Fraud theories

In the global economy, the phenomenon of accounting fraud and corporate fraud can be observed. Numerous fraud theories explain fraud as a factor of human behaviour. This paper examines fraud in light of the prevailing fraud theories.

Cressey examined the first theory of fraud (1953), where Cressy establishes a link between fraud and human behaviour. The ‘fraud triangle theory’ has been established as Cressey’s (1953) model in which three conditions that can lead to fraud are analysed; Opportunity, pressure, and rationalization. Fraud is essentially motivated by external forces. The motivation for managers to commit fraud is ‘pressure’. Cressey (1953), Albrecht et al. (2006), Lister (2007), and Manurung and Hadian (2013) all support the notion that non-financial and financial pressures motivate fraud by managers. In addition, according to SAS No. 99 ( Ramos and West, 2003), external pressure, financial goals, personal financial need, and financial stability are motives for fraud by managers. According to Cressey (1953), Wilson (2018), Wolfe and Hermanson (2004), and Kelly and Hartley (2010), a fraudster’s opportunity lies in his or her capabilities. The fraudster can identify the structural weaknesses of the company and then break the trust. Inadequate corporate governance, a lack of internal control, and a lack of control by responsible personnel are conditions for opportunity development. According to SAS No. 99, there are three categories of accounting fraud, which are distinguished by organizational structure, sector and inadequate supervision. The final component of the fraud triangle is rationalization. Rationalization is an attitude based on a set of moral beliefs. The actions of fraudsters are justified by their beliefs. In summary, Cressey (1953) stated that the three components of pressure-chance-rationalization constitute the behavioural characteristic of a cheater.

Wolfe and Hermanson (2004) added a new component, ‘capability’, to the fraud triangle theory. Wolfe and Hermanson (2004) assumed that ‘capability’ refers to the personal skills and characteristics that lead administrators to commit fraud. The new fraud theory is known as the ‘Diamond Fraud’ theory. According to Wolfe and Hermanson (2004), the fraud diamond theory suggests that ‘opportunity’ opens the door to fraud. Motive and justification determine the individual who commits fraud. Therefore, the individual must have the ‘capability’ to recognize and see the ‘opportunity’ to commit fraud.

Furthermore, Crowe (2011) extends the fraud triangle to include ‘arrogance’ and ‘competence’. The new fraud theory is referred to as the ‘Fraud Pentagon Theory’ and consists of five elements: ‘Opportunity - Pressure - Rationalization - Competence – Arrogance’. Crowe’s (2011) theory extends Wolfe and Hermanson (2004) by adding two new components ‘competence’ and ‘arrogance.’ More specifically a person’s ‘competence’ is his or her measure to perform a capability, in this cause, fraud. ‘Arrogance’, on the other hand, is a human behavioural trait that dominates an organization and overrides internal rules. This person is able to develop fraud strategies and has the goal of enriching themselves and winning.

Voussinas developed the latest fraud theory (2019). This fraud theory is known as the ‘Fraud Hexagon Theory’ and consists of the following six components: stimulus, opportunity, rationalization, capability, collusion, and ego. According to Vousinas (2019), the explanation of these components is as follows: ‘Stimulus’ is the incentive for a leader to commit fraud; ‘Capability’ is the ability required to commit fraud; ‘Opportunity’ is the opportunity to commit fraud; ‘Ego’ is a characteristic of human behaviour; and ‘Collusion’ is the agreement of people to commit fraud and justify their actions, which is ‘rationalization’.

We searched for documents with the fraud theories within as keywords in article titles, abstracts, and keywords. We perform a keyword search for ‘Fraud Triangle’, OR ‘Fraud Diamond’ OR ‘Fraud Pentagon’, OR’ Fraud Hexagon’ as each of the key theories previously discussed. Also, we search all open access types available to the Scopus database and all citation information. In addition, we select bibliographical information as affiliations, serial identifiers, publishers, editors, and some other information as tradenames and manufacturers, conference information, including references and accession numbers. Also, in this research, we include all authors who have examined fraud theories and all the subject areas, for example, ‘business management and accounting and ‘social sciences’.

Furthermore, we include all document types like articles, conference papers, reviews, book chapters, conference reviews, notes and data papers. Also, we include documents in the final publication stage and articles in the press. We include countries such as the United States, Indonesia, Malaysia, Australia, United Kingdom, Canada, China, Ghana, France and Germany. The investigation was limited to papers published only in English from 2004 to 2022. We did not include in our research the current year 2023, as documents are still being published. The last research was done at the end of January. The results from the above criteria are to collect 302 papers.

In continuing for each document of 302 papers, we use the author’s keywords of each paper and the abstracts. Then we export the data in the CSV file and use the VOS viewer program for our bibliometric analysis.

After importing the CVS file into the VOS viewer program, we choose the type of analysis and counting method. From this section, we analyse all the keywords of selected articles. The results of selected keywords appeared in Table 1. In this research, we created a network synthesis of keywords from 302 published articles. In creating the network, we only summarized the keywords that the researchers indicated occurred more than twice in the articles. Since authors sometimes use different words with similar meanings, we removed keywords with synonymous meanings. In more detail, keywords with synonymous meanings were excluded, such as ‘internal control’ and ‘internal controls’, ‘financial statement fraud’ and ‘fraudulent financial statements’, as shown in Table 1. Then a word cloud was created with the most frequent keywords from 2004 to 2022, as shown in Figure 1.

Then we classify the keywords into clusters as shown in Table 5. A group of keywords is called a cluster and is included in the map. Clusters are not overlapping in the VOS viewer program. Clusters do not necessarily fully cover all keywords on a map. Consequently, there may be keywords that do not fit into any cluster. Thus, a keyword may fit into only one cluster. Clusters are identified by cluster numbers.

Furthermore, we create a map based on text data. More specifically, we select all abstracts of the 302 papers related to fraud theories, and we use the binary country method to choose the threshold. We use the option to create a term co-occurrence map based on text data. Then we choose the field of abstracts and the most frequent words in analysing abstracts. The results are shown in Table 6. Also, we create clusters in the VOS viewer program as shown in Table 7.

We use multidimensional scaling to create the bibliometric map to increase the relatedness and similarity of elements (keywords) in a low-dimensional space. According to Van Eck et al. (2006) and Van Eck and Waltman (2007), the association strength for similarities (sij) is calculated by equation 1. AS ij = C ij C i C j (1)

For every set of keywords (item) i and j, VOS involves as input a similarity s _ij ( s _ij ≥ 0). The similarities s _ij as weights calculated on a ratio scale. The locales of keywords (items) in a map minimizing by equation (2). V x 1 … … … … x n = ∑ i < j s ij x i − x j 2 (2)

Subject to: 2 n n − 1 ∑ i < j x i − x j = 1 (3)

The purpose is the weighted sum to reduce the sum of the squares’ distance among every set of keywords (items). The weighted similarity among items is the squared distance between a pair of items. The average distance between two items equals one; to avoid solutions, all items have the exact location—the objective function given by equation 4, which is the ideal location of an item. The equation for the ideal location is shown in equation 4. x i ∗ = ∑ J ≠ i s ij x j ∑ J ≠ i s ij (4)

The ideal position of item i is defined as the weighted average of the positions of all other items. The ideal position of a keyword should occupy the most natural position that an item can have. Thus, the items appear to have the most desirable position that is closest to the ideal position described in Equation 2. The constraint in Equation 3 assumes that the positions of all keywords except the keyword (item i) are static and ignored. Equation 4 also describes the minimization of the objective function.

If the positions of all keywords i are static and the constraint is ignored, the minimization of the objective function has the effect of placing keyword i at its best position. Since the keywords have no static position and the solution is defined by the objective function and the constraint, we assume that the elements are not in the ideal position. So, according to the objective function, the keywords (elements) are placed no less than at the ideal position.

So, a map was created based on the counting technique and co-occurrence analysis with all keywords from 302 articles on fraud theories. The counting technique and co-occurrence analysis are used to determine the threshold value. The number of keywords to be selected is 48, and the lowest number of keywords is five out of the 988 keywords, 48 is the threshold. For each of the 48 keywords, the total strength of co-occurrence links with other keywords is estimated. The keywords with the largest total link strength are selected and listed in Table 4 below.

Metrics and methods of human behaviour

External pressures arise when managers want to respond to third-party demands. Managers may increase debt to external sources to remain competitive under these conditions. Another measure of external pressure in the fraud accounting literature is leverage and the ratio of total liabilities to total assets. Tiffani and Marfuah (2015), Manurung and Hardika (2015), and Skousen et al. (2009) suggest that the total debt/total assets ratio has a positive impact on financial statement falsification. Aprilia (2017) supports that a high debt to total assets ratio of a company can put pressure on the management team. When the company’s debt level is high, managers feel threatened by bankruptcy, which leads to pressure on management. External competitive pressure can cause managers to increase the company’s debt in order to stay in the market. Pressure can be referred to as another fraud factor.

The pressure/stimulus metrics for human behaviour created by Supri et al. (2018) and Sunardi and Amin (2018) provide evidence that financial goals can pressure management to achieve financial goals. In general, financial stability is a key issue for management. When financial stability is affected by economic conditions, managers feel pressured, which can lead to falsified financial statements. In the literature on accounting fraud, financial instability is measured by differences in total assets. If the differences in total assets are frequent in each year, it means that the company is in an unstable situation. Tiffani and Marfuah (2015), Manurung and Hardika (2015), and Skousen et al. (2009) also associate financial stability with falsification in financial reporting.

Personal financial needs are another human behavioural factor that can lead a company’s management to commit fraud. Tiffani and Marfuah (2015) pointed out that when the role of management is unclear, such as when the manager becomes part of the shareholders or participates in the board of directors’ commission, personal financial needs increase and may influence financial reporting. The ratio that can be used to measure own financial needs is the sum of shares held by insiders/total ordinary shares outstanding.

Return on equity (ROE) has been shown by Summers and Sweeney (1998), Brazel et al. (2006), Okoye et al. (2009), and Dechow et al. (2011) that it affects the percentage of fraud. A lower ROE means that the company has not attracted the attention of investors. In other words, the company management can edit the financial reports to attract more investors.

Setiawati and Baningrum (2018) found that companies with negative return on assets (ROA) tend to have poor economic performance. In another study, Emalia et al. (2020) supported that negative ROA does not meet financial objectives and that a company’s management team can manage financial reporting to improve the company’s financial performance. The human behaviour capability studied by Beasley (1996) supports that external board members lead to more effective management oversight and serve as a fraud prevention measure. Albrecht et al. (2010) also confirmed Beasley’s (1996) statement.

According to the literature, the last pressure factor is when a company seems to have liquidity problems. Persons (1995) proved that low current ratio leads to lower liquidity of a company’s assets and consequently a company is more vulnerable to short-term solvency.

According to fraud theories, capability is a component that can lead to fraud. Wolfe and Hermanson (2004) and Manurung and Hardika (2015) show that board changes can be a temporary period that causes stress and can lead to fraud. In addition, effective monitoring is one way for organisations to reduce the likelihood of fraud. Beasly et al. (2010) support that an audit committee with independent members can reduce the likelihood of fraud. Tiffani and Marfuah (2015) also suggest that an effective monitoring mechanism can reduce the risk of fraud. The measure of effective monitoring is the percentage of independent committee members.

Skousen et al. (2009) and Miller et al. (2012) investigated the possibility of human behaviour and demonstrated that claims have high inherent risk and can be manipulated. Their research was along the lines of Loebbecke et al. (1989), who reached the same conclusion. Thus, a high accounts receivable ratio may indicate manipulation in financial statements. Moreover, as described above, effective monitoring may measure the opportunity component of fraud theories.

Sykes and Matza (1957) suggested that a firm’s violation of internal regulations can be justified by rationalisation techniques. According to Skousen et al. (2009), rationalisation can be a measure when the auditor’s opinion affects the appearance of financial reporting. When the auditor’s opinion on the presentation of financial reporting includes the comment that it is not fair because it does not follow Generally Accepted Accounting Principles (GAAP), the management team tends to justify it ( Lokanan and Sharma, 2018).

Another characteristic of human behaviour is integrity. Integrity can be measured by earnings management, sales history, and earnings growth. Revenue management occurs when management fails to meet financial goals. It is then possible to manipulate accounting principles (GAAP) to produce financial reports. Also, managers can exploit the flexibility of accounting rules and manipulate the level of earnings ( Manurung and Hadian, 2013). In addition, integrity can be measured by sales performance. Assets support sales, so an increase in sales leads to an increase in assets ( Brigham and Houston, 2006). If the sales growth ratio is positive, the company is viable. The sales growth ratio is also indicated as the market share or position of the company in the market. There are many ways for managers to manipulate sales, and sales growth ratio has increased in fraudulent financial reporting. Another measure of integrity is the earnings growth rate - usually measured by ROE (return on investment) or earnings per share. Generally, the profit growth rate measures the manager’s valuation. Managers can affect the cost of goods and sales sold, interest and operating expenses, and income tax. According to agency theory and the various conflicts, managers can manipulate all of the above ratios and present a company’s financial performance differently.

The metric used in the accounting literature for ego is how many images and information of the CEO are related to the company’s performance. According to Apriliana and Agustina (2017), the CEO applies the internal rules of a company. Thus, if a CEO is characterized by arrogance, this may act as a fraud factor.

Collusion is described in the Hexagon Theory and supports the link between fraudulent financial reporting and collusion. Collusion occurs when more than two individuals engage in deceptive practices to achieve a goal for their benefit - the members’ agreement involves money or other benefits to facilitate their work. The law opposes collusion because it is only for personal gain. When collusion exists, the likelihood of fraud increases. The main characteristics of collusion are, first, payments to management or employees to gain money or goods, and second, the presence of an intermediary for the acquisition of services and goods.

Table 3 summarizes all of the above measures used by researchers to detect fraud according to fraud theories. Many researchers have used regression models to examine the relationship between human behavioural characteristics and fraud theories.

We use the counting method and co-occurrence analysis to answer the first two research questions, “Do ethical issues conflict with sustainability issues in a company?” In addition, the second research question is “How can fraud be detected by studying human behaviour using fraud theories?”

Total link strength indicates the number of articles in which the two keywords appear. The total link strength is a positive mathematical value. The larger this value, the stronger the link. According to our results, rationalization, opportunity, corruption, and pressure are the most important human characteristics that can lead to corporate fraud. Ethics also has a relatively high total link strength.

Then we classify the keywords into clusters as shown in Table 5. A group of keywords is called a cluster and is included in the map. Clusters are not overlapping in the VOS viewer program. Clusters do not necessarily fully cover all keywords on a map. Consequently, there may be keywords that do not fit into any cluster. Thus, a keyword may fit into only one cluster. Clusters are identified by cluster numbers.

Nodes denote the keywords of each research article in a network diagram. A link in a network is one of the links between nodes in Figure 1. Keywords are nodes and have different attributes. Weight (link, total link strength, and occurrence) are key attributes. When keywords have been assigned to clusters, cluster numbers are an instance of an attribute. Numeric values represent these attributes. Weight attributes are non-negative values. Thus, the weight of a keyword should indicate the importance of the keyword. A keyword with a higher weight is interpreted as more significant than a keyword with a lower weight. There are also two weighting attributes in cluster analysis: The overall link strength and the links attribute. For a given keyword, the total link strength and links attributes indicate the number of links of a keyword with other keywords and the total strength of links of a keyword with other keywords.

Thus, according to our results, human behaviour and ethics are associated with fraud. The column of link attribute shows the number of keyword links with fraud theories. The total link strength attribute shows the total strength of keyword links of fraud theories.

The lowest number of occurrences of a keyword is 10 out of the 5982 entries; 182 meet the threshold. A relevance score is estimated for each of the 182 items. Based on this score, the most relevant keywords are selected and presented in Table 6 below. The number of keywords to be selected is 109.

Using the same procedure as described above, we classify the abstracts into clusters as shown in Table 7 - a set of keywords from the abstracts form clusters in the map. According to our results in Table 7, the keywords from the abstracts form four different clusters. The first cluster contains 13 keywords, “effect, ability, necessity, accounting fraud, board of directors, fraud diamond, financial target, financial stability, accounting fraud, effectiveness, external pressure, arrogance”, which are related to human behaviour and accounting fraud. It is also clear from the first cluster that financial goals and stability are key motivations for fraud. The second cluster establishes a link between the social impact of fraudulent financial statements and the prevention of fraud by auditors. The third cluster includes six keywords – ‘author, crime, corruption, cressey, financial fraud, fraud, behaviour’-and links human behaviour to corruption and crime. The last cluster contains only two terms - keywords: ‘employee fraud, asset misappropriation’ and connects asset misappropriation with employee fraud. So, from all the above clusters, we can conclude the relationship between human behaviour characteristic and fraudulent financial statements.