The financial crisis and the effect of audit quality on earnings management

(1)

Amsterdam Business School

The financial crisis and the effect of audit quality on earnings management

Robin Gerssen- 10681000

Obdam, June 2015

Word count: 13502

MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam Supervisor: Dr. Sanjay Bissesur

(2)

Abstract

The objective of this study is to examine the effect of audit quality on earnings management during the financial crisis. I find a significant relationship between audit quality and earnings management during the financial crisis. This relation does not significantly differ from the relationship between audit quality and earnings management in the pre-crisis period.

Statement of Originality

This document is written by student Robin Gerssen 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.

(3)

Earnings management is a concept which has been widely researched in the past years. In the past the existence of earnings management has been proven for example by examining discretionary accruals (Jones, 1991). The effect of audit quality on earnings management has been the subject of research published in journals for an extensive amount of times. Significant evidence has been found that higher audit quality reduces earnings management (Becker et al, 1998).

The financial crisis creates a unique situation in which the relationship between audit quality and earnings management can be tested under specific circumstances. For example in the financial crisis, companies have common incentives, to attract potential investors by using high-quality financial reporting (Cimini, 2014). The financial crisis has also been proven to be an interesting case from the perspective of the auditor. Attention has been drawn to the limits and limited capabilities of auditing and the need to reform aspects of practice and regulations (Humphrey et al., 2009). On the other hand, the typical questioning in the aftermath of major banking collapses as to ‘’where were the auditors’’ has been less prevalent than in the past (Humphrey et al., 2009).

The goal of this thesis will be to examine the effect of audit quality on earnings management during the financial crisis. The main research question for this thesis will be:

What is the effect of audit quality on earnings management during the financial crisis?

This thesis will contribute to science as it gives new insight on the effects of audit quality on earnings management. There has never been a unique event that had such a significant impact on society and the economic landscape as the financial crisis. This thesis will contribute to the discussion if during a unique event such as the financial crisis, the effect of audit quality on earnings management is still the same.

To answer the above stated research question, a database study will be conducted. The sample will consist of US firms. Using the discretionary accruals model, earnings management will be identified. Audit quality will be measured by Big Four vs non Big Four.

The result of this research is that audit quality reduces earnings management measured by discretionary accruals during the financial crisis. Big Four auditors still reduce earnings management significantly more than non-Big Four auditors. This effect in the financial crisis period is not significantly different compared to the pre-crisis period.

These results contribute to literature as they show the effect of an unique event like the financial crisis on well-known relationships like audit quality on earnings management. It shows that the expertise of BigFour auditors ensures that even during the financial crisis BigFour auditors reduce earnings management significantly more than non-BigFour auditors.

(5)

earnings management and audit quality is discussed. Chapter three will describe the research methodology that is used to perform the research. In chapter four the results from the regressions are discusses.

(6)

2 Literature review

In this section, the theoretic foundation of the thesis will be outlined. I will start by explaining the basis on which this thesis is written. As the information in this thesis will be indirectly derived from financial statements the process that feeds financial statements, being financial reporting will be described. Secondly I will discuss the theory on witch this thesis will be based. Then I will discuss prior research literature on earnings quality, earnings management, audit quality and the relation between audit quality and earnings management. Furthermore, I discuss the effect of the financial crisis on earnings

management

2.1 The objective of financial reporting

Financial reporting or accounting serves two objectives (roles): decision usefulness (the valuation role) and stewardship (the efficient contracting role)(Scott, 2011).

The first objective of financial reporting is decision usefulness (valuation role). Decision usefulness is past oriented and helps financial statement users to better make investment decisions. Decision usefulness is adopted by major standard-setting bodies like IASB and FASB. Both the IASB and the FASB show a clear recognition of the role of financial reporting in providing useful information for investors (Scott, 2011). For information to be useful to base decisions on, information needs to be relevant. However, information can only be useful when it is also reliable.

For decision usefulness the balance sheet approach is adopted. This implies that the balance sheet should reflect the true fundamental value of the firm. Fair value is used to measure assets and liabilities and earnings represent changes in firm wealth (Scott, 2011).

The second objective of financial reporting is stewardship (efficient contracting). Stewardship is defined as ‘’a feature of the principal-agent relationship whereby the agent is assumed to safeguard the resource of the

principal’’ (Belkaoui & Pavlik, 1992) Financial reporting is an important communication device between

management and the suppliers of equity or debt. Financial reporting is used to evaluate the work performed by the managers (stewards) of a company. The stewardship role assumes that financial reporting is used to incentivize and monitor managers. To measure performance of managers,

information needs to be reliable and not necessarily relevant. The income statement approach is linked to the stewardship role (Scott, 2011). The goal of the income statement approach is that financial reporting provides an unbiased measurement of firm performance. Balance sheet accounts are the residual of income statement accounts and represent the liquidation value of the firm.

The above mentioned two perspectives have been the subject of political discussion as to what perspective is most relevant (Dichev, Graham, Harvey, & Rajgopal, 2013) By conducting a survey with 169 CFOs of public companies, the perspective from the companies itself was examined by Dichev et al., (2013). The company CFOs stated that the purpose of accounting is to give investors a basis for valuing

(7)

a company. This is consistent with the decision usefulness perspective. As stated above decision usefulness is also adopted by the major regulating authorities (IASB and FASB).

Although a significant part of literature and regulating authorities adopted the decision usefulness perspective (balance sheet approach), there has been criticism on this perspective. Holthausen and Watts (2001) state that the valuation models provide little to no theory that can be used in accounting. The assumption that accounting numbers provide information for valuation by itself, provides very little in the way of a theory for accounting as it cannot explain components of income for example (Holthausen & Watts, 2001) This critical view of the valuation role is challenged by Barth et al. (2001). The possible contracting and other uses of financial statements do not in any way demise the importance of value relevance research as value relevance research provides insights into questions of interest to standard setters and other non-academic constituents (Barth et al., 2001).

2.2 Agency theory

To get a basis for identifying the variables that together constitute the research question, the Agency theory is discussed. The Agency theory will give the point of view from which this thesis will be written. The Agency theory describes the principal-agent relationship that exists between owners of a firm and the board of directors that run the firm (Jensen & Meckling, 1976). The Agency theory indicates that there is a conflict of interest between the agent who runs the firm (management) and the principal (the owner). This problem only exists when management and owner are not the same person. Typically there are two types of agency relationships: shareholder-manager contracting and shareholder-debtholder contracting. Because of the conflict in interest and the fact that the principal cannot monitor every move of the agent, information asymmetry arises. Information asymmetry is the difference in information between the principal and agent. Information asymmetry can be explained by two ‘’problems’’ being adverse selection and moral hazard (Scott, 2011).

Adverse selection states that one of the parties involved (the agent) has an information advantage in a potential business transaction, which can be exploited at the expense of the other party being the principal. Adverse selection is an ex ante information asymmetry problem as it occurs before signing contract of making an investment decision.

Moral hazard states that one of the parties (the agent) is able to observe their actions but the principal cannot. This means that the agent can take actions that benefit the agent but not the principal. Moral hazard is an ex post information asymmetry problem as it occurs after signing a contract of making an investment decision.

The principal can limit divergences from his interest caused by moral hazard of adverse selection by incurring monitoring costs or using incentives (Jensen & Meckling, 1976). An example of monitoring

(8)

costs, is hiring an high quality auditor.

The Agency theory is applicable to my thesis because earnings management is a typical agency problem. The agent tries to manage earnings to suit the needs of the agent. The principal cannot monitor this managing of earnings. Therefore an auditor is hired to reduce earnings management. In a time of crisis, earnings management can still be a problem for principals. If for example a company has a bad year, the agent might want to take all the losses in that year (big bath accounting), the principal cannot monitor this, therefore the influence of an auditor is a still expected to be a relevant variable.

2.3 Earnings quality

Stockholders make investing decisions on a daily basis. These decisions are fore the most part based on the performance of companies. The bottom-line measurement for performance is earnings. For investors to make investing decisions or for commissioners to monitor the performance of executives, earnings need to be of high quality. High quality earnings is defined as follows:

‘’Higher quality earnings provide more information about the features of a firm’s financial performance that are relevant to a specific decision made by a specific decision-maker.’’ (Dechow et al., 2010, p.344)

According to Dechow et al. (2010) the definition of earnings quality is meaningless unless it is seen together with the following three features. First, the definition of earnings quality depends on the decision-relevance of the information. Earnings quality is defined only in the context of a specific decision model. Second, a reported earnings number is only of high quality if it is informative about the firm’s financial performance. Many of the aspects of firm performance are unobservable. Third, earnings quality is determined jointly by the ability of the accounting system to measure performance and by the relevance of the measured performance to the decision.

To measure earnings quality a number of empirical proxies are summarized by Dechow et al. (2010). These proxies are based on prior research and are as follows:

Persistence

Firms that present more stable and persistent earnings have a more ‘’sustainable’’ cash flow/earnings stream. This results in a more useful input into DCF-based equity valuations.

Magnitude of accruals

Extreme accruals are perceived to be low quality as they represent a persistent component of earnings. Residuals from accrual models

Residual accruals arising from accrual models represent management discretion of estimated errors. Both reduce decision usefulness and therefore earnings quality.

(9)

Smoothness

Smoothing transitory cash flows lead to an improvement of earnings persistence and earnings

informativeness. However managers can attempt to smooth permanent changes in cash flows. This will lead to a less timely and less informative earnings number.

Timely loss recognition

Timely loss recognition is needed to counter the natural optimism of managers. The timely recognition of considered to represent high quality earnings.

Benchmarks

Unusual clustering in earning distributions is an indicator for earnings management around targets. Observations slightly above of at targets have low quality earnings.

Earnings response coefficient

If investors respond to information, this implies that this information has a lot of value. A higher

correlation with value means that earnings better reflect fundamental performance and thus are of higher quality.

External indicators of earnings management

This proxy is based on the assumption that firms that had errors like restatements or Accounting and Auditing Enforcement Releases (AAERs) by the SEC, provide low quality earnings.

All of the above mentioned proxies are used in prior research. In chapter three the choice for the proxy that is used in this thesis will be made.

2.4 Earnings management

To understand the concept of earnings management, the following definition can be considered:

a purposeful intervention in the external financial reporting process, with the intent of obtaining some private gain ( as opposed to, say merely facilitating the neutral operation of the process) (Schipper, 1989)

According to Beneish (2001) in earnings management two perspectives can be recognized: the opportunistic perspective, that assumes that managers tent to mislead investors, and the information perspective (first explained by Holthausen & Leftwich, (1983)) , that assumes managers use managerial discretion to reveal to their investors their private expectations. There has been much prior work that has tested the opportunistic perspective, however little research has been done on the information perspective (Beneish, 2001).

(10)

2.4.1 Accrual earnings management

As mentioned in the paragraph above, earnings management is a purposeful intervention of the external reporting process. Accruals are used to shift results from one period to another, to give the ‘’best’’ representation of earnings at year end. Without accruals, earnings would fluctuate greatly from year to year. This would reduce the usefulness of the presented figures for users of the annual report. Investors are generally more interested in the earnings that created value within a specific period rather than cash flows collected. By using estimations and accruals to reflect these estimations, the relevance of the numbers increase, but the reliability decreases (Johnson, 2005). The reliability of the numbers decreases due to the fact that there is need for discretion in determining the accruals to be included in the annual reports.

Managers can use discretion to increase income numbers and make or break earning targets. Also the increasing of earnings can be done to get a higher compensation. The marginal benefits of earnings management are the most when earnings are managed from a small decrease in earning to a small increase (Burgstahler & Dichev, 1997). The same benefits apply when a small loss is managed to a small profit. To explain this phenomenon Burgstahler and Dichev (1997) provide two explanations. First, a firm that reports an earnings decrease or loss bears sharply higher costs when engaging in transactions with

stakeholders than when the firm had reported an increase in earnings or profit. The second explanation is based on that stakeholders use heuristics to determine the terms of transactions with the firm. When it is costly for stakeholders to retrieve and process information about earnings for all of the firms that they interact with, Burgstahler and Dichev, (1997) assume that some stakeholders use heuristic cutoffs when there are zero changes in earnings or zero earning all together.

Based om the literature above it can be stated that managers have incentive to manage earnings through accruals or through real earnings management.

2.4.2 Real earnings management

After the implementation of the Sarbanes-Oxley act in 2002 there was an significant switch from accrual based earnings management to real earnings management (Cohen et al, 2008). Although there was a switch between accrual earnings management to real earnings management, earnings management in total remained at the same level as in the pre-sox period (Cohen et al., 2008). In their research Cohen et al (2008) identified three methods that can be classified as real earnings management:

 Increasing sales volume by offering discounts or by making credit terms more lenient  Increasing production to report lower cost of goods sold per unit

 Decreasing discretionary expenses such as advertising, research and development and SG&A expenses.

(11)

Real earnings management has a direct impact on cash flows as opposed to accrual earnings management which only affects the numbers in the annual report. Real earnings management has higher costs than accrual earnings management (Cohen & Zarowin, 2010). This can be explained by the fact that real earnings management is not in the best interest of the company and costs, assets and activities of the company are changed in a way that could hurt the company in the long term.

There are at least two reasons managers use real earnings management more easily than accrual earnings management. Fist, accrual based earnings management is more likely to be detected by the auditor. Second, it is risky for a manager to rely on accrual based earnings management alone. The difference between unmanaged earnings and the desired threshold can be bigger than the amount by which it is possible to manipulate accruals after year end. If this is the case, managers are left with no other option than to use real earnings management before year end, as it is not possible to use real earnings management after the end of the fiscal period.

2.5 Detecting earnings management

From prior research, three approaches have been designed to evaluate the existence of earnings management: Firstly calculating expected and unexpended accruals, secondly focusing on specific accruals like the provision for bad debt and thirdly investigating discontinuities in the distribution of earnings (Beneish, 2001).

2.5.1 Accrual models

In this study, a method to identify and measure earnings management needed to be found. In prior and modern literature a number of models is frequently used, these models were first discussed by Dechow, Sloan and Sweeney (1995), The models that were discussed in the research of Dechow et al. (1995) were:

 The Healy model  The DeAngelo Model  The Jones Model

 The Modified Jones Model  The Industry Model

The models stated above measure two types of errors:

 Type one: The null hypothesis that earnings are not systematically managed is rejected, when in fact this is true.

 Type two: The null hypothesis that earnings are not systematically managed is accepted, when in fact this is false.

(12)

The models mentioned above will be discussed in this paragraph. Furthermore the developments in estimating discretionary accruals using the Jones model after the creation of the model will be discussed. The Healy model

The Healy model tests for earnings management by comparing the mean total accruals as scaled by lagged total assets, across the earnings management partitioning variable (Healy, 1985). The study of Healy (1985) differs from other earnings management studies as the prediction is made that systematic earnings management occurs in every period. Healy (1985) divided his sample in three groups were earnings are predicted to be managed upwards for one group and downward for the other two. To detect earnings management a comparison was made of the mean total accruals in the group were earnings were predicted to be managed upwards to the mean total accruals for each of the groups where earnings is predicted to be managed downwards. This leads to the following formula:

The DeAngelo model

The model that was created by DeAngelo (1986) calculates discretionary accruals by comparing total accruals of the year tested with total accruals of the year before. DeAngelo (1986) assumes that all of the changes in accruals are contributable to earnings management. The model is very similar as the Healy model as they both use total accruals from the estimation period as the proxy for discretionary accruals. The formula for this model is:

Dechow et al. (1995) find the assumption that nondiscretionary accruals are constant unlikely. The nature of the accrual accounting process gives rise to the need for changes in the level of nondiscretionary accruals as they should reflect changes in economic circumstances (Kaplan, 1985) By not including changes in economic circumstances in the model, inflated standard errors will be caused due to the

omission of relevant variables. In addition, when firms experience severe changes in economic conditions, this will lead to biased estimates of earnings management. (Dechow et al., (1995).

(13)

The Jones Model

Jones (1991) conducted research on earnings management by managers during import relief investigations. To conduct this research, Jones (1991) developed a model that estimates ‘’normal’’ (expected) total accruals. This model allows for changes in nondiscretionary accruals that are caused by the change in economic conditions. Jones, (1991) found that during import relief investigations, managers use income-decreasing accruals to manage their earnings. The formula used is as follows:

Jones (1991) uses a time series approach to estimate the firm specific parameters for normal accruals. The Jones model estimates the longest time series of observations prior to year t-1 for each firm. By using a long time series the estimation efficiency is improved but the likelihood of structural change in the estimation period increases (Jones, 1991).

Another way of estimating the firm specific parameters is by using a cross-sectional estimation (DeFond & Jiambalvo, 1994). This estimation method creates coefficients in a given year for a specific industry. This avoids the assumption from the time-series approach that coefficients are stable across years (DeFond & Jiambalvo, 1994).

An important assumption in the Jones Model is that revenues are nondiscretionary (Dechow et al., 1995). If management uses discretion to accrue revenues at year end this will lead to an increase in total accruals. The Jones Model uses total accruals with respect to revenues and will therefore extract the discretionary component of accruals if earnings are managed true revenues. This causes the estimate of earnings management to be biased.

The Modified Jones model

To remove the chance of bias due to managing earnings through revenues, Dechow et al. (1995) created the Modified Jones model. The only difference from the original jones model is that the change in revenues is adjusted for the change in receivables in the event period. The modified version of the Jones model assumes that all of the changes in credit sales in the even period result from earnings management (Dechow et al., 1995). The reasoning behind this assumption is that it is easier to manage earnings by exercising discretion of the recognition of revenue on credit sales than to manage earnings by exercising

(14)

discretion over the recognition of revenue on cash sales (Dechow et al., 1995). The following formula applies for the Modified Jones model:

∆REC stands for the net receivables in year T less net receivables in year T-1. The Industry model

The Industry model is developed by Dechow and Sloan (1991). The Industry model, like the Jones model also holds into account that nondiscretionary accruals change over time. The model assumes that

variation in the determinants of nondiscretionary accruals are the same for firms in the same industry. The Industry model for nondiscretionary accruals is:

Shortcoming of the industry model is that if changes in nondiscretionary accruals largely reflect responses to changes in firm-specific circumstances, the Industry model will not extract all these nondiscretionary accruals from the discretionary accruals proxy (Dechow et al., 1995)

Developments of the Jones model

Ever since the development of the Jones model in 1991, research has been conducted to improve the model or test the model for shortcomings. McNichols (2002) states that there are many reasons to suspect that the discretionary accruals from the Jones model also reflect nondiscretionary aspects other than pure management discretion. This is mostly due to the fact that the Jones model assumes that lagged and future changes in sales are not relevant. The estimation results from prior research by Dechow and Dichev (2002) imply that including cash flows in the Jones model ‘’might’’ reduce the extent to which the model omits variables that correlate with firm economic fundamentals (McNichols, 2002). The findings from the research of McNichols (2002) indicate that researchers should consider the implications of incorporating cash flows in the model used for measuring discretionary accruals to create a more powerful model.

Kothari, Leone and Wasley (2005) examined the power of tests based on performance-matched discretionary accruals with the power of traditional accrual models like the Jones model. Dechow et al. (1995) conclude that ‘’all the models reject the null hypothesis of no earnings management when applied to samples of firms with extreme financial performance’’. This assumes that all firms that have extreme financial performance use earnings management. Therefore traditional accrual models might be

(15)

misspecified when they are applied to a sample of firms with extreme performance. This is caused by the mechanical relationship between performance and estimated discretionary accruals (Kothari et al., 2005)

Matching performance using the Kothari model might reduce Type I errors but based on the research of Keung and Shih (2013) the frequency of Type II errors increases. Keung and Shih (2013) find that matching firms using ROA causes discretionary accruals to be underestimated. This results in an increase of Type II errors. Using the estimated discretionary accruals as an dependent or independent variable in regression, can bias the regression coefficient toward zero (Keung & Shih, 2013).

Jones et al. (2008) conducted research on the ability of popular discretionary accrual models to detect extreme cases of earnings management. These extreme cases were defined as fraudulent earnings and non-fraudulent restatements of financial statements. Jones et al. (2008) examined a total of 10 measures of earnings management including the Jones model, Modified Jones model, Modified Jones model including ROA and models based and the accrual estimation error models from Dechow and Dichev 2002 and McNickols 2002. Jones et al., (2008) found that only the discretionary accruals

estimated from cross-sectional models of working capital changes on past, present, and future cash flows can predict fraud and non-fraudulent restatements of earnings.

Any accrual based earnings management in one period, must reverse in another period. (Dechow et al., 2012). If researchers have reasonable priors concerning the period in which the hypothesized earnings management is expected to revers, the power of the test for earnings management can be significantly improved if these reversals are incorporated (Dechow et al., 2012). Dechow et al., (2012) found that incorporating reversals in the accrual model can increase the explanatory power of the model by 40%, although this can only be done if a researcher has reasonable priors concerning the timing of these reversals (Dechow et al., 2012).

2.5.2 Specific accruals

As there has been criticism on using aggregate accruals, researchers also use a specific accrual, such as the provision for bad debt (McNichols & Wilson, 1988). According to McNickols and Wilson (1988) the provision for bad debt is a proxy for earnings management that is more powerful and less likely to be biased than the proxies used in other studies. When using the provision for bad debt as a proxy, McNickols and Wilson (1988) found evidence that firms manage their earnings by choosing income-decreasing accruals when income is unusually high.

Another specific accrual that is used to detect earnings management is the loss reserve for property casualty insurers (Beaver & McNichols, 1998). This accrual reflects the information of

management about future cash flows that are needed to settle claims on policies written as of the balance sheet date. Because of the substantial judgement needed to estimate the loss reserve, this accrual is

(16)

subjected to management discretion and can be biased (Beaver & McNichols, 1998). By examining the SEC required disclosures about revisions in the loss reserve, Beaver and McNickols (1998) found that there is substantial correlation in year to year loss reserve development. This suggests that management uses discretion over reported loss reserves and that reported loss reserves do not fully reflect available information (Beaver & McNichols, 1998).

2.5.3 Distribution of earnings

The models used in the studies of Jones (1991) and McNickols and Wilson (1988) are widely used in earnings management research. Research by Burgstahler and Dichev (1997) suggests an alternative measure for earnings management. Burgstahler and Dichev (1997) measured earnings management by using a pooled cross-sectional distribution approach. In practice, this means investigating the

discontinuities in the distribution of reported earnings around three thresholds (zero earnings, last year’s earnings and this year’s analysts’ expectations). The research provided significant empirical evidence that earnings decreases and losses are frequently managed away (Burgstahler & Dichev, 1997). They found that 8% to 12% of the firms that have small pre-managed earnings decreases exercise discretion to report earnings increases (Burgstahler & Dichev, 1997). For 30% to 44% of the firms that had slightly negative pre-managed earnings Burgstahler and Dichev (1997) found that those firms exercise discretion to report positive earnings.

Developing a model to measure discretionary accruals is still the most used method to research earnings management in modern literature. This can be explained by three likely causes. First, if earnings management is an unobservable component of accruals, it is less likely that investors are able to detect the effect of earnings management on reported earnings (Beneish, 2001). Second, by studying accruals the problems that are associated with the inability to measure the effect of various accounting choices on earnings are reduced(Watts & Zimmerman, 1990). Third, accruals are the product of Generally Accepted Accounting Principles, and, if earnings will be managed, it is more likely that managers choose the accrual component rather than the cash flow component.

In addition to the above stated, earnings management is difficult to detect. Until now the best method and still the most used method in modern literature for measuring earnings management is by using a discretionary accruals model.

2.6 Audit quality

Audit quality is believed to be directly linked to firm size. DeAngelo (1981) researched the effect of firm size on audit quality. In this research audit quality was measured by the probability that a given auditor will discover a breach in the client’s accounting system and report the breach. DeAngelo (1981) has found

(17)

that the larger the auditor as measured by the number of current clients and the smaller the client as a fraction of the auditors’ total quasi-rents, the higher the perceived quality of the audit. Indirect this means that the audit quality of Big Four auditors, due to the fact that they have more clients, is believed to be higher than non-Big Four auditors.

Francis and Krishnan (1999) have found direct evidence for the increase in audit quality when the auditor size increases. Francis and Krishnan (1999) measured the audit quality as the rate that auditors give audit report modifications to high-accrual firms. This is interpreted as the amount of conservatism that an auditor uses in forming his opinion (Francis & Krishnan, 1999). Francis and Krishan (1999) found that in their sample of US listed companies, only Big Six1 auditors show evidence of conservatism.

Khurana and Raman (2004) researched the difference in audit quality between Big Four and non-Big Four auditors in ASEAN countries and compared those results with auditors in the US. Khurana and Raman (2004) also used conservatism as the proxy for audit quality. Evidence was found that the relation between auditor size and audit quality, contrary to US auditors, was not significantly different for Big Four in comparison with non-Big four auditors in ASEAN countries (Khurana & Raman, 2004)

Another well researched proxy for audit quality is tenure. The debate if tenure leads to less independence of the auditor and therefore lower audit quality is ongoing. A recent development is the mandatory rotation of OOB (organizations of public interest) in the Netherlands.

Carey and Simnett (2006) focused their research on audit partner tenure. The propensity to issue a going-concern modified opinion diminishes over the audit partner’s tenure. This suggests that the audit quality decreases as the tenure of the audit partner increases (Carey & Simnett, 2006). When audit quality is measured by the just beating (or missing) of earnings benchmarks, the findings stay the same. A lower proportion of clients miss break-even for long tenure observations .This suggests that clients who have long tenure audit partners have a greater ability to manage earnings in order to report a profit (Carey & Simnett, 2006). Both of the above findings were only applicable to non-big six auditors.

Opposed to the findings of Carey and Simnett, (2006) Johnson, Khurana, and Reynolds (2002) find that the lack of sufficient client-specific knowledge caused by short tenured audits reduces audit quality as it reduces the probability that auditors detect material errors and misstatements. When the auditor tenure increases, firm specific expertise ensures that auditors rely less on managements estimates and more on their own knowledge of the firm and environment (Solomon, Shields, & Whittington, 1999). This increases the quality of the audit.

2.7 The effect of audit quality on earnings management

In the previous two paragraphs audit quality and earnings management was discussed. In this paragraph

(18)

the effect of audit quality on earnings management is discussed.

‘’In comparison to low-quality auditors, high quality auditors are more likely to detect questionable accounting practices and

object to their use of quality the audit report’’ (Becker et al., 1998, p.17)’’ Becker et al. (1998) examined the relation

between audit quality and earnings management. For their research they made the assumption based on prior research that Big Six auditors are of higher quality than non-Big Six auditors (Becker et al., 1998). Significant evidence was found that clients of Big Six auditor’s report discretionary accruals that increase income relatively less than the clients of non-Big Six auditors (Becker et al., 1998).

Chen et al. (2011) conducted research on the effect of audit quality on earnings management in the Chinese market. Chen et al. (2011) examined the effect of audit quality on earnings management for two classes of firms in China, state-owned enterprises (SOEs) and non-state owned enterprises (NSOEs). Chen et al. (2011) found that NSOEs exhibit a greater reduction in earnings management relative to SOEs when both of the categories employ high-quality auditors. The findings were in line with their expectation that the impact of higher quality auditors would be higher for NSOEs than for SOEs.

Most of the evidence from previous literature supports the notion that Big Four (or Big Six/Eight) auditors produce higher quality audits and reduce earnings management. An example of a study that proves the opposite, is the study of Antle et al.(2006). The research of Antle et al. (2006) focused on UK firms from 1994 through 2000 and also tested the robustness of the results by using US data. Antle et al. (2006) found significant evidence of a positive effect of audit fees on abnormal accruals in both the US and the UK. Antle et al. (2006) explained this finding by describing this is consistent with the concept that higher audit fees lead to more acceptance of abnormal accruals and supports the

unconscious influence or bias theory in the behavioral literature. A difference between the research of Antle et al. (2006) and Becker et al. (1998) as well as Chen et al. (2011) is that Antle et al. (2006) measures audit quality by the height of the audit fee, as to Becker et al. (1998) and Chen et al. (2011) measure audit quality by Big Four vs non-Big Four.

In contrary to the findings of Antle et al. (2006), Larcker and Richardson (2004) find that auditors are less likely to allow abnormal accruals for firms where they have the greatest financial interest, ergo higher audit fees. This is caused by auditors wanting to protect their reputation in respect to limiting unusual accounting choices of client firms (Larcker & Richardson, 2004).

2.7 Financial crisis and earnings management

In recent studies the effect of the financial crisis on earnings management was examined. Cimini (2014) conducted research on the effects of the financial crisis on earnings management for European listed firms. Their goal was to examine whether and how the financial crisis affects the tendency of insiders to manipulate the substance of annual reports by varying out earnings management (Cimini, 2014). The general expectation of the study was that there would be a reduction in earnings management during the

(19)

crisis, because of the increase of conditional conservatism (Cimini, 2014). From their event study, Cimini (2014) found evidence of decreases in earnings management after 2008. These findings were explained by the fact that companies have common incentives, during a crisis, to attract potential investors by using high-quality financial reporting.

Habib et al. (2013) researched the effect of financial distress on earnings management choices for New Zealand firms. They concluded that financially distressed firms manipulate earnings downwards. The association found by Habib et al. (2013) was not significantly different in the financial crisis period as to compared to the pre-crisis period. This implies that the level of earnings management during the financial crisis does not significantly differ from before the crisis.

The influence of the choice of auditor on earnings management during a crisis situation was examined in the research of Yew Ming et al. (2007). In the paper of Yew Ming et al. (2007) the presence of negative earnings management activities in service-oriented public listed companies was researched. The setting Yew Ming et al. (2007) used was the Asian financial crisis. The research was conducted on service-oriented public-listed companies in Singapore. During the Asian financial crisis, the selected companies engaged in significantly less earnings management activities (Yew Ming et al., 2007). The evidence of the reduction in earnings management during the Asian financial crisis was only found for companies that were audited by Big-6 auditing firms. This finding implies that during the Asian financial crisis Big-6 audit firms worked as a deterrent to earnings management, which enhanced the quality of the earnings reported by their clients. (Yew Ming et al., 2007)

The influence of audit quality on earnings management for five EU countries during the financial crisis has been researched by Iatridis and Dimitras (2013). The purpose of the study by(Iatridis and Dimitras (2013) was to examine the consequences of the financial crisis on European countries in relation with earnings management. Focus of the research was financially distressed companies that were audited by a Big Four auditor during recession years. Results were different between the five countries. Three of the countries that were audited by a big-four auditor displayed a negative relation with earnings management. Two of the five counties displayed a positive relation with earnings management (Iatridis & Dimitras, 2013). Differences in the levels of earnings management between the five countries were contributed to the differences in regulation and code-law versus common-law characteristics (Iatridis & Dimitras, 2013).

(20)

3 Research methodology

To investigate the influence of audit quality on earnings management during the financial crisis, the following research question is used:

What is the effect of audit quality on earnings management during the financial crisis?

To investigate the above stated research question, two hypothesis are formulated. These hypothesis are explained in the following two paragraphs.

3.1 Hypothesis one

Based on the literature, I expect that for the effect of audit quality on earnings management, even during the financial crisis, a high quality audit reduces earnings management. This expectation is supported based on the research of Cimini (2014) that earnings management will still be present during the financial crisis (although less than before the crisis). And that audit quality generally reduces earnings management (Becker et al., 1998). To examine if this effect is still present during the financial crisis and my expectation can be scientifically supported, hypothesis one is formed:

H1: Audit quality reduces the amount of earnings management used by managers during the financial crisis.

3.1 Hypothesis two

Because of the unique characteristics of the financial crisis, as compared to a setting without an financial crisis, my expectation is that the effect of audit quality on earnings management will be less during the crisis. This expectation is based on the fact that even firms that regularly deliver high quality audits do not have knowledge of such a unique event as the crisis. The expectation is still that firms that deliver high quality audits reduce earnings management more than firms that deliver less quality audits (refer to hypothesis one), only the expectation is that this effect will be significantly less during the financial crisis as compared to pre-crisis. To examine the above stated, hypothesis two is formed.

H2: The effect of audit quality on earnings management is less during the financial crisis, compared to the years before the financial crisis.

3.1 Research model

To investigate the first hypothesis, a regression analysis is used to examine the effect of audit quality on earnings management. A manager can use accruals to ‘’manage’’ their earnings from year to year.

Therefore the proxy for earnings management will be abnormal discretionary accruals. This assumption is confirmed by the research of Dechow et al. (1995). The model to measure discretionary accruals is the modified jones model as originally designed by Jones (1991) and later modified by Dechow et al. (1995).

(21)

NDAt = α1(1 / At -1) + α2(∆REVt - ∆RECt) + α3(PPEt)

where: NDAt is nondiscretionary accruals in year t scaled by lagged total assets;

∆REVt is revenues in year t less revenues in year t - 1;

∆REC stands for the net receivables in year t less net receivables in year t-1.; PPEt is gross property plant and equipment at the end of year t;

At - 1 is total assets at the end of year t - 1; and

a1, a2, a3 are cross sectional industry estimates

As total accruals minus non-discretionary accruals equals discretionary accruals, the following formula is used to calculate discretionary accruals:

TAt = a1(1 / At -1) + a2(∆REVt - ∆RECt) + a3(PPEt)+ vt

Where: TA stands for total accruals scaled by lagged total assets and vt for the residual being discretionary

accruals .

To determine the expected level of accruals for a firm, the expected level of accruals based on firm fundamentals is estimated with the cross-sectional method for every industry group with at least 20 firms in a given year. The choice to use industry groups is based on the research of Kothari et al., (2005) and justified by the notion that any group of firms can be used for the estimation sample (Ecker et al., 2013) To calculate total accruals the same formula that has been used by Dechow et al. (1995) is used:

TA= (∆CAt-∆CLt--∆Casht+∆STDt-Dept)/ (At-1)

Where ∆CAt is change in current assets;

∆CLt is change in current liabilities;

∆Casht is change in cash and cash equivalents;

∆STDt is change in debt include in current liabilities;

Dep is depreciation and amortization expense; A stands for total assets at t-1.

As it is believed that audit quality is directly linked to firm size (Francis & Krishnan, 1999), the proxy for audit quality is Big Four vs non-Big Four. To answer the first hypothesis the following regression formula is used:

(22)

The above stated formula is derived from Doukakis (2014). Prior literature documents that earnings management practices are affected by factors such as auditor type, growth, firm size, financial leverage and profitability (Doukakis, 2014). Doukatis (2014) bases this formula on the research on Becker et al. (1998), Bartov, Gul, and Tsui (2000), Klein (2002) and Bowen, Rajgopal, and Venkatachalam (2008)

GROWTH is the annual percentage change in sales. Managers can use income-increasing accruals when growth slows down (Summers & Sweeney, 1998). To raise the value of their shares even more, growth companies might engage in income-increasing earnings management and because of the riskiness of the investment be more likely to engage in window-dressing activities (Doukakis, 2014).The growth variable is included to control for the impact of growth on earnings management.

SIZE is included to control for the effect of firm size on earnings management and is the natural logarithm of market value of equity. Larger firms have greater analyses following and investor scrutiny than smaller firms and may face higher political costs when engaging in earnings management (Watts & Zimmerman, 1978). Contradicting effect of size is that larger firms may have more opportunities to manage earnings because of the complexity of their operations(Lobo & Zhou, 2006).

As debt covenant restrictions influence accounting choices (DeFond & Jiambalvo, 1994) LEVERAGE is included to recognize debt-contracting incentives for earnings management and is measured as the ratio of total liabilities to last year’s total assets.

ROE (return on equity) is calculated as net income divided by last year’s total assets. ROE controls for the relationship between profitability and earnings management (Doukakis, 2014).

CFO: The absolute value of cash flows from operations scaled by lagged total assets is added to the regression. This is done to control for the fact that in a period of extreme financial performance, estimated discretionary accruals will be to large (Dechow et al., 1995) .

BIG4 represents a dummy where value one represents a company with a Big Four auditor and zero otherwise.

As a selection variable the dummy Crisis will be used. As my expectation is that the effect of audit quality on earnings management will be higher in the pre-crisis (2003-2007) period, this will be labeled as 1 and otherwise 0 (crisis period, 2008-2012). The crisis period of 2008-2012 is based on the research of Cimini (2014).

To investigate the second hypothesis, the following formula will be used:

DAi,t= a0+ a1GROWTHi,t + a2SIZEi,t + a3LEVERAGEi,t + a4ROEi,t+ a5CFOi,t + a6CRISISi,t +

(23)

To examine if the effect of audit quality on earnings management is less during the crisis, the dummy CRISISxBIG4 is used.

3.2 Sample selection

The data that is used for investigating the first hypothesis is a sample of US firms for the years 2008-2012. The data is be derived from Compustat using Wharton Research Data Services (WRDS). To process the data SPSS is used. The choice to use US firms is based on the fact that for US firms there is relatively more data than for non-US firms.

To investigate the second hypothesis, the same regression that is used to answer hypothesis one will be used. The sample will consist of the same US firms as used in answering hypothesis one, with the selected firm years 2003-2007. To measure if the influence of audit quality on earnings management is significantly less during the financial crisis, a dummy variable is created to separate the firm years 2003-2007 from 2008-2012.

Sample Description

123.263 Compustat North American firms 2002-2012 (no financial sector) -49.391 Merging with Audit Analytics

-44.029 Controlling for 11 firm years -201 Deleting Canadian firms

-5.801 Deleting blanks Jones Model variables -1.872 Winsorizing Jones Model variables -3.045 Deleting missing observations

18.924 Final sample

Table 3.1 Sample breakdown

The initial sample consists of 123.263 North American non-financial firm years. The selection of only non-financial companies is based on earlier earnings management research. Banks, insurance companies, other financial holdings companies have to be excluded when measuring discretionary accruals since specific regulations apply to these industries that could affect discretionary accruals (Becker et al., 1998). To get the data for audit quality, the data from Compustat is merged with data from Audit Analytics. Deleting blank cases reduces the sample with -49.391 to 73.872. To calculate year to year variance variables like ∆REV11 firms years need to be included for every firm. Firms with less than 11 years were deleted. This results in an reduction of -44.029 firm years. As the focus of this thesis is on the influence of audit quality on earnings management in the United States, Canadian firm years are deleted. This results in a reduction of -201 firm years. Deleting the blanks for variables included in the Jones Model reduces the sample by -5.801. Because of the fact that the Jones Model variables compute the dependent variable

(24)

discretionary accruals, the top 1% and bottom 1% are deleted. This is done to treat outliers in the

dependent variable. This deletion results in an reduction of the sample by -1.872. The last reduction of the sample is caused by the deletion of missing observations for the input variables of my regression model. The final reduction consist of -3.045 firm years and brings the sample to 18.924 firm years.

(25)

4 Results

In this chapter the results of the performed regression analysis are described. In the first paragraph the constructing of the dependent variable discretionary accruals will be discusses. The second paragraph will show the descriptive statistics and correlations for the variables that are used to test the two hypotheses. Finally in the third paragraph, the results of the performed regression and the impact on the two hypothesizes is discussed.

4.1 Discretionary accruals

The dependent variable in the research model that is used to test the two formed hypotheses is

discretionary accruals. As mentioned in paragraph 3.1, the Modified Jones model developed by Dechow et al. (1995) is used. To give insight in the variables that together form the proxy for earnings

management, the descriptive statistics are shown below.

Table 4.1 Descriptive statistics Jones Variables

Variable Statistic Pre-Crisis Crisis (Crisis=Pre-crisis) Tests of null

Total Accruals Mean -0,029 -0,034 t-value 4,666

Median -0,027 -0,032 Sig. (2-tailed) 0,000

Std. Deviation 0,078 0,074

Minimum -0,435 -0,433

Maximum 0,332 0,328

1 / At - 1 Mean 0,016 0,012 t-value 5,593

Median 0,002 0,001 Sig. (2-tailed) 0,000

Minimum 0,000 0,000

Maximum 0,657 0,646

∆REV Mean 0,102 0,040 t-value 20,827

Minimum -0,691 -0,714

Maximum 1,296 1,296

∆REC Mean 0,028 0,007 t-value 24,410

Minimum -0,162 -0,163

Maximum 0,351 0,344

PPE Mean 0,458 0,444 t-value 2,071

Minimum - -

Maximum 2,075 2,060

(26)

As is shown in table 4.1 total accruals are slightly more negative in the crisis period than in the pre-crisis period. Scaled total assets is slightly lower in the crisis period than in the pre-crisis period. Intuitively this makes sense, as companies are more likely to decrease in size in a crisis than to increase. The growth in revenues and also receivables is less in the crisis than before. This is to be expected as the crisis constrains growth and puts pressure on the results of a company. Property plant and equipment is slightly lower in the crisis period as to compared to the pre-crisis period. Intuitively this can be explained by the fact that in a crisis period companies tend to slow down investments but depreciation of PP&E continues.

The overall image that is created in table 4.1 is that the financial crisis has a negative effect on all the variables. This is to be expected as the financial crisis has a negative effect on growth, results and the general financial health of a company.

The variables stated in table 4.1 are the basis of the discretionary accruals which are used as the proxy for earnings management. The variables are regressed using a linear regression in SPSS.

The Pearson correlation test can be used to determine correlations between variables. These correlations can be used to give insight in the explanatory power of a model. The correlations of the variables that are used in the Modified Jones Model are showed in table 4.2 As can be expected based on the research of Jones (1991) and Dechow et al. (1995) all of the variables have a significant correlation with total accruals, meaning the variables affect total accruals either positively or negatively. The Spearman correlations show the same image except that lagged total assets does not have a significant influence on total accruals when measured as the Spearman correlation.

Table 4.2 Correlations Jones Model

_AccrualsTotal 1 / At - 1 ∆REV ∆REC PPE

Total Accruals -0,039** _0,136** _0,316**_-0,267** 1 / At - 1 -0,002 0,027** _-0,022**_0,022**

∆REV 0,082** _0,065** _0,364**_0,083**

∆REC 0,314** _0,002 _0,416** _-0,085**

PPE -0,319** _0,037** _0,192** _-0,083**

** Indicates correlation is significant at the 0.05 and 0.01 level (2-tailed). Above diagonal contains Pearson correlation, below contains Spearman correlation.

Refer to appendix 1 for variable definitions.

4.2 Descriptive statistics and correlations

Before the discussion of the regression results in paragraph 4.3, in this paragraph the descriptive statistics and correlations of the variables that are included in my research model are discussed. By

(27)

analyzing the descriptive statistics and correlations an early indication concerning the outcome of the regression model can be obtained.

Table 4.3 Descriptive statistics research model

Variable Statistic Pre-crisis Crisis (Crisis=Pre-crisis) Tests of null

DA Mean -0,003 -0,006 t-value 3,386

Median -0,002 -0,004 Sig. (2-tailed) 0,001

Minimum -0,433 -0,414

Maximum 0,352 0,362

ROE Mean 0,141 -0,150 t-value 0,547

Std. Deviation 54,135 7,369 Minimum -2.722,657 -286,827

Maximum 4.677,657 544,180

LEVERAGE Mean 0,623 0,617 t-value 2,044

Minimum - -

Maximum 22,265 46,550

SIZE Mean 6,295 6,059 t-value 6,853

Minimum -4,162 -2,901

Maximum 12,760 12,436

GROWTH Mean 0,592 0,196 t-value 1,052

Minimum -165,300 -25,056

Maximum 3.701,467 474,812

CFO Mean 0,059 0,056 t-value 0,51

Minimum -4,964 -2,187

Maximum 2,390 2,395

BIG4 Mean 0,763 0,678 t-value 12,300

Minimum - -

Maximum 1,000 1,000

(28)

Table 4.3 shows that most of the variables in the model change negatively in the crisis period as to compared to the pre-crisis period. The mean of return on equity changes from 0,1 positive to -0,2

negative, although this change is not significant. The size of companies decreases significantly from 6,3 to 6,0 and growth decreases from 0,6 to 0,2 (non-significant). These changes are to be expected as the economic reduces company growth and puts pressure on results.

Leverage shows a minor decrease, this implies that companies have not changed their way of financing during the crisis. Cash flows divided by lagged total assets also only shows a non-significant minor decrease, implying cash flows are not greatly impacted. This can be explained by the fact that company sales decreases but due to cost cutting, also do the costs.

Discretionary accruals only show a minor decrease but is still present during the financial crisis. Big Four is a dummy variable and has little explanatory power when only using descriptive statistics. What can be stated is that during the pre-crisis period 76% of the companies had an Big Four auditor compared to 68% in the crisis period.

To examine how all of the variables used in the research model effect one another, correlation table 4.4 is included.

Table 4.4 Correlations Research model

DA GROWTH SIZE LEVERAGE ROE CFO CRISISxBIG4 CRISIS BIG4 DA 0,008 -0,018* -0,032** 0,005 -0,091** -0,004 0,025** -0,036** GROWTH 0,054** 0,009 0,003 0 -0,008 0,009 0,007 0,005 SIZE -0,023** 0,155** 0,019* -0,008 0,287** 0,276** 0,050** 0,561** LEVERAGE -0,023** -0,037** 0,127** -0,003 -0,141** -0,002 0,016* -0,018* ROE 0,127* 0,143** 0,253** 0,148** 0,001 0,004 0,004 0,002 CFO -0,149** 0,206** 0,405** -0,167** 0,351** 0,069** 0,006 0,152** CRISISxBIG4 0,026 0,186** 0,281** 0,21** 0,011 0,118** 0,764** 0,527** CRISIS 0,026** 0,219** 0,043** 0,035** 0,004 0,029** 0,764** 0,089** BIG4 -0,035** 0,075** 0,576** 0,022** 0,053** 0,223** 0,527** 0,089**

*. Correlation is significant at the 0.05 level (2-tailed). **. Correlation is significant at the 0.01 level (2-tailed).

Above diagonal contains Pearson correlation, below contains Spearman correlation. Refer to appendix 1 for variable definitions.

(29)

Table 4.4 shows that size, leverage and CFO have a significant correlation with discretionary accruals for both the Pearson and Spearman test. This is in line with the literature as stated in paragraph 3.1.

When using the Pearson correlation, growth and return on investments are not significantly correlated with discretionary accruals, this is contrary to the research of Doukakis (2014). Intuitively this can be explained by the fact that both growth and return on equity are negatively affected by the financial crisis and therefore the correlation with discretionary accruals weakens. Although this intuitive

explanations is not substantiated by the correlations results as shown in table 4.4, as the crisis variable does not have significant correlation with either growth or return on equity. When Spearman correlation is used, the effects of growth and return on investment on discretionary accruals are significant.

Crisis and Big Four are the two dummy variables. In table 4.4 it shows that the financial crisis is positively correlated with discretionary accruals. Big Four is negatively correlated with discretionary accruals, this is in line with the research of Becker et al. (1998).

The dummy that CRISISxBIG4 does not have significant correlation with discretionary accruals. This is an early indication that companies from the sample that are in midst of the financial crisis and are audited by a Big Four auditor, do not show less earnings management than the rest of the sample. 4.3 Regression results

To research the two hypothesis stated in chapter tree and below, two regressions will be used. The first regression is used to answer the following hypothesis:

H1: Audit quality reduces the amount of earnings management used by managers during the financial crisis.

To test this hypothesis the research model as explained in paragraph 3.1 is used:

DAi,t= a0+a1GROWTHi,t + a2SIZEi,t + a3LEVERAGEi,t + a4ROEi,t+ a5CFOi,t+ a6BIG4i,t +

ε

i,t

Refer to paragraph 3.1 for the definition of the variables and the motivation to include these variables in the research model. The above stated model is regressed using SPSS. To select only the cases that are subject to the financial crisis, the dummy CRISIS is used as a selection variable.

(30)

Table 4.5 Coefficients H1 t Sig. B Std. Error (Constant) -0,007 0,002 -3,210 0,001*** GROWTH 0,000 0,000 2,882 0,004*** SIZE 0,002 0,000 4,433 0,000*** LEVERAGE _-0,006 _0,001 _-5,893 _0,000*** ROE 0,000 0,000 0,767 0,443 CFO _-0,051 _0,004 _-11,730 _0,000*** BIG4 -0,006 0,002 -3,100 0,002*** R2 adj N 0,020 8.406

*, **, and *** stand for statistical significance at 10%, 5%, and 1% levels, respectively. Refer to appendix 1 for variable definitions.

Table 4.5 shows an significant positive influence of growth on earnings management. This is in line with the results of Summers and Sweeney (1998) who state that managers use income-increasing accruals when growth slows down. As growth slows down due to the financial crisis, this result was expected.

Size also has a significant influence on earnings management during the financial crisis this is in line with the study of Watts and Zimmerman (1978). This indicates that even during the financial crisis, larger companies due to the greater analyses following and investor scrutiny show higher earnings management.

Leverage exhibits an significant influence on earnings management during the financial crisis. Intuitively this seems illogical due to the fact that during the economic crisis the risk of not meeting requirements in debt contracts because of bad results caused by the financial crisis is high. This might incentivize managers to manage their earnings to meet these requirements, but from the results found, the opposite seems true as leverage has a negative influence on earnings management.

ROE does not have an significant influence on earnings management. This is in contrast with the results found in Doukakis (2014). Intuitively this can be explained by the fact that companies are less profitable during the financial crisis. The relationship between profitability and earnings management then weakens.

CFO shows a significant impact on earnings management. This variable was included to control for the fact that in periods of extreme performance, estimated discretionary accruals will be to large(small) (Dechow et al., 1995). As the financial crisis can be logically described as a period with extreme (negative) performance, the significant impact was to be expected.

The dummy variable BIG4 is the variable that in the research model used, answers the first hypothesis. As can be seen in table 4.5, Big Four has an significant impact on earnings management. This

(31)

implies that companies that during the financial crisis are audited by a Big Four accountant, use less earnings management as measured true discretionary accruals then companies that are audited by non-Big Four auditors. This finding is in line with the research of Yew Ming et al. (2007) who found the same results for the Asian crisis. Based on the above stated, hypothesis one is supported. Audit quality in fact reduces earnings management during the financial crisis.

H2: The effect of audit quality on earnings management is less during the financial crisis, compared to the years before the financial crisis.

To second hypothesis will be answered in two steps. Firstly the same regression that was used to answer hypothesis one will be used but only the years prior to the crisis will be selected. Secondly a regression that covers all of the cases will be used including a dummy variable that separates the cases that are not subject to the financial crisis and are audited by a Big Four auditor.

As stated above, first the same regression as used to answer hypothesis one is used. Only

difference compared to hypothesis one is that this time only the cases that are not subject to the financial crisis will be selected. This is done by using the selection variable crisis. The results of this regression are shown in table. 4.6 Table 4.6 Coefficients H2 t Sig. B Std. Error (Constant) 0,004 0,002 1,943 0,052* GROWTH 0,000 0,000 0,551 0,581 SIZE 0,000 0,000 1,130 0,258 LEVERAGE -0,006 0,001 -3,832 0,000*** ROE 0,000 0,000 0,673 0,501 CFO -0,025 0,003 -7,434 0,000*** BIG4 -0,007 0,002 -3,387 0,000*** R2 adj 0,008 N 10.135

*, **, and *** stand for statistical significance at 10%, 5%, and 1% levels, respectively. Refer to appendix 1 for variable definitions

Just like in the crisis sample, ROE does not have a significant influence on earnings management. Although ROE was already non-significant in the crisis sample, size is only non-significant in the pre-crisis sample. These findings a in contrast with the findings of Lobo and Zhou (2006).

Growth is also only non-significant in the pre-crisis period. This in in contrast with the findings of Summers and Sweeney (1998). Intuitively this can be explained by the fact that in the years before the crisis the economic climate was more friendly then during the crisis. Before the crisis managers of

(32)

growing firms may have had less incentive to manager earnings, because growth was more easy to realize. Leverage shows little difference with the crisis sample. This implies that incentives to manage earnings to meet debt covenants requirements, do not differ substantially between the crisis and pre-crisis period.

Cash flows divided by total assets are slightly lower in the pre-crisis period as compared with the crisis period. This can be explained by the fact that CFO controls for the impact of extreme firm

performance on earnings management. During the financial crisis extreme (negative) firm performance is more likely than before the financial crisis.

Big four has a slightly stronger negative effect on earnings management in the pre-crisis period as to compared with the crisis period. This is in line with hypothesis two.

To examine if the difference between the effect of audit quality on earnings management in the crisis period compared to the pre-crisis period, the following regression formula from paragraph 3.1 is used:

DAi,t= a0+ a1GROWTHi,t + a2SIZEi,t + a3LEVERAGEi,t + a4ROEi,t+ a5CFOi,t + a6CRISISi,t +

a7BIG4i,t + a8CRISISxBIG4i,t +

ε

i,t

Table 4.7 Coefficients significance H2

t Sig. B Std. Error (Constant) -0,004 0,002 -2,293 0,022** GROWTH 0,000 0,000 0,937 0,349 SIZE 0,001 0,000 3,697 0,000*** LEVERAGE -0,006 0,001 -6,781 0,000*** ROE 0,000 0,000 0,786 0,432 CFO -0,035 0,003 -13,097 0,000*** CRISIS 0,008 0,002 3,228 0,001*** BIG4 -0,005 0,002 -2,667 0,008*** CRISISxBIG4 -0,003 0,002 -1,517 0,129 R2 adj 0,012 N 18.540

*, **, and *** stand for statistical significance at 10%, 5%, and 1% levels, respectively. Refer to appendix 1 for variable definitions

As becomes clear in table 4.7, the variable CRISISxBIGFOUR has little influence on earnings management measured by discretionary accruals. This implies that there is no significant difference between the effect of audit quality on earnings management for firms audited by an Big Four auditor in the pre-crisis period compared with the rest of the sample. Based on the above found results, hypothesis two is rejected. The effect of audit quality on earnings management is not significantly less for firms in the

The financial crisis and the effect of audit quality on earnings management

The financial crisis and the effect of audit quality on earnings management

Robin Gerssen- 10681000

Table of Contents

ε

ε