Help with article

Journal of Forensic and Investigative Accounting
Volume 9: Issue 1, January–June, 2017
Application of Forensic Tools to Detect Fraud: The Case of Toshiba
Anupam Mehta ∗
Ganga Bhavani
Introduction
References to fraudulent financial statements (FFS) have increased in frequency in the last several
years. FFS primarily consists of manipulating elements by overstating assets, sales and profit or
by understating liabilities, expense or losses (Charalambos T., 2002). “The auditor has a
responsibility to plan and perform the audit to obtain reasonable assurance about whether the
financial statements are free of material misstatement, whether caused by error or fraud”-SAA 99
and SAS 113. However, during the past several years, financial and accounting fraud has
appeared in the headlines of mainstream news worldwide. Although accounting fraud is not a
new phenomenon, recent cases involve much larger sums than previously (Clements, 2016). The
present study tests the effectiveness of three popular forensic tools in detecting FFS by Toshiba
Corporation from 2008‒2014. The three tools are the Beneish Model, the Altman Z-Score and
Benford’s Law. The comparison of the results and discussion of the tools’ relative effectiveness
provide direction for investigators about the selected tools’ effectiveness for detecting FFS.
Every tool has its advantages and limitations. By using only one forensic tool to detect fraud, an
auditor cannot adequately judge the financial statements of any corporation. This study highlights
the weaknesses of the selected forensic tools as well as their areas of application. Thorough
application of these tools to Toshiba’s financial statements revealed that the three tools did not
give the same results. In addition, it was not possible to use them with the same input.
The present study’s focus was to detect fraud in the financial statements of Toshiba Corporation
of Japan during seven years, from 2008‒2014, as evidence exists that fraud took place in the
company during those years. To detect the fraud, the selected forensic tools were applied to
Toshiba’s financial statements for the sample period. The study compared the results of the three
tools, discussed their limitations and suggested which was best for the purpose. To our
knowledge, no prior research has used all three forensic tools in one study, particularly not in the
case of Toshiba.
Toshiba’s Fraud
Toshiba Group is a widely-acclaimed Japanese-based company with ¥10.12 billion in business
market capitalisation. The organization, which has a 140-year history, had been undertaking an
orderly, ¥152 billion (USD$1.2 billion) expansion of benefits over the course of the 2008 to 2014
budgetary years. The FFS surfaced after examinations prompted the renunciation of the
organization’s main eight administrators, including the CEO, who assumed full responsibility for
the misrepresentation (The Economist, 2015).
692
The authors are, respectively, Associate Professor at Institute of Management Technology, Dubai, Adjunct
Faculty at Institute of Management Technology, Dubai
∗
Journal of Forensic and Investigative Accounting
Volume 9: Issue 1, January–June, 2017
About the Company
Toshiba Group includes Toshiba Corporation, which has 598 combined auxiliaries, with main
operations in Energy and Infrastructure, Community Solutions, Healthcare Systems and Services,
Electronic Devices and Components and Lifestyle Items and Services. Toshiba Group’s products
are manufactured and sold worldwide. As of March 31, 2015, the organization’s budget and stock
information included a basic load of ¥439.901 million, and the quantity of shares issued was
4,237,600,000 (Toshiba Group Annual Report, 2014).
This paper is organised as follows: next section presents a review of the selected forensic tools.
Then, the paper describes the study’s methodology. Next, it presents and discusses the study’s
results. Finally, the paper presents conclusions and suggestions.
Literature Review
Detecting FFS
Ultimately, the prevention and detection of FFS is not only the responsibility of internal and
external auditors but the collective responsibility of all stakeholders in an organisation.
According to a report from the Central Audit Quality (CAQ, 2010), if corporate executives
exchange information, inconsistencies in financial reporting will be brought to the fore, and the
opportunity to perpetrate FFS will be curbed. However, rapid asset growth, increased cash needs
and external financing all increase the likelihood of fraud (Christopher et al., 2008).
Per research by Beasley et al., (1999), FFS frequently involves the overstatement of revenues and
assets. Intentional misstatement in financial statements is noted much more frequently in
revenues than is misappropriation of assets. Beasley et al., noted that on an overall, cumulative
basis, the average fraud was USD$25 million, and the median fraud was USD$4.1 million. In
addition, Cynthia. H (2005) expressed a similar opinion on preventing and detecting manipulated
financial statements, noting that detecting FFS using normal audit procedures is extremely
difficult, not only for auditors but for all stakeholders. There are three main reasons for this,
according to Fanning et al., (1998). First is, a lack of knowledge concerning the characteristics of
fraud management. Second is, auditors lack the experience necessary to detect manipulated
financial statements. Third is, managers derive new techniques to mislead auditors and investors.
Fraud is very common currently. Of the various types of fraud, financial fraud causes huge losses,
not only to investors but for the country’s economy as a whole. Therefore, it is important to
prevent and detect fraud before it causes the business to collapse, devastating investors and
damaging the economy. There are various methods for detecting FFS. The models selected for
this study were the Beneish Model, the Altman Z-Score and Benford’s Law.
The Beneish Model
The Beneish Model is a mathematical model created by Professor Messod Daniel Beneish, who
formulated several analysis ratios with variables to identify the occurrence of financial fraud or
the tendency to engage in earnings manipulation. The model’s variables are constructed from the
data in the organization’s financial statements and, once computed, they create an M-Score, which
shows the degree to which earnings have been manipulated. The model’s efficiency has been
tested by various researchers. Muntari M (2015) used the model on Enron Corporation and found
that the company’s FFS could have been recognized as early as 1997, significantly before it
petitioned for insolvency in 2001. Normah Omar et al., (2014) applied the Beneish Model and
Ratio Analysis to Megan Media Holdings Berhad (MMHB), finding that the company
693
Journal of Forensic and Investigative Accounting
Volume 9: Issue 1, January–June, 2017
manipulated its earnings. Its operating-efficiency ratios showed that the company recorded
fictitious revenue, proving that the Beneish Model can detect FFS. Drabkova (2014) tested five of
the many statistical and mathematical models available for FFS detection: the Beneish M-Score
Model, the TATA – Total Accruals to Total Assets in the t-period, Three Jones Nondiscretionary
Accruals, and the Altman Z-Score Model. The results showed that the Altman and Beneish
Models were able to identify the financial health of the selected case study. Many researchers
have applied the Beneish Model to the popular corporate scandals of WorldCom and Enron
Corporation to identify their financial statement manipulations. Joost (2010) applied the Beneish
M-Score and Logit Score models to WorldCom, and the results showed that the status of this
company as a going concern should have been changed to that of a clean concern. Using these
statistical models, Joost concluded that WorldCom depended significantly on external financing,
implying that this need for credit may have been the reason for the company’s earnings
manipulations.
However, certain studies show that the Beneish Model is not the ultimate detector of fraud. The
ratios used in the model can only help to flag the problematic areas for auditor review. In a study
by Cynthia (2005), they did not prove to be consistent indicators of problems. In addition,
Ugochukwu (2015) compared use of the Beneish Model’s eight-variable and five-variable
versions on relevant items in the financial reports of 11 selected manufacturing companies in
Nigeria for the period from 2008‒2013. The results showed that the five-variable version
appeared to be more effective in predicting genuine, existing risks of material misstatement. A
study conducted by Amoa (2014) applied both the Altman and Beneish models to FFS by Anglo
Gold Ashanti and found that the Altman Model was more efficient at both predicting bankruptcy
and detecting FFS than the Beneish Model. The Beneish Model found no financial statement
manipulation in the company, whereas the Altman Model found four financial distresses the firm
had gone through during the years investigated.
Similarly, a recent study by Edmond (2016) noted that the Beneish M-Score and the Altman ZScore both detected FFS in Enron Corporation in 1998, 2000, and 2001. Both models were used
to analyse data retrieved from Enron Corporation’s annual reports, and each displayed flaws.
Both suffered from the effects of defining the metrics used to perform the financial analysis.
Hence, each model produced different values for some of the metrics used to calculate the ratios.
This can result in differing predictions of a company’s default risk and earnings manipulations.
The Beneish M-Score is like Altman Z-Score except that the M-Score focuses on assessing the
degree of profit control as opposed to deciding when an organization may reach bankruptcy. Few
studies have tried to apply two statistical models, but of those that have, most have used the
Beneish Model as one of the two models used. Nooraslinda et al., (2013) compared the use,
process and application of Benford’s Law and the Beneish Model in detecting accounting fraud,
concluding that both techniques appeared to have benefits in detecting and preventing fraud.
Altman’s Z-Score
Altman’s Model has been used in various industries to predict bankruptcy, and researchers have
also used it to detect FFS. According to Altman (1968), his model correctly predicts financial
failure for ninety-five percent of firms one year prior to their demise. Two years prior to
insolvency, accuracy decreases to seventy-two percent, and three years out, to fifty-two percent.
In addition, a study by Hawariah et al., (2014) found that Z-Scores, which measure the probability
of bankruptcy, are sufficient to detect FFS. They compared Z-Scores to other individual variables
that were expected to return negative figures, as firms with poorer financial conditions (and,
694
Journal of Forensic and Investigative Accounting
Volume 9: Issue 1, January–June, 2017
therefore, smaller Z-Scores) are more likely to engage in fraudulent financial reporting. A study
conducted by Charalambos (2013) used Z-Scores and other techniques on published data from
seventy-six firms, finding that Z-Scores can detect FFS. Charalambos found that Z-Scores
classified the entire sample with accuracy rates of more than eighty-four percent, and their general
indicators were associated with the FFS in the selected firms.
Mehta et al., (2012) found the Z-Scores model had a high probability of detecting FFS in a sample
company. The Altman Z-Score model includes the following variables: 1) the ratio of Inventory
to Sales; 2) the ratio of Total Debt to Total Assets; 3) the ratio of Net Profit to Total Assets; and 4)
financial distress (the Z-Score). The researchers found that the model efficiently predicted
variables, with an overall accuracy of 81.28%. In general, the indicators entered in the model
were associated with the firm’s FFS. Per the results, companies with high Inventories with
respect to Sales, high Debt with respect to Total Assets, low Net Profit with respect to Total
Assets and low Z-Scores were more likely to misrepresent their financial statements.
Gnyana (2015) applied Altman’s Z-Score to predict corporate bankruptcy in five selected fast
moving consumer goods (FMCG) companies during five years, from 2011‒2015.The author
concluded that by applying the Z-Score and selecting liquidity ratios, investors can use the model
to analyse the financial positions of companies. The Z-Scores of all selected FMCG companies
for the years in question showed sound financial positions. In addition, the study suggested that
companies should regularly estimate their Z-Scores when strategizing to improve their financial
positions.
Despite the fact that Altman’s Z-Score is easy to apply and includes various financial ratios, it has
been criticized for not incorporating all the important financial ratios. In addition, the model was
built based on accrual-basis balance sheets and income statements and does not take into account
cash-flow information. Stepanyan (2014) highlighted a new angle in Altman’s Z-Score Model in
his research on the bankruptcy chances of seven large US airlines, using Z-Scores for six
consecutive years. He noted that over the past thirty years, many tests have found Altman’s
bankruptcy prediction model to be roughly eighty to ninety percent accurate in predicting
corporate default two years prior to bankruptcy filing.
Benford’s Law
The Big Four accounting firms use Benford’s Law to conform to the fraud-detection
recommendations in the Financial Statements of the Statement of Auditing Standards No. 99,
which highlights the importance of Benford’s Law to assessing the possibility of financial
misstatement. The first author to thoroughly research and recommend Benford’s law was Nigrini.
According to Nigriniet et al., (1997), Benford’s Law can test the authenticity of lists of numbers
by comparing their actual and expected digital frequencies. The non-conformity of the results can
indicate FFS in a company.
However, some studies in the literature are cautious about the effectiveness of Benford’s Law in
detecting fraud. A study conducted by Hayes (2012) found Benford’s Law useful as an early
indicator of the possibility of FFS and possibly of use as a warning sign of bankruptcy. In
addition, the study found that Benford’s Law alone cannot detect FFS and that deviations from
Benford’s Law can cause an analyst to question the validity, accuracy or completeness of the
numbers. However, Benford’s Law can still be an appropriate method to detect the possibility of
fraud. It is a different way of looking at numbers. In conjunction with other audit tools, it can
help auditors minimize the expectation gap by increasing their chances of finding fraud and can
695
Journal of Forensic and Investigative Accounting
Volume 9: Issue 1, January–June, 2017
help companies’ bottom lines by finding inefficiencies and errors. In addition, Benford’s Law
improves sampling so auditors can concentrate on fraudulent or otherwise suspicious areas (Gogi
Overhoff, 2011). Durtschi et al., (2004) noted that Benford’s Law has been promoted as a simple,
effective tool for detecting fraud. They cited an actual example in which Benford’s law succeeded
in identifying fraud in a volume of accounting data. In addition, they noted that digital analysis
based on Benford’s Law is most effective and that there are areas where auditors should exercise
attention. The study indicated that certain limitations to the law. Likewise, Etteridge et al.,
(1999) cautioned that a data set that, when tested, does not conform to Benford’s Law may show
only operating inefficiencies or flaws in accounting and reporting systems, rather than fraud.
Need and Significance of the Study
According to the American Institute of Certified Public Accountants SAS No.82 (1997) and the
U.S. Government Accountability Office (2004), there are two types of financial misstatements.
First are financial misstatements due to FFS. Second are misstatements resulting from employee
fraud or defalcation. Fraudulent financial reporting frequently involves the overstatement of
revenues and assets (Beasley et al., 1999). Financial analysts, investors and management have
developed various forensic indices to aid forensic accountants in assessing the probability of
earnings manipulation. Each tool/model has its flaws and impediments to providing accurate
results, and therein lies the confusion, which affects auditors and stakeholders, regarding the best
model to use to detect various types of financial misstatements. After thorough examination of
the literature, the present case study chose three statistical techniques: the Beneish Model MScore (both five- and eight-variable), the Altman Z-Score and Benford’s Law. The reasons for the
selections included popularity, usage and applicability. First, a list of thiryt-six fraudinvestigation techniques was developed using common fraud and forensic-accounting texts
(Albrecht et al., 2015).Most of these tools and techniques are common in practice and used not
only for fraud detection but other purposes as well.
The present study tested the abilities of the three selected models to detect FFS in Toshiba
Corporation, the most recent of the large accounting and financial statements scandals. Although
the Toshiba scandal involved years from 2008‒2014, the study’s scope was from 2004‒2014.
This study contributes to filling the gap in the available literature on the application and
effectiveness of forensic tools in detecting FFS. To our knowledge, no prior research has used all
three forensic tools in one study, particularly not in the case of Toshiba.
The Objectives of the Study
1. To test the efficacy of the Beneish M-Score, the Altman Z-Score and Benford’s Law in
detecting FFS in Toshiba Corporation.
2. To compare the results of the three tools and suggest which is most useful to the present
purpose.
Hypothesis Development
Based on the above objectives, the following three hypotheses were developed regarding the three
tools.
696
Journal of Forensic and Investigative Accounting
Volume 9: Issue 1, January–June, 2017
H0 (1): The Beneish eight-factored and five-factored
variables will not effectively detect the Toshiba FFS.
Fraudulent
Financial
Statements of
Toshiba
H0 (2): The Altman Z-Score cannot be effectively useful
in detection of fraud in the Financial Statements of
Toshiba
H0 (3): The Benford’s Law Model cannot be effectively
useful in detection of fraud in the Financial Statements of
Toshiba.
Figure 1: Hypotheses of the study
Methodology
•
Apply the Beneish Model with both five- and eight-factor variables to Toshiba’s financial
statements.
•
Apply the Altman Z-Score to Toshiba’s financial statements.
•
Apply Benford’s Law to Toshiba’s financial statements.
•
Analyse each of these applications. Each of the three tools has a different procedure for
application. The methodologies of the tools are discussed below.
The Beneish Model
The Beneish M-Score is a mathematical model with two versions, one with five variables and one
with eight variables, both of which can identify financial fraud in earnings manipulations. The
Beneish Model has been acclaimed as being more sophisticated than ratio analysis (Cynthia,
2005; Roxas, 2011; Ugochukwuet et al., 2013). Aside from the high comprehensibility they
maintain, the eight-variable and five-variable versions of the model are both quite simple for
auditors to use (Beneish et al., 2008). The model incorporates the recommended ratio and trend
analysis common among preparers of financial statements, financial analysts and fraud examiners
by comparing the relationships between key financial-statement items for signs of earnings
manipulation (Ugochukwu et al., 2015). The Beneish Model is similar to the Altman Z-Score
Model, except that it does not predict bankruptcy.
Steps for application of Beneish Model
1. Calculate the eight variables or the five variables of the M-Score Model.
2. Enter the variables used into the model equation to calculate the M-Score. The present study
used Microsoft Excel to do this.
3. After calculating the M-Score and getting the results, categorize the company as a manipulator
if the M-Score >-2.22.
Then, the variables shown in Table I were applied to the function of the M-Score: The equation
for calculating the M-Score using eight variables is as follows.
697
Journal of Forensic and Investigative Accounting
Volume 9: Issue 1, January–June, 2017
M-Score = -4.84 + (0.92 * DSRI) + (0.528 * GMI) + (0.404 * AQI) + (0.892 * SGI) +
(0.115 * DEPI) – (0.172 * SGAI) + (4.679 * TATA) – (0.327 * LVGI).
Table I: Beneish (1999) and Rationale of the Variables
The equation for calculating the M-Score using five variables excludes SGAI, LEVI and TATA,
which were found not to be significant to the original Beneish Model. The equation for
calculating it is as follows.
M = –6.065 + 0.823 * DSRI + 0.906 * GMI + 0.593 * AQI + 0.717 * SGI + 0.107 * DEPI
According to Beneish (1999), an M-Score greater than -2.22 indicates that the company is
involved in FFS.
The Altman Z-Score
In 1968, Edward Altman developed a bankruptcy-prediction model using Multiple Discriminant
Analysis (MDA). The Z-Scores that it generates can be used to predict the potential of
bankruptcy two years prior to the actual filing.
698
Journal of Forensic and Investigative Accounting
Volume 9: Issue 1, January–June, 2017
Steps to use the Altman Z-Score
1. Calculate all five variables in the Z-Score Model.
2. Enter all of five variables into the model’s equation and calculate the Z-Score. The present
study used Microsoft Excel to do this.
3. After calculating the Z-Score and getting the results, categorize the selected company per the
benchmark standards of the Z-Score, which are given below.
Z-Score Benchmark Standards
Financially sound if greater than
Caution required if between
Likely to go bankrupt within two years if between
Likelihood of bankruptcy is high if below
Average for non-bankrupt companies
Average for bankrupt companies
2.99
2.77–2.99
1.8–2.7
1.88
5.02
-0.29
Table II: Altman Z-Score and Rational of the variables
Then, the variables shown in Table II were applied to the function of the Z-Score as follows.
These Z-Scores, which combine five financial ratios of a publicly traded firm, are generated using
the formula below.
Z-Score = 1.2 X1+ 1.4X2+ 3.3 X3+ 0.6 X4+ 1.0 X5.
699
Journal of Forensic and Investigative Accounting
Volume 9: Issue 1, January–June, 2017
Benford’s Law
According to research by Nigrini (1997), the original Benford’s Law included 5 tests in the areas
of accounting and auditing. The model converts digits into calculations, which is why it is also
called the Digits test. The five tests are: First Digit, Second Digit, First Two Digits, First Three
Digits, and Last Two Digits. Each test has its own purpose. Both the First and Second Digit tests
are high-level tests used to check the general reasonableness of data. They identify only identify
obvious anomalies. To get efficient, effective results, the input data must be massive. Using less
data does not enable comparison of patterns.
In the present study, Toshiba’s financial statements for 2007 through 2014, years during a known
period of fraud, were obtained from the company’s website, providing a sufficient volume of data
to enable comparison of patterns. A similar study conducted by Haynes (2012) compiled six
years of data from three U.S. municipalities and found non-conforming results, suggesting that
Benford’s Law can be used to find financial misstatements.
Although Benford’s Law might not accurately detect fraud, it can still indicate the possibility of
fraud. Non-conformities to Benford’s Law are red flags indicating possible irregularities, thereby
directing an auditor’s attention to the financial statements that merit further attention. The
following steps were taken to analyse Toshiba’s financial statements using Nigrini’s rules (1997).
Steps to use Benford’s Law
1. Perform digital analysis of each data set using a software program called NigriniCycle.xlsx,
which is an Excel program created by Nigrini.
2. Analyse the numbers from Toshiba’s published, comprehensive annual financial reports.
3. Compile the numbers for all seven years to get sufficiently massive data.
4. Omit numbers such as page numbers, dates, the numbers of notes, references to time (e.g.,
depreciation over ten years or ninety-day notes).
5. Omit numbers that were sub-totals or totals that did not convey any new information. For
example, subtotals of total current assets or total current liabilities can be omitted. Since these
subtotals and totals are the sums or differences between items and do not reflect any new
information, they cannot be manipulated.
6. To assess each digit test’s conformity to Benford’s Law, a test called the Mean Absolute
Deviation (MAD) is used, as per NigriniCycle.xlsx. By referring to a range of MAD values,
which is given on a table, the results can be evaluated for conformity to Benford’s Law to indicate
the degree of possible fraud. The higher the MAD value, the larger the difference between the
actual and expected values and the higher the chances of fraud.
The other benchmark for conformity used in this model is the Z-Statistic, which is automatically
generated after the test is conducted. Per GogiGogi Overhoff (2011) that the Z-Statistic of
Benfod’s law measures the size of the deviations between the expected and the actual values. The
larger the Z-Score (commonly 1% at 2.58, 5% at 1.96, or 10% at 1.65), the less likely it is that the
result is due to chance. According to Benfod’s law, after analysing the test results the conclusions
will be given in the following order.
700
Journal of Forensic and Investigative Accounting
Volume 9: Issue 1, January–June, 2017
Analysis
The Beneish Model
Table III shows that Toshiba’s overall M-Score results for 2008‒2014 are less than the benchmark
of -2.22, signifying that, overall, Toshiba was not manipulating earnings in the years under
review. Although Toshiba’s FFS for 2008‒2014 has been proved by the Japanese government
and various authorities with access to the evidence, the Beneish Model did not detect this fraud.
Using the eight-variable version of the model, whose outcome was comparatively weighed against
that of the five-variable version, the present study did not detect a possible risk of material
misstatement in Toshiba’s published figures/financial data for the years examined. As Table III
shows, the M-Score indicators for 2008‒2014 (-2.75, -2.50, -2.76, -2.83, -2.58, -2.49 and -2.58,
according to the eight-variable model and -3.02, -2.75, -2.93, -2.96, -2.73, -2.83 and -2.87,
according to five-variable one) did not indicate that the company was involved in material
misstatement in any of the years studied.
701
Journal of Forensic and Investigative Accounting
Volume 9: Issue 1, January–June, 2017
Table III:
However, the following is an analysis of the individual scores.
DSRI: DSRI is above 1.0 in the years of 2010, 2012, and 2013, indicating that the ratio of
Accounts Receivables to Sales increased in these years. In 2014, there was a slight decrease from
2013, from 1.105 to 0.964, indicating that the previously inflated revenue was reduced in the
current year.
GMI: The ratio of Sales to Cost of Goods Sold remained almost the same from 2010‒2014. The
GMI values for 2008 and 2009 were approximately the same, and thereafter, GMI values were
almost same from 2010‒2014.
AQI: This was lesser than 1.0, signifying a reduction in Asset Quality. However, Toshiba’s AQI
for the seven selected years never crossed the crucial mean of 1.254.
SGI: These scores were inconsistent over the seven years studied. In 2008, SGI was 1.079, but in
2009 and 2010, it fell, reaching 1.135 by the end of 2014.
DEPI: These results indicated an increase in value of the Depreciation Index from 2008‒2014.
This is the only variable that exceeded the mean index of 1.077, barely crossing the threshold that
indicates possible manipulation, which is 1.0767. The increases indicated growth in income that
was the result of decreasing depreciation. The value of this index clearly depicted earnings
manipulations by Toshiba for the years studied.
SGAI: The trend in SGAI crossed the 1.0 standard of the Beneish Model in 2008, 2009, 2012, and
2013, indicating increased in Sales and General and Administrative Expenses, which should raise
suspicion about Toshiba’s administrative efficiency. However, in 2014, the SGAI decreased to
0.984.
702
Journal of Forensic and Investigative Accounting
Volume 9: Issue 1, January–June, 2017
LVGI: The most important indicator is the Leverage Index. This variable showed the relationship
among outside liabilities in the form of both long-term and short-term liabilities to total assets.
An increase in the Leverage Index clearly indicated that the company was prone to earnings
manipulation. In 2009 and 2012, it exceeded 1.0, reaching 1.024 and 1.123, respectively. In all
other years, this variable was stable.
TATA: Total Accruals to Total Assets is useful for calculating the income from continuing
operations and cash flows from operations. In 2009, the TATA was 0.003, but in all other years,
this variable had negative values, indicating that the company was not receiving profits from any
sources other than their main ones.
Table IV: Toshiba Corporation
Applying the Beneish Model to Toshiba’s financial statements indicated that the company was not
manipulating its earnings. The calculations in the last two columns in Table IV represent the
model’s findings and categorize the company into one of two groups, non-manipulators and
manipulators. As Table IV shows, Toshiba scored close to the threshold for being in the
manipulators category in only one variable of the eight used: DEPI. A close consideration of the
indicators included in the eight-variable version of the model shows that except for the DEPI,
none appear to indicate risks of material misstatement.
Altman Z-Score
In 2008, the Z-Score was 1.970, indicating that the firm was going to go bankrupt within the next
two years. Except for 2008, the Z-Scores for all other years, from 2009‒2014, indicated that
Toshiba was not sound and would not long continue in the market. These lower Z-Scores, 1.237,
703
Journal of Forensic and Investigative Accounting
Volume 9: Issue 1, January–June, 2017
1.641, 1.799, 1.596, 1.541 and 1.567 respectively, showed that the chances of the company filing
for bankruptcy were very high.
These Z-Scores rightly indicated that the company was not sound financially and they also
indicated that there were financial misstatements by Toshiba.
However, the following is an analysis of the individual scores.
X1: As Table V shows, low Z-Score values were conditioned by the ratio of Working Capital to
Total Assets, which was either negative or very low for all the years examined, a possible
indicator that the company had liquidity problems. This component of the Z-Score model
indicates liquidity problems that increase the possibility of bankruptcy. The values improved
slightly over the years, except in 2008 and 2009, which had negative results, -0.0095 and -0.0637
respectively. From 2010‒2014, the results, 0.0501, 0.056, 0.0591, 0.0688 and 0.0689 respectively
were essentially stable. The results in 2013 and 2014 were almost the same.
X2: The ratio of Retained Earnings to Total Assets implied that Toshiba had not been able to
accumulate and reinvest profits during the period studied. Profits were used to cover the
accumulated losses incurred in prior years. Nonetheless, low values of the ratio of Retained
Earnings to Total Assets generally indicate low profitability. From 2011‒2013, values for this
variable, 0.103, 0.103 and 0.104 respectively were stable.
X3: The ratio of Earnings before Interest and Taxes to Total Assets, which reflects profitability
and operating efficiency, was generally low for the years studied, which, again, speaks a low level
of operating profitability and operating efficiency before taxes and financial leverage. In other
words, this ratio represents the Return on Assets (ROA) measure. Only in 2009 did the variable
show a negative value, -0.045; the results for all other years were both positive and stable.
X4: Although in 2009 the result for this variable decreased to 0.1813, for all other study years, the
value was stable.
X5: In 2013, the value for this variable decreased slightly to 0.9719, indicating decreased
effectiveness of asset use to generate revenue. In 2008, the result was 1.3365, the highest value
during the seven years studied.
Table V:
704
Journal of Forensic and Investigative Accounting
Volume 9: Issue 1, January–June, 2017
Benford’s Law
As Table VI shows, the first digit’s test for Toshiba’s annual financial statement data showed
MAD of 0.035757, which exceeded the 0.015 critical values for non-conformity by a wide
margin. Figure 2 shows the difference between the actual and expected proportions of first digits
from Benford’s Law. Since there was overall non-conformity to Benford’s Law in the first digit’s
test, this signals that the data set may have had abnormal duplications and anomalies. This result
shows that deviation from the actual and Benford’s values was greater than the accepted level of
standard. However, the digits 1‒8 as given in Table VII of the first digit’s test did not return a ZStatistic higher than 1.96, meaning that the individual differences in the actual and expected
frequencies were not significant. The digit 9 returned a Z-Statistic of 2.006, which is higher than
1.96, confirming that there was a manipulation in this digit place. As Table VI shows, the second
digits’ test showed a MAD of 0.02833, which exceeded the 0.012 critical value for nonconformity by a wide margin. Figure 3 shows the difference in the actual and expected
proportions of second digits from Benford’s Law. Since there was overall non-conformity to
Benford’s Law in the second digits’ test, this signals that the data set may have had abnormal
duplications and anomalies. This result shows that deviation from actual and Benford’s values
were greater than the accepted level of standard. However, none of the second digits had a ZStatistic higher than 1.96, meaning that the individual differences in the actual and expected
frequencies were not significant. Table VIII shows the actual and Benford’s results of the first
digits 1‒9 and the second digits 0‒9.
Table VI:
Table VIII:
Table VII: Results of Z-Test
First
Digit
Place
Second Place
0
1.889991
1
0.533751
0.503298
2
1.499288
0.638015
3
0.156113
0.161703
4
2.002138
0.383824
5
1.172223
1.305913
6
1.195333
1.240557
7
0.407402
0.061803
8
1.199708
1.049519
9
2.005956
2.04154
705
Journal of Forensic and Investigative Accounting
Volume 9: Issue 1, January–June, 2017
Figure 2: The results of First Digits
text from 1-9
Figure 3: The results of Second Digits Test 0-9
Conclusion
The primary of objective of this study was to examine the efficacy of the Beneish M-Score, the
Altman Z-Score and Benford’s Law in detecting FFS by Toshiba Corporation. The study found
that the null hypothesis of the Beneish Model was accepted, meaning that this is model was not
effective in detecting FFS at Toshiba. Comparative application of the five-variable version of the
model on the same financial data showed results that were slightly lower than those of the eightvariable model, strengthening the study’s results by further supporting that there was no material
misstatement in Toshiba’s financial statements. These results are consistent with those of a
similar study conducted by Karikari (2014) on Anglo Gold Ashanti. The author used the Beneish
M-Score and the Altman Z-Score on the selected company, and the results of the Beneish Model
did not indicate financial distress, but those of the Altman Z-Score did.
In the present study, the null hypothesis regarding Altman’s Z-Score was rejected, meaning that
the Altman’s Z-Score was useful in detecting FFS by Toshiba. These results are consistent with
those of studies conducted by Hawariah et al., (2014), Mehta et al., (2012) and Charalambos
(2002). These authors also found that Z-Scores that measured the probability of bankruptcy were
effective at detecting FFS. The present study found that unlike to the Beneish M-Score, the
Altman Z-Score was very effective in identifying FFS.
In the present study, the null hypothesis regarding Benford’s Law was rejected, meaning that
Benford’s Law was useful in detecting FFS by Toshiba. These results were consistent with those
of studies conducted by Gogi Overhoff (2011), Durtschi (2004) and Hayes (2012).
Like any other forensic tool, all three of the models tested have limitations. According to Nigrini
(2011), Benford’s Law can identify only digits manipulation, and while it can give an indication
of the probability of fraud, it cannot give its exact location. The massive volume of input data
required by this model increases the possibility that it contains errors.
Discussion and Suggestions
706
Journal of Forensic and Investigative Accounting
Volume 9: Issue 1, January–June, 2017
One objective of this research was to suggest which of the three tested forensic tools was most
useful for detecting FFS. The results of the present study support using more than one forensic
tool to detect FFS, because each model has shortcomings. To apply the Beneish Model variables,
one must consider the financial values in the target corporation’s financial statements. The
model’s results will be more accurate when the scope of the study is more than five years and the
financial values in the financial statements are large. The Beneish Model is a probabilistic model,
so it will not detect manipulation with 100% accuracy (Beneish et al., 1999). The results of the
present study support that statement, showing that this model failed to detect the fraud in
Toshiba’s financial statements, returning an M-Score of less than the threshold value of -2.22.
Altman’s Z-Score is very simple to use and rapidly provides a snapshot of the target corporation’s
financial position. The present study found that the Z-Score was the most accurate model of the
three tested. The study’s results suggested that all forensic tools are not useful with regard to
financial statements. For example, Benford’s Law is useful for detecting digits fraud, so it must
be applied to the target company’s day-to-day transactions, check collections and cancellations
and debt collections, rather than to financial statements. However, all three forensic tools used in
the study were useful for indicating red flags regarding the scope of the fraud at Toshiba, although
none could pin point the exact location or area of the fraud.
707
Journal of Forensic and Investigative Accounting
Volume 9: Issue 1, January–June, 2017
References
Albrecht, C., Holland, D., Malagueño, R., Dolan, S., and Tzafrir, S. (2015). The role of power in
financial statement fraud schemes. Journal of Business Ethics, 131(4), 803–813.
AICPA(1997), Statement on Auditing Standards (SAS) No. 82: Consideration of Fraud in a
Financial Statement Audit. American Institute of Certified Public Accountants, New
York.
Altman, E. (1968) Financial ratios. Discriminant analysis and the prediction of corporate
bankruptcy. The Journal of Finance 23(4), 589–609.
Beasley, S. Mark, Joseph v. Carcello & Dana R. Hermanson (1987–1997). Fraudulent Financial
Reporting: An Analysis of U.S. Public Companies. Research Commissioned by the
Committee of Sponsoring Organizations of the Treadway Commission.
Beneish, Messod (1999). A Note on Wiedman’s (1999). Instructional Case: Detecting Earnings
Manipulation. Indiana University issues in Accounting Education Vol. 14, No.
Beneish, M.D (2013). Earnings Manipulation and Expected Returns. Journal of Financial
Analysts Volume 69.
Beneish, M.D., Charles M.C. Lee, and D. Craig Nichols (2012). Fraud Detection and Expected
Returns. Electronic copy available at: http://ssrn.com/abstract=1998387
Centre for Audit Quality (2010). Deterring and Detecting Financial Reporting Fraud: A Platform
for Action, from: http://www.thecaq.org/anti_fraudinitiative/CAQanti-fraudreport
Christopher J. Skousen, Kevin R. Smith, and Charlotte J. Wright (2008). Detecting and predicting
financial statement fraud: the effectiveness of the fraud triangle and SAS no. 99.
Electronic copy available at: http://ssrn.com/abstract=1295494
Chrysovalantis Gaganis, Fotios Pasiouras, and Charalambos Spathis (2013).Regulations and Audit
Opinions:Evidence from EU Banking Institutions. Computational Economics 41, 387–
405.
Charalambos T. Spathis (2002). Detecting false financial statements using published data: some
evidence from Greece. Managerial Auditing Journal 17/4, 179–191.
Cindy Durtschi,William Hillison, and Carl Pacini Utah, (2004). The Effective Use of Benford’s
Law to Assist in Detecting Fraud in Accounting Data’, Journal of Forensic Accounting,
1524-5586/Vol V (2004), 17–34.
Cynthia, H. 2005, “Analysis ratios for detecting Financial Statement fraud,” ACFE Fraud
Magazine.
Hawariah Dalniala, Amrizah Kamaluddina, Zuraidah Mohd Sanusia, and Khairun Syafiza
Khairuddin (2014). Accountability in financial reporting: detecting fraudulent
firms.Procedia—Social and Behavioural Sciences 145, 61–69.
Etteridge M. L. and R. P. Srivastava (1999).Using digital analysis to enhance data integrity. Issues
in Accounting Education. 14(4): 675–690.
708
Journal of Forensic and Investigative Accounting
Volume 9: Issue 1, January–June, 2017
Edmond Ofori (2016). Detecting Corporate Financial Fraud Using Modified Altman Z-Score and
Beneish M-Score. The Case of Enron Corporation. Research Journal of Finance and
Accounting, ISSN 2222-1697 (Paper) ISSN 2222-2847 (Online) vol.7 No.4 www.iiste.org
Edward Altman (1968). Financial ratios, discriminant analysis and the prediction of corporate
bankruptcy. Journal of Finance, 23(23):589–609.
Fanning K. Cogger (1998). Neural Network Detection of Management Fraud Using Published
Financial Data. Journal of Intelligent Systems in Accounting, Finance and Management 7,
21–41 Freed.
Gnyana Ranjan Bal (2015). Prediction of financial distress using Altman Z-Score: a study of
select FMCG Companies. Indian Journal of Applied Research Volume : 5 | Issue : 9 |
September 2015 | ISSN – 2249-555X.
Gogi Overhoff, CFE 2011, “The impact and reality of fraud auditing Benford’s law: why and how
to use it,” Association of certified fraud Examiners―Certified Fraud Examiner and the
ACFE Logo are trademarks owned by the Association of Certified Fraud Examiners
Incorporation.
Hawariah Dalniala, Amrizah Kamaluddina, Zuraidah Mohd Sanusia, and Khairun Syafiza
Khairuddina (2014). Accountability in financial reporting: detecting fraudulent firms.
ScienceDirect Procedia—Social and Behavioral Sciences Elsevier Ltd , 145, 61–69.
Hawariah Dalniala, Amrizah Kamaluddina, Zuraidah Mohd Sanusia, and Khairun Syafiza
Khairuddina (2014). Detecting Fraudulent Financial Reporting through Financial
Statement Analysis. Journal of Advanced Management Science 17–22.
Haynes H Allyn (2012). Detecting Fraud in Bankrupt Municipalities Using Benford’s Law.
ScrippsSeniorTheses. Paper 42. http://scholarship.claremont.edu/scripps_theses/42
Joost Impink (2010). Earnings manipulation and bankruptcy: WorldCom. Third version, Bachelor
paper, Betzabeth Ignacius 5754992.
Karikari Amoa-Gyarteng (2014). Analysing a Listed Firm in Ghana for Early Warning Signs of
Bankruptcy and Financial Statement Fraud: An Empirical Investigation of AngloGold
Ashanti. European Journal of Business and Management, ISSN 2222-2839 (Online)
Vol.6, No.5.
Mehta Ujal, Patel Amit, Patel Hiral, and Purohit Rajan (2012). Detection of Fraudulent Financial
Statement in India: An Exploratory Study. Proceedings of the 2nd International
Conference on Enterprise Systems & Accounting (ICESA 2005), GFMJR, January—June
Vol.No.4.
Mark S. Beasley, Joseph V. Carcello , and Dana R. Hermanson (1999). Fraudulent Financial
Reporting—An Analysis of U.S. Public Companies. This project was commissioned by
the Committee of Sponsoring Organizations of the Treadway Commission (COSO).
Mark Warshavsky (2012). Analysing Earnings Quality as a Financial Forensic Tool, contributing
author to Financial Forensics Body of Knowledge, FVLE issue 39.
Muntari Mahama (2015). Detecting Corporate fraud and financial distress using the Altman and
Beneish models the case of Enron Corporation. International Journal of Economics,
709
Journal of Forensic and Investigative Accounting
Volume 9: Issue 1, January–June, 2017
Commerce and Management United Kingdom, Vol. III, Issue 1, Jan 2015
http://ijecm.co.uk/ ISSN 2348 0386.
Nigrini, M., Miller, S (2009). Data Diagnostics Using Second-Order Tests of Benford‘s Law,
available at: http://www.tcnj.edu/~business/documents/DataDiagnosticsUsingSecon
Nigrini, Mark J (1997). The Use of Benford’s Law as an Aid in Analytical Procedures’ Auditing,
Journal of Practice & Theory 16.2 (1997): 52–67.
Normah Omar, Ridzuan Kunji Koya, Zuraidah Mohd Sanusi, and Nur Aima Shafie (2014).
Financial Statement Fraud: A Case Examination Using Beneish Model and Ratio
Analysis. International Journal of Trade, Economics and Finance Vol. 5, No. 2.
Nooraslinda Abdul Aris, Rohana Othman, Siti Maznah Mohd Arif, Mohamad Affendi Abdul
Malek, and Normah Omar (2013). Fraud Detection: Benford’s Law vs. Beneish Model.
Symposium on Humanities, Science and Engineering Research (SHUSER).
Roxas, M. (2011). Financial Statement Fraud Detection Using Ratio and Digital Analysis, Journal
of Leadership, Accountability and Ethics, Vol. 8, No. 4, 56–66.
Stepanyan, A. (2014). Altman’s Z-Score in the Airline Business. Case Study of Major U.S.
Carriers. Are they Potential Bankruptcy Candidates? International Journal of Advances
in Management and Economics Vol.3 Issue 116–24. Available online at:
www.managementjournal.info
Ugochukwu Nwoye J., Okoye Emma I Oraka Azubuike O (2013). Beneish Model as Effective
Complement to the Application of SAS No. 99 in the Conduct of Audit in Nigeria’,
Academy of Business & Scientific Research. Management and Administrative Sciences
Review, ISSN: 2308-1368 Volume: 2, Issue: 6, 640–655.
710