Indicating Financial
Statement Fraud – A test of
three indicators
Thesis MSc Accountancy 2018-2019
University of Groningen – Faculty of Economics and BusinessAuthor: Ruben Noorda Student Number: 2344548 Email: r.noorda@student.rug.nl Address: Couperusstraat 115 City: Groningen Supervisor: dr. K. Linke Date: 11-3-2019
Abstract
This examines whether there are patterns observable in financial statements that can indicate fraudulent reporting behavior. Therefore, we will look at sales growth, gross profit margin and days-sales-outstanding ratio. Using an independent samples t-test consisting of a sample of 105 fraud firms, we show that fraud firms report significantly higher sales growth in the three-year period before fraud and during the fraud period itself. Furthermore, we find that fraud firms have a higher days-sales-outstanding ratio in the period before fraud and during fraud. The study also attempts to find a relation between gross profit margin and financial reporting fraud. However, no significant relation could be found in this study.
Keywords: Financial statement fraud, Sales Growth, Gross Profit Margin, Days Sales
Table Of Contents
1. Introduction ... 4
2. Theory and Hypothesis Development ... 7
Agency Theory... 7
Incentives ... 7
The Slope towards Financial Statement Fraud ... 8
Beneish model ... 9 Hypothesis development ... 10 3. Research Method ... 13 Sample ... 14 Statistical model ... 15 Dependent variable... 15 Independent variables ... 16 Control Variables ... 16 4. Results ... 18 Descriptive statistics ... 18 Correlation analysis ... 20 Hypothesis testing ... 22 Control Variables ... 24 5. Discussion... 25 6. Conclusion ... 26 References ... 29
1. Introduction
Back in 2011, a large accounting scandal came to the light at the company Olympus. Management of Olympus succeeded to hide $1.7 billion of losses over a 13-year time period1.
Three years later, another recent comprehensive financial statement fraud case that revealed is the Tesco Accounting scandal, where management reported an amount of over £250 million non-existing profits in 20142. These examples are just a selection of recent accounting fraud,
serving as a proof that extensive financial statement fraud cases do still exist these days. Several measures like anti-fraud legislation and changes in the International Auditing
Standards did not result in financial reporting fraud risks to be mitigated (Silver et al., 2008). In fact, the extent in which financial reporting fraud risks develop is increasing (Beasley et al., 2010).
Financial statements are very important for different types of stakeholders, because they include information about financial performance and financial positions. Stakeholders like investors, creditors or shareholders rely on the information provided in financial
statements in business decision making (Beasley et al., 2010). If financial statements contain inaccurate or even manipulated information, stakeholders will rely on wrong information when making business decisions. Therefore, it is essential that firms provide correct information in their financial statements to inform stakeholders about the true financial performance and positions. Correct information in financial statement is an important condition to make securities markets work efficiently (Beaver, 1996; Beasley et al., 2010). Moreover, financial statement fraud will result in lower financial market efficiency,
increasing transaction costs (Perols & Lougee, 2011) and will result in significant damage to stakeholders (Chen et al., 2013).
Empirical studies in the area of financial reporting fraud already found multiple determinants of financial statement fraud. These determinants include corporate governance characteristics (Beasley, 1996), CEO compensation (Erickson et al., 2006) or audit quality (Becker et al., 2010). However, despite their usefulness to improve internal and external monitoring, these determinants lack usefulness to act as a red flag for indicating financial
1 More information: https://www.bbc.com/news/business-39741921
2 More information:
statement fraud as they are often lagging indicators and difficult to measure. Therefore, this research focusses on the development of fraud predictors being useful for practitioners such as auditors, analysts or investors to predict overvalued equity as described by Jensen (2005). Financial statement fraud and earnings management literature already managed to develop several detection models such as the F-score model (Dechow, 2011), the M-score model (Beneish, 1997) and the Jones-model (Jones, 1995). These models have higher usefulness as they could be calculated almost directly from financial statements. However, these models do not differentiate whether they measure earnings management or financial statement fraud. More recent research approaches financial statement fraud as a process developing over a longer period of time (Schrand & Zechman, 2012). In addition, Welsh et al. (2015) found that fraud develops over a larger period of time in a slippery slope from small unethical decision into larger fraudulent behavior. Following, Welsh et al. (2015), it could be useful to examine how these fraud paths could be characterized. Hence, developing fraud detection models where these characteristics will be used could improve existing models significantly.
Therefore, this research will examine whether there are differences in fraud indicators in these two periods.
Identifying how these fraud paths look like is useful by designing decision aids for detecting financial statement fraud (Dechow et al., 2011). Examples of decision aids are regression models, (Beneish, 1999), datamining (Green and Choi, 1997) or checklists (Asare and Wright, 2004). A big advantage of decision aids to detect financial reporting fraud is that they could be used by a broad amount of stakeholders and not only auditors or other
specialists. Decision aids are necessary in detecting financial statement fraud, as Humpherys et al. (2011) showed that human only have a slightly higher chance of discovering fraudulent financial reporting than random luck. Furthermore, auditors are less experienced in
discovering fraud because only a small percentage of them will get in touch with financial reporting fraud (Dechow et al., 2011). In addition, Beasley et al. (2010) found that auditors are actively involved in fraudulent behavior in 23 percent of financial reporting frauds, and decision aids could be used to close this detection gap. Lastly, the use of decision aids results in higher quality fraud risk assessment for auditors compared to auditors not using decision aids (Hogan et al., 2008).
Managers face pressure to maintain performance at the level expected by investors and targets of analysts (Jensen, 2005). Managers are willing to meet these expectations by using within-GAAP earnings management, because not meeting short-term earnings targets will
result in long-term punishment in the stock price of firms due to reputation loss (de Jong et al., 2013). In order to meet short-term benchmarks and not lose reputation, managers are willing to shift from within-GAAP earnings management to outside-GAAP earnings management (de Jong et al., 2013), resulting in managers to engage in a transgression from small unethical behavior towards fraud (Welsh et al., 2015). When looking at financial statement fraud, more revenues than costs are manipulated (Beasley et al., 2010).
Interestingly, literature in financial statement fraud indicators do not differentiate whether such predictors indicate fraud or indicate earnings management. Similarities or differences found between these two can be used to create models used in financial reporting fraud decision aids (Dechow et al., 2011). Therefore, how fraud could be indicated and modelized in financial statements is still a gap in recent financial statement fraud detection literature. This leads to the research question used in this thesis:
“How could reporting fraud be indicated directly from financial statements ?”
This study contributes to the literature in multiple ways. First, the study examines a broader period than existing studies in the are of financial reporting fraud literature. Where other studies do not include both the period during fraud and before fraud in their research or do not differentiate between these two (Beasley, 1996; Perols & Lougee, 2011; Dechow et al., 2011), this study does. As a result, this research provides evidence for the existence of patterns in the period before fraud and during fraud in financial statements. Second, this study finds an association between the days-sales-outstanding ratio and the presence of fraud in financial statements. Where many studies do not look further than linking accruals or accrual quality with financial statement fraud (Cohen et al., 2008; Wiedman & Hendricks, 2013), this study shows evidence that there also is a relation between the days-sales-outstanding ratio and financial statement fraud. Third, this thesis takes a different theoretical approach to financial statement fraud than existing research in the area of financial statement fraud research. Where other studies are anchored in using the fraud triangle of Cressey (1953) (Skousen et al., 2009; Finnerty et al., 2016), this study takes different approach with the use of the Agency Theory (Jensen and Meckling, 1976) and the Agency theory of overvalued equity theory of Jensen (2005).
In the remainder of this thesis, the hypothesizes will be elaborated supported by background theory. Thereafter, the research model and its results will be described, followed by
2. Theory and Hypothesis Development
Agency Theory
The Agency Theory (Jensen and Meckling, 1976) is used to explain the contractual relationship in which one party (the principal) hires another party (the agent) to delegate responsibility to the agent. An often used example of this contractual relationship, is when the shareholder (principal) hires a CEO or CFO (agents) to make important business decisions or run daily activities. Another situation is when a business-unit manager (agent) needs to do work on behalf of a senior executive (principal). The agent and the principal are expected to cooperate to accomplish the best outcomes for the firm, but this may not always be the case. The agent as well as the principal may have self-interest in business decision making
(Eisenhardt, 1985). Activities being low risk or even risk-free for managers themselves will result in self-interested behavior, obliging managers with paying subsequent costs as a result from managerial behavior (Fama and Jensen, 1983). This phenomenon is known as the agency problem and occurs when the agent and principal have conflicting goals towards certain situations. Prior research regarding the Agency theory already found that the agency problem leads to managers making operating decisions that are not in line with the firms’ best interest (Harrell and Harrison, 1994; Booth and Schulz, 2004). Information asymmetry
provides agents with the possibility to act in self-interest (Harrell and Harrison, 1994). Most of these decisions end up with a disadvantage for the principal, as the agent has the incentive and the possibility to bend decision making to an advantageous situation in perspective of the agent. In financial reporting decision making, this situation is not different. Many
organizational forces act as an incentive for managers to apply overly aggressive financial reporting strategies (Jensen, 2005). Therefore, agency theory provides a conceptual framework to explain financial reporting decisions (Baiman, 1990).
Incentives
When a firm is performing well, managers could genuinely share optimistic insights in financial statements, matching the expectations of market participants like shareholders. In this situation, the optimistic reflection in financial statements corresponds with underlying firm performance and investment decisions. However, at some point in time, management will be challenged to meet the exceptions with reported results in the financial statements of
the firm (Jensen, 2005). In this situations, managers start with “Gaming the System” and is the beginning of the overvaluation of a firms’ equity in financial statements (Jensen, 2005). The agency theory implies several motives for taking financial reporting decisions in managers’ self-interest. First, managers have direct benefit of continued growth and
overvaluation because of higher bonus rewards and higher valuated stock options (Cheng and Warfield, 2005; Healy, 1985). Second, managers are willing to perform well in the labor market to obtain future jobs. Good performance in the past is therefore very important for managers. Declining firm performance and disappointed shareholders will diminish managers position in the job market (Burns and Kedia, 2006). Thus, managers will try to prevent
shareholders from being disappointed and will be incentivized to overvalue equity (Burns and Kedia, 2006). Third, managers could be aware that there is information asymmetry in the financial reporting process and are therefore harder to monitor by shareholders and investors. In this way, managers could utilize information asymmetry to match expectations between shareholders and results as reported in financial statements. In the remainder of this section, we will describe how managers achieve matched expectations.
The Slope towards Financial Statement Fraud
A tool corporate managers could use to meet earnings expectations is earnings management. With this tool, revenues will be recognized earlier and expenses will be transferred to the future through accrual management or real earnings management (Jensen, 2005). Povel et al. (2007) found that firms with high financial performance feel pressurized by investors to maintain high financial performance in the future. When being pressurized by the capital markets, the main goal corporate managers have when managing earnings is to take away uncertainty by investors (Jensen, 2005). In order to achieve this, managers will try to show a constant firm growth in their financial statements. As a results of managing earnings, firms achieve higher shareholder value in comparison with firms not engaging in earnings
management strategies. Povel et al. (2007) states that firms do apply earnings management, even when reported earnings is not justified by investment decisions or expected future performance. Finnerty et al. (2016) finds evidence that management is incentivized to “let the good times rolling”. Earnings management has the characteristic that it is only effective at beating earnings expectations on a short term, as the inevitable accrual reserve reaction will decrease earnings. Using accruals to beat earnings expectations will lead to even bigger
expectation gaps in the future when firm growth is not sufficient (Allen et al., 2013) and consequently results in overvalued equity (Jensen, 2005).
When using earnings management is not sufficient to align earnings expectations between corporate managers and capital markets because within-GAAP flexibilities comes to an end, managers could make use of a second tool, which is outside-GAAP earnings management or financial statement fraud (Perols & Lougee, 2011; Jensen 2005). Jensen (2005) compares the use of earnings management with the use of managerial heroin. For corporate managers, it feels great when media, analysts and investors praise their achievements, but it works addictive (Jensen, 2005). Therefore, managers will continue to produce what these organizational forces like to see, namely continuous growth. To meet earnings targets, managers must violate the generally accepted accounting principles and exceed the limits of within-GAAP flexibility (Black et al., 2017). Jensen (2005) states that in this phase, equity is overvalued and performance needed to justify the reported equity can never be produced, except by pure luck. However, managers still believe that the required performance will be produced in the future due to managerial overconfidence (Schrand & Zechman, 2012). A slippery slope is created where incentives from the capital markets pushes management from managing earnings towards financial reporting fraud.
Beneish model
“Balance Sheet, Income Statement and the Statement of Cash Flows are interrelated and that is why certain numbers show up in financial statements when don’t make sense” (Joseph, 2001). This is exactly what Beneish (1997) intends to utilize in their earnings management classification model. The Beneish model is a mathematical model using eight different financial ratios to detect earnings management and financial statement fraud. The ratios used in this research are derived from the Beneish model (Beneish, 1997). The outcome of the model is the M-score, indicating the likelihood that a firm manipulated earnings. The ratios are directly constructed from firms’ financial statements. Furthermore, not only the M-score model could be used to make an indication of earnings management, the ratios used in the Beneish (1999) model could also be used in decision aids. The ratios in the Beneish model include sales growth, gross margin, and receivables versus sales. These three variables will be used in our thesis, for the reason that Beneish (1997) shows that they are dependent from each other when reporting sales in financial statements. When sales are manipulated, gross profit margin and accounts receivables should show deviant values as well (Beneish, 1997).
Hypothesis development Sales growth
Since the sales account is a primary target when managing earnings or committing financial statement fraud, sales patterns are also useful for the detection and prediction of financial statement fraud (Blocher & Cooper, 1988). Beneish (1997) argues that sales growth is useful in indicating earnings management. Growth firms suffer (or take advantage from it in a manager perspective) from higher information asymmetry providing managers the
opportunity to manipulate growth rates (Johnson et al., 2009). In addition, firms with rapid growth may suffer from a decrease in internal control effectiveness because they outgrow their internal control systems and may therefore be harder to monitor (Ettredge et al., 2010; Stice, 1991).
Summers and Sweeney (1998) found that growth differs significantly among fraud firms and non-fraud firms, because growth firms are under pressure from market participants like shareholders and investors to report continuous growth. Periods where growth is declining will therefore be manipulated to persuade market participants that firm growth is stable and meet optimistic shareholder expectations (Dechow 2011; Loebbecke et al., 1989; Bell et al., 1991). In addition, Stice et al. (1991) found that managers may manipulate sales growth with use of fraud or earnings management to maintain previous trends. The addictive feeling managers gain when managing earnings and satisfying investors will result financial fraud to let revenues stay constant, even when firm performance is not sufficient to do so (Jensen, 2005). Considering all the above, the following three hypothesis will be tested in this thesis.
H1: Fraud firms have a higher sales growth rate in the period before fraud H2: Fraud firms have a higher sales growth rate in the period during fraud H3: Fraud firms show a higher sales growth rate before and during fraud
Gross Profit Margin
Similar firms should have a similar gross margin percentage, as inventories, sales and cost of goods sold should be corresponding with each other (Fanning & Cogger, 1998). Deviant gross profit margins should therefore be an indication of financial statement fraud. Other studies also suggest that gross profit margin could be an indication of financial statement fraud for example Ravisanker et al. (2011) find that gross profit margin is often inflated by fraudulent firms to make financial statements look more impressive. Inconsistent gross profit margin ratios are a sign of fraudulent financial reporting behavior (Trotman & Wright, 2012).
Tonge at al. (2003) state that earnings is the most important determinant of firm value for shareholders nowadays. When earnings are at the level expected by shareholders, stock prices rise. While gross profit margin is a popular measure to not only calculate current earnings, but also predicting future earnings, it is therefore likely that managers manipulate earnings by increasing sales and lowering cost of goods sold resulting in a higher gross margin (Nichols, 2002; Jensen, 2005). Managers motivation to engage in earnings manipulation is to signal good financial health and future earnings towards shareholders and investors (Warshavsky, 2012). Manipulating earnings will be achieved with help of earnings management or financial statement fraud when GAAP-flexibility comes to an end (Jensen, 2005). As a result, the next three hypothesis are drawn up.
H4: Fraud firms have a higher gross profit margin in the period before fraud H5: Fraud firms have a higher gross profit margin in the period during fraud
H6: Fraud firms show an increased gross profit margin in the period before and during fraud
Days-sales-outstanding ratio
Firms with high accruals are higher valued by investors than similar firms with lower accruals due to the naïve fixation on accruals by shareholders and investors (Sloan,1996). Earnings persistence is therefore misjudged and firms with high accrued revenues may be overvalued by the market (Hirshlefier and Teoh, 2003). In addition, Sloan (1996) and Pincus et al. (2008) found that investors do not differentiate between cash-based earnings and accrual based-earnings. Managers may be aware of this phenomenon which is called “accrual anomaly” by Sloan (1996). Furthermore, the asymmetry between earnings and accruals will decrease
transparency in financial statements. Accrual anomaly may be prevalent due to information asymmetry between managers and shareholders. Following the accrual anomaly principle, investors possibly act in the same way concerning the relation accounts receivables and sales, giving managers the opportunity to increase sales through earnings management or financial statement fraud.
The incentive to increase accounts receivables could be explained by firm growth, as
mentioned earlier in this chapter. Therefore, days-sales-outstanding should not be considered as a separate target for financial statement fraud, but as a part of the reporting cycle where sales is part of. The asymmetry between the recognition of sales and the cash transaction should provide manager with the opportunity to commit fraud (Wu et al., 2010). The constant pressure by shareholders on managers to report growth could lead to reporting higher
accounts receivables (Zhang, 2007; Wu et al., 2010). Jensen (2005) states that managers use receivables to increase growth and beat market expectations. Hence, the last set of three expectations are as following.
H7: Fraud firms have a higher days-sales-outstanding ratio before fraud H8: Fraud firms have a higher days-sales-outstanding ratio during fraud
H9: Fraud firms show an increasing days-sales-outstanding ratio in the period before and during fraud
(Jensen & Meckling, 1976)
(Beneish, 1997)
Conflicting
goals
(Incentive)
Information
asymmetry
(Opportunity)
Sales
Growth
Gross
Margin
DSO
3. Research Method
This study will have a quantitative approach towards financial statement fraud. The reason behind the choice for this approach is the availability of data. Literature in the field of financial statement fraud indicators already made significant progress. Also, data availability is not a problem. Therefore, a quantitative approach towards the research question is chosen. The litigation release database of the SEC (Securities and Exchange Commission) has been selected as our data source for quantitative fraud data for several reasons. First, most of the financial statement fraud cases took place in the United States (Linke, 2012). Also, the data in the SEC database is publicly available, which makes our study verifiable and repeatable and prevents data availability problems from occurring. Furthermore, a SEC litigation release is often published with a complaint which contains detailed information about the fraud case. Lastly, the firms mentioned in the SEC litigation database could be considered as a fraud firm in a consistent way because they are alleged for committing financial statement fraud. In other words, the type 1 error (when a firm is considered as fraud firm while actually being
innocent) in the SEC litigation release database is very low (Perols and Lougee, 2011). Using SEC data is a method used by prominent fraud literature, for example Beasley (1999), Lennox & Pittman, (2010) and Dechow (2011). Beasley et al.,(2010) even states that SEC database is the most reliable and comprehensive data source available concerning financial statement fraud data.
The SEC litigation release database is our primary source for determining the fraud sample and is preferred over several other sources. For example (1) the SEC Accounting and Auditing Enforcement Releases (AAER), because SEC does not guarantee its completeness (Linke, 2012). The Financial Statement Restatement Database (2) of the Government Accountability Office (GOA) is also not used in this this study, despite its high amount of observations and overlap with the SEC database. The GOA database consists of firms with financial restatements which also could be misinterpretations of accounting rules and
unintentional misstatements (Beasley et al., 2010). Also, the GOA database doesn’t take any significance or materiality in consideration.
All other data (Sales Growth, Gross Profit Margin, and Days-Sales-Outstanding ratio, Auditor) will be collected through Compustat Global as provided by WRDS.
Sample
Our sample consists of hand collected data from firms in the SEC litigation release database initiating fraud in the period between 2000 and 2014. Due to a period of three years before fraud and the fact that fraud duration includes multiple the total observation period is between 1997 and 2015. This three-year period is consistent with Dechow et al. (2011) finding that fraud firms report abnormal high growth in the three years period before the initiation of fraud. Starting with a sample of 183 fraud firms, many firms in the SEC database where not suitable for our research. First, many samples did not contain the full string of observations during the fraud period or in the period three years before the initiation of the fraud, resulting in a sample of 129 fraud firms. Due to firms not having an appropriate match in the
Compustat database or the matched sample having incomplete data, a final sample of 105 fraud firms is achieved. This sample size is similar to other studies using similar data
(Beasley, 1996; Feroz et al., 2000; Perols & Lougee, 2011) and should therefore be sufficient to perform the research. We will add a matched sample for each sample in the SEC database. That will bring our total sample size to 210. The matching process will be based on the variables firm size and industry type, where the industry must be an exact match on all four digits of the SIC-industry codes and the nearest match on total assets wich may not be greater than 100% of the fraud firm and not be smaller than 50% of the fraud firm. This matching procedure is confirmed to be appropriate by Ettredge et al. (2006), Beneish (1997) and Bayley and Taylor (2007). Also, Dechow (2011) claims that industry and firm size are appropriate matching criteria in financial statement fraud research. The fraud period as indicated in the SEC complaints will be used to determine the observation period for the fraud firm as well as the observation period for the corresponding matched sample.
The definition of Beasley et al. (2010) will be used when selecting financial statement fraud samples from the SEC litigation release database. This definition is as following:
“…the intentional material misstatement of financial statements or financial disclosures or the perpetration of an illegal act that has a material direct effect on the financial statements or financial disclosures.”
This means any cases where there is sign of self-enrichment should be excluded from our data sample.
Collecting the fraud samples was a shared effort among our thesis group. A list of 230 SEC complaints provided by University of Groningen was divided over five students. We checked
all complaints whether they met de conditions described above. If not, they were removed from the sample, resulting in a sample of 183 fraud firms.
Statistical model
The following statistical model will be used in this study. The model is estimated using the statistical methods shown in table 1.
FRAUD = b0 + b1 SALESG_Y + b2 SALESG_Q + b3 SALESG_QY + b4 GROSSM_Y + b5 GROSSM_Q + b6 DSO + b7 BIG4 + b8 FINCRIS + e
bi represents a variable from our research model and e represents the error. Each variable is explained hereafter in this section.
Hypothesis Type Data Method
FRAUD Dependent Dummy -
SALESG_Y H1, H2, H3 Independent Interval Independent samples t-test
SALESG_Q H1, H2, H3 Independent Interval Independent samples t-test
SALESG_QY H1, H2, H3 Independent Interval Independent samples t-test
GROSSM_Y H4, H5, H6 Independent Interval Independent samples t-test
GROSSM_Q H4, H5, H6 Independent Interval Independent samples t-test
DSO H7, H8, H9 Independent Interval Independent samples t-test
BIG4 Control Dummy Logistic Regression
FINCRIS Control Dummy Logistic Regression Table 1: Statistical methods used
Dependent variable
Financial Statement Fraud
Financial statement fraud will be our dependent variable. The variable will be measures as a dummy, 0 will indicate there is no financial statement fraud within a firm, and 1 will indicate a fraud firm. The variable is shown in the tables as FRAUD.
Independent variables
Sales Growth
Sales growth is measured in three different periods. These include annual growth
(SALESG_Y), quarterly growth (SALESG_Q) and quarterly growth relative to the previous year (SALESG_QY). The quarterly and annual measurement are consistent with Dechow, (2011) creating an annual and quarterly dataset for each variable.
Gross Profit Margin
Gross Profit Margin is calculated by dividing gross profit by revenues. This variable is measured in two different periods, namely annual (GROSSM_Y) and quarterly
(GROSSM_Q). Again, quarterly and annual variables are derived from Dechow (2011).
Days-sales-outstanding
Days-sales-outstanding ratio is calculated by dividing accounts receivables at the end of the fiscal year by the total amount of annual sales. The variable is shown in tables as DSO.
Control Variables
BIG4 auditor
BIG4 auditors have a higher incentive to detect fraudulent financial statement from their clients due to BIG4 auditors to prevent reputation damage and litigation costs
(DeAngelo,1981). In addition, Francis (2004) found that BIG4 audit firms provides provide better audit services compared to non-BIG4 audit firms. Lennox and Pittman (2010) found a negative significant relation between BIG4 audits and the occurrence of financial statement fraud. In this study, this effect will be controlled by using BIG4 as a control variable. Each firm is coded with a zero (0) for not being audited by a BIG 4 audit firm (KPMG, PwC, EY, Deloitte, Arthur & Andersen or one of its predecessors). BIG 4 audit firms are coded with a 1 for the variable BIG4.
Financial crisis
The second control variable in our study is the financial crisis. During the financial crisis (2008-2009) performance indicators where declining rapidly (Moore, 2016). During such financial distress firms high are more likely to have financial statements (Liou, 2008).
Performance indicators like sales, assets, profit or expenses where manipulated to the desired value to satisfy shareholders (Doukakis, 2010). Due to the crisis years 2008-2009 being part
of our research sample, we control for this effect. Observations are labeled with a dummy variable 1 when reported during times of financial crisis, and 0 when not reported during the financial crisis. This variable is coded with FINCRIS in tables.
4. Results
In the results section, the descriptive statistics, a correlation analysis and the hypothesis testing will be presented.
Descriptive statistics
The minimums, maximums, sample size, means and standard deviations of the total sample are presented in Table 2. Means and standard deviations are reported in the period before fraud, during fraud and during both periods combined. In this way, descriptive statistics correspond with the hypothesis structure as shown earlier in this thesis. In Table 3, descriptive statistics of the fraud and non-fraud sample are separated and compared to each other. Again, three different time periods are reported. Furthermore, the descriptive values of non-fraud and fraud firms are separated to reveal differences between these two samples if present.
Table 2: Descriptive statistics in the period before fraud1, during fraud2 and both periods together3 TOTAL
SAMPLE N MIN MAX MEAN ST.DEV.
FRAUD 210 0 1 0,50 0,5 SALESG_Y 210 -0,988 8,693 0,334 1 0,1942 0,2773 0,9151 0,7242 0,8443 SALESG_Q 210 -8,311 8,455 0,101 1 0,0652 0,0863 0,6241 0,4562 0,5603 SALESG_QY 210 -8,017 12,705 0,367 1 0,2341 0,3073 1,2671 1,0442 1,1723 GROSSM_Y 210 -12,567 0,988 0,283 1 0,3302 0,3023 0,9841 0,6952 0,8773 GROSSM_Q 210 -13,823 5,320 0,232 1 0,2892 0,2553 1,2421 0,7832 1,0753 DSO 210 0 280,2 69,4 1 63,12 66,73 61,31 56,512 59,43 BIG4 210 0 1 0,75 0,432 FINCRIS 210 0 1 0,11 0,311
Table 3: Descriptive statistics of the sample separated in non-fraud and fraud firms in the period before fraud1,
during fraud2 and both periods together3
NON-FRAUD FIRMS
FRAUD FIRMS
VARIABLES N MEAN N MEAN DIFFERENCE
FRAUD 105 0 105 1 0 SALESG_Y 105 0,2241 0,1352 0,2113 105 0,4451 0,2522 0,3663 0,2211 0,1172 0,1553 SALESG_Q 105 0,0711 0,0572 0,0653 105 0,1291 0,0742 0,1063 0,0581 0,0172 0,0413 SALESG_QY 105 0,2321 0,1842 0,2113 105 0,4971 0,2842 0,4033 0,2651 0,1002 0,1923 GROSSM_Y 105 0,282 0,3032 0,2913 105 0,2841 0,3572 0,3133 0,0041 0,0542 0,0223 GROSSM_Q 105 0,2621 0,2712 0,2663 105 0,2011 0,3072 0,2453 -0,0611 0,0362 -0,0213 DSO 105 64,71 59,82 62,73 105 73,91 66,22 70,83 7,21 6,4,2 8,13 BIG4 105 0,82 105 0,68 -0,14 FINCRIS 105 0,11 105 0,11 0
Correlation analysis
The correlation matrix is shown in Table 4. The correlation method used is the Pearson Correlation Coefficient (two-tailed).
SALESG_Y shows positive significant relation with FRAUD at the 0,01 level before fraud, during fraud and in both periods together. SALESG_Q and SALESG_QY show positive significant relations with FRAUD in the period before fraud and the total observation period. The significance levels of SALESG_Q are at the 0,05 level while SALESG_QY shows a significance level of 0,01 in both periods. GROSSM_Y and GROSSM_Q do show varying directions and not significant relations with FRAUD. DSO, on the other hand shows positive significant relations in all periods at the 0,01 level.
The control variables BIG4 shows a negative significant relation with FRAUD in all three periods at the 0,01 level, in contrast to FINCRIS which doesn’t show any significant relation with FRAUD.
Multi collinearity problems do not occur in this model. Tabachnik & Fidell (2013) state that as long as correlation values stay below 0,9, the assumption that multicollinearity problems do not arise could be made.
FR A U D S A L E S G _Y SA L E SG _Q SA L E SG _Q Y G R O SS M _Y G R O SS M _Q DS O BI G 4 FI N C R IS FRAUD Pre During Total 1 1 1 SALESG_Y Pre During Total ,121** ,081** ,105** 1 1 1 SALESG_Q Pre During Total ,046* ,018 ,036* ,308** ,298** ,306** 1 1 1 SALESG_QY Pre During Total ,105** ,048 ,082** ,610** ,753** ,661** ,304** ,259** ,290** 1 1 1 GROSSM_Y Pre During Total ,001 ,039 ,013 -,099** -,326** -,163** -0,38 -,171** -,074** -,018 -,254** -,099** 1 1 1 GROSSM_Q Pre During Total -0,025 ,022 -,010 -,165** -,302** -,201** -,029 -,106** -,050** -,009 -,199** -,068** ,877** ,867** ,872** 1 1 1 DSO Pre During Total ,074** ,057* ,068** ,177* ,060* ,140** ,087** ,007 ,064** ,088* ,057* ,079** -,146** -,041** -,113* -,162** -,060** -,132** 1 1 1 BIG4 Pre During Total -,184** -,140** -,166** -,163** -,188** -,170** -,069** -,055* -,064** ,100** -,142** -,116** ,071** ,099** ,079** ,056** ,125** ,076** -,121** -,116** -,120** 1 1 1 FINCRIS Pre During Total -,004 ,005 ,000 -,080 -,094** ,000 ,041 ,001 ,022 ,105** -,080** ,017 -,022 ,949 -,011 ,039 ,015 -,016 -,014 -,056* -,036* ,026 -,031 -,028 1 1 1
Hypothesis testing
In table 5, the independent samples t-test is presented showing the hypothesis testing of all variables. According to these results, the hypothesis will be rejected or accepted.
Hypothesis 1: Our results show significant differences between fraud and non-fraud firms in
sales growth percentages. With M=0,45 for fraud firms and M=0,22 for non-fraud firms for annual sales growth (SALESG_Y) under the conditions; t(210)=5,72, p=0,000 the hypothesis is accepted when looking at annual data. The hypothesis is also accepted for the quarterly data with M=0,13 for fraud firms and M=0,07 for non-fraud firms under the conditions;
t(210)=2,168, p=0,030 a significant difference is shown in our sample. Lastly, the hypothesis is also accepted when looking at quarterly data relative to previous years. With M=0,50 for fraud firms and M=0,23 for non-fraud firms, fraud firms show significantly more sales growth with conditions t(210)=4,72, p=0,000.
Table 5: Results of the independent samples t-test. *=significant at the ,10 level, **=significant at the ,05 level, ***= significant at the ,01 level.
VARIABLES
t
df
Sig. (2-t)
S.E.
SALESG_Y Before During Total 5,714 3,172 6,479 210 210 210 ,000*** ,002*** ,000*** ,039 ,037 ,027 SALESG_Q Before During Total 2,168 ,717 2,226 210 210 210 ,030** ,473 ,026** ,026 ,023 ,018 SALESG_QY Before During Total 4,721 1,908 4,937 210 210 210 ,000*** ,057* ,000*** ,056 ,052 ,039 GROSSM_Y Before During Total ,036 1,527 -,781 210 210 210 ,971 ,127 ,527 ,041 ,035 ,029 GROSSM_Q Before During Total -1,182 ,899 -,663 210 210 210 ,237 374 ,435 ,052 ,039 ,034 DSO Before During Total 3,499 2,221 4,158 210 210 210 ,000*** ,026** ,000*** 2,601 2,880 1,934
Hypothesis 2: Significant differences are found for the fraud and non-fraud samples in the
period during fraud. The annual sales growth shows M=0,25 for fraud firms and M=0,16 for non-fraud firms under conditions t(210)=3,172, p=0,002. Therefore, hypothesis 2 will be accepted for annual sales growth. Quarterly sales growth (SALESG_Q) does not show a significance difference with M=0,06 for fraud firms and 0,06 for non-fraud (t(210=0,72, p=0,473). In contrast, quarterly sales growth relative to the previous year (SALESG_QY) does show significant results (M=0,28 for fraud firms, M=0,18 for non-fraud firms,
conditions t(210=1,908, p=0,057) and therefore hypothesis 2 is accepted when analyzing this data.
Hypothesis 3:In the total period (before fraud and during fraud periods combined), differences
are significant for the annual sales with M=0,37 for fraud firms and M=0,21 for non-fraud firms (t(210)=6,479, p=0,000. Therefore, hypothesis 3 will be accepted regarding annual data. Hypothesis 3 will also be accepted for quarterly data (SALESG_Q) with M=0,11 for fraud firms, M=0,07 for non-fraud firms, conditions (t(210)=2,23, p= 0,026, and for
quarterly data relative to the previous year (SALESG_QY) with M=,040 for fraud firms and M=0,21 for non-fraud firms with conditions t(210)= 4,97, p=0,000.
Hypothesis 4: No significant difference in our sample in the period before fraud is found for
annual gross profit margin (GROSSM_Y) with M=0,28 for fraud firms and M=0,28 for non-fraud firms (t(210=1,53, p=0,971). The same occurred with quarterly gross profit margin with M=0,20 for fraud firms and M=0,26 for non-fraud firms. As a result, hypothesis 4 will be rejected for both GROSSM_Y and GROSSM_Q.
Hypothesis 5:In the period during fraud, no significant differences are found for annual gross
profit margin and quarterly gross profit margin. GROSSM_Y shows M=0,36 for fraud firms and M=0,30 for non-fraud firms under conditions t(210)=,036, p=0,127. GROSS_Q shows M=0,31 for fraud firms and M=0,27 for non-fraud firms (t(210)=0,90, p=0,127. Therefore, hypothesis 5 will be rejected for both GROSSM_Y and GROSSM_Q.
Hypothesis 6: With M=0,31 for fraud firms and M=0,29 and conditions t(210)=-0,72,
p=0,527, no significant difference in our sample is found for annual gross profit margin (GROSSM_Y) in the total period. The same applies for the quarterly gross profit margin with M=0,25 for fraud firms and M=0,27 for non-fraud firms under conditions t(210)=-0,55 and p=0,435. Hypothesis 6 will therefore be fully rejected for both GROSSM_Y and
Hypothesis 7: Days-sales-outstanding ratio (DSO) shows significant difference in the period
before fraud with M=73,9 for fraud firms and M=64,7 for non-fraud firms under conditions t(210)=3,50 and p=0,000. As a result, hypothesis 7 will be accepted.
Hypothesis 8:In the period during fraud, a significant difference in means for fraud firms
(M=66,2) and non-fraud firms (M=59,8) is found under conditions t(210)=2,22 and p=0,026. Therefore, hypothesis 8 will be accepted.
Hypothesis 9: With M=70,8 for fraud firms and M=62,7 for non-fraud firms and conditions
t(210)=4,16 and p=0,000, a significant difference in means is found for days-sales-outstanding ratio (DSO). Hence, hypothesis 9 will be accepted.
Control Variables
Hereafter, the control variables BIG4 and FINCRIS are tested for significance. Results are shown in table 6.
Variable
B
NAGELKERKE-R-SQUARE
Sig.
BIG4 -,788 ,037 ,000*** FINCRIS -,032 ,000 ,755Table 6: Logistic regression
The control variable BIG4 shows a very significant difference between fraud firms (p > 0,01) with FRAUD. This significance supports the results found in the correlation matrix which shows a positive significant relation with FRAUD. The relation between BIG4 and FRAUD is not surprising as it can be supported by a great amount of literature in the area of fraudulent reporting. For example, Lennox & Pittman (2010) and Lin et al. (2010) found a negative relation between BIG4 audits and financial statement fraud. Also, Paik et al. (2018) found the same effect when using BIG4 auditor as a control variable. The other control variable
5. Discussion
The main goal of this research was to examine whether fraud could be indicated directly from financial statements in the period before fraud and during fraud. Three fraud indicators were tested for significance, namely sales growth, gross profit margin and days-sales-outstanding ratio. The main research question was as following:
“How could reporting fraud be indicated directly from financial statements ?”
Our findings show that sales growth is a significant indicator of financial statement fraud in the period before fraud and during the fraud period itself. Firms with high sales growth are harder to monitor and have lower internal control effectiveness, providing growth firms with the opportunity to manipulate sales growth (Ettredge et al., 2010). Furthermore, managers are under pressure from shareholders and investors to report continuous growth in financial statements (Loebbecke, 1989; Finnerty et al., 2016) . As a result, sales growth rates are higher in the period before fraud and the period during fraud in relative to with non-fraud firms. This finding is consistent with results of Finnerty et al., (2016) finding that firms report higher stock returns before fraud “to keep the good times rolling”. This implication could mean that managers are under pressure to satisfy shareholders. Therefore, shareholders and investors should be aware that higher sales growth rates are not by definition a good sign. High sales growth can indicate that managers are under pressure to report high sales growth to satisfy stakeholders. Stakeholders should be careful by interpreting high sales growth, as our results show that such high sales growth (our results show that fraud firms report sales growth two times as high relative to non-fraud firms) could be an indication of financial statement fraud. Financial statement fraud will result in overvalued equity and may cause destruction of shareholder value (Jensen, 2005). Furthermore, our paper shows that both before fraud and during growth sales growth are significantly higher concerning annual sales growth. This finding implies that managers do not have different intentions before fraud and during fraud, and financial statement fraud may be the consequence of the slippery slope (Schrand & Zechman, 2012). Therefore, our results could be seen as criticism on the use of (aggressive) earnings management strategies. Our results show that reported sales growth rates before and during fraud are very close to each other and significantly higher at fraud firms. While using earnings management, managers may be triggered to look for the edge of what is possible with the use of earnings management and may suffer from overconfidence (Schrand & Zechman, 2012).
In this thesis, no differences in the reporting of gross profit margins between fraud firms and non-fraud firms show up. A possible explanation for the absence of the relation between gross profit margin and fraud may be that firms are aware that patterns may be visible too obvious, resulting in managers trying to conceal this pattern by also increasing costs (Trotman & Wright, 2012). In this way, any suspicious behavior is hidden, making the detection of fraud less likely. However, the literature seems to provide reason to assume that the relation does exist (Trotman & Wright, 2012; Fanning & Cogger, 1998). Other variables regarding gross profit margin could therefore be able to help finding significant relations between fraud and gross profit margin. Due to the inevitable accrual reverse response, high relative changes (both positive or negative), or higher relative differences between the lowest and highest gross profit margin may be better in indicating financial statement fraud instead of absolute growth. It is still an empirical question whether gross profit margin could act as fraud indicator.
Lastly, this study examined DSO-ratio as an indicator for financial statement fraud. Both the period before and the period during fraud show significant differences in comparison to non-fraud firms. The accounts receivable account makes it possible to inflate revenues, for example the sales account. Therefore, a higher DSO-ratio is consistent with the finding that sales growth is increased at fraud firms. As stated earlier, in financial statements numbers may pop-up when they don’t make sense and that is exactly what happens with DSO-ratio (Wells, 2001). Managers have the opportunity to do so because of the asymmetry in recognition of revenues (Sloan, 1996). Noted should be that the DSO-ratio should not be recognized as a separate target in financial statement fraud, but as a tool to increase sales growth. This parallel is shown in our results as both DSO-ratio and sales growth show significant higher levels at fraud firms. Therefore, DSO-ratio will work as a fraud indicator.
6. Conclusion
The aim of this thesis was to find evidence for three different indicators of financial statement fraud in firms’ financial reports. These indicators include sales growth, gross profit margin and days-sales-outstanding ratio. These indicators are derived from the Beneish model (1997). Additional research to fraud indicators is needed, as financial statement fraud is still a
problem causing potential damage to stakeholder (Chen et al., 2013). In order to detect patterns of fraudulent behavior at firms, a sample of 105 fraud firms is tested in the period 1999-2014 using a matched pair design. Our results show that sales growth is higher before
fraud and during the fraud period. Also, days-sales-outstanding ratio is higher in both the period before fraud and during the fraud period.
There are multiple limitations in this research. First, this study is likely to suffer from a selection bias. This could be a problem, as a selection bias may result in unreliable statistics due to missing data (Hernan et al., 2004). The selection bias originates from two sources. The first source is the SEC database. As the SEC wants to let the court punish individuals and firms for committing fraudulent behavior, it is plausible that they search for the cases with the most amount of evidence and visibility of fraud because the SEC does not have enough recourses to investigate all financial statement fraud cases. This means that not all fraud cases are included in this study, only the ones that are identified by the SEC. The characteristics of the missing fraud cases could not be observed and may differ from the results shown in our study, which may affect the generalizability. Second, because data of Compustat is used, not all firms may be represented in our dataset because Compustat does not contain data from smaller firms. Therefore, data from smaller firms may be missing in our dataset which may again affect the generalizability of our study. Another limitation of our study is the
subjectivity in the data collection. Despite the use of definitions by selecting fraud cases, data collection is done by multiple persons and therefore multiple interpretations of fraud may be used by the collection of samples.
Theoretical and practical implications
This thesis contributes to the literature in multiple ways. First, this thesis finds additional evidence for the M-score model (Beneish, 1997). Second, the paper managed to demonstrate that earnings management and financial statement fraud indicators in financial statement do not show differences. This has been achieved by differentiating between the fraud period itself and the period prior to the fraud. Prior literature in financial statement fraud did include the period prior to the fraud and the period during fraud, however the literature didn’t
examine whether there are different characteristics or similarities in fraud indicators during these two periods. Therefore, this paper adds to the literature showing how fraud could be indicated in the period before fraud and during fraud.
Financial statement fraud is still an ongoing issue in the current financial market. Therefore, the outcomes of this study might be useful for multiple stakeholders to help detecting financial statement fraud and prevent financial losses.
First, investors could take advantage of this study. With help of our results, it is possible to declare early warnings or red flags to the investment markets. Potential investors could evaluate potential fraud risks and ask themselves the question whether the investment is very lucrative or too good to be true. Furthermore, current shareholders could evaluate the fraud risks in their investment portfolio and decide if fraud is likely to be prevalent.
Second, the audit firms could use our results to red flag potential fraud firms. With help of data analytics, fraud firms should pop up when firms report suspicious numbers in
comparison to industry average or earlier reported numbers. Due to human limitations and the often limited fraud recognition experience of auditors, our findings might help to increase the fraud detection rate dramatically.
Future research
To improve generalizability of the results shown in this research, further research is needed. First, this paper only tested US samples. To generalize results, future research should include outside-US firms, for example European firms or Asian firms. Second, the scale of the study could be improved. Now that there is evidence that sales growth and DSO-ratio could predict fraud, it could be worth the effort to gather a larger amount of samples to improve
generalizability.
Furthermore, this sample only studied three out of eight predictors for financial statement fraud from the M-score model (Beneish, 1997). The other five predictors are still not tested whether they indicate fraud in the period before fraud and during fraud. Studying the
remaining five predictors from the M-score model will provide additional contribution to our main research question.
References
Allen, E., Larson, C., Sloan, R. (2013). Accrual reversals, earnings and stock returns. Journal
of Accounting and Economics, 56(1), 113-255.
Asare, S., Wright, A.M. (2004). The Effectiveness of Alternative Risk Assessment and Program Planning Tools in a Fraud Setting. Contemporary Accounting Research, 21, 325-352.
Baiman, S. (1990). Agency research in managerial accounting: A second look. Accounting,
Organizations and Society, 15(4), 341-371.
Bayley, L., Taylor, L.S. (2007). Identifying Earnings Overstatements: A Practical Test. SSRN
Electronic Journal.
Beasley, M. S. (1996). An Empirical Analysis of the Relation between the Board of Director Composition and Financial Statement Fraud. The Accounting Review, 71(4), 443-465.
Beasley, M. S., Carcello, J. V., Hermanson, D. R., & Committee of Sponsoring Organizations of the Treadway Commission. (1999). Fraudulent financial reporting: 1987-1997: An analysis of U.S. public companies : research report.
Beasley, M. S., Carcello J. V., Hermanson, D. R., Neal T. L. (2010). Fraudulent financial reporting 1998 – 2007: An analysis of U.S. public companies. Committee of Sponsoring Organizations of the Treadway Commission (COSO).
Beaver, W. (1996). Financial ratios as predictors of failure. Journal of Accounting Research (4), 71-111.
Becker, C. L., Defond, M. L., Jiambalvo, J., Subramanyam, K. (2010). The Effect of Audit Quality on Earnings Management. Contemporary Accounting Research, 15(1), 1-24. Bell, T. B.; Szykowny, S., Willingham, J. J. (1991). Assessing the likelihood of fraudulent financial reporting: a cascaded logit approach. working paper, KPMG Peat Marwick, Montvale, NJ
Beneish, M. (1997). Detecting GAAP violation: Implications for assessing earnings management among firms with extreme financial performance. Journal of Accounting and
Public Policy, 16, 271−309.
Beneish, M. (1999). The Detection of Earnings Manipulation. Financial Analysts
Journal, 55(5), 24-36.
Black, E. T., Christensen, T., Schmardebeck, R. (2017). The relation between earnings management and non-GAAP reporting. Contemporary Accounting Research, 34(2), 750-782. Blocher, E., J. C. Cooper. (1988). A Study of Auditors' Analytical Review Performance.
Booth, P., Schulz, A.K.D. (2004). The Impact of an Ethical Environment on Managers' Project Evaluation Judgements under Agency Problem Conditions. Accounting,
Organizations and Society. (29), 473-488.
Burns, N., Kedia, S. (2006). The impact of performance-based compensation on misreporting.
Journal of Financial Economics (79), 35-67.
Chen, J.; Cumming, D.; Hou, W.; Lee, E. (2013). Executive integrity, audit opinion, and fraud in Chinese listed firms. Emerging Markets Review. (15), 72-91.
Cheng, Q., Warfield, T.D. (2005). Equity incentives and earnings management. The
Accounting Review (80), 441-476.
Cohen, D.A., Dey A., Lys T.Z. (2008). Real and Accrual-Based Earnings Management in the Pre- and Post-Sarbanes-Oxley Periods. The Accounting Review: 83(3), 757-787.
Cressey, D. R. (1950). The criminal violation of financial trust. American Sociological
Review, 15 (6), pp.738-743, December, pp.1-15.
DeAngelo, L. (1981). Auditor size and audit quality. Journal of Accounting and Economics 3(3), 183–99
Dechow, P., G, W. L., Sloan, C., R.G. (2011). Predicting material accounting misstatements.
Contemporary Accounting Research, 42(1), 1-36.
Doukakis, L. (2010). The persistence of earnings and earnings components after the adoption of IFRS. Managerial Finance. 36(11), 969-980.
Eisenhardt, K. M. (1989). Agency theory: An assessment and review. Academy of
Management Review 14: 57-74.
Erickson, M., Hanlon, M., Maydey, E. L. (2006). Is There a Link between Executive Equity Incentives and Accounting Fraud? Journal of Accounting Research, 44(1), 113-143.
Ettredge, M., Li, C., Sun, L. (2006). The impact of SOX Section 404 internal control quality assessment on audit delay in the SOX era. Auditing: A Journal of Practice & Theory, 25(2), 1-23.
Ettredge, M., Scholz, S. , Smith, K. R., Sun, L. (2010). How Do Restatements Begin?
Evidence of Earnings Management Preceding Restated Financial Reports. Journal of Business Finance & Accounting, 37: 332-355.
Fama, E.F. and Jensen, M.C. (1983). Separation of Ownership and Control. Journal of Law
and Economics. (26), 301- 325.
Fanning, K., Cogger, K. (1998). Neural network detection of management fraud using published financial data. International Journal of Intelligent Systems in Accounting, Finance
Liou, F.M. (2008). Fraudulent financial reporting detection and business failure prediction models: a comparison. Managerial Auditing Journal, 23(7), 650-662.
Feroz, E. H., Kwon, T. M., Pastena, V. S., Park, K. (2000). The efficacy of red flags in predicting the SEC's targets: an artificial neural networks approach. Intelligent Systems in
Accounting, Finance & Management, 9(3), 145-157.
Finnerty, J., Hegde, S., Malone, C. (2016). Fraud and firm performance: keeping the good times (apparently) rolling. Managerial Finance (42), 151-172.
Francis, J. R. 2004. What do we know about audit quality? The British Accounting Review 36(4): 345–68
Green, Brian & H. Choi, J. (1997). Assessing the Risk of Management Fraud Through Neural Network Technology. Auditing A Journal of Practice & Theory. 16(1), 18-28.
Harrell, A. and Harrison., P.D. (1994). An incentive to shirk, privately held information, and managers’ project evaluation decisions. Accounting, Organizations and Society 19, 569-577. Healy, P.M. The Effect of Bonus Schemes on Accounting Decisions (1985). Journal of
Accounting & Economics, 7, 85–107.
Hernán, M. A., Hernández-Díaz, S., Robins, J. M. (2004). A structural approach to selection bias. Epidemiology, 615-625.
Hirshlefier, D., Teoh, S., (2003), Limited attention, information disclosure, and financial reporting. Journal of Accounting & Economics 36, 337-386.
Hogan, C.E., Rezaee, Z., Riley, R.A., Velury, U. (2008), Financial Statement Fraud, Insights from the academic literature. Auditing: A Journal of Practice & Theory, 27, 231- 252. Humpherys, S. L., Moffitt, K.F., Burns, M.B., Burgoon, J.K., Felix., W.F. (2011).
Identification of fraudulent financial statements using linguistic credibility analysis. Decision
Support Systems 50(3), 585-594.
Jensen, M. (2005). Agency Costs of Overvalued Equity. Financial Management, 34(1), 5-19. Jensen, M. C., Meckling, W. H. (1976). Theory of firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics, 3(4), 305-360.
Johnson, S.A., Ryan, H.E., Tian, Y.S. (2009). Managerial Incentives and Corporate Fraud: The Sources of Incentives Matter, Review of Finance, 13(1) 115–145.
Jones, J.J. (1991) Earnings Management during Import Relief Investigations. Journal of Accounting Research, 29, 193-228.
Jong de, A., Mertens, G., Poel van der, M., & Dijk van, R. (2013). How does earnings management influence investor’s perceptions of firm value? Survey evidence from financial analysts. Review of Accounting Studies, 19(2), 606-627
Lennox, C., Pittman, J. A. (2010). Big Five audits and accounting fraud. Contemporary
Accounting Research, 27(1), 209-247.
Lin, J. W., Hwang, M. I. (2010). Audit Quality, Corporate Governance, and Earnings Management: A Meta-Analysis. International Journal of Auditing, 14: 57-77.
Linke, K. G. B. (2012). Finance middle managers and financial statement fraud Groningen: University of Groningen, SOM research school
Loebbecke, J. K., Eining, M. M. ,Willingham, J. J., (1989). Auditors’ experience with material irregularities: frequency, nature, and detectability auditing. Journal of practice
theory, 9, 1-28
Moore, T., Mirzaei, A. (2016). The impact of the global financial crisis on industry growth. The Manchester School, 84(2), 159-180.
Nichols, M. E. (2002). Discussion of "The quality of accruals and earnings: The role of accrual estimation errors. The Accounting Review, 11 (Supplement), 61-69
Paik, D. G., Kim, T., Krumwiede, K., Lee, B. B. (2018). The Association Between Audit Fees and Accounting Restatement Resulting from Accounting Fraud and Clerical Errors. Journal
of Forensic & Investigative Accounting, 10(3), 332-356.
Perols, J. L., Lougee, B. A. (2011). The relation between earnings management and financial statement fraud. Advances in Accounting, 27(1), 39-53.
Pincus, M., Rajgopal, S., Venkatachalam, M. (2007). The accrual anomaly: International evidence. The Accounting Review, 82(1), 169-203.
Povel, P., Singh, R., Winton, A. (2007). Booms, busts, and fraud. Review of Financial
Studies, 20(4), 1219-1254.
Ravisankar, P., Ravi, V., Rao, G. R., Bose, I. (2011). Detection of financial statement fraud and feature selection using data mining techniques. Decision Support Systems, 50(2), 491-500 Schrand, C. M., Zechman, S. L. (2012). Executive overconfidence and the slippery slope to financial misreporting. Journal of Accounting and Economics, 53, 311-329.
Silver, S. E., Fleming, A.S., Riley, R.A. (2008). Preventing and Detecting Collusive Management Fraud. The CPA Journal, 78(10), 46-48.
Skousen, C.J., Smith, K.R., Wright, C.J. (2009). Detecting and predicting financial statement fraud: The effectiveness of the fraud triangle and SAS No. 99. Advances in Financial
Economics. 13, 53-81
Sloan, R., (1996). Do Stock prices fully reflect information in accruals and cash flows about future earnings? The Accounting Review, 71, 289-315.
Stice, J. (1991). Using financial and market information to identify pre- engagement market factors associated with lawsuits against auditors. The Accounting Review, 66(3), 516–533.
Summers S. L., Sweeney J. T., (1988), Fraudulently Misstated Financial Statements and Insider Trading: An Empirical Analysis. The Accounting Review, 73(1), 131-146 . Tabachnick, B. G., Fidell, L.S. (2013): Using Multivariate Statistics.
Tonge, A., Greer, L. Lawton, A. (2003). The Enron Story: You Can Fool Some of the People Some of the Time... Business Ethics: A European Review, 12(1), 4–22.
Trotman, K.T., Wright, W.F. (2012). Triangulation of audit evidence in fraud risk assessments. Accounting, Organizations and Society, (37)1, 41-53.
Warshavsky, M. (2012). Analyzing Earnings Quality as a Financial Forensic Tool. Financial
Valuation and Litigation Expert Journal, (39), 16–20.
Wells, J.T. (2001). Irrational ratios. Journal of Accountancy. 192(2), 80
Welsh, D. T., Ordonez, D. G., Snyder and Christian, M. S. (2015). The slippery slope: How small ethical transgressions pave the way for larger future transgressions. Journal of Applied
Psychology, 100, 114-127.
Wiedman, C. I., Hendricks, K. B. (2013). Firm Accrual Quality Following Restatements: A Signaling View. Journal of Business Finance & Accounting, 40, 1095-1125.
Wu, J., Zhang, L., Zhang X., (2010). The Q-theory Approach to Understanding the Accrual Anomaly, Journal of Accounting Research 48, 177-223.
Zhang, X.F., (2007). Accruals, Investment, and Accrual Anomaly. The Accounting Review 82, 1333–1363.
