Audit firm tenure and audit quality:
Evidence from the US in a post-SOX Era
Master Thesis
Martijn Ooijevaar 5800366
MSc : Accountancy & Control Thesis Supervisor: Dr. J.J.F. van Raak Submission date: 12-10-2014
Abstract
“Above all, emphasis should be placed on identifying and reacting appropriately to risk, and on establishing counterweights to circumstances that could detract from the ultimate goal of obtaining a high level of assurance that the financial statements are free of material misstatement.”- James Doty, PCAOB Chairman (December 6th, 2012)
The starting point of my thesis is the Sarbanes-Oxley Act as a regulatory reaction to the financial scandals at the beginning of the 21th century and the associated debate concerning auditor independence and auditor rotation. As the results of these regulations become increasingly available, it is important that research is undertaken in order to provide evidence on the desired effects of these regulations have been achieved. Furthermore the driving arguments behind these regulations can be tested for validity.
In this study, I test if the length of the audit firm’s tenure has an effect on the audit quality. I hypothesize that companies that have a firm auditing them with a medium tenure, are associated with higher audit quality, as short tenure causes knowledge issues (Knapp, 1991) and long tenure has familiarity issues (Myers et al, 2003). This study aims to find evidence whether or not audit tenure is still an issue after implementation of the Sarbanes-Oxley Act. Using the abnormal working capital accruals model by Defond and Park (2001), and the current accruals model as used by Myers et al. (2003), different cut off points are used for short, medium and long tenure in order to search for differences in accrual quality.
The main results indicate that short or long audit firm tenure have no significant influence on audit quality, when compared to medium tenure. However, after adjusting the cut-off date for long tenure to more than 9 years, I find that long audit firm tenure has a negative impact on abnormal working accruals, contrasting the initial hypothesis.
Table of contents
1. Introduction ……….……….………. 4
2. Literature review ……….……….…………. 7
2.1 The need for auditing and auditor independence ...…………. 7
2.2 Audit Quality …………..…………....…………..……… 8
2.2.1 Measures of Audit Quality ….…………..……….. 9
2.2.2 Audit Rotation and Audit Quality ……….……… 9
3. Hypotheses Development ………..……….…...……..………… 11
4. Methodology ……….……… 12
4.1 Sample collection ……….………. 12
4.2 Research method ………..………. 13
4.3 Control variables …..……….……… 15
5. Emperical results….. ……….…….………….. 17
1. Introduction
In the United States, the need for auditor rotation has been debated for almost half a century (Myers et al, 2003). The most commonly heard argument for the implementation of such regulations is that long tenure may lead to complacency on the part of the auditor. This complacency may result in a lessened tendency to issue qualified statements (Vanstraelen, 2010) or independence issues due to economic desire to retain the client (Geitzman and Sen, 2002), leading to lower quality of financial reporting. After the accounting scandals of Enron, Worldcom and others around the years 2001-2003, the debate around auditor tenure was refuelled.
In 2002, amidst the scandals, the U.S. Congress passed the Sarbanes-Oxley Act (SOX) to improve the quality of financial reporting and to restore investor confidence in the reliability of financial statements. An important aspect of SOX are its internal control reporting requirements, which allow investors to be informed about the quality of a firm’s internal controls. Specifically, Section 302 of the Act requires management to evaluate and report the effectiveness of disclosure and control procedures (SEC, 2002), while Section 404 requires the auditor to form an opinion on the effectiveness of the internal controls.
Furthermore, SOX mandated audit partner rotation at least every 5 years. Before the implementation of SOX this was set at 7 years. The key motivation behind this measure was the battling of the aforementioned complacency; improving the quality of the audit with a fresh set of eyes (Hatfield et al, 2007).
SOX also contained a provision that the U.S. Comptroller General initiated a study investigating the possible effects of a mandatory audit firm rotation system. The conclusion at the time was that such a mandatory system may not be the best way to improve auditor independence; it could address issues around auditor independence in appearance, but heavy costs would be involved (GAO, 2003). However, the general accounting office did suggest that if SOX did not improve audit quality, the system might be necessary (Carcello & Nagi, 2004).
accounting firms and the listed companies, writing over 700 comment letters (PCOAB, 2011). After the closing of the comment period, the chairman of the PCAOB stated the continued intention of the PCAOB to examine restrictions on firm tenure in order to safeguard auditor independence. Contrary to this statement, PCAOB board member Jay Hanson stated on a personal note that he saw “hurdles” when it comes to implementing a limit, clearly arguing that research proving a clear correlation between audit quality and audit firm tenure has not yet been done (Accountingweb, 2012).
Audit quality is also a point of debate in the European Union. Before the financial crisis, the 8th company law directive was already in place to regulate the audit profession, requiring audit partner rotation every 7 years. In October 2010, the European Commission issued the Green Paper “Audit Policy: Lessons from the Crisis”. This Paper started a debate on the role of the auditor and subsequent regulation to be implemented. This culminated in proposed regulation by European Commissioner Barnier, which aims to introduce mandatory audit firm rotation every 6 years, or 9 in the case of a joint audit (EU, 2011). This proposal has also met great resistance from the professional community. On May 27th 2014, the European Union has published a new guideline regarding audit rotation. This guideline states that audit partners are still required to rotate every 7 years. Furthermore, A maximum audit firm tenure has been set at 20 years, with a mandatory public tender after the first 10 years (eur-lex.europa.eu, 2014).
In light of these developments in the area of auditor rotation, and the accompanying debate surrounding these developments, the research question which this thesis aims to answer is as follows: “To what extent could the possible implementation of audit firm rotation improve audit quality for US listed companies in the post SOX-era?”
The data sample collected to examine the effect of audit firm tenure on audit quality consists of US listed companies for the fiscal years 2011 and 2012. This sample provides a representative group of companies which will be directly affected by the implementation of a mandatory audit firm rotation under the SOX regulation. In order to examine the effect of a possible mandatory rotation, the effects of the length of the auditor-client relationship on audit quality is examined. Audit quality will be measured in terms of two proxies. These proxies will be accrual-based. Accruals are thought to be under managements discretion, and thus indicative of earnings quality and audit quality. The first proxy is the abnormal working
capital accruals model, introduced by Defond and Park (2001), the second proxy is the current accruals models as used by Myers et al (2003).
This study contributes to the literature on audit quality and audit firm rotation in the U.S. By showing a positive relationship between long audit firm tenure and audit quality, this study provides an argument to the side of the debate which deems further regulation unnecessary. On the side of short tenure however, no conclusive evidence has been found.
The remainder of the paper is organized as follows: In the next chapter, literature on audit firm rotation and -tenure, and audit quality will be discussed. In chapter three the hypothesis will be formulated. Chapter four will see a depiction of the sample selection and research method. Chapter five will contain the descriptive statistics of the study, chapter six will contain the conclusion and discussion.
2. Literature review
In this chapter, the relevant issues surrounding the research question will be discussed. First, the need for audit quality and auditor independence will be explained. In the second section, literature surrounding audit quality research is discussed. In the third section, a summary of the current state of the academic debate surrounding these issues will be given. Following these addresses, the hypotheses are developed in chapter three.
2.1 The need for audit quality & auditor independence
DeAngelo (1981) provides one of the most commonly used definitions of audit quality, derived from the Watts and Zimmerman (1981) model: “audit quality is the joint probability that an auditor will discover a breach in the client’s accounting system, and subsequently report that breach”.
Watts and Zimmerman (1981) use the market theory in their article to stress the importance of auditor independence. Using a hypothetical situation where there are no auditing regulations, they use market theory to prove the need for the auditing profession as well as independence.
In this situation, the manager of a company will only engage in a contract with an auditor if he feels that the increase in market value of the company exceeds the costs of contracting. Below is shown how the market perceives this added value.
Figure 1
Formula for the need for auditing
(a) (b) (c) (d)
The equation shows that market value of the company can increase, if the auditor reports breaches made by the manager to the market, if they exist (a).
This equation is broken down into the probability of three components. First off, the chance that the manager breaches his contract has to be higher than zero; if no breach exists, an auditor will add no value, only the cost of the audit(b).
If a breach exists, the discovery of this breach is dependent on the expertise and the amount of time or effort the auditor puts into the audit (c). Finally, if discovered, the final step is to report the breach to the market. This last component can be viewed as the independence of the auditor (d). If there are no incentives lowering the independence, the auditor will always report a discovered breach.
The product of (c) and (d) can never be 1. As ultimate expertise would require such an investment, that the benefits of the audit will be outweighed by the costs. Absolute independence is also unachievable, as the fact that the manager pays the auditor will always constitute a form of financial dependency. Therefore, using transaction cost theory in a competitive market, the auditor will try to be as competent and independent as possible, against acceptable cost.
Weber et al. (2008) go on to discuss the incentives audit firms have to provide high audit quality. These incentives are the insurance and the reputation rationale. The insurance rationale entails the idea that wealthier audit firm have a stronger incentive to provide high quality audits, in order to prevent litigation. This is because they are thought to be more likely to be sued, as they have ‘deeper pockets’ than small audit firms. The reputation rationale argues that firms with a good reputation are more likely to provide high quality audits. In the event of a low quality audit occurring for these firms, the publicity may result in loss in fee premiums and possible client loss. The Big 4 (EY, PWC, KPMG Deloitte) are thought to be influenced by these two rationales, as is underlined by the research of Becker et al. (1998). Becker et al. find that financial statement audited by The Big 4 have less abnormal accruals than those audited by non-Big 4 firms. Abnormal accruals are thought to be indicative of audit quality, as is shown in the next section.
2.2 Literature on audit quality
DeAngelo (1981) goes on to state that the evaluation of audit quality involves significant costs, and therefore proxies for audit quality have to be developed. These proxies have to be based on observable variables which are correlated with audit quality. Francis (2004, 2011) provides an overview of the audit quality literature and provides a clear positive association between earnings quality and audit quality, constituting earnings quality measures as measures of audit quality. In this segment the state of research around these proxies on audit quality is discussed.
Lys & Watts (1994) investigate audit quality through lawsuits filed against auditors, finding that higher audit fees are associated with a greater likelihood of being sued, as auditor independence is undermined by greater audit fees. Also client size is an indicator of the likelihood of being sued, which is in line with deep pockets theory.
Lennox (1999) uses the information in auditor reports in cases of financial distress. In his research he finds that bigger auditors more effectively report on financial distress, indicating a higher level of audit quality. These results are in line with the theory of Weber et al (2008) mentioned in the previous section.
Accruals as a measure of earnings quality are often used in the literature (Johnson et al., 2002; Myers et al. 2003, Carey and Simnett, 2006). As management can influence financial statements through the use of accruals, the ability of the auditor to constrain this phenomenon is thought to be a good measure of audit quality.
Johnson et al. (2002) investigate the quality of financial reporting by using the absolute value of discretionary accruals. Also Johnson et al. divide audit firm tenure into groups in order to make a comparison between groups. They find that these accruals are more prevalent as tenure is short (between 2 and 3 years) than in the medium group (between 4 and 8 years). For longer tenure, no significant results were found. This leads to the conclusion that audit quality is lower for the first year of the auditor-client relationship.
2.2.2 Audit rotation and audit quality
Audit quality In order to thoroughly define audit rotation, I need to first distinguish between the two types of audit rotation. First off, there is audit partner rotation. Audit partner rotation
entails that if a firm has a client for a long time, that the partner responsible for the engagement changes. As mentioned in the introduction, SOX Audit firm rotation entails the switching of the entire firm after a certain amount of time (Kramer et al, 2011). Kramer finds an indication that mandatory firm rotation may have a positive impact on the amount of earnings conservatism, as it is found that earnings conservatism increases after a firm has rotated.
Dopuch et al. (2001) found that in a situation of mandatory audit firm rotation, auditors are likely to produce less biased reports. Davis et al. (2003) find that discretionary accruals increase with auditor tenure and conclude that management is allowed more reporting flexibility by the auditor when tenure is high.
Ghosh and Moon (2005) find using the earnings response coefficient, as a proxy for audit quality, that investors’ perception of audit quality increases as audit tenure is higher. Also, one-year ahead forecasts are more associated with earnings as audit tenure is higher. This leads them to conclude that both information intermediaries and investors perceive audit quality as higher as tenure grows longer.
Myers et al. (2003) use absolute abnormal and current accruals to investigate the need for mandatory audit rotation. They find that long audit firm tenure is associated with higher audit quality, providing an argument against mandatory audit rotation.
Johnson et al. (2002) investigate the quality of financial reporting by splitting audit tenure up in three groups of length. The lower length group (2 to 3 years) is associated with lower quality financial reports, whereas the group with the longest tenure (9 years or more) does not show a significant decrease in reporting quality. Johnson contributes this lower quality to the learning effect of the auditor; the auditor is less likely to detect errors in new clients and may also be subject to management pressure in the first years of the engagement.
On the whole, previous literature seems to suggest that a long auditor-client relationship has a positive influence audit quality, contrary to the legislative intentions of the PCAOB as mentioned in the introduction.
3. Hypothesis development
In order to identify whether there is an association between audit quality and audit tenure, I use a measure of audit quality. Following Carey & Simnet (2006), I use the abnormal working accruals model to investigate a possible relationship between audit tenure and audit quality. This model is used to test periods of short, medium and long tenure. As shown in the literature review, audit quality is thought to be influenced by constraining factors as tenure is short, and possible factors as tenure is long. The medium group will therefore be compared with the long tenure group in order to test for deterioration of the audit quality over time. The medium group will also be compared to the short group, in order to test for improvement of audit quality in the first years of the audit. The idea in these comparisons is that audit quality will be higher for the medium group.
This leads to the formulation of the following two hypotheses:
H1: Short audit firm tenure is associated with higher levels of abnormal working accruals than medium tenure.
H2: Long audit firm tenure is associated with higher levels of abnormal working accruals than medium tenure.
If H1 is supported as a hypothesis, this means that there is an indication of lower audit quality during the first years of the audit firm-client relationship, implying that there are heavy costs associated with the establishment of a mandatory audit firm rotation. The increase in audit firm switches will increase the amount of audit firm-client relationships which are in their first three years, leading to a greater loss of client specific knowledge on the whole. Costs will have to be incurred to compensate this loss in audit quality.
Alternatively, if H2 is supported, this means that the audit quality will decrease over time, and as such, that long audit firm-client relationships are economically undesirable due to the decrease in audit quality over time, providing an argument for the implementation of mandatory audit firm rotation.
4. Methodology
In this chapter, the sample collection is discussed as well as the research method used to investigate the research question. The first section contains the sample selection description. The second section contains the chosen research method, wherein the abnormal current working accruals model (AWCA) and current accrual model (CA) are explained, followed by the selected control variables.
4.1 Data collection
The data collection for this research involves the review and analysis of financial information published by companies listed at the U.S. stock exchange, derived from the COMPUSTAT North America fundamentals annual Database. I selected the years from 1995 up until 2012 in order to determine the audit firm tenure for 2011 and 2012. This is done to ensure that companies belonging to the LONG group are rightfully classified as such. Going back to 1995 provides room for further analysis. Companies with no data before 2003 are excluded from the sample due to failure to determine with certainty the length of the audit firm tenure.
As mentioned earlier, the data is collected for the years 2011-2012 from COMPUSTAT. I selected these years as these are the most recent years on which full
Table 1
DATA Items COMPUSTAT
ACT (Current Assets –Total)
AT (Assets – Total)
CHE (Cash and Short-Term Investments) CSTK (Common/Ordinary Stock
DD1 (Long-Term debt due in One Year) DLC (Debt in Current Liabilities)
LCT (Current Liabilities – Total) LT (Liabilities – Total)
DVPD (Cash Dividends Paid)
NI (Net Income
SALE (Sales/Turnover
information is available. This is important to mitigate any effects surrounding the implementation of the Sarbanes-Oxley legislation. The data items as shown on the previous page in table 1 have been used.
4.1.1 Data exclusion
An initial 16,422 firm year observations were retrieved from the database. From these observations, 2,311 were from the years 2011 and 2012, the remainder was only used for audit tenure and AWCA/CA calculation. After deleting observations with missing values for all variables, The initial dataset comprised of 895 firm year observations. Using the company SIC codes, firms in the financial services industries have been excluded from the sample (SIC Codes 6000-6299). This exclusion is based on the fact that different regulation applies to this industry, making their total asset base and derived financial structure incomparable to the remaining sample (Carey and Simnett, 2006). This removed 70 observations and brought the number down to 824. Lastly, for the variables suffering from outliers (AWCA, ROA, LEV, GROWTH), I removed the top and bottom 0,5 of the observations, ending up with a total of 792 firm year observations.
4.2 Research method
In this section the AWCA model, as used by Defond and Park (2001), and the CA model as used by Myers et al. (2003) are set out. By applying two different accrual models, I aim to better reflect on the relationship between audit quality and audit firm tenure.
4.2.1 AWCA Model
Abnormal working capital accruals are defined by Defond and Park (2001) as the difference between realized working capital in the current year subtracted by an expected working capital level, based on the level of working capital in the previous year, moderated by the level of sales. The model is therefore formulated as follows:
AWCAt= WCt– [(WCt-1/St-1) * St]
Where:
WCt = non-cash working capital accruals from current fiscal year computed as
(current assets – cash and cash equivalents) – (current liabilities – short term debt)
WCt-1 = non-cash working capital accruals from prior fiscal year
St = sales from current fiscal year
St-1 = sales from previous fiscal year
4.2.2 CA Model
The current accruals model, as used by Myers et al (2003), is defined as the change in current liabilities minus the change in short term notes and current portion of long term debt, deducted from the change in current assets minus the change in cash. This second model is therefore formulated as follows:
Current Accruals = ((ΔCA - ΔCash) – (ΔCL - ΔSTD))
Where:
ΔCA = change in current assets
ΔCash = change in cash and cash equivalents ΔCL = change in current liabilities
4.3 Control variables
Following Johnson et al. (2002), the audit firm tenure is translated into three samples: short tenure (2-3 years), medium tenure (4-8 years) and long tenure ( >8 years). This is translated into the regression model by adding two dummy variables (SHORT and LONG). Medium tenure is not included, as the comparison of short and long tenure with medium are sufficient. The regression model to examine the effect of audit firm tenure on audit quality is therefore formulated as follows:
AWCAt/CAt= β0 + β1SHORTt + β2LONGt + β3BIG4t + β4ROAt + β5LEVt+
β6GROWTHt + β7SIZEt + β8MARKET_CAPt+ β9RETURNt
Where:
AWCAt = abnormal working capital accruals from current fiscal year scaled by total
assets
CAt = current accruals from current fiscal year scaled by total assets
SHORTt = 1 if the audit-client relationship is two or three years at the
current fiscal year, and 0 otherwise
LONGt = 1 if the audit-client relationship is more than or equal to nine years at the
current fiscal year, and 0 otherwise
BIG4t = 1 if the audit firm from current fiscal year is one of the Big 4 audit firms
(Deloitte, Ernst & Young, KPMG or Pricewaterhousecoopers), and 0
otherwise
ROAt = the return on assets (net income divided by total assets) from current fiscal
year
GROWTHt = the change in sales during the current fiscal year divided by the sales from
prior fiscal year
SIZEt = the natural logarithm of total assets of current fiscal year
MARKET_CAPt = the natural logarithm of the market capitalization at the end of
current fiscal year
RETURNt = the stock return during the current fiscal year (cash dividend paid from
current fiscal year plus the change in market capitalization during the current fiscal year divided by the market capitalization from prior fiscal year)
Johnson et al (2002) mention other characteristics which could influence audit quality, beside audit firm tenure. Firstly, it is broadly investigated whether Big 4 Firms (BIG4) have an incentive to provide higher audit quality due to reputational issues, as discussed in the literature review (Becker, 1998; Lennox,1999) . Johnson et al. also see the return on assets as being of influence on audit quality. As companies are faced with diminishing or negative returns on assets (ROA), earnings management may occur. Johnson et al also argue that larger companies have more sophisticated financial reporting systems, which can be of influence on audit quality. Firms with high leverage (LEV) are prone to the use of earnings management, as they are more likely to be in (near) violation of debt covenants (DeFond and Jiambalvo, 1994). Alternatively, a firm’s growth (GROWTH) is associated with accruals, thus implicitly having an effect on abnormal working accruals (Carey and Simnett, 2006).
5. Emperical results
In this chapter the descriptive statistics and results are presented.
5.1 Univariate analysis
The sample consists of 792 firm-year observations for the years 2011 and 2012. The initial sample size has been reduced due to data availability constraints and as a result of outliers. Firms with abnormal levels of leverage and ROA have been excluded from the dataset. Also, the dependent variable has been winsorised at the 1% percent level due to clear outliers.
Table 1 contains the number of observations, mean, standard deviation, minimum value and maximum value per variable. The number of observations as equal amongst the different variables, as only firm observations with full data availability have been selected.
Table 1 Descriptive statistics
Variable Observations Mean Std. Dev. Min Max
AWCA* 792 -0.006 0.107 -0.976 0.640 CA* 792 -0.002 0.108 -1.125 0.459 DUMMY_SHORT 792 0.076 0.265 0.000 1.000 DUMMY_LONG 792 0.697 0.460 0.000 1.000 BIG_4 792 0.672 0.470 0.000 1.000 ROA 792 -0.002 0.379 -4.902 4.009 LEV 792 0.533 0.484 0.000 7.990 GROWTH 792 0.152 1.480 -1.000 37.605 SIZE 792 2.680 1.035 -0.804 5.308 MARKET_CAP 792 2.597 1.090 -1.281 5.365 RETURN 792 0.072 0.498 -0.951 3.430 *= Dependent variable
It is shown in Table 1 that 7.6 percent of the sample has an audit firm tenure of 2 to 3 years, and 69.7 percent of the sample has an audit firm tenure of 9 years or longer. By way of exclusion it can therefore be determined that the remaining 22.7 percent has a medium audit firm tenure of between 4 – 8 years. An explanation for this distribution is the relative narrow window for short tenure (2 years), where medium tenure (4 years) and long tenure (possibly indefinite) have greater windows. Also the average tenure within the group is set at 12.2 years, consistent with the idea that a switch of audit firms is a costly matter. The choice for a two-year period of investigation also becomes clear in this table, as the amount of SHORT observations is limited to 60 ( 792 * 0.076). Only examining the year 2012 would have left a SHORT sample group of under 30 . Also, the variable BIG_4 shows that the greater part of the market, 67.2% of the sample, is audited by a Big4 auditor.
5.2 Pearson correlation matrix
Table 2
Pearson correlation matrix
AWCA
DUMMY_~ T
DUMMY_
~G BIG_4 ROA LEV GROWTH SIZE
MARKET ~P RETURN AWCA 1.000 DUMMY_SHOR T 0.009 1.000 0.805 DUMMY_LONG -0.031 -0.434 1.000 0.386 0.000* BIG_4 0.033 -0.135 0.346 1.000 0.350 0.000* 0.000* ROA 0.266 -0.004 0.024 0.145 1.000 0.000* 0.911 0.500 0.000* LEV -0.296 0.009 0.035 -0.040 -0.489 1.000 0.000* 0.802 0.330 0.262 0.000* GROWTH 0.046 0.096 -0.096 -0.067 0.018 -0.080 1.000 0.194 0.007* 0.007* 0.059 0.609 0.025* SIZE 0.068 -0.175 0.298 0.660 0.254 -0.046 -0.023 1.000 0.056 0.000* 0.000* 0.000* 0.000* 0.192 0.518 MARKET_CAP 0.085 -0.168 0.309 0.636 0.272 -0.116 -0.010 0.916 1.000 0.017* 0.000* 0.000* 0.000* 0.000* 0.001* 0.778 0.000* RETURN 0.022 0.064 0.016 -0.020 0.094 -0.069 0.081 0.009 0.110 1.000 0.535 0.074 0.661 0.566 0.008* 0.054 0.023* 0.796 0.002* * = significant at 5 percent
based on the size of the company. This could mean that there are problems with multicollinearity. This also explains the correlation with BIG_4, as larger companies are more likely to be audited by a Big 4 auditor. The correlation between the two size-based variables and ROA can be explained by the fact that the sample has been based on two consecutive years, in which recurring companies for the second year that have grown, will most likely have seen a ROA that is higher than other companies in the sample. However, due to sample size issues surrounding the group of short tenure, the choice for a two year sample remains. A Pearson correlation matrix for the CA model offered practically the same outcome on significance level.
In order to test if the magnitude of multicollinearity is an issue in the AWCA and CA models, I calculate the Variance Inflation Factor (VIF) for the independent variables. The results are tabulated as follows:
Table 3
Variance Inflation Factors
Variable VIF 1/VIF
SIZE 7,13 0,140 MARKET_CAP 6,95 0,144 BIG_4 1,89 0,528 ROA 1,43 0,701 DUMMY_LONG 1,4 0,714 LEV 1,37 0,729 DUMMY_SHORT 1,26 0,796 RETURN 1,09 0,917 GROWTH 1,03 0,972 Mean VIF 2,62
A common rule of thumb is to consider a VIF for any single variable higher than 5 there is a high level of multicollinearity present in the model. However, Kutner (2004) suggests 10 as a threshold for high multicollinearity. I choose to follow Kutner (2004), and exclude no variables from the models.
5.3.1 Multivariate Analysis (AWCA)
In this section, the results for the AWCA regression model are presented. In reminder, the regression was formulated as follows:
AWCAt= β0 + β1SHORTt + β2LONGt + β3BIG4t + β4ROAt + β5LEVt + β6GROWTHt+
β7SIZEt + β8MARKET_CAPt+ β9RETURNt
Table 4
Multivariate OLS regression of AWCA
Coef. Std. Err. t P>|t| [95% Conf. Inter.]
DUMMY_SHORT 0,000 0,015 0,02 0,984 -0,030 0,030 DUMMY_LONG -0,008 0,009 -0,81 0,417 -0,026 0,011 BIG_4 -0,002 0,011 -0,19 0,851 -0,023 0,019 ROA 0,043 0,011 3,80 0,000* 0,021 0,065 LEV -0,047 0,009 -5,46 0,000* -0,065 -0,030 GROWTH 0,002 0,002 0,70 0,485 -0,003 0,007 SIZE 0,002 0,009 0,18 0,855 -0,017 0,020 MARKET_CAP 0,002 0,009 0,23 0,819 -0,015 0,019 RETURN -0,002 0,008 -0,32 0,752 -0,017 0,012 _cons 0,017 0,012 1,36 0,174 -0,007 0,041 Number of observations 792 F-statistic 10,633 Prob F 0,000 Adjusted R-squared 0,109 * significant at 5 percent
In Table 4 the results of the multivariate OLS regression of AWCA are tabulated. The F-statistic is significant at the 1 percent level, amounting to 10.63, allowing for the acceptance of the model as a whole having a correlation with AWCA. R-squared is 10.9 percent, which is comparable to the Johnson et al (2002) result. The coefficient for DUMMY_SHORT is set at 0.00, with a p-value of 0.984. These results offer no support for the first hypothesis. The coefficient for DUMMY_LONG is -0.01, with a p-value of 0.417. However slightly stronger
5.3.2 Multivariate Analysis (CA)
In this section, the results for the CA regression model are presented. In reminder, the regression was formulated as follows:
CAt= β0 + β1SHORTt + β2LONGt + β3BIG4t + β4ROAt + β5LEVt + β6GROWTHt+
β7SIZEt + β8MARKET_CAPt+ β9RETURNt
Table 5
Multivariate OLS regression of CA
Coef. Std. Err. t P>|t| [95% Conf. Inter.]
DUMMY_SHORT 0,011 0,015 0,74 0,461 -0,019 0,041 DUMMY_LONG 0,000 0,009 0,03 0,974 -0,018 0,018 BIG_4 0,000 0,010 -0,05 0,963 -0,021 0,020 ROA 0,066 0,011 5,87 0,000* 0,044 0,088 LEV -0,035 0,009 -4,04 0,000* -0,052 -0,018 GROWTH 0,003 0,002 1,19 0,234 -0,002 0,008 SIZE -0,016 0,009 -1,78 0,076 -0,034 0,002 MARKET_CAP 0,021 0,009 2,44 0,015 0,004 0,038 RETURN 0,006 0,007 0,75 0,452 -0,009 0,020 _cons 0,004 0,012 0,35 0,723 -0,019 0,028 Number of observations 792 F-statistic 14,552 Prob F 0 Adjusted R-squared 0,144 * significant at 5 percent
In Table 5 the results of the multivariate OLS regression of the CA model are tabulated. The F-statistic is significant at the 1 percent level, amounting to 14.55, allowing for the acceptance of the model as a whole having a correlation with current accruals. R-squared is 14.4 percent, significantly higher than the AWCA-model. The coefficient for DUMMY_SHORT is set at 0.11, with a p-value of 0.461. These results offer no support for the first hypothesis. The coefficient for DUMMY_LONG is 0, with a p-value of 0.974. These results also offer no findings. In summary, again neither of the hypotheses is supported.
5.4 Alternative cut-off dates
In this section, I perform two more AWCA and CA models using varying cut-off dates. In the first section the focus will lie on finding significant results for the SHORT dummy, While in the second section the focus will lie on finding significant results for the LONG dummy. As Johnson et al (2002) indicated, the wrong cut-off date can be of influence on the power of a statistical test. In this study, I choose to decrease the cut-off dates by one year in the first test, and increasing it by one year in the second test.
5.4.1 Cut-off 2 and 8
In the first alternative test, I move both cut-off dates back to 2 and 8 years. The results for AWCA are tabulated below in Table 6. The R-squared of the model increases slightly to 10.94%. However, the results for the SHORT variable still do not reach a significant value, with a P-value of 0.306. The only significantly correlated control variables still remain ROA and LEV.
Table 6
Multivariate OLS regression of AWCA (2-8) Coef. Std. Err. t P>|t| [95% Conf. Interval] DUMMY_SHORT -0,011 0,018 -0,58 0,560 -0,047 0,025 DUMMY_LONG -0,010 0,009 -1,02 0,306 -0,028 0,009 BIG_4 -0,003 0,010 -0,26 0,794 -0,023 0,018 ROA 0,043 0,011 3,82 0.000* 0,021 0,066 LEV -0,047 0,009 -5,44 0.000* -0,064 -0,030 GROWTH 0,002 0,002 0,74 0,460 -0,003 0,007 SIZE 0,002 0,009 0,19 0,846 -0,016 0,020 MARKET_CAP 0,002 0,009 0,21 0,831 -0,015 0,019 RETURN -0,002 0,008 -0,3 0,761 -0,017 0,012 _cons 0,020 0,012 1,59 0,111 -0,005 0,044 Number of observations 792 F-statistic 10,670 Prob F 0,000
Table 7
Multivariate OLS regression of CA (2-8)
Coef. Std. Err. t P>|t| [95% Conf. Inter.] DUMMY_SHORT 0,004 0,018 0,21 0,831 -0,032 0,040 DUMMY_LONG -0,002 0,009 -0,27 0,791 -0,021 0,016 BIG_4 0,000 0,010 0,00 1,000 -0,020 0,020 ROA 0,066 0,011 5,88 0,000* 0,044 0,088 LEV -0,035 0,009 -4,03 0,000* -0,052 -0,018 GROWTH 0,003 0,002 1,20 0,231 -0,002 0,008 SIZE -0,017 0,009 -1,81 0,071 -0,035 0,001 MARKET_CAP 0,021 0,009 2,45 0,014* 0,004 0,038 RETURN 0,006 0,007 0,80 0,426 -0,009 0,021 _cons 0,007 0,012 0,59 0,557 -0,017 0,031 Number of observations 792 F-statistic 14,494 Prob F 0 Adjusted R-squared 0,143 * significant at 5 percent
Above, table 7 shows the results for the CA model with cut-off dates of 2 and 8 years. With p-values at 0,831 and 0,791 for DUMMY_SHORT and DUMMY_LONG respectively, no support is found for either of the hypotheses. Concluding, moving the cut-off dates back offers no additional insight into the relationship of audit firm tenure and audit quality.
5.4.2 Cut-off 4 and 10
In the second alternative test, I move both cut-off dates forward to 4 and 10 years. For the AWCA model, as shown on the next page in table 8, this increases the R-squared of the model greatly to 11.37%. For the SHORT variable, this still offers no additional insights, as could be expected based on the main results. However, setting the cut-off date for the LONG dummy to more than 9 years offers an interesting result. With a Pvalue of 0.040 and a coefficient of -0.0192, This analysis shows that firms being audited for more than 9 years have relatively lower levels of AWCA than firms that have been audited by the same for more than 4 but less than 10 years. This result is in line with the second hypothesis. It also indicates that the cut-off date of the LONG group may have been initially chosen wrongly for the preliminary results.
Table 8
Multivariate OLS regression of AWCA (4-10) Coef. Std. Err. t P>|t| [95% Conf. Interval] DUMMY_SHORT -0,003 0,013 -0,21 0,833 -0,029 0,024 DUMMY_LONG -0,019 0,009 -2,05 0.040* -0,038 -0,001 BIG_4 0,002 0,011 0,21 0,833 -0,019 0,023 ROA 0,043 0,011 3,76 0.000* 0,020 0,065 LEV -0,047 0,009 -5,39 0.000* -0,064 -0,030 GROWTH 0,002 0,002 0,63 0,530 -0,003 0,006 SIZE 0,001 0,009 0,1 0,922 -0,017 0,019 MARKET_CAP 0,003 0,009 0,36 0,721 -0,014 0,020 RETURN -0,002 0,008 -0,28 0,778 -0,017 0,013 _cons 0,020 0,012 1,64 0,101 -0,004 0,045 Number of observations 792 F-statistic 11,145 Prob F 0,000 Adjusted R-squared 0,114 * significant at 5 percent
On the next page, table 9 show the results for the CA model with cut-off dates at 4 and 10 years. Here, p-values have also increased, to 0,508 and 0,503 for DUMMY_SHORT and DUMMY_LONG respectively, but still do not corroborate the findings of the AWCA model.
Coef. Std. Err. t P>|t| [95% Conf. Inter.] DUMMY_SHORT 0,009 0,013 0,66 0,508 -0,017 0,035 DUMMY_LONG -0,006 0,009 -0,67 0,503 -0,024 0,012 BIG_4 0,001 0,011 0,13 0,899 -0,020 0,022 ROA 0,066 0,011 5,84 0,000* 0,044 0,088 LEV -0,035 0,009 -4,01 0,000* -0,052 -0,018 GROWTH 0,003 0,002 1,12 0,265 -0,002 0,008 SIZE -0,016 0,009 -1,77 0,077 -0,034 0,002 MARKET_CAP 0,022 0,009 2,50 0,013* 0,005 0,038 RETURN 0,006 0,007 0,75 0,454 -0,009 0,020 _cons 0,006 0,012 0,47 0,637 -0,018 0,030 Number of observations 792 F-statistic 14,686 Prob F 0 Adjusted R-squared 0,145 * significant at 5 percent 5.5 Robustness check
In order to test for the robustness of the results of the AWCA model, I perform a robustness check. This robustness check is estimated by the following models:
AWCAt= β0 + β1TENUREt + β2BIG4t + β3ROAt + β4LEVt + β5GROWTHt + β6SIZEt+
β7MARKET_CAPt+ β8RETURNt
In this model, the dummy variables SHORT and LONG have been replaced by one variable, TENURE, representing the years of tenure. Essentially, this replaces the three comparative groups by one variable which should show the trend of decreasing levels of AWCA as TENURE increases.
The results for AWCA are shown in the table below:
Table 10
Robustness check (AWCA) Coef. Std. Err. t P>|t| [95% Conf. Interval] TENURE -0,001 0,001 -2,05 0,041* -0,003 0,000 BIG_4 0,001 0,010 0,06 0,955 -0,020 0,021 ROA 0,043 0,011 3,78 0,000* 0,021 0,065 LEV -0,047 0,009 -5,44 0,000* -0,064 -0,030 GROWTH 0,001 0,002 0,60 0,551 -0,003 0,006 SIZE 0,002 0,009 0,20 0,845 -0,016 0,020 MARKET_CAP 0,003 0,009 0,33 0,741 -0,014 0,020 RETURN -0,002 0,007 -0,33 0,744 -0,017 0,012 _cons 0,025 0,012 2,02 0,044 0,001 0,049 Number of observations 792 F-statistic 12,458 Prob F 0,000 Adjusted R-squared 0,113 * significant at 5 percent
As is seen in the table above, TENURE is negatively associated with the level of AWCA present in a firm. This is not consistent with the initial hypotheses, however it does supports the main findings.
The U.S. Comptroller General concluded in 2003 that mandatory audit firm rotation was not the most effective way to increase audit quality in general. However, if the Sarbanes Oxley did not improve audit quality sufficiently, the system may be necessary. As the PCAOB re-opened the debate surrounding this topic, the lack of consensus between the proponents and opponents indicated that further research was necessary. This thesis has tried to provide current evidence surrounding the relationship between audit quality and audit firm tenure in the US by attempting the following research question: “To what extent could the possible implementation of audit firm rotation improve audit quality for US listed companies in the post SOX-era?”
The sample of this thesis is derived from the COMPUSTAT North America fundamentals annual Database for the years 2011 and 2012. This is to a great extent exemplary for the companies affected by a possible implementation of audit firm rotation under U.S. regulation. The impact of the length of the audit firm’s tenure at the client on audit quality was investigated by quantifying tenure through historical data and proxying audit quality by the AWCA and CA models.
The prediction of the hypotheses was that medium tenure would have the highest level of audit quality, and thus lower levels of AWCA or CA, when compared to the short and long tenure group.
In the main results, based on cut-off periods used in previous studies, the AWCA nor the CA model provided significant results. No difference in AWCA is found between the short and medium tenure group. Between medium and long tenure no significant relation was found either. This means no relation between audit quality and audit firm tenure was found to support H1 or H2..
In further analysis however, moving the cut-off for long tenure to 10 or more years and for short tenure to 4 or less, it is found that long tenure has a negative relation with AWCA. This indicates that long tenure has lower levels of AWCA than medium tenure, which in this model was stated as being between 5 and 9 years. For short tenure in this case, a significant relationship was still not found. Moving in the other direction, making the long group longer than 8 and the short group 2 years, no relationship was found on either side. For the CA model, still no significant findings were found.
Taken together, the results of the study have contradictory implications for the debate surrounding audit firm rotation. The results fail to provide support for the idea that short audit firm tenure is associated with lower audit quality. This is inconsistent with the argument against a mandatory rotation system that the first years of the audit-client relationship are associated with lower levels of audit quality. On the other hand, there are limited findings in this study that there is a negative relation of long tenure with AWCA compared to medium tenure, inconsistent with the arguments of proponents of firm rotation, who claim familiarity and complacency negatively influence audit quality in the long run.
This thesis contains several limitations. First off, the results are offset against a hypothetical future introduction of mandatory rotation in which incentives are thought to remain the same. In fact, decisions surrounding abnormal working accruals made by auditors and managers may be influenced by the new incentives of the mandatory rotation system. Furthermore the use of a proxy for audit quality in itself is a limitation, as there may be other constructs and control variables which are able to better reflect reality. In future research I suggest to further build on the models present within the existing literature in order to better reflect reality.
References
Accountingweb.(2012) Pcaob member sees hurdles mandatory firm rotation. Available at:
http://www.accountingweb.com/article/pcaob-board-member-sees-hurdles-mandatory-firm-rotation/220423
Becker, C. L., DeFond, M. L., Jiambalve, J. J. and K.R. Subramanyam. (1998). The effect of Audit quality on earnings management. Contemporary Accounting Research, 15, (1), 1-24
Carcello, J.V. and A.L. Nagy (2004). Audit Firm Tenure and Fraudulent Financial Reporting
Auditing: A Journal of Practice and Theory, 23, (2), 55-69.
Carey, P., and R. Simnett (2006). Audit Partner Tenure and Audit Quality. The Accounting Review, 81, (3), 653-676.
Davis, L. R., B. Soo, and G. Trompeter. (2003). Auditor Tenure, Auditor Independence and Earnings Management. Working paper, Boston College.
DeAngelo, L. (1981). Auditor Size and Audit Quality. Journal of Accounting and
Economics, 3, (1), 183-99.
Defond, M.L., and J. Jiambalvo (1994). Debt Covenant Violation and Manipulation of Accruals. Journal of Accounting and Economics, 17, (1-2), 145-176.
Defond, M.L., and C.W. Park (2001). The Reversal of Abnormal Accruals and the Market Valuation of Earnings Surprises. The Accounting Review, 76, (3), 375-404.
Dopuch, N., King, R., & Schwartz, R. (2001). An Experimental Investigation of Retention and Rotation Requirements. Journal of Accounting Research, 39, (1), 93-117. European Union (2011) Proposal for a regulation of the European Parliament and of the
Council on specific requirements regarding statutory audit of public-interest entities.
http://ec.europa.eu/internal_market/auditing/docs/reform/regulation_en.pdf
European Union (2006) Company Law 8th Directive. Available at:
http://eur-lex.europa.eu/legal-content/NL/TXT/PDF/?uri=CELEX:32014R 0537&from=EN
Financial Times. (2012). Regulator weakens audit tendering reform, Available at:
http://www.ft.com/cms/s/0/c7063ce6-8a10-11e1-a0c8-00144feab49a.html#axzz33
rHs7y1m
Francis, J.R. (2004)“What do we know about Audit Quality?” The British Accounting
U.S. General Accounting Office (GAO). (2003). Public Accounting Firms: Required Study
on the Potential Effects of Mandatory Audit Firm Rotation. Washington DC:
Government Printing Office.
Ghosh, A. and D. Moon (2005). Auditor Tenure and Perception of Auditor Quality. The
Accounting Review, 80, (2), 585-612.
Hatfield, R.C., Jackson, S.B., & S.D. Vandervelde (2007). The effects of auditor rotation and
client pressure on proposed audit adjustments.Working paper, University of Alabama.
International Auditing and Assurance Standards Board (2010). International Standard on
Auditing 200: Overall Objectives of the Independent Auditor and the Conduct of an Audit in accordance with International Standards on Auditing.
Johnson, E., Khurana, I.K. and J.K. Reynolds (2002). Audit-Firm Tenure and the
Quality of Financial Reports. Contemporary Accounting Research, 19, (4), 637-660. Knapp, M. 1991. Factors that audit committees use as surrogates for audit quality. Auditing:
A Journal of Practice and Theory,10, (1): 35-52.
KPMG, (2011). Further EC Proposals for Reform of the Audit Market. Available at:
http://www.kpmg.com/Global/en/services/Audit/EU-AuditReform/Documents/further-ec-audit-reform-proposals-in-the-headlines.pdf
Kramer, S., Georgakopoulos, G., Sotiropoulos, I. & K. Vasileiou (2011). Audit Firm Rotation, Audit Firm Tenure and Earnings Conservatism. International Journal of
Economics and Management. 6, (8), 44-57.
Kutner MH, Nachtsheim CJ, Neter J, Applied Linear Regression Models, 4th edition, McGraw-Hill Irwin, 2004.
Leone, J. (2007). Factors related to internal control disclosure:A discussion of Ashbaugh, Collins, and Kinney (2007) and Doyle, Ge, and McVay (2007). Journal of Accounting
and Economics, 44, (1), 224-237.
Myers, J., Myers, L., & T. Omer (2003) Exploring the term of the auditor-client relationship and the quality if earnings: A case for mandatory auditor rotation? The Accounting
Review, 78, (3), 779-799.
PCAOB (2011) Comments for Docket 037. Available at:
http://pcaobus.org/Rules/Rulemaking/Pages/Docket037Comments.aspx
Sarbanes-Oxley Act (2002) Public Law No. 107-204. Washington, DC: Government Printing Office
Watts, R.L. en J.L. Zimmerman (1981). The Markets for Independence and Independent Auditors, Rochester University, unpublished.
Weber, J., M. Willenborg, and J. Zhang (2008). Does Auditor Reputation Matter? The Case of KPMG Germany and ComROAD AG. Journal of Accounting Research, 46, (4), 941-972.
