural logarithm of the assets and ln GDP stands for the natural logarithm of the gross domestic product. The other variables are dummy variables, they take on the value 1 or 0. SOX takes on the value of 1 if the year is 2002 or later (until 2007) and 0 if it is between 1997-2001. Pharma stands for the pharmaceutical industry and takes on the value 1 if the firm is in this industry, otherwise 0. The same goes for the other industry dummies. Min stands for the mining industry, Constr stands for the construction industry,
TransCom stands for the transportation and communication industry, Whole stands for the
wholesale industry, Ret stands for the retail industry, Serv stands for the services industry and
Manu stands for the manufacturing industry. SOX*Pharma is the interaction variable between SOX and Pharma which takes on the value of 1 if the company is in the pharmaceutical industry
and when it is the year 2002 or later and takes on 0 if one of them is not 1. The ε stands for the error term.
The hypothesis is that the instalment of the SOX Act had a negative impact on risk taking in the pharmaceutical industry when compared to other industries.
To perform this test the Chi-squared tests were done and the results were interpreted on a 5% significance level.
4. Results
This section discusses the results. In table 4.1 you can find the regression outcome for capital expenditures as dependent variable. The random effects regression has been done (as you can see in the output in the appendix) and there is a Chi-squared outcome. The Chi-squared significance level is less than 0.05, so the model jointly explains the dependent variable on a significant level. If you look at the variables themselves, you can see that lnassets, sox and
soxpharma are the only significant coefficients (yellow highlighted). The overall explanatory
power of the model, the R squared, is 0.63 so more than half of the model is explained by the variables who are included in the model. However, note the limitation that this does not indicate that there is also a causal relationship between the variables.
The coefficient SOX is negative which implies that when SOX was installed the capital expenditures of non-pharmaceutical firms decreased. The results show that the interaction variable SOX*Pharma is significantly positive. This suggests heterogeneous effects between pharmaceutical and non-pharmaceutical firms. As the coefficient of the interaction term
SOX*Pharma is larger than the coefficient of SOX, the results suggest that the effect of the
instalment of the SOX Act led to lower capital expenditures of non-pharmaceutical firms but to higher capital expenditures of pharmaceutical firms. This result is remarkable, since it is not in line with the previous literature and the hypothesis. A reason for the slight increase could be that in the pharmaceutical market the capital expenditures and R&D expenditures are so crucial for running the daily business that the instalment of the SOX Act has made almost no effect.
However, only the SOX and SOX*Pharma variables are significant. The overall significant effect therefore is that there is an effect close to zero on risk-taking for the pharmaceutical firms and for the non-pharmaceutical firms it would be more a negative effect.
In table 4.2 you can see the regression with the R&D expenditures as the dependent variable. In this case the Chi-squared is also significant so the model jointly explains the dependent variable on a significant level again. Also the same variables as in the other model are significant (lnassets, sox and soxpharma). The R squared is a little bit lower now, namely 0.55. The big difference here is that there are three more industry dummies omitted from the model. The reason for this could be that there is multicollinearity between them. The results of
this regression are similar to the ones found with the capital expenditures as dependent variable. Again, the results show that the interaction variable SOX*Pharma is significantly positive so there is the suggestion of heterogeneous effects between pharmaceutical and non-pharmaceutical firms. As the coefficient of the interaction term SOX*Pharma is here also larger than the coefficient of SOX, the results suggest that the effect of the instalment of the SOX Act led to lower capital expenditures of non-pharmaceutical firms but to higher capital expenditures of pharmaceutical firms.
To see what the overall effect is of SOX without the connection between Pharma and
SOX, I run another regression but without the interaction variable SOX*Pharma. The outcomes
can be found in table 4.3. As one can see, the overall effect of SOX is now negative and significant when a 10%-level is used (green highlighted). This is in line with the other discussed literature.
Table 4.1 Overview STATA output panel data regression random effects with capital expenditures as the dependent variable
Outcome P > | z |
Chi-squared 216.10 0.000
SOX + SOX * Pharma effect size 8.194 -
R squared 0.6312 - Coefficient P > | z | Ln GDP 58.9846 0.591 Ln Assets 188.7289 0.000 SOX*Pharma 109.2934 0.004 SOX -101.099 0.006 Pharma 190.9326 0.213 Min 175.0758 0.518 Constr 78.53604 0.659 TransCom -114.4962 0.748 Whole 3.097386 0.988 Ret 181.6078 0.315 Serv 151.3373 0.416
Table 4.2 Overview STATA output panel data regression random effects with R&D expenditures as the dependent variable
Table 4.3 Overview STATA output panel data regression random effects with capital expenditures as the dependent variable and without the interaction variable SOX*Pharma
Outcome P > | z |
Chi-squared 209.62 0.000
SOX effect size -57.99 0.084
R squared 0.632 - Coefficient P > | z | Ln GDP 44.14 0.69 Ln Assets 194.56 0.385 SOX -57.99 0.084 Pharma 246.30 0.102 Min 163.91 0.541 Constr 77.98 0.658 TransCom -117.43 0.739 Whole -2.06 0.992 Ret 187.31 0.294 Serv 159.84 0.385 Outcome P > |z| Chi-squared 181.69 0.000
SOX + SOX * Pharma effect size 639.30
R squared 0.55 - Coefficient P > | z | Ln GDP 531.2532 0.326 Ln Assets 511.3109 0.000 SOX*Pharma 1025.405 0.000 SOX -386.1051 0.044 Pharma -60.13707 0.899
Min 0 (omitted) (omitted)
Constr 0 (omitted) (omitted)
TransCom 0 (omitted) (omitted)
Whole -1016.551 0.113
Ret -28.23401 0.959
Serv 126.3299 0.834
Outcome P > | z |
Chi-squared 181.69 0.000
SOX + SOX * Pharma effect size 639.30 -
5. Conclusion
In this section I will summarize my findings, state the limitations of my research and discuss how possible future research can be done.
5.1 Conclusive points
In this paper I tried to answer the research question if the instalment of the SOX Act has been an influence in excessive risk taking in the pharmaceutical market. I stated the hypothesis that the instalment of the SOX Act has been of a negative influence on risk taking in the pharmaceutical market. To answer the research question I performed a literature review and two multiple regressions. The models I tested had a proxy for risk taking, which were capital expenditures and research and development expenditures. It contained the natural logarithm of assets which accounted for the firm size, the natural logarithm for the gross domestic product, a dummy variable for SOX, some industry dummy variables and an interaction variable between SOX and the pharmaceutical industry dummy. I performed regressions on this model using longitudinal data, also known as panel data with the random effects variant. I found that there was a slight positive effect after the instalment of SOX for the pharmaceutical companies but actually close to zero. For the non-pharmaceutical companies I found a negative effect after SOX was installed which was in line with the other articles of Bargeron et al. (2010) or Cohen et al. (2007).
5.2 Limitations
A possible explanation for the obtained results is that I have only looked at big, medium and small sized firms in America. However, they still had a market capitalization of over a 100 million US $. Maybe if I also looked at the very small firms there would be another outcome to my research.
Also, I have only looked at a few generalised industries. One could go further into detail about the industries to get a better view of the impact of SOX per industry. The proxies for risk taking, capital expenditures and research and development expenditures, could also be worked out in more detail by adding other variables which are similar with the proxies used here to get Figure 5.2: Results with R&D expenditures as the dependent variable
a more precise image of how much risk a firm takes (think for instance of Debt, if a firm has a lot of debt it probably will take more risk).
Another limitation could be that I am looking at proxies for risk taking even though risk taking itself is a proxy for excessive risk taking. Excessive risk taking cannot be precisely identified so it is a subjective view, it depends on which proxy one would choose.
This research is not externally valid since the sample is probably not large enough to account for whole North America since I have only used around 80 firms. If one would increase the sample with valid firms for the research, the external validity would increase.
5.3 Possible future research
For further research on the effect of the instalment of the SOX Act for pharmaceutical companies I would recommend that it would be conducted for a larger sample and all kinds of sizes of companies. Also, by looking at each industry specifically I think there can be different outcomes than the one found in the used literature.
List of tables and figures
Tables
Table 3.1.1 – Summary statistics of the variables 10
Table 4.1 – Overview STATA output panel data regression 15 random effects with capital expenditures as the
dependent variable
Table 4.2 – Overview STATA output panel data regression 16
random effects with research and development expenditures as the dependent variable
Table 4.3 – Overview STATA output panel data regression 16 random effects with capital expenditures as the
dependent variable and without the interaction variable SOX*Pharma
Table A.1 – Overview SOX titles 22
Figures
Figure B.1 – Average capital expenditures of pharmaceutical and 23 non-pharmaceutical companies
Figure B.2 – Average R&D expenditures of pharmaceutical and 23 non-pharmaceutical companies
Bibliography
Ashbaugh-Skaife, H., & Collins, D., & Kinney Jr, W., & Lafond, R. (2008). The effect of SOX internal control deficiencies on firm risk and cost of equity. Journal of Accounting Research, 47(1), 1-43.
Bargeron, L., & Lehn, K., & Zutter, C. (2010). Sarbanes-Oxley and corporate risk-taking. Journal of
Accounting and Economics, 49, 34-52.
Benston, George J. (November 6, 2003). The Quality of Corporate Financial Statements and their Auditors Before and After Enron.
Cohen, D., & Dey, A., & Lys, T. (2007). The Sarbanes-Oxley Act of 2002: implications for compensation contracts and managerial risk-taking. Northwestern University working paper.
Dey, A. (2009). The chilling effect of Sarbanes Oxley: A discussion of Sarbanes-Oxley and corporate risk taking. Journal of Accounting and Economics, 49(1), 53-57.
Engel, E., & Hayes, R. (2006). The Sarbanes-Oxley Act and firm’s going-private decisions. Journal
of Accounting and Economics, 47(1), 116-145.
Gerety, M., Lehn, K. (1997). The Causes and Consequences of Accounting Fraud. Managerial and
Decision Economics. 18, 587-599.
Hochberg, Y., & Sapienza, P., & Vissing-Jorgensen, A. (2009). A lobbying approach to evaluating the Sarbanes-Oxley act of 2002. Journal of Accounting Research. 47(2), 519-583.
Jahmani, Y., Downling, William A. (2008). The Impact of Sarbanes-Oxley Act. Journal of Business
New York Times (2013, 7 May). Enron shareholders look to SEC for support. New York
Times.
http://www.nytimes.com/2007/05/10/business/worldbusiness/10iht-enron.1.5648578.html?_r=0
Solomon, D., Bryan-Low, C. (2004). Companies Complain about Cost of Corporate-Governance Rules. Wall Street Journal.
Wood, David A. (2012). "Internal Audit Outsourcing and the Risk of Misleading or Fraudulent Financial Reporting: Did Sarbanes-Oxley Get It Wrong?". Contemporary Accounting
Appendix A) Tables
1. Table overview SOX titles
Title Subject Regarding
1 Public Company Accounting Oversight Board (PCAOB)
Provide independent oversight of public accounting firms providing audit services and enforcing compliance with the specific mandates of SOX.
2 Auditor Independence Establishes standards for external auditor independence
to limit conflicts of interest
3 Corporate Responsibility Senior executives take individual responsibility for the accuracy and completeness of corporate financial reports 4 Enhanced Financial Disclosures Describes enhanced reporting requirements for financial
transactions including off-balance-sheet transactions, pro-forma figures and stock transactions of corporate officers 5 Analyst Conflicts of Interest Includes measures designed to help restore investor
confidence in the reporting of securities analysts
6 Commission Resources and Authority Defines practices to restore investor confidence in securities analysts
7 Studies and Reports Requires the Comptroller General and the SEC to perform
various studies and report their findings
8 Corporate and Criminal Fraud Accountability Describes specific criminal penalties for manipulation, destruction or alteration of financial records or other interference with investigations, while providing certain protections for whistle-blowers
9 White Collar Crime Penalty Enhancement Increases the criminal penalties associated with white-collar crimes and conspiracies, adds failure to certify corporate financial reports as a criminal offense
10 Corporate Tax Returns States that the Chief Executive Officer should sign the company tax return.
11 Corporate Fraud Accountability Identifies corporate fraud and records tampering as criminal offenses and joins those offenses to specific penalties
0 200 400 600 800 1000 1200 1400 1600 Non-pharmaceut ical companies Pharmaceut ical companies 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 Non-pharmaceuti cal companies Pharmaceuti cal companies B) Figures C) STATA output
Summary statistics all companies
Figure B.1: Average capital expenditures of
pharmaceutical and non-pharmaceutical companies
Figure B.2:Average R&D expenditures of
pharmaceutical and non-pharmaceutical companies
soxpharma 649 .2126348 .409487 0 1 lnassets 625 7.405003 2.076774 2.539079 11.72548 lngdp 649 16.22838 .1624458 15.96826 16.48812 manu 649 .1016949 .3024798 0 1 serv 649 .1186441 .3236186 0 1 ret 649 .1355932 .3426202 0 1 whole 649 .0677966 .2515905 0 1 transcom 649 .0169492 .1291805 0 1 constr 649 .1355932 .3426202 0 1 min 649 .0338983 .181107 0 1 pharma 649 .3898305 .4880879 0 1 sox 649 .5454545 .4983137 0 1 gdp 649 1.13e+07 1841728 8608515 1.45e+07 year 649 2002 3.164717 1997 2007 rdexp 437 1051.084 1953.718 0 12183 capex 620 395.2743 635.3285 .297 3008.199 assets 625 9532.889 18483.16 12.668 123684 firm 648 39735.69 54757.49 1078 264391 Variable Obs Mean Std. Dev. Min Max
Summary statistics pharmaceutical companies
Summary statistics non-pharmaceutical companies
soxpharma 253 .5454545 .4989166 0 1 lnassets 246 8.291687 2.151672 2.539079 11.72548 lngdp 253 16.22838 .1626423 15.96826 16.48812 manu 253 0 0 0 0 serv 253 0 0 0 0 ret 253 0 0 0 0 whole 253 0 0 0 0 transcom 253 0 0 0 0 constr 253 0 0 0 0 min 253 0 0 0 0 pharma 253 1 0 1 1 sox 253 .5454545 .4989166 0 1 gdp 253 1.13e+07 1843956 8608515 1.45e+07 year 253 2002 3.168546 1997 2007 rdexp 213 1892.678 2312.773 .731 12183 capex 244 663.3707 788.5322 .772 3008.199 assets 246 17662.62 25354.71 12.668 123684 firm 253 36270.96 48666.21 1078 223098 Variable Obs Mean Std. Dev. Min Max soxpharma 396 0 0 0 0 lnassets 379 6.829476 1.809797 2.675183 11.05295 lngdp 396 16.22838 .1625259 15.96826 16.48812 manu 396 .1666667 .3731494 0 1 serv 396 .1944444 .3962731 0 1 ret 396 .2222222 .4162656 0 1 whole 396 .1111111 .3146672 0 1 transcom 396 .0277778 .1645434 0 1 constr 396 .2222222 .4162656 0 1 min 396 .0555556 .2293512 0 1 pharma 396 0 0 0 0 sox 396 .5454545 .4985595 0 1 gdp 396 1.13e+07 1842636 8608515 1.45e+07 year 396 2002 3.166278 1997 2007 rdexp 225 249.7031 1026.417 0 9742 capex 376 221.2969 431.564 .297 2496 assets 379 4256.073 8738.145 14.515 63130.2 firm 396 41909.97 58210.65 1704 264391 Variable Obs Mean Std. Dev. Min Max
Regression output with capital expenditures as dependent variable
Regression output with research and development as dependent variable
.
rho .66865493 (fraction of variance due to u_i)
sigma_e 224.35245 sigma_u 318.70698 _cons -2051.796 1732.18 -1.18 0.236 -5446.805 1343.214 manu 0 (omitted) serv 151.3373 186.1774 0.81 0.416 -213.5637 516.2382 ret 181.6078 180.5931 1.01 0.315 -172.3481 535.5637 whole 3.097386 213.3784 0.01 0.988 -415.1165 421.3113 transcom -114.4962 356.2283 -0.32 0.748 -812.6908 583.6985 constr 78.53604 178.0679 0.44 0.659 -270.4706 427.5426 min 175.0758 270.8109 0.65 0.518 -355.7038 705.8554 soxpharma 109.2934 38.27514 2.86 0.004 34.27547 184.3113 pharma 190.9326 153.4573 1.24 0.213 -109.8381 491.7034 sox -101.099 36.58156 -2.76 0.006 -172.7976 -29.4005 lnassets 188.7289 17.38907 10.85 0.000 154.6469 222.8108 lngdp 58.9846 109.9048 0.54 0.591 -156.4248 274.394 capex Coef. Std. Err. z P>|z| [95% Conf. Interval] corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 Wald chi2(11) = 216.10 overall = 0.6312 max = 11 between = 0.7094 avg = 10.5 within = 0.1419 min = 6 R-sq: Obs per group:
Group variable: firm Number of groups = 59 Random-effects GLS regression Number of obs = 620
_cons -11375.5 8585.852 -1.32 0.185 -28203.46 5452.459 manu 0 (omitted) serv 126.3299 604.1612 0.21 0.834 -1057.804 1310.464 ret -28.23401 551.1923 -0.05 0.959 -1108.551 1052.083 whole -1016.551 641.3636 -1.58 0.113 -2273.601 240.4982 transcom 0 (omitted) constr 0 (omitted) min 0 (omitted) soxpharma 1025.405 187.8375 5.46 0.000 657.2502 1393.56 pharma -60.13707 471.4831 -0.13 0.899 -984.2269 863.9528 sox -386.1051 192.1315 -2.01 0.044 -762.6758 -9.534301 lnassets 511.3109 68.00493 7.52 0.000 378.0236 644.5981 lngdp 531.2532 540.6669 0.98 0.326 -528.4344 1590.941 rdexp Coef. Std. Err. z P>|z| [95% Conf. Interval] corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 Wald chi2(8) = 181.69 overall = 0.5547 max = 11 between = 0.6884 avg = 10.4 within = 0.2043 min = 3 R-sq: Obs per group:
Group variable: firm Number of groups = 42 Random-effects GLS regression Number of obs = 437
Regression output without interaction variable SOX*Pharma
Regression output with added variable nonpharma
rho .65999749 (fraction of variance due to u_i)
sigma_e 226.19584 sigma_u 315.1481 _cons -1630.417 1734.042 -0.94 0.347 -5029.076 1768.243 manu 0 (omitted) serv 159.8437 184.1847 0.87 0.385 -201.1518 520.8391 ret 187.3077 178.67 1.05 0.294 -162.879 537.4945 whole -2.062821 211.0556 -0.01 0.992 -415.7241 411.5985 transcom -117.4321 352.3428 -0.33 0.739 -808.0114 573.1471 constr 77.98795 176.1252 0.44 0.658 -267.2111 423.187 min 163.9123 267.8847 0.61 0.541 -361.1321 688.9567 pharma 0 (omitted) sox -57.99144 33.55818 -1.73 0.084 -123.7643 7.781388 nonpharma -246.3022 150.5402 -1.64 0.102 -541.3556 48.75119 lngdp 44.14071 110.5152 0.40 0.690 -172.4652 260.7466 lnassets 194.5605 17.28071 11.26 0.000 160.6909 228.4301 capex Coef. Std. Err. z P>|z| [95% Conf. Interval] corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 Wald chi2(10) = 209.62 overall = 0.6322 max = 11 between = 0.7125 avg = 10.5 within = 0.1275 min = 6 R-sq: Obs per group:
Group variable: firm Number of groups = 59 Random-effects GLS regression Number of obs = 620 rho .65999749 (fraction of variance due to u_i)
sigma_e 226.19584 sigma_u 315.1481 _cons -1876.719 1742.908 -1.08 0.282 -5292.756 1539.319 lngdp 44.14071 110.5152 0.40 0.690 -172.4652 260.7466 lnassets 194.5605 17.28071 11.26 0.000 160.6909 228.4301 manu 0 (omitted) serv 159.8437 184.1847 0.87 0.385 -201.1518 520.8391 ret 187.3077 178.67 1.05 0.294 -162.879 537.4945 whole -2.062821 211.0556 -0.01 0.992 -415.7241 411.5985 transcom -117.4321 352.3428 -0.33 0.739 -808.0114 573.1471 constr 77.98795 176.1252 0.44 0.658 -267.2111 423.187 min 163.9123 267.8847 0.61 0.541 -361.1321 688.9567 pharma 246.3022 150.5402 1.64 0.102 -48.75119 541.3556 sox -57.99144 33.55818 -1.73 0.084 -123.7643 7.781388 capex Coef. Std. Err. z P>|z| [95% Conf. Interval] corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 Wald chi2(10) = 209.62 overall = 0.6322 max = 11 between = 0.7125 avg = 10.5 within = 0.1275 min = 6 R-sq: Obs per group:
Group variable: firm Number of groups = 59 Random-effects GLS regression Number of obs = 620
Robustness check with capital expenditures as dependent variable, regression with fixed effects
Robustness check with research and development expenditures as dependent variable, regression with fixed effects
F test that all u_i=0: F(58, 557) = 22.95 Prob > F = 0.0000 rho .75073647 (fraction of variance due to u_i)
sigma_e 224.35245 sigma_u 389.3545 _cons -3714.719 1785.397 -2.08 0.038 -7221.653 -207.7848 manu 0 (omitted) serv 0 (omitted) ret 0 (omitted) whole 0 (omitted) transcom 0 (omitted) constr 0 (omitted) min 0 (omitted) pharma 0 (omitted) sox -99.97977 36.16259 -2.76 0.006 -171.0115 -28.94805 soxpharma 121.3626 37.98654 3.19 0.001 46.74817 195.9769 lnassets 129.6312 24.14217 5.37 0.000 82.21041 177.0521 lngdp 195.943 115.755 1.69 0.091 -31.42668 423.3127 capex Coef. Std. Err. t P>|t| [95% Conf. Interval] corr(u_i, Xb) = 0.5336 Prob > F = 0.0000 F(4,557) = 23.92 overall = 0.6139 max = 11 between = 0.6959 avg = 10.5 within = 0.1466 min = 6 R-sq: Obs per group:
Group variable: firm Number of groups = 59 Fixed-effects (within) regression Number of obs = 620
rho .55166489 (fraction of variance due to u_i) sigma_e 959.05666 sigma_u 1063.8504 _cons -15614.8 9044.357 -1.73 0.085 -33396.45 2166.858 manu 0 (omitted) serv 0 (omitted) ret 0 (omitted) whole 0 (omitted) transcom 0 (omitted) constr 0 (omitted) min 0 (omitted) pharma 0 (omitted) sox -381.2747 191.7384 -1.99 0.047 -758.242 -4.307357 soxpharma 1061.878 188.3765 5.64 0.000 691.5203 1432.235 lnassets 354.6497 120.1449 2.95 0.003 118.4387 590.8606 lngdp 857.3611 585.0502 1.47 0.144 -292.8766 2007.599 rdexp Coef. Std. Err. t P>|t| [95% Conf. Interval] corr(u_i, Xb) = 0.3582 Prob > F = 0.0000 F(4,391) = 25.53 overall = 0.5270 max = 11 between = 0.6528 avg = 10.4 within = 0.2071 min = 3 R-sq: Obs per group:
Group variable: firm Number of groups = 42 Fixed-effects (within) regression Number of obs = 437
