Master Thesis.
Does an increase in a legal regime affect the price that
auditors charge for auditing intangible assets?
A research on the effect and implications of an increase in a legal regime,
the implementation of the Sarbanes-Oxley act, on the audit fees of
intangible assets.
Jelle J. D. Peters 10332200
Thesis supervisor: Dhr. Dr. P. Ghazizadeh Final version, date: 23-06-2017
Word count: 14.938
MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam
Table of Contents
Acknowledgements. ... 2
Abstract. ... 2
1. Introduction. ... 3
2. Literature review and hypothesis development. ... 6
2.1 Intangible assets and the risk for the auditor. ... 6
2.1.1 Intangible assets. ... 6
2.1.2 Risk of intangible assets for the auditor. ... 8
2.2 Audit fee model. ... 11
2.3 SOX, the effect of SOX on audit pricing, particularly on intangible assets. ... 13
2.3.1 SOX. ... 13
2.3.2 The effect of SOX on audit pricing. ... 16
2.3.3 The implications of SOX on intangible assets. ... 18
2.4 Hypothesis development. ... 19
3. Methodology and data. ... 20
3.1 Development of the regression model. ... 20
3.2 Sample selection. ... 22
3.3 Descriptive statistics and manipulations of the variables. ... 24
3.3.1 Main regression ... 25
3.3.2 Sub regression with R&D added ... 31
3.3.3 Sub regression with R&D instead of intangible assets. ... 32
4. Results... 34
4.1 Results. ... 34
4.2 Sub regressions. ... 37
4.2.1 Sub regression with R&D added. ... 37
4.2.2 Sub regression with R&D instead of intangible assets. ... 38
5. Discussion. ... 40
6. Conclusion. ... 41
Table of figures.
Table 1, descriptive statistics of the variables of the main regression ... 26 Table 2, Correlation between the independent variables of the main regression ... 29 Table 3, regression results of the main regression, sub regression 1 and sub regression 2. ... 35
Abstract.
An increase in a legal regime could have unexpected consequences. This thesis shows that intangible assets have an incremental sensitivity towards an increase in a legal regime, which is important since it affects the M&A process in a country. The process is affected because there is more risk involved for the auditors and they will charge higher fees for auditing intangible assets. This will be studied by an empirical study with multiple regressions on the audit fees that auditors charge. It turns out that the innovation process, R&D expense, is not affected by the implementation of SOX but the M&A process is affected since auditors charge 8,9% higher audit fees, on average 1300 dollars. I question if this will stop M&A since this is nog a significant amount if compared to total amount of goodwill payed, on average 25 million dollars.
Acknowledgements.
I want to thank Pouyan Ghazizadeh for always guiding me in the right direction and correcting me and my thesis during this process. I also want to thank Hilde Beune for correcting my grammar and for all of the support. Also I would like to thank Mark Smeenk for helping me out as soon as I was stuck with my regression. Lastly, I would like to thank KPMG for the opportunity to write my thesis at their office and Robin van Broekhoven for offering me guidance and discussing topics, implications and possible results of my thesis. I could not have written a thesis like this without your help!
Statement of Originality
This document is written by student Jelle Peters who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
1. Introduction.
“A ship is always safe at the shore but that is not what it is built for.” (Albert Einstein). An auditor is used to bearing risk, that is born with the profession, providing reasonable assurance that there are no material misstatements in financial statements. But as soon as the risk increases an auditor would want to be compensated for this increase (Simunic, 1980). After the implementation of the Sarbanes-Oxley act (here-after: SOX) the risk of an audit increased (Asare, Cunningham, & Wright, 2007). Intangible assets are a risky type of assets for the auditors because it has more discretion than normal assets (Kohlbeck, Cohen, & Holder-Webb, 2009). This study will take a look at the effect on the audit pricing of intangible assets after an increase in the legal regime, the implementation of SOX, and discusses the possible effects and implications.
Intangible assets is a group of assets which are hard to define and measure. They include (non-exclusive) recognized R&D, patents, trademarks, copyrights, intellectual property and goodwill (Tsai, Lu, Hung, & Yen, 2016). It is difficult for a part of the intangible assets, mostly research and development expenses (here-after: R&D), to determine if they could be recognized as an asset, for what price it could be recognized and at which moment in the year it should be recognized.
The auditor has to make an assessment for all the intangible assets to what extent the value is correct and if they already can be recognized. This is a challenging and time consuming procedure. For some intangible assets the auditor even has to hire a specialist (Yao, Percy, & Hu, 2015). All of this increases the audit fees which the auditor would like to receive from the client. Because of this the auditor faces a lot of risk as well. It is riskier because it is easier to make mistakes because the intangible assets are more complicated than other assets. This is shown by the fact that companies with more intangible assets have more analyst coverage, who can and want to elaborate on the information in the financial statements (Barth, Kasznik, & McNichols, 2001). Regular assets are straightforward and easy to audit but there could be an incremental sensitivity of changes like SOX on the intangible assets.
The possible impact of the incremental sensitivity regarding the intangible assets are firstly important for the regulators and companies themselves. If the pricing of audit fees for intangibles increases, many companies will re-assess if investing in intangibles is worth the costs. The regulators increased the legal environment but if SOX has a severe effect on the audit fees it could lead to a lack of innovation in a nation as well, because intangible assets are a driver of innovation and value creation (Pike, Roos, & Marr, 2005). This could also have implications for companies with merger and acquisition strategies (here-after: M&A) since goodwill is a major component of intangible assets. These companies will pay less or stop their acquisitions due to an increase in price. The innovation aspect is mostly caught by the R&D expense, since it is only possible to recognize computer related expenses under US GAAP (Deloitte, 2017). The potential incremental sensitivity also has implications for auditing firms, if audit fees for intangibles are rising, a good audit of intangible assets could be a competitive advantage. Overall the implications for a rise in the audit fees of intangible assets could have severe outcomes.
Since the implementation of SOX, the audit agency is personally liable and beares more risk for the financial statements they sign off (US House of representatives, 2002). The Public Company Accounting Oversight Board (here-after: PCAOB), which has been created with the implementation of SOX, can deprive the license of the agency and there is a bigger chance of litigation by investors since SOX. This means that there is an even bigger risk for the auditors after SOX, because there is a chance of litigation as soon as they miss-assess any of the intangible assets. I assume that auditors would like to be compensated for this risk, which leads me to the following research question in this thesis:
‘To what extent did the implementation of the Sarbanes-Oxley act affect the price auditors charge for auditing intangible assets’.
My thesis contributes to the current literature since it focuses on the implications of the implementation of SOX on the audit fees of a specific type of assets. There has been already a lot of studies describing the audit fees, the implications of SOX and companies who beare more risk after the implementation of SOX (Hay, Knechel, & Ling, 2008), (Raghunandan & Rama, 2006) & (Lenard, Petruska, Alam, & Yu, 2012). But there has been no research yet to the implications of SOX for intangible assets.
This is relevant because of the discretion management has regarding intangible assets and the potential incremental sensitivity with its possible implications, which was mentioned earlier.
I will try to discover the possible effect of auditors charging more for intangible assets by performing multiple regression analyses with the Audit fee model of Simunic (1980) as a starting point. Additionally, I will use a few different more recent papers to develop my own audit fee model, this will be explained in section 2.2. Eventually, I will include these variables: Audit fees, firm size, complexity of the audit, auditor risk, intangible assets, SOX dummy and an interaction variable. A R&D and interaction 2 variable will be added in the two sub regressions. In the regressions the audit fees will be the dependent variable and the others are independent variables.
The regression will be performed with data of financial statements from the pre- and post-SOX period and are retrieved from the Compustat Fundamentals Annual, Compustat Historical Segments and Audit Analytics databases. First I look at the impact for companies with a M&A strategy in the main regression with just intangible assets. Subsequently there are two sub regressions performed, one with R&D expense included in the main regression and one with R&D expense instead of intangible assets with only companies with R&D expense included. This is performed to get an insight in the effect of SOX on the innovation process.
The outcome of the main regression suggests that the M&A process is affected, since auditors charge 8,9% higher audit fees for auditing intangible assets after the implementation of SOX. This means that the price for acquisitions for companies went up as well. But this will not affect the M&A process since companies have to pay approximately 1300 dollars more for auditing their intangible assets after the implementation of SOX while the average company has 25 million dollars of intangible assets. For auditors the higher audit fees could be important though since they can get a competitive advantage by focusing on intangible assets.
The two sub regressions did not have significant results for interaction variables with R&D expense. This means that the implementation of SOX did not affect the audit fees charged for auditing R&D expense. So the innovation process did not increase in price after the implementation of SOX. No increase in price could be due to the fact that the R&D expense is not recognized so there is less discretion by management in determining the value. This means that the implementation of SOX did not have severe effects for society at large since it only affects companies with M&A strategies slightly. In the next section, section two, I will discuss the relevant literature. At the end of section two the hypothesis will be developed after a summary of the relevant literature. The third section of this thesis will describe the methodology and discuss which data is used. The results are presented in the fourth section while there is a discussion about those results in the fifth section. I will end the thesis with the conclusion in section six.
2. Literature review and hypothesis development.
The literature review is divided in four sections. In the first section I will discuss what intangible assets are in general and what the risks of the intangible assets are for the auditor. In the second section the audit fee model will be discussed; which shows how auditors determine what the audit fee will be. The third section will discuss to what extent SOX has an effect on the auditors, especially on the intangible assets. This part will elaborate the effect of being personally liable as an audit agency on the audit pricing as well. In the last section I will summarize and use the conclusions of the literature review and by that I will develop my hypothesis.
2.1 Intangible assets and the risk for the auditor. 2.1.1 Intangible assets.
Before I describe what the risks of the intangible assets are for the auditors in section 2.1.2, I will explain what intangible assets are in general. Intangible assets are assets which cannot be touched, so include knowledge, information, intellectual property, goodwill and recognized R&D (Tsai et al., 2016). For all of these assets it is hard to assess the price and determine if they already can be recognized. However according to Tsai et al. (2016) the intangible assets are a big driver of profitability if used correctly with the right assets, so intangible assets are very important for a company.
Intangible assets are getting more important for companies, which is shown by the change from a focus on tangible to a focus on more intangible assets in the United States industry (more than doubled from 1977 to 2006) (Nakamura, 2010). But intangible assets are also essential for value creation for companies where reputation is important (Pike, Roos, & Marr, 2005). An explanation can be found in the notion that they are important for the R&D process (Pike et al., 2005). This is shown by the fact that companies with a lot of know-how and human capital are almost fully dependent on intangible assets (Axtle-Ortiz, 2013). So intangible assets are important for a lot of companies and because intangibles are important for the R&D process, they are important in the innovation process for companies.
This is emphasized by Kramer, Marinelli, Iammarino, & Revilla Diez (2011), who describe that for companies in the knowledge-intensive industries intangible assets could be a competitive advantage. Besides that intangible assets provide a competitive advantage, a type of intangibles, intellectual property, give an advantage in the innovation process too (Hayton, 2005). A problem of intangible assets is that they are hard to valuate and internally generated R&D cannot always be recognized using US GAAP (Hayton, 2005) & (Deloitte, 2017). The valuation issue will be elaborated next. An important characteristic of intangible assets is that they cannot be separated from tangible assets and create synergies for the company with the tangible assets (Basu & Waymire, 2008). The intangible assets are often undervalued because of these synergies and they are worth more to the company then it states on the balance sheet. This happens because the intangible assets that are created with R&D, especially in the United States, are valued at historical cost if they could be recognized (Basu & Waymire, 2008). This is also because it is hard to assess the added value of the intangible assets. But they are also hard to measure because it is not clear who invests and who profits exactly (Nakamura, 2010). This shows that intangible assets are hard to valuate and are underestimated in general, this could be troublesome for the auditors, which will be elaborated in section 2.1.2.
Intangible assets are a big driver for profitability. Especially for companies with a lot of R&D and innovation the intangible assets are crucial. Besides being crucial for a company, intangibles are crucial for the innovation process as well. The intangible assets provide a competitive advantage for most companies, mainly because of the synergies with the tangible assets. Yet the value of these synergies and the intangible assets itself are hard to asses and cannot always be recognized, therefore the intangible assets are often undervalued.
2.1.2 Risk of intangible assets for the auditor.
This section will elaborate why intangible assets are riskier and harder to audit. The implications for this research are that the complexity and risk of the intangible assets for the auditor could create an incremental sensitivity regarding new regulations, like SOX.
Companies with more intangible assets have more coverage by analysts (Barth et al., 2001). This heightened coverage is because of the risk associated with intangible assets, which creates an information breach where the analysts think they can elaborate on the pricing. There is more analyst coverage since there is more risk and less informative prices due to the intangible assets on the balance sheet of these companies (Barth et al, 2001). The analysts want to clarify and inform investors about the actual risk and prices. Investors in general experience problems while assessing the market value of intangible assets as well (Tsai et al., 2016). Which means that there is a clear opinion from the analysts as well as the investors that the valuation of intangible assets is hard and often unclear. The auditors carry a lot of risk by auditing intangible assets because the valuation is hard and often unclear.
But investors do think intangible assets are important because companies with more unrecognized intangible assets have significantly higher returns (Kohlbeck & Warfield, 2007). This is also important for the valuation of the company and the intangible assets, because auditors have to account for these intangibles. Eventhough those intangible assets are not even on the balance sheet. This could be troublesome.
For example insurers can receive the highest profit if they invest in risky assets, but if they suffer significant losses with their investment, their (unrecognized) intangible assets like brand name wil decline in value (Yu, Lin, Oppenheimer, & Chen, 2008). Most of these intangible assets can not be recognized, but should be considered during the audit.
So unrecognized intangible assets are troublesome for the auditors while they are not even on the balance sheet. This also shows a risk due to the chance of litigation as soon as the auditors make a mistake or missasses anything of these intangible assets. Especially, because investors take the unrecognized intangible assets in account in their investment decisions.
The intangible assets are the most important assets of the firm (Kanodia, Sapra, & Venugopalan, 2004). However there is a difference in market valuations of the company as soon as the intangible assets are recognized or just expensed. What is remarkable is that it is only relevant to recognize intangible assets for your stock price if it is a major component of the company (Kanodia et al., 2004). This is due to the fact that most investors already take the intangible assets into account like described earlier. Intangible assets that are voluntarily recognized are valued higher by investors than intangible assets that have to be recognized, like goodwill after the purchase of a company (Wyatt, 2005). This demonstrates that investors value the discretion of management highly. But there are mixed reactions by investors as soon as the brand name is recognized as an asset (Kallapur & Kwan, 2004).
These mixed reactions are due to the fact that the amount that is recognized is not always in line with the current market capitalization. This means that management will and can use their discretion to their advantage, which could be troublesome for auditors as they may have to scrutinize it to get the amounts correct. So there is discretion in the valuation of intangible assets, but managers use this discretion carefully, shown by the fact that they only recognize intangibles that are highly correlated with underlying economic factors (Wyatt, 2005).
US GAAP tries to resolve this by focussing on conservatism regarding the intangible assets. This is emphasized by the fact that only goodwill is recognized at fair value under US GAAP (Deloitte, 2017). The other intangible assets are valued at historical cost. Remarkable is the treatment of the R&D assets under US GAAP (Deloitte, 2017). The R&D expenses should always be expensed when incurred except for a few assumptions mostly relating to computer software and hardware (Deloitte, 2017). A company is also allowed to capitalize the R&D expenses if they can be used for a different intention.
This conservatism is maybe due to the fact that US GAAP wants to deter opportunism within companies (Kothari, Ramanna, & Skinner, 2010). This is because of the decrease in managerial discretion which arises with fair value accounting. However I assume that a side effect of this conservatism in US GAAP could be that it also withholds innovation within companies like investments in R&D.
Like discussed before, the intangible assets could be valued at historical cost or at market value which gives often a lot of discretion. But there are other options for the auditor if neither of those methods are available. According to McClure (2012 A) the intangible assets could be valuated with the comparable profits method and the residual profit split method. Both are based on the predicted returns that the intangible assets will generate in the future. There is a lot of criticism surrounding these methods, because it does not show the systematic risk involved with it (Mcclure, 2012 B). What is clear is that there is a debate surrounding the valuation of intangible assets and that it is hard to assess the correct value. But there is a lot of discretion for management and options to value the intangible assets as well. This makes the intangibles troublesome and harder for auditors.
The debate surrounding the valuation of intangible assets is proved by the fact that even regulators are struggling with how to handle the valuation (Eckstein, 2004). Surrounding the implementation of SOX, regulators were trying to harmonize and clearify the rules surrounding the valuation of the intangible assets.
Especially, because there are differences between US GAAP en the rules in the United Kingdom (Eckstein, 2004). This is hard for an auditor if he has clients who have to report in both countries. A bigger discussion about the regulations outside the United States is not relevant for this thesis since it focusses on an increase in a legal regime in the United States.
So Intangible assets are more troublesome for the auditors then other assets for a few reasons. First of all, companies with a lot of intangible assets have more analyst coverage. Secondly, the intangible assets are getting more important in the current society, but the value is hard to assess and the asset itself is not always recognized. Thirdly, there are a lot of ways to valuate and there is a lot of discretion for management in valuating the intangible assets. The auditors have to assess whether these assumptions and valuations are correct. Lastly, there is a debate surrounding the valuation of intangible assets with a focus in US GAAP on conservatism to deter the opportunism of management.
2.2 Audit fee model.
This section describes the audit fee model and to what extent some parts and developments are relevant for this thesis and the regression. The audit fee model has been developed heavily throughout the years. It started off with the original audit fee model from Simunic (1980). He focused mainly on the auditee size, the amount of resources consumed by the auditor and the internal accounting system of the auditee. Eventually, he focused as well on the receivables, inventories, profit, complexity of the auditee and the time that the auditor has been auditing the client (Simunic, 1980). This gives a good starting point of the audit fee model which I will use and which has been developed heavily throughout the years.
Simunic (1980) mentions a risk premium that the auditor wants for auditing the statements of a client as well. The audit fee will rise if there is more inherent risk during the audit. This risk is associated with the litigation risk that the auditor will endure during and after the audit. The auditor wants to be compensated for this litigation risk, due to the possible chance of a damaged reputation and incurring litigation payments (Simunic, 1980).
More recently there have been a few additions to the audit fee model which makes it more accurate. Hay, Knechel & Wong (2006) have summarized these improvements until that moment and describe almost every component (+/-150) that could influence the audit fee. Overall there is in this study a focus on auditor risk, complexity of the client, leverage of the client, strategy/market of the client and auditor specific attributes. The relevance for my study is that companies with M&A strategies, high R&D costs and major shareholding companies have higher audit fees. These are all companies or strategies which involve a lot of intangible assets, because of goodwill (resulting from a M&A and major shareholding strategy) and recognized R&D.
As soon as intangible assets are measured at fair value instead of historical cost, the audit fees will rise (Yao et al., 2015). This means that an increase in complexity and risk will result in higher audit fees. It is more complex because intangible assets at fair value are harder to valuate and more risky because it is easier to missasses the value. Generally, Fields, Fraser & Wilkins (2004) note that audit fees are driven by complexity and risk. When intangible assets are used as a proxy for M&A strategies, because of the goodwill involved in this strategy, there is a significant correlation between the amount of intangible assets and the audit fees (Fields et al., 2004). So, I assume that the intangible assets are very complex and risky and therefore have a correlation with the audit fees.
This notion is supported by the fact that risk is a main component of the audit fees (Jubb, Houghton, & Butterworth, 1996). Risk can be divided in business risk and auditor risk. However business risk is not significantly correlated with the audit fees, auditor risk is significantly correlated though. Auditor risk is the risk that an auditor beares during the audit and similar to litigation risk. As soon as the auditor risk increases, the fees will go up (Jubb et al., 1996). This risk can be quantified and measured by the amount of unexplained audit fees (Hribar, Kravet, & Wilson, 2014). The unexplained audit fees are a sign of accounting quality which has to increase if the audit risk is higher (Hribar et al., 2014). So risk is a major part of the audit fees but can be shown by an unexplained increase in the audit fees.
As soon as a legal regime in a country increases the audit fees will increase significantly (Choi, Kim, Liu, & Simunic, 2008). An increase in a legal regime means a scrutiny for the rules surrounding auditors or exposing auditors to more risk during an audit. Likewise happened with the implementation of SOX. So complex assets, like intangibles become more risky and therefore I assume that the audit fees will rise after the increase in the legal regime.
Together these studies show that as soon as the legal regime increases, which is a scrutiny for the rules like with the implementation of SOX (see section 2.3), that the auditor beares more risk. Because the risk an auditor beares is a major determinant of the audit fees, the audit fees will probably rise after the increase in the legal regime. The increase in risk can be quantified by the rise in the audit fees for the part that is not explained by other variables of the audit fee model. Intangible assets are risky assets for the auditors, like described in section 2.1.2, but could have, as mentioned, an incremental sensitivity to the increase in the legal regime.
So if you take the intangible assets in mind, the major components of an audit fee would be the company size, the complexity of the company, the auditor risk and the intangible assets in general. In section 3.1 I will explain to what extent these components will be used in the regression and how they will be operationalized.
2.3 SOX, the effect of SOX on audit pricing, particularly on intangible assets. This section describes the overall implementation of SOX and the effects of SOX, divided in three parts. The first part describes the implementation of SOX and elaborates as to what SOX is in general. The second part explains the effects of SOX on the audit fees, while the third part focuses on the effects of SOX on intangible assets specifically.
2.3.1 SOX.
Due to the major collapse of multiple companies, like Enron and WorldCom, in the beginning of this century the Sarbanes-Oxley act (SOX) was implemented (Bather & Burnaby, 2006). This was a reaction by the American government, who blamed the senior executives and the external auditors for the collapses. SOX would regulate and scrutinize the work of the auditor to ensure that similar collapses will not happen again.
SOX was implemented in the summer of 2002 and would be active half a year later (180 days) at the end of January 2003 (US House of representatives, 2002). This section will discuss what SOX is in general and what its effects were for the accountants in general.
After the implementation of SOX, the quality of the financial reporting in the United States improved (Srinivasan & Coates, 2014). This is partly due to the increased regulation, but caused by the more cautious attitude of the clients and the auditors as well. Besides the direct costs, like audit fees, the indirect costs and other benefits ten years after the implementation of SOX cannot be measured (Srinivasan & Coates, 2014). So besides the improved financial reporting quality the benefits of SOX cannot be measured yet.
With the implementation of SOX the PCAOB has been created and an audit firm has to register with the PCAOB if it wants to audit publicly traded companies (US House of representatives, 2002). The PCAOB has been given a lot of power to scrutinize auditors. Audit firms have to act more careful and have to perform more work, especially because of the new In Control Of Financial Reporting (here-after: ICOFR) report described in section 404.
According to the new SOX law there is a greater opportunity on litigation by the investors to sue the client as well as the auditor (US House of representatives, 2002). Due to these implications auditors will be more careful and beare more risk than before the implementation of SOX.
Before the PCAOB was created the AICPA was creating the rules for the independence, which was effectively just self-regulation (Bather & Burnaby, 2006). The society thought that the collapse of Enron was due to problems with independence of the auditors. Therefore, the main focus of the PCAOB is to keep the auditors independent of their clients. The other main focus that the PCAOB had was that they can give severe sanctions to audit agencies. These sanctions have to be reviewed by the Securities and Exchange Commission (here-after: SEC) (Bather & Burnaby, 2006). However the SEC will only act if the sanction is excessive or unnecessary. So the PCAOB has been given a lot power as soon as it was created.
Until 2011 the PCAOB disciplined 63 CPA’s or firms after reviewing the audits (Herron & Gilbertson, 2011). This shows that the PCAOB is not afraid to show their power and act on it. Most of these cases the sanction was due to an act of omission, which means that the auditor or audit firm did not collect enough audit evidence for a highly material account. Because of these mistakes 13 audit firms, which were registered with the PCAOB, got their license revoked permanently (Herron & Gilbertson, 2011). The implication for the auditors is that there is more risk involved and they should audit more careful after the implementation of SOX.
After the implementation of SOX, a lot of non-audit services performed by the auditor are no longer allowed, or by specific approval by the audit committee of the client (Krishnan & Yu, 2011). Therefore, a lot of knowledge spillovers, the effect that the audit is cheaper because the auditor also performs non-audit services and knows the company better, would be lost. Remarkably, it turns out that the knowledge spillovers got bigger after the implementation of SOX (Krishnan & Yu, 2011). This could be due to the specific tasks that the auditor is still allowed to do that created more knowledge spillovers. Or this could be due to the fact that the auditor is more aware and efficient with information. What is relevant for this thesis is that the implementation of SOX has at least no impact on the audit fees because of a decrease in knowledge spillovers. Section 805 of the Sarbanes-Oxley act states that the auditor is personally liable for the financial statements they sign off (US House of representatives, 2002). The auditors and the audit firm are only personally liable, to be sued, as soon as it is clear that there was a mistake due to severe negligence (US House of representatives, 2002). The implications is that only after there was severe negligence in the audit that there is a chance of litigation towards the auditors. SOX was published on 30th of July 2002 and would be active 180 days after publication (27th of January 2003). This means that auditors in the United States are personally liable for all the financial statements they sign off from 27th of January 2003. For this thesis I assume that is for every financial statement that is signed of in 2003 because there are hardly any sign offs in the first 4 weeks in January. But also because audit firms are already under pressure and clients would not accept if audit firms would evade new laws.
The financial reporting quality improved after the implementation of SOX. Thereby with the implementation of SOX the legal regime increased because the PCAOB has been created and it can and will scrutinize auditors. Besides this implication, the auditors are personally liable since the implementation for the financial statements that they sign off. Lastly, after the implementation of SOX the knowledge spillovers did not decrease significantly which could not have an effect on the audit fees. The next section will describe more implications of SOX but focuses explicitly on the audit pricing. 2.3.2 The effect of SOX on audit pricing.
Besides all of the general effects of SOX on the audit there are a lot of effects on the pricing of the audit. This section will clarify these effects and explain to what extent it affects the intangible assets and this thesis. In section 3.1 the effect on the regression will be discussed.
Since auditors beare more risk after the implementation of SOX, auditors will act more careful (Habib, Jiang, Bhuiyan, & Islam, 2014). The implications of possible litigation for clients as well as for auditors made them more careful during the audit. Investors therefore assume that discretionary accruals, for example, are more informative (Habib et al., 2014).
Auditors act more careful but often do not ask a premium in the audit fees. What happens instead is that the audit agencies will refuse to audit the risky clients (Habib et al., 2014). There are mixed results regarding this topic though, since auditors are asking for a premium in the audit fee as soon as the personal assets of management are excluded of litigation (Chung, Hillegeist, & Wynn, 2015). This is a sign of a risky client, because management will behave more opportunistically if they have nothing to lose. What is relevant for this thesis is that it shows how audit agencies act regarding risky clients, or clients with risky assets, after the implementation of SOX. Audit firms either refuse risky clients or ask a premium in the audit fees.
After the implementation of SOX the fees for auditors rose (Ghosh & Pawlewicz, 2007). The rise is due to the expected auditor legal liability. Due to a higher chance of legal liability, auditors will perform more work on a client and they want to be compensated for the higher risk that they endure after the implementation of SOX (Ghosh & Pawlewicz, 2007). The notion of acting more careful is confirmed by the fact that auditors find more significant deficiencies and less material deficiencies (Asare, Cunningham, & Wright, 2007). A significant deficiency is a mistake with impact while a material deficiency is a severe mistake with more impact. This effect can be explained by more work performed by the auditor and avoiding legal liability since the auditor is legally involved as soon as he assesses a deficiency material. So both the extra work as well as the higher risk will lead to a rise in the audit fees and has effect on the audit. There is a rise of 51% of the audit fees for companies which are complying with SOX (Cosgrove & Niederjohn, 2008). There are two reasons for this rise, firstly this is due to the increased litigation risk for the auditors. Secondly the rise is due to compliance with the SOX 404 report (Cosgrove & Niederjohn, 2008). The second reason is confirmed by the fact that the first group of companies who complied with the SOX 404 report were paying between 65% and 86% more for an audit compared to the companies who did not yet comply (Bhamornsiri, Guinn, & Schroeder, 2009) & (Raghunandan & Rama, 2006). So the higher audit fees after the implementation of SOX is due to the increased litigation risk for auditors and the SOX 404 report. What increases the audit fees even more is when you comply with the SOX 404 report and there are material weaknesses found. The audit fees are between 35% and 43% higher when there is a material weakness found in the internal control system of the auditee than with companies that do not have material weaknesses (Hogan & Wilkins, 2008) & (Raghunandan & Rama, 2006). Even the pervasiveness and severity of material weaknesses affect the audit fees significantly (Mitra, 2009).
An increase in the audit fees before SOX was most of the time a sign that the earnings quality of the financial statements increases (Mitra, Deis, & Hossain, 2009). This increase was due to more hours in the audit to scrutinize management discretion. Mitra et al. (2009) found that this phenomenon, although less heavy, still exists after the implementation of SOX.
This means that if there is an unexplained rise in the audit fees of a particular company that this can be explained by the auditors who is scrutinizing managerial discretion (Mitra et al., 2009). So intangible assets have a lot of managerial discretion in the valuation and recognizing process, like described in section 2.1 and higher audit fees could be due to the fact that the auditors want to scrutinize this discretion.
There is a notion that a better corporate governance could lead to less external auditing work and lower audit fees (Hay, Knechel, & Ling, 2008). So the implementation of SOX, which would strengthen the corporate governance would lead to lower audit fees eventually. However it turns out that a strengthened corporate governance actually is a complementary of external auditing (Hay et al., 2008). This means that the strengthened corporate governance will actually lead to higher audit fees because of a complementary relationship between internal controls and demand for external controls. So after the implementation of SOX and the increase in corporate governance legislation would the audit fees increase because of a higher demand for external controls.
Thus SOX had two important implications for the rise of the audit fees. Firstly the auditor legal liability, where auditors beare more risk for which want to be compensated. They have more work as well due to the legal liability because auditors want to be more cautious than before the implementation of SOX. However they do refuse more risky clients as well. Secondly the SOX 404 report, which resulted in way more work. But a rise in the audit fees could also be due to the fact that the auditor is scrutinizing managerial discretion or improving the quality of the financial statements. I will explain more about the variables, the appropriate proxies and to what extent I will use them in my regression in section 3.1.
2.3.3 The implications of SOX on intangible assets.
SOX did not have specific regulations for the intangible assets, besides the general effects like described above. This does not mean that the intangible assets are not affected at all, because they are. However the intangible assets are mostly riskier for the auditor after the implementation of SOX because of the complexity of the intangibles themselves. There was no new regulations regarding the valuation or auditing of the intangible assets specifically.
2.4 Hypothesis development.
The literature review described a few important topics which I will summarize in this section to develop my hypothesis. Firstly, intangible assets are a driver of profitability and in some industries even a competitive advantages (Kramer et al., 2011). However firms that have more intangible assets have higher analyst coverage (Barth, Kasznik, & McNichols, 2001). Which means that analysts think that they can elaborate an information breach because there are more intangible assets in the company.
What also emphasizes the complexity is a higher change on mistakes and litigation for the auditor by auditing intangible assets because they are that complex (Wyatt, 2005). The litigation risk and the analyst coverage is a sign that the intangible assets are very complex and risky for the auditors.
Second topic is that the risk component is very important in the audit fee model (Simunic, 1980). This means that auditors want to be compensated more for risk they endure during the audit. Other studies describe that auditors want to be compensated more if intangible assets are valued at fair value as well.
Lastly, auditors are more sensitive to risk since the implementation of SOX. It is easier to sue your auditor for mistakes in the financial statements and the PCAOB has opportunities to scrutinize the auditors. This means that the intangible assets, which were already risky, are even more risky for the auditors after the implementation of SOX. I think auditors want to be compensated for this risk. Therefore my hypothesis is:
H1: The implementation of SOX affected the pricing of audit fees for intangible assets upward more heavily than other assets because auditors became more sensitive for risk
Multiple regressions will be performed to investigate this. A main regression with just intangible assets included, a sub regression with intangible assets and R&D expense included and a sub regression with just R&D expense included with only companies who invest in R&D. The methodology and an explanation of how the regression will be performed will be discussed in the next section.
3. Methodology and data.
This section describes the methodology, the data and what the regression model will be. Subsequently the manipulations that are performed to get the variables fitting for the regression will be discussed. The first part will elaborate on the variables and, if necessary, the proxies used to operationalize them. Eventually the first part will show the actual regression model. The second part will describe the sample selection. Especially where the data was retrieved and which assumptions were used by retrieving and adapting the data and the general manipulations that were performed for the complete dataset. Section 3.3 discusses the descriptive statistics, correlation and the manipulations performed separately for every variable.
3.1 Development of the regression model.
As described in section 2.2, if you take intangible assets in account, these are the main components of the audit fee model: Size, complexity of the client, auditor risk and intangible assets specifically. So therefore these components will be included as variables in the regression of this thesis.
For the variable of intangible assets I do not need a proxy as it can be quantified. I do correct skewness by scaling intangible assets by dividing them with the total assets of a company. I do this to correct for the correlation of the intangible assets and total assets and to get a better view of the effect of SOX on the intangible assets relatively to total assets. The variable of size will be quantified by the total assets of a company. But I will use the logarithm of the total assets to eliminate the skewness created by the largest and smallest companies. Especially the skewness created by the large companies could have a severe effect on the regression because a few major companies concern a material percentage of the total assets. For the variables of complexity of the client and the auditor risk I do need proxies, which I will interpret and elaborate next. Auditor risk is included because, as described in 2.2, it is a major part of the audit fee model even before SOX. The increase in risk after the implementation of SOX will be captured by the SOX variable.
A similar, but more focused on the managerial implications of SOX, study like this thesis has used the variable of complexity and auditor risk as well (Lenard, Petruska, Alam, & Yu, 2012). The study used the number of business segments as a proxy for complexity and a loss occurred in the past three years as a proxy for auditor risk (Lenard et al., 2012). In my opinion these are good proxies, because a lot of business segments makes an audit very complex and a loss occurred in the past three years means there is pressure on the management of the firm which maybe wants to manipulate or fool the auditor, which makes it risky. I will include operational segments as well with the number of business segments because I think that this will increase the complexity of an audit too. I think that it is even more relevant to look at a loss at the current year because auditors are aware there is more risk involved and will perform more work on the client.
Another similar study uses the same variables as well (Huang, Raghunandan, & Rama, 2009). In the regression the study used the debt to assets ratio of the company that is audited as a proxy for the auditor risk (Huang et al., 2009). Next to the loss occurred in
the current year, I think that they both explain the auditor risk more completely for the
effect on the audit fees because a debt to asset ratio shows companies which has been under pressure or in distress for a while. This gives a good mix of companies who are suddenly in distress and in distress for a long time.
Like described in section 2.3, the main effects of SOX on the audit fees are the heightened legal liability and the compliance with the SOX 404 report. I will use a dummy variable in the regression so I do not need any control variables for the effects of SOX and the severe effects mentioned will be caught by the dummy variable. To measure the specific effects of the intangible assets on SOX, I will be using an interaction variable in which I will be multiplying the intangible assets with the dummy variable of SOX. This will show exactly to what extent SOX had an effect on the audit fees for auditing intangible assets. The audit fees will be used as the dependent variable. In section 2.2 the audit fee model of Simunic was described, just as the additions.
The regression will therefore look like this:
, ∗ ∗
∗ ∗ ∗
∗ ∗
3.2 Sample selection.
I will perform an archival research for my thesis. The WRDS database is used to retrieve the data, which has the audit fees from major companies from 1-1-2001 (fiscal 2000) onwards. The data that is used is from two years before the implementation of SOX (2001 and 2002) and two years after the implementation of SOX (2003 and 2004) to get a complete view of what the effect was of SOX.
Only companies with total assets of more than 100.000 dollars are included. This is because only companies who have their shares publicly traded have to comply with SOX. The minimum requirement in the US to publicly trade your shares is total assets of more than 100.000 dollars (US House of representatives, 2002). I will not exclude any other smaller companies, because the results are specifically applicable to smaller companies, as will be explained next.
Any other smaller companies will not be excluded since an incremental sensitivity of the intangible assets regarding an increase in the legal regime will not matter a lot to major companies. The increase in costs regarding intangible assets are not relevant for them in for example the innovation or a M&A process because it is for them not a big impact in the total audit fees. For smaller companies a small increase in the audit fees and therefore the cost of the innovation process is relatively way more costly. Because of an increase in the cost of the innovation or M&A process which is relatively high for smaller companies they could decide to innovate less or even stop with innovating and to stop acquiring particular companies or pay less for those companies. If they pay less it could also lead to less acquisitions as the owners of the company would still want to receive the same price as before SOX. So the implications of this study are specifically relevant to smaller companies and therefore they will be included.
The audit fees are retrieved from the Audit Analytics Database. The data of the firm size, loss in current year, debt to asset ratio, amount of intangible assets and the SOX variable are retrieved from the Compustat Fundamentals Annual database. I took the logarithm of the audit fees and total assets to reduce the skewness in the variables caused by the few major companies which concern a material percentage of the total assets and would therefore create skewed variables in the regression. I have scaled the intangible assets to reduce skewness and to get a better view of the effect of the implementation of SOX on intangible assets relatively to total assets.
The loss in current year is a dummy variable and has a value of ‘1’ if the net income was negative in that specific year and a value of ‘0’ if the net income is positive. For the debt to asset ratio I’ve retrieved the total debt and the total assets and divided them through each other. The SOX variable is a dummy variable and has a value of ‘1’ for every audit performed in 2003 and 2004 because SOX was active since 27th of January 2003. As I mentioned before, I assume that all the audits in 2003 are complying with SOX although SOX was not yet implemented in the first 26 days of 2003. The SOX dummy has a value of ‘0’ if the audit was performed in 2001 and 2002.
The Audit Analytics and Compustat Fundamentals Annual databases were merged with the Company fkey (Audit Analytics) and the Global Company Key (Compustat), which are the same. Together with the year of the observation they are a unique combination and the databases could be merged by this combination.
The number of segments were retrieved from the Compustat Historical Segments database. Only the operational segments and business segments were used, to eliminate double counting and to have the data keep making sense, as will be explained in section 3.3.1. The operational segments are seen as part of business segments for the purpose of this thesis because they increase the complexity of the audit. All the double segments were eliminated just as the segments where there were nog business activity performed anymore. The accuracy of the data was determined by comparing the total revenues of the business segments of a specific company with the total revenues from that specific company.
The total revenues of those specific companies were retrieved from the Compustat Fundamentals Annual database. Approximately 200 companies that did not match within 2 percent were eliminated. After this procedure the data was merged with the other data by matching them on the Global Company Key and the year of the specific observation.
I have started with 71.939 observations from Compustat Fundamentals Annual where the other two databases were merged into. After that I have eliminated several observations. First of all, only observations with values in every database were included, observations with a missing value in either of the databases were eliminated. Subsequently, every observation with no audit fees was eliminated because this is the dependent variable and it needs to have a value for the regression to make sense. Companies with total assets below 100.000 Dollar were eliminated as well since those companies do not have to comply with SOX. To eliminate finally double observations, observations with no business segments and observations with illogical data since these are probably all errors in the database and would only create noise in the regression. After these eliminations there were 15.465 observations left.
That almost 80% of the observations are dropped in the eliminations and that there is only 15.465 observations available for the regression is remarkable. This is due to the fact that of a lot of the companies the audit fees are not available or have assets below 100.000. After the merge of the three databases a lot of the companies had either missing values for business segments or missing audit fees and therefore a lot of the observations were dropped as well. Although there were 80% of the observations eliminated, the 15.465 observations left for the regression are sufficient and nice for a regression.
3.3 Descriptive statistics and manipulations of the variables.
The manipulations used to get every variable fitting for the regressions are described in this section. The descriptive statistics, correlation, autocorrelation and homogeneity to see if the variables are fitting for a regression are discussed in this section as well. First of all the manipulations and descriptive statistics of the main regressions are discussed in 3.3.1.
The descriptive statistics of the first sub regression with R&D added will be discussed in section 3.3.2. In 3.3.3 the descriptive statistics and the manipulations of the sub regression where R&D replaces intangible assets will be discussed.
The two sub regressions will be conducted since intangible assets include hardly any R&D expense, so the innovation part is not included in the main regression. This is due to the fact that R&D expense can only be recognized under US GAAP if it is software or website design related and can be used for different purposes as well (Deloitte, 2017). I think that these are only a small part of the R&D expense and I assume therefore that R&D expense cannot be recognized and are not included in the intangible assets and main regression. The R&D expense are included in the two sub regressions to show the impact of SOX on the innovation in the United States. I assume that the main part of the intangible assets is goodwill so the main regression will clarify mainly the effect of SOX on M&A strategies in the United States.
3.3.1 Main regression
These are the variables, like discussed in section 3.1, that will be used in the regression: Audit fees, Size, Number of segments, Loss in current year, Debt to asset ratio, Intangible assets, SOX and an Interaction variable between SOX and intangible assets. I will discuss the variables by the order of the manipulations in this section.
For an acceptable limit of skewness a value of two will be used (Trochim & Donnelly, 2006). If the skewness is higher, the data will be manipulated or there will be explained why the variable still can be included in the regression. In the descriptive statistics the number of observations, the mean, variance, skewness, kurtosis, the lowest value, the value of the percentile 1, 5, 50, 95, 99 and the highest value of every variable is shown. The first manipulation was to reduce the skewness of the variables of audit fees and total assets. The skewness is reduced by using the logarithm of the observations. This was necessary since the five percent biggest companies contain a material percentage of the total assets and therefore create skewness of the variables in the regression.
26
Descriptive statistics of the variables in the main regression. The p1, p5, p50, p95, p99 are the value of the first, fifth, fiftieth, ninetyfifth and nineteenineth percentile of the variable. The value of min and max are the minimal and maximal observation of the variable. The Log Total audit fees (DLR) is the logarithm of the total audit fees in dollars paid by a company in a specific observation. The Log Total assets (MLN) is the value of the amount of the total assets in millions owned by a company in a specific observation. The number of segments are the number of business segments that a company has in a specific observation. The loss in current year (dummy) has a value of one if the company has made a loss and otherwise a zero. The debt to asset ratio is a ratio of the total debt divided by the total assets of a company. The Total intangible assets (MLN), scaled are the total intangible assets of a company divided by the total assets of that company. The SOX dummy has a value of one is the Sarbanes-Oxley act was applicable during the audit of the observation and zero if it was not yet applicable. The interaction variable is an interaction between the SOX dummy and the Total intangible assets. For the manipulations and sample selection I refer to section 3.2 and 3.3.
For example with the total assets, the biggest five percent of the companies have assets between 14 billion and 117 billion while the mean is 6,7 billion dollars. This created a skewness before the manipulations of approximately 13 (audit fees) and 15 (total assets). This skewness means that the variables are not normally distributed and the mean and the data in general is heavily affected by the few major companies. This results in a wrong view of the variable as a whole. As can be seen in the descriptive statistics in Table 1, the distribution of the audit fees and total assets variables are very neatly after using the logarithm of the observations. The skewness is reduced to 0,18 (audit fees) and -0,06 (total assets) and the variables are normally distributed. Therefore they are fit to include in the regression. Besides the general manipulations discussed in section 3.2, there were no other manipulations performed on the variables of audit fees and total assets.
The intangible assets variable are very skewed as well. This is due to the same problem as with the total assets and audit fees. I resolve this by scaling the intangible assets by dividing it by the total assets. This way the intangible assets variable makes more sense due to the fact that it is completely corrected for size by the total assets. The impact of SOX is now shown better as it is relatively shown by the percentage of intangible assets that a company has of their total assets and the audit fees charged for that.
The number of segments is the proxy for the complexity of the audit variable and the descriptive statistics after the manipulations is shown in Table 1. Before I could manipulate the data, I had to do a few sanity checks and drop illogical data. All the companies with zero segments, segments with no revenues, segments without any assets and segments without a name, SIC code or type were dropped, because these are observations which do not have any business activity. I assume that if there is no business activity in a segment or the segment does not even has a type that it will not add any complexity to the audit. To avoid double counting of the same segments only operational and business segments were used. The geological and state segments are not significantly increasing the accuracy, but do often increase double counting of the same segments so are not included to keep the data more accurate. To avoid double counting all the corporate segments, discontinued segments, segments with the name ‘other’ and duplicates were dropped as well.
This is due to the fact that a corporate headquarters does not make an audit more complicated but just consolidates the other segments. Finally the total revenues of all the segments combined were compared with the total revenues of the company as a whole. The few companies, approximately 200, with a difference of more than two percent were dropped.
After these manipulations the data was merged with the Compustat Fundamental Annuals data (master data) because the segment data was retrieved from the Compustat Historical Segments database. The observations were merged with the Global Company Key and year to get unique observations. The companies that were not included in both of the datasets, approximately 1300, were dropped. All of these manipulations increased the accuracy and decreased the double counting, but also increased the skewness and kurtosis of the data. This is not a problem since almost all the companies have between one and five segments from which almost the half of them only has one segment and there are no observations below one as can be seen in Table 1. This is not surprising since a normal company always has at least one segment (the company itself) but does not often have way more segments since there are a lot of small companies included as well. A high skewness and kurtosis was expected and are not surprising under these circumstances. This will not affect the audit, audit fees and therefore the regression though.
The loss in current year is the proxy for the audit risk variable and the descriptive statistics after the manipulations are shown in Table 1. This is a dummy variable and is a ‘1’ if the net income in a specific year was negative. Almost half of the companies (46%) in the sample had a loss in the current year. This is relatively high, but not extreme, because in the US 30-40% of the companies on the stock market report a loss every year (Ro, 2014). The debt to asset ratio is a proxy for the audit risk variable as well. The descriptive statistics after the manipulations are also shown in Table 1. The debt to asset ratio in general shows a different part of the companies who are in distress because they are highly leveraged but have not made a loss this year. This ratio is calculated by dividing the total debt, long term plus short term, by the total assets of a company. The skewness and kurtosis of the manipulated data (3,2 and 17,12) is remarkably high, even after winsorizing the top 1%.
29
Table 2, Correlation between the independent variables of the main regression
Correlation between the independent variables of the main regression. The Total assets (MLN) is the value of the amount of the total assets in millions owned by a company in a specific observation. The number of segments are the number of business segments that a company has in a specific. The loss in current year (dummy) has a value of one if the company has made a loss and otherwise a zero. The debt to asset ratio is a ratio of the total debt divided by the total assets of a company. The Total intangible assets, scaled; are the total intangible assets of a company divided by the total assets of that company in a specific year. The SOX dummy has a value of one is the Sarbanes-Oxley act was applicable during the audit of the observation and zero if it was not yet applicable. The interaction variable is an interaction between the SOX dummy and the Total intangible assets.
This high skewness is due to the fact that this is a ratio and most of the companies are not in distress (95% of the companies has more assets than debt) and that there are no observations below zero. But only a few companies are that highly leveraged that they have more debt than assets and a very few have a multitude of debt against their assets. This is typical in an ordinary market and therefore a high skewness in a debt to asset ratio is normal.
The SOX variable is a dummy variable and the descriptive statistics after the manipulations are shown in Table 1 as well. This dummy variable has a value of ‘1’ if SOX was applicable (2003 and 2004) and ‘0’ if SOX was not applicable yet (2001 and 2002). Like discussed in section 2, SOX was applicable after one month in 2003 and I assume that every financial statement that was signed off in 2003 complies therefore with SOX.
The interaction variable is a combination of the SOX variable and the Intangible assets variable and the descriptive statistics after manipulations are the last shown in Table 1. This variable effectively gives the scaled intangible assets of a company at the moment that SOX is applicable. The skewness and kurtosis follow approximately the same pattern as with the intangible assets themselves but are exaggerated by the SOX dummy because 46% has a value of zero.
Table 2 describes the correlation between the independent variables of the main regression. I assume that all the values between -0,5 and 0,5 are acceptable, and above 0,75 and lower then -0,75 are problematic like described in Huber & Stephens (1993). Every value between -0,75 and -0,5 and between 0,5 and 0,75 will be explained to what extent this impacts the regression. Huber & Stephens (1993) describe that a tolerance (1 - correlation) is acceptable as long as it is higher than 0,25 or lower than -0,25. There is only one pair of variables that have correlation with each other in this main regression that should be explained and there are no correlations that are unacceptable.
There is correlation between the intangible assets and the interaction variable which was expected since it consists partly of the intangible assets. This correlation was counted for as it is an interaction variable and therefore this will not impact the regression.
There has not been a check for autocorrelation since there are no variables in this regression that could auto correlate with each other. There has been a check for homogeneity and the regression has homogeneity and is therefore fitting for the regression. The next section will describe the manipulations and descriptive statistics for the first sub regression where the R&D expense is added.
3.3.2 Sub regression with R&D added
This section will discuss the descriptive statistics and the manipulations performed for the first sub regression. For this sub regression the R&D expense of every company is added to the main regression. Another interaction variable (interaction 2 variable) will be created to give a good assessment of the impact of SOX on the pricing of the audit fees charged for auditing companies with high R&D expense. The first sub regression will therefore look like this:
log , ∗ ∗
∗ ∗ ∗
∗
∗ ∗
∗ ∗
This regression now shows the impact of SOX on the innovation part of a company (R&D expense) and the M&A part of a company (intangible assets) since the intangibles mostly include goodwill. I will not discuss the manipulations performed on the variables that are included in the main regression since that has been discussed in section 3.3.1 and there are only non-significant differences with the descriptive statistics of this regression. For the descriptive statistics is refer to table 1.
The R&D variable is one of the variables that is added in this sub regressions. The variable is scaled by dividing it by the total assets. Therefore the results of this sub regression give a similar view of the relative impact of SOX on the audit fees charged. This means that it gives a better view on the percentage that is charged in comparison to the size of the company and therefore it shows the impact of SOX on the audit fees for companies more clearly.
It is remarkable that almost half of the companies, after the general manipulations described in 3.3, do not invest in R&D (43%). Which is good when there is an increase in the audit fees after SOX, the effect will be more clear because half of the companies do not suffer from an increase of the audit fees. If there is an incremental sensitivity for the R&D expense, this results of this regression will show this. The interaction 2 variable is a interaction of the SOX dummy and the R&D variable. Just like the regular interaction variable it follows the same pattern as the underlying variable (R&D in this case) which is exaggerated by the SOX dummy.
Like discussed in section 3.3.1 the correlation is acceptable if the value is between 0,5 and -0,5 and should be explained if it is between, 0,5 and 0,75, and, -0,5 and -0,75 (Huber & Stephens, 1993). In this sub regression there is still correlation that should be explained between the first interaction variable and the intangible assets. But there is another correlation that should be explained between the independent variables which is similar. The interaction 2 variable correlates with the R&D variable with a value of 0,65. This is expected and will not impact the regression. There are no unacceptable levels of correlation in this sub regression.
There has not been a check for autocorrelation since there are no variables in this regression that could auto correlate with each other. There has been a check for homogeneity and the regression has homogeneity and is therefore fitting for the regression. The next section will describe the descriptive statistics of the second sub regression where the R&D expense replaces the intangible assets and there are only companies included who have R&D expense.
3.3.3 Sub regression with R&D instead of intangible assets.
For this sub regression the intangible assets are replaced with the R&D expense of the companies if you compare it to the main regression. The interaction variable is now the interaction between the R&D expense and the SOX dummy instead of the intangible assets and the SOX dummy. This sub regression is performed because it shows the direct impact of SOX on innovation excluding goodwill, which is the other main component of the intangible assets under a normal regime.
