مجله پیشرفت های حسابداری
سال نهم شماره 2 (پیاپی 73، پاییز و زمستان 1396)

Recent scandals and corporate failures have shaken the confidence of investors as accounts which were purported to reflect a “true and fair view” of businesses have been misleading. In an era with evolutionary financial frauds, computer assisted automated fraud detection mechanisms will be more effective and efficient with specialized domain knowledge. Statistics and machine learning based technologies have been shown to be an effective way to deter and defect fraud. Therefore, this study deals with the identification of factors related to fraudulent financial statements and investigating the effectiveness of data mining techniques in detecting firms that issue fraudulent financial statements.

 Research Questions

In order to achieve the objectives of this research, the following questions are developed:1. Can the fraudulent and non-fraudulent financial statements be detected from the text of annual corporate audit report fillings?
2. Can a quantitative fraud detection model be developed that will provide an analytical procedure for automating detection of potential fraud?

Data mining is used in many domains, including finance, engineering and biomedicine. There are two categories of data mining methods: unsupervised and supervised. Identifying financial statements fraud can be regarded as a typical classification problem. Data mining proposes several classification methods derived from the fields of statistics and artificial intelligence. Three methods, which enjoy a good reputation for their classification capabilities, are employed in this research study. These methods are Decision Trees, Support Vector Machines and Boosting algorithm. In this research, nineteen red flags of fraud extracted from auditing standard No. 240 along with data mining techniques such as principal components analysis and modified K-Means clustering were used to discriminate financial statements fraud cases. We consider data mining based financial fraud detection techniques such as Decision Trees, Support Vector Machines and Boosting approach in order to identify fraud. Based upon the fraudulent financial reporting related literature, 40 predictor variables are selected as input variables. These variables represent measurement proxies for a firm’s attributes of financial leverage, profitability, asset composition, liquidity, efficiency, size, growth, overall financial position, audit firm size, audit firm tenure and auditor change. These predictor variables are collected from the financial statements of listed companies in Tehran Stock Exchange (TSE) for the period of 2003-2015.

This research uses a quantitative approach on textual data to discriminate fraud cases using text mining techniques. The first research question was: can the fraudulent and non-fraudulent financial statements be detected from the text of annual corporate audit report fillings? Modified k-Means clustering demonstrated a good ability to discriminate fraud cases from non-fraud cases and thus it answered the first research question. the successful implementation of the Decision Trees, Support Vector Machines and Boosting algorithms also answers the second research question: can a quantitative fraud detection model be developed that will provide an analytical procedure for automating detection of potential fraud? In comparative assessment of the models’ performance, Support Vector Machines achieved the best performance to correctly classify validation sample in a 5-fold cross validation procedure. Decision trees and boosting algorithms also achieve a satisfactorily high performance.

The present study contributes to auditing and accounting research by examining the suggested variables that can best discriminate cases of financial statements fraud. These results suggest that there is validity for detecting financial statement fraud using data mining techniques such as Decision Trees, Support Vector Machines and Boosting approach. Our analysis provides insight for auditors, taxation authorities, investors, the stock exchange and banking system.

Understanding corporate information environment and factors affecting it has an important role in analyzing financial reports and setting accounting standards and laws regarding financial reporting. One of the phenomena studied in financial markets is intra-industry information transfer. Prior research on intra-industry information transfer shows that industry peers’ reports provide informative information about individual firms and documented significant stock price responses to industry peers’ announcement of earnings, (e.g. foster, 1981) management earnings forecasts, (e.g. Han, Wild and Ramesh, 1989) dividend, (Laux, Starks and Yoon, 1998) bankruptcy (Lang and Stulz, 1992) and earnings restatements. (Gleason, Jenkins and Johnson, 2008) Furthermore, the evidence in prior studies suggests that industry peers’ performance is more informative in some industries than in others. (e.g. Freeman and Tse, 1992) considering different theories suggesting effects of intra-industry information transfer phenomenon on manager’s behavior and demands for information, an important question that arises is whether financial reporting and its quality is different in industries with high intra-industry connectedness and firms operating in correlated environments.
In this study, we examined the relationship of intra-industry connectedness with firm’s earnings quality, management forecasts accuracy, firm’s disclosure quality and information asymmetry.
Research Questions or Hypothesis: In industries with high intra-industry connectedness industry peers’ reports are likely to provide useful information to evaluate the performance of individual firms. Under information perspective, since there is more useful information from industry peers in industries with high intra-industry connectedness, demand for voluntary disclosure is likely to be lower and it leads to less voluntary disclosure. On the other hand, herding perspective suggests two opposite effects on corporate information environment in these industries. First, firms operating in correlated environments are more likely to exhibit herding behavior (Demski and Sappington, 1984; Scharfstein and Stein, 1990). Scharfstein and Stein (1990) show that in correlated environments, managers mimic the decisions of their peers, even at the expense of ignoring their own private information. Hence, it’s likely that in industries with high intra-industry connectedness firms manage their earnings so that their earnings are closer to industry peers’ earnings. On the other hand, under herding perspective in industries with high intra-industry connectedness firm managers may have stronger incentives to herd in providing voluntary disclosures to enhance their reputation (Graham, Harvey, and Rajgopal, 2005).
Based on the above-mentioned theories, we can expect both positive and negative effects on corporate information environment. In this study we examined the following hypotheses:H1: There is an association between intra-industry connectedness and earnings quality.
H2: There is an association between intra-industry connectedness and management forecasts accuracy.
H3: There is an association between intra-industry connectedness and disclosure quality.
H4: There is an association between intra-industry connectedness and information asymmetry.
In this study we used panel data regression analyses for hypothesis testing. Our sample includes 178 firms listed in Iran stock markets operating in industries with at least 10 firms. We measured intra-industry connectedness as the absolute value of the covariance of sales change across all firms in the same industry for each industry. This covariance factor is calculated based on decomposition of the aggregate volatility of sales, presented and used in prior studies, (e.g. Comin and Philippon, 2005; Kim and Kwon, 2016; Pollack, 2013; Chiu, 2014). Our measure of earnings quality is based on the Dechow and Dichev (2002) model. Following Francis et al (2005) and Chiu (2014), we used the standard deviation of firm’s residuals over prior five years as a measure of earnings quality. Disclosure quality is measured based on scoring witch have been published by Tehran Stock Exchange each year and information asymmetry is measured by the bid-ask spread.
Our results show a significant negative relation between intra-industry connectedness and earnings quality, a positive relation between intra-industry connectedness and management forecasts accuracy, a negative relation between intra-industry connectedness and disclosure quality and a negative relation between intra-industry connectedness and information asymmetry.
Discussion and The negative relation between intra-industry connectedness and earnings quality can be explained from herding perspective. In industries with high intra-industry connectedness, managers have intensives to mimic other managers and bring their earnings closer to those of other firms by earnings management. This herding explanation can be backed up by results of Kedia et al (2015) study which shows firms are more likely to begin managing earnings after the public announcement of a restatement by another firm in their industry. This herding behavior can happen about management earnings forecast too; when a manager mimics others in declaring goals and earning forecast, he would try to achieve that earnings by earnings management if it’s not possible to achieve them normally. The tendency to bring earnings closer to management forecasts explains positive relation between intra-industry connectedness and management forecasts accuracy. The negative relation between intra-industry connectedness and disclosure quality can be explained by information perspective. In these industries providing useful information from industry peers can reduce demand for disclosure by individual firms and lead to lower disclosure quality. Overall, results of first three hypotheses can explain the negative relation between intra-industry connectedness and information asymmetry. It would not be surprising to find out the tendency to earnings managements and low-quality disclosure in industries with high intra-industry connectedness lead to low information asymmetry.
Findings of our research can be used as a warning to investors, accounting standard setting board and lawmakers to be careful and take in account the tendency of managers to mimic others in industries with high intra-industry connectedness.

Economic theory suggests that firm performance determined by industry fundamentals is relatively long lasting. On the other hand, performance that deviates from industry norms tends to dissipate more quickly. Investors tend to fixate on reported earnings, however, not fully recognizing differences in the persistence of its components. As a result, we expect the market to underreact to the higher persistence of the industry-wide component of earnings.
Research Hypothesis: The aim of the study is to investigate the persistence of components of industry and firm earnings.
H1. The industry-wide component of earnings is more persistent than the firm-specific component.
H2. Investors fail to fully appreciate difference persistence of industry-wide component of earnings and firm-specific earnings.
H3. Investors underreact to the persistence of industry-wide earning.
H4. Investors overreact to the persistence of firm-specific earning.
H5. Industry wide cash flows is the most persistent component of earnings.
H6. Firm-specific accruals is the least persistent component of earnings.
H7. Investors underreact to industry-wide cash flows component.
H8. Investors overreact to firm-specific accruals component.
To achieve the purpose of this study, a sample consisting of 113 companies listed in Tehran Stock Exchange during the year 2005 to 2015 were selected. In order to test the hypothesis, model of Mishkin approach was applied. The Mishkin test was conducted using non-linear least square estimation. The cross-equation restrictions were tested using a Wald-based test with robust Wald statistics.
The results of research indicate that industry-wide component of earnings is more persistent than the firm-specific component and investors fail to fully appreciate difference persistence of industry-wide component of earnings and firm-specific earnings; also they underreact to the persistence of industry-wide earning and overreact to the persistence of firm-specific earning. The results show that industry wide cash flows is the most persistent component of earnings and firm-specific accruals is the least persistent component of earnings. Research data show investors underreact to industry-wide cash flows component and they overreact firm-specific accruals component.
We find that industry-wide earnings are significantly more persistent than firm-specific earnings. However, investors fail to fully distinguish this differential persistence as stock prices place similar weights on these two earnings components in forecasting one-year-ahead earnings. In other words, the market underreacts to the persistence of industry-wide earnings and overreacts to the persistence of firm-specific earnings. Finally,we find that industry-wide cash flow is the most persistent component of earnings while firm-specific accruals is the least persistent. Results indicates that investors underreact to industry-wide cash flows component while they overreact to firm-specific accruals component.

Stock price crash risk in the market is a key concern of investors, and research in this field is important for the capital markets. Increasing the phenomenon of stock price crash will result in investor’s pessimism about investing in the stock market, and this could finally force the investors to withdraw their investment. On the other hand, investment decisions are widely dependent on information asymmetry and agency problems. Therefore, the management ability which is a qualitative component that captures the intrinsic talent of managers by varying circumstances, can increase investment efficiency by reducing information asymmetry and agency costs, and thus provide the optimal use of resources, accountability, transparency, fairness and rights of all stakeholders companies.
Hypothesis: The aim of this study is to investigate the relationship between efficient investment and management ability with stock price crash of listed companies on the Tehran Stock Exchange.
Data of 152 listed companies on the Tehran Stock Exchange during the period of 2006 to 2015 is investigated. To measure the stock price crash risk from the regression model Hutton et al. (2009) and the investment efficiency from the extended version of Biddle et al. (2009) model and the management ability, Demerjian et al.(2013) model, based on accounting variables is used. Data are analyzed by using panel data approach and multiple regressions model.
The results show that there is not a significant relationship between management ability and the investment efficiency; there is a significant and positive relationship between management ability and stock price crash risk. One of the reasons that led to such results in the Iran capital market may refer to the inability to analyze managers from the current and future conditions of the company and industry, and to utilize their intelligence and talent to achieve personal returns, raising information asymmetry and agency costs between stokholders, creditors and managers, lack of voluntary disclosure of information in accounting reportings, ups and downs of the capital market in recent years and the existence of a dominant government structure over Iran's economic system
There are two views regarding the impact of management ability on investment efficiency: An optimistic view, which states that a competent manager can use economic assets in order to increase shareholder's wealth in an appropriate time; and based on this view, there is a significant positive relationship between the ability of managers and investment efficiency. The pessimistic view argues that, in addition to the agency theory and conflict of interest between managers and owners, managers may use their talents in order to achieve their personal interests; or may act with high confidence and overestimate the return on investments because of their high intelligence, so that this fact will also cause losses. Our research results support pessimistic view. Therefore, the stock exchange should provide a mechanism to prevent opportunistic management practices for hoarding information and creating price bubbles and investor's uncertainty about such information.

Importance of macroeconomic variables is clear to everyone and announcements of them are seen and carefully scrutinized by different groups of users; however, initial estimates and economic forecasting of macro variable is raised as a serious challenge in economic planning. In this context, little or no evidence has been provided for exploring the relationship between accounting and economics (Macro Accounting) in developing countries like Iran. The idea of macro accounting was based on the idea that accounting variables such as aggregate accounting earnings convey information about Macroeconomics. This paper presents the use of fundamental accounting variables as the best leading indicators of macroeconomics variables.
Research Questions: The main questions of this paper are as follows:Can the combination of Elman neural network and particle swarm optimization improve models prediction in comparison to others?
How can fundamental accounting variables improve the predictive power of the model?
In this study we rely on predictive power of various models including Elman neural networks and particle swarm optimizationn. An Artificial Neural Network (ANN) is an information processing paradigm that is inspired by the way biological nervous systems. There are many types of artificial neural networks such as Elman Networks. Elman Networks are a form of recurrent neural networks which have connections from their hidden layer back to a special copy layer. Particle swarm optimization (PSO) is a population based stochastic optimization technique developed by Eberhart and Kennedy in 1995, inspired by social behavior of bird flocking or fish schooling. PSO shares many similarities with evolutionary computation techniques such as Genetic Algorithms (GA). We construct portfolio based on 88 largest firms in Tehran Stock Exchange. The sample period covers 20 semi annuals from 1385 to 1395. For this purpose, fundamental accounting variables (including net income, gross income, inventory, accounts receivable, administrative, general and sales expense, capital expenditures, debt and tax costs) have been chosen and their explanatory power in predicting macro.
Taking into consideration more alternative measures for accounting can decrease model's errors. As mentioned in previous subsection, using fundamental accounting variables would enable producing more reliable and accurate results. Our findings suggest that fluctuations in accounting information including net income, gross income, inventory, account receivables, general and sales expense and capital expenditure are a leading indicator of macroeconomic variables. Results show that fundamental accounting variables have predictive power in predicting GDP growth and unemployment rate for the next one and two quarters respectively. Also the empirical results from combination of artificial intelligence models show that optimization of Elman artificial network with particle swarm optimization improves effectiveness of model in comparison to Elman artificial network.
Discussion and This study distinguishes itself from previous papers with the introduction of key variables that have not been studied previously in macro accounting subject such as fundamental accounting variables. Prior studies mostly address accounting earnings in general neglecting predictive power of fundamental accounting variables. The main consequences of this study are effective links between accounting and economic information that must be included in the economic and financial decisions. So we recommend studies in application of accounting numbers in modeling macro. Overall the consequences of this paper introduce a new idea that the informativeness of accounting variables is not only in the micro level, but also in macro economy level.

Importance of macroeconomic variables is clear to everyone and announcements of them are seen and carefully scrutinized by different groups of users; however, initial estimates and economic forecasting of macro variable is raised as a serious challenge in economic planning. In this context, little or no evidence has been provided for exploring the relationship between accounting and economics (Macro Accounting) in developing countries like Iran. The idea of macro accounting was based on the idea that accounting variables such as aggregate accounting earnings convey information about Macroeconomics. This paper presents the use of fundamental accounting variables as the best leading indicators of macroeconomics variables.
Research Questions: The main questions of this paper are as follows: Can the combination of Elman neural network and particle swarm optimization improve models prediction in comparison to others?
How can fundamental accounting variables improve the predictive power of the model?
In this study we rely on predictive power of various models including Elman neural networks and particle swarm optimizationn. An Artificial Neural Network (ANN) is an information processing paradigm that is inspired by the way biological nervous systems. There are many types of artificial neural networks such as Elman Networks. Elman Networks are a form of recurrent neural networks which have connections from their hidden layer back to a special copy layer. Particle swarm optimization (PSO) is a population based stochastic optimization technique developed by Eberhart and Kennedy in 1995, inspired by social behavior of bird flocking or fish schooling. PSO shares many similarities with evolutionary computation techniques such as Genetic Algorithms (GA). We construct portfolio based on 88 largest firms in Tehran Stock Exchange. The sample period covers 20 semi annuals from 1385 to 1395. For this purpose, fundamental accounting variables (including net income, gross income, inventory, accounts receivable, administrative, general and sales expense, capital expenditures, debt and tax costs) have been chosen and their explanatory power in predicting macro.
Taking into consideration more alternative measures for accounting can decrease model's errors. As mentioned in previous subsection, using fundamental accounting variables would enable producing more reliable and accurate results. Our findings suggest that fluctuations in accounting information including net income, gross income, inventory, account receivables, general and sales expense and capital expenditure are a leading indicator of macroeconomic variables. Results show that fundamental accounting variables have predictive power in predicting GDP growth and unemployment rate for the next one and two quarters respectively. Also the empirical results from combination of artificial intelligence models show that optimization of Elman artificial network with particle swarm optimization improves effectiveness of model in comparison to Elman artificial network.
Discussion and This study distinguishes itself from previous papers with the introduction of key variables that have not been studied previously in macro accounting subject such as fundamental accounting variables. Prior studies mostly address accounting earnings in general neglecting predictive power of fundamental accounting variables. The main consequences of this study are effective links between accounting and economic information that must be included in the economic and financial decisions. So we recommend studies in application of accounting numbers in modeling macro. Overall the consequences of this paper introduce a new idea that the informativeness of accounting variables is not only in the micro level, but also in macro economy level.