Research on Distinguish the Accounting Information Distortion Based on the Principle Components-logistic Regression Model
Zhenjie Li
School of Economics and Management, Yantai University
ytmysky@163.com
Abstract
Nowadays, the accounting information distortion of listed company is generally common in the market. It has caused adverse effects to the enterprise itself and even the development of the securities market. In order to solve this problem, domestic and foreign scholars have done a series of researches from different perspectives and create a lot of detection model to identify financial reporting fraud more correctly. As far as we are concerned, these models’ index selection, calculation, prediction and application are not so satisfying and few efficient recognition models can be applied generally. In this paper, we combine the method of principal component analysis with logistic regression method. Then we select variables from the financial data that reflect the profitability, turnover, the establishment of the enterprise and some other perspectives. This accounting information distortion detection model is created by improving the method and index selection which has a higher correct recognition rate. We have chosen the 2012 financial statements from 56 firms for sample and the forecasting accuracy of the model reached 92.86%. We can get that it has obvious advantages compared to the predicted results from simple logistic regression model.
Keywords: Financial fraud, Principal component analysis, Logistic regression
1. Introduction
Financial reporting fraud has been a common phenomenon in the securities market. For instance, in the international market, in 2001 American Enron event shocked the whole world; in 2002 the world communication and Lantian Limited by Share Ltd set bad examples again. In the domestic market, Qiong Minyuan event happened in 1998; Yinguangxia event happened in 2001; Tianyi Science and Technology was caught financial fraud in 2004; Unisplendour Guhan Group Corporation Limited was also got punished in 2008 and Wanfu Biotechnology event was exposed in 2012. Among these fraud cases, some are typical report fraud and some cases may finished by cooperating with the auditing department. Anyway the financial reporting fraud caused very bad influence to the securities market. Therefore, the financial reporting fraud detection is one of the most urgent problems to the auditing department and the securities regulatory authorities. It must be controlled strictly to promote our securities market and national economic development.
The financial reporting fraud can be divided into financial data fraud and non-financial data fraud. It can also be divided into annual report fraud, interim report fraud, (letting) prospectus fraud, not timely disclosure notice on important matters and other information misrepresentation. We can find that the financial data fraud occupies a larger proportion of financial fraud events from the occurrence of financial fraud. In addition, according to a research of 2009 [1], in the past ten years the financial reporting fraud happened 20 times annually which accounted for 49.21% of the total fraud cases. These companies falsify the
International Journal of Security and Its Applications
Vol.8, No.4 (2014)
38 Copyright ⓒ 2014 SERSC
financial statements by making up or inflating income, concealing or not timely disclosing of important matters. In this paper, the study of financial reporting fraud points at the annual report fraud and the sample data is from annual report.
As the carrier of an enterprise to send its business operation in formation, financial statements are provided to reflect its financial situation, operating results and cash flow synthetically. The financial report of each issue should fairly reflect the financial condition, profitability, growth status and so forth. Many enterprises tamper their important data in the financial statements for concealing or overstating. Financial reporting fraud is the result of internal factors and external factors. For the enterprise itself, it is generally due to the information asymmetry, interest driven, corporate governance failure and the independence of audit deficiency. We can also catch the external factors like in adequate supervision, laws and regulations deficiency. Although China has a series of laws, they need to be reinforced compared to those in other countries. We usually pay too much attention to punish the enterprises while ignore the main persons in charge who only get some public condemnation or penalty. Financial reporting fraud has a continual damage effect on the company's development and also the securities market. Therefore, strengthening the internal and external audit as well as the correct identification of the fraud is an urgent task in the development of the securities market. In order to correctly identify the fraudulent financial statements, this paper will select all kinds of influencing factors of financial reporting fraud as indicator variables. We will create a recognition model by the combining principal component analysis and logistic regression method. The results will be compared with previous research achievements and the model will be found out the advantages and disadvantages.
The domestic and foreign researches on the financial fraud were mainly focused on the motivation, influencing factors and identification methods. Financial fraud detection can be classified into financial fraud signal judgment and recognition model creation. Khanh Nguyen [2] presented the identification of financial reporting fraud, such as providing specific information on the financial report analysis, financial data comparative analysis for previous years, some index ratio of relatively large impact analysis. Kinney [3] thought the financial reporting fraud can be more likely to appear in those financial distressed companies because these enterprise managers have stronger motivation to cover up their temporary financial difficulties. Albrecht [4] thought that problems about interest burden and cash flow turnover will increase the possibility of fraud, such as the issue of income needs, descent of earning quality and high levels of debt. Different scholars have different financial judgments so we don’t have a centralized, perfect system about these conjectures. We cannot apply all these methods when judge the financial fraud and that is not functional. Then the detection models appeared. Green and Choi [5] used artificial neural network technology to construct the financial reporting fraud detection model which was built based on the original financial data. It was founded to be very effective when applied in the sample. Benish [6] proposed to detect financial fraud report by searching whether there is any false accounting data. 74 companies from 1957 to 1993 in USA by CSRC for accounting fraud sample and other listed company for normal samples, he used 8 financial indicators to establish the Probitregression prediction model of which the accurate prediction rate is 75% and was applied actually. Summers and Sweeny [7] found that usually higher inventory turnover, rapid sales growth and rate of return on total assets occurred in the previous year. Efstathins Kirkosa [8] chose 38paired healthy enterprises and unhealthy ones from manufacturing industry of Greek and used 10financial indicators as input variables to establish a decision tree, neural network and Bias belief network model. Two of the overall mean variance hypothesis testing and Logistic regression equations are used by Lou Quan [9] to study the characteristics of financial reporting fraud companies. Chen Guoxin [10] uses China's listed company data to establish the 4 indicators
International Journal of Security and Its Applications
Vol.8, No.4 (2014)
Copyright ⓒ 2014 SERSC 39
of fraud. Cheng Liang [11] established the one-way ANOVA recognition model based on
fraudulent financial information fraud detection and the overall accuracy was
86%.MaoDaowei, Zhu Min [12]selected 136 samples and used analytical review methods to
establish a one-way ANOVA model, multivariate, discriminate model, linear probability
model and Logistic regression model. They used the models to forecast and the Logistic
regression model prediction accuracy was up to 80%. Wang Ya and Yuan Quan [13] selected
35 samples and 35 control samples, chose five indicator variables to establish a Logistic
regression model to quantify the financial reporting fraud identification, the model’s accuracy
is up to 84%.
Based on principal component analysis and logistic regression methods, we select the
appropriate indicators from each respect of the enterprise such as the financial situation and
the basic aspects to build recognition model which has a higher accuracy compared with
ordinary single-factor model and logistic regression model. In this paper, the second and third
part will explain the basic idea of Logistic regression method and the establishment of
ideological principal component logistic regression model in detail. A simple logistic
regression model and principal component-logistic regression model will be created in the
fourth part. We will use the two models to predict and compare the effect and make a
conclusion.
2. The Basic Idea of Logistic Regression Method
Logistic regression model, a multiple-variable analysis category, was proposed by J.
Berkson in 1944 [14]. It is a common method of sociology, biostatistics, clinical,
quantity psychology, econometrics, marketing and other statistical empirical analysis. It
is most widely used in medicine, and then gradually expands the scope of application.
logistic regression is mainly used to predict the dependent variable and the relationship
between a set of discrete explanatory variables. The most commonly used is binary
logistic regression and the value of the variable contains only two categories, for
example: good or ba
วิจัยแยกแยะความผิดเพี้ยนข้อมูลบัญชีแบบจำลองการถดถอยโลจิสติกส่วนประกอบหลักZhenjie ลี่โรงเรียนของเศรษฐศาสตร์และการจัดการ มหาวิทยาลัยไถytmysky@163.comบทคัดย่อในปัจจุบัน ความผิดเพี้ยนข้อมูลบัญชีของบริษัทจดทะเบียนได้ทั่วไปโดยทั่วไปในตลาด จึงได้ทำส่งผลต่อองค์กรเองและแม้แต่การพัฒนาของตลาดหลักทรัพย์ เพื่อแก้ปัญหานี้ นักวิชาการภายในประเทศ และต่างประเทศได้ทำชุดงานวิจัยจากมุมมอง และสร้างมากรุ่นตรวจสอบเพื่อระบุรายงานทุจริตทางการเงินอย่างถูกต้องมากขึ้น เท่าที่เรามีความกังวล ของโมเดลเหล่านี้เลือกดัชนี คำนวณ คาดเดา และโปรแกรมประยุกต์ไม่ให้ความพึงพอใจ และสามารถใช้รูปแบบการรับรู้ที่มีประสิทธิภาพน้อยโดยทั่วไป ในเอกสารนี้ เรารวมวิธีการวิเคราะห์ส่วนประกอบหลัก ด้วยวิธีถดถอยโลจิสติก แล้ว เราเลือกตัวแปรจากข้อมูลทางการเงินที่แสดงผลกำไร หมุนเวียน การจัดตั้งองค์กรและมุมมองอื่น ๆ บัญชีข้อมูลบิดเบือนตรวจสอบรูปแบบนี้จะถูกสร้างขึ้น โดยการปรับปรุงการเลือกวิธีและดัชนีที่มีอัตราการรับรู้ที่ถูกต้องสูง เราได้เลือกที่งบการเงินจากบริษัท 56 ตัวอย่างและความแม่นยำที่คาดการณ์ของแบบจำลองถึง 92.86% เราจะได้รับว่า มีข้อได้เปรียบที่ชัดเจนเมื่อเทียบกับผลการคาดการณ์จากแบบจำลองการถดถอยโลจิสติกอย่างคำสำคัญ: การเงินทุจริต หลักการวิเคราะห์องค์ประกอบ การถดถอยโลจิสติก1. บทนำรายงานทุจริตทางการเงินได้รับปรากฏการณ์ทั่วไปในตลาดหลักทรัพย์ ตัวอย่าง ในประเทศ ในปีค.ศ. 2001 เหตุการณ์อเมริกัน Enron shocked โลกทั้งโลก ใน 2002 สื่อสารโลกและ Lantian จำกัด โดย จำกัดการใช้งานร่วมกันได้อย่างดีอีก ในตลาดภายในประเทศ คองหยวน Minyuan เหตุการณ์ที่เกิดขึ้นในปี 1998 Yinguangxia เหตุการณ์ที่เกิดขึ้นในปีค.ศ. 2001 เทียนยี่วิทยาศาสตร์และเทคโนโลยีที่ถูกจับทุจริตทางการเงินในปี 2004 ยังไม่ได้โทษ Unisplendour Guhan กลุ่มคอร์ปอเรชั่น จำกัดในปี 2551 และเหตุการณ์เทคโนโลยีชีวภาพว่านฟู่แฟมิได้เปิดเผยในปี 2012 ในกรณีฉ้อโกง มีทุจริตรายงานทั่วไป และบางกรณีอาจเสร็จ โดยประสานงานกับฝ่ายตรวจสอบ อย่างไรก็ตาม การทุจริตรายงานทางการเงินเกิดอิทธิพลที่ดีไปยังตลาดหลักทรัพย์ ดังนั้น ตรวจสอบทุจริตรายงานทางการเงินเป็นหนึ่งในปัญหาเร่งด่วนที่สุดฝ่ายตรวจสอบและหน่วยงานกำกับดูแลหลักทรัพย์ จะต้องถูกควบคุมอย่างเคร่งครัดเพื่อส่งเสริมการพัฒนาเศรษฐกิจแห่งชาติและตลาดหลักทรัพย์ของเราThe financial reporting fraud can be divided into financial data fraud and non-financial data fraud. It can also be divided into annual report fraud, interim report fraud, (letting) prospectus fraud, not timely disclosure notice on important matters and other information misrepresentation. We can find that the financial data fraud occupies a larger proportion of financial fraud events from the occurrence of financial fraud. In addition, according to a research of 2009 [1], in the past ten years the financial reporting fraud happened 20 times annually which accounted for 49.21% of the total fraud cases. These companies falsify theInternational Journal of Security and Its ApplicationsVol.8, No.4 (2014)38 Copyright ⓒ 2014 SERSCfinancial statements by making up or inflating income, concealing or not timely disclosing of important matters. In this paper, the study of financial reporting fraud points at the annual report fraud and the sample data is from annual report.As the carrier of an enterprise to send its business operation in formation, financial statements are provided to reflect its financial situation, operating results and cash flow synthetically. The financial report of each issue should fairly reflect the financial condition, profitability, growth status and so forth. Many enterprises tamper their important data in the financial statements for concealing or overstating. Financial reporting fraud is the result of internal factors and external factors. For the enterprise itself, it is generally due to the information asymmetry, interest driven, corporate governance failure and the independence of audit deficiency. We can also catch the external factors like in adequate supervision, laws and regulations deficiency. Although China has a series of laws, they need to be reinforced compared to those in other countries. We usually pay too much attention to punish the enterprises while ignore the main persons in charge who only get some public condemnation or penalty. Financial reporting fraud has a continual damage effect on the company's development and also the securities market. Therefore, strengthening the internal and external audit as well as the correct identification of the fraud is an urgent task in the development of the securities market. In order to correctly identify the fraudulent financial statements, this paper will select all kinds of influencing factors of financial reporting fraud as indicator variables. We will create a recognition model by the combining principal component analysis and logistic regression method. The results will be compared with previous research achievements and the model will be found out the advantages and disadvantages.The domestic and foreign researches on the financial fraud were mainly focused on the motivation, influencing factors and identification methods. Financial fraud detection can be classified into financial fraud signal judgment and recognition model creation. Khanh Nguyen [2] presented the identification of financial reporting fraud, such as providing specific information on the financial report analysis, financial data comparative analysis for previous years, some index ratio of relatively large impact analysis. Kinney [3] thought the financial reporting fraud can be more likely to appear in those financial distressed companies because these enterprise managers have stronger motivation to cover up their temporary financial difficulties. Albrecht [4] thought that problems about interest burden and cash flow turnover will increase the possibility of fraud, such as the issue of income needs, descent of earning quality and high levels of debt. Different scholars have different financial judgments so we don’t have a centralized, perfect system about these conjectures. We cannot apply all these methods when judge the financial fraud and that is not functional. Then the detection models appeared. Green and Choi [5] used artificial neural network technology to construct the financial reporting fraud detection model which was built based on the original financial data. It was founded to be very effective when applied in the sample. Benish [6] proposed to detect financial fraud report by searching whether there is any false accounting data. 74 companies from 1957 to 1993 in USA by CSRC for accounting fraud sample and other listed company for normal samples, he used 8 financial indicators to establish the Probitregression prediction model of which the accurate prediction rate is 75% and was applied actually. Summers and Sweeny [7] found that usually higher inventory turnover, rapid sales growth and rate of return on total assets occurred in the previous year. Efstathins Kirkosa [8] chose 38paired healthy enterprises and unhealthy ones from manufacturing industry of Greek and used 10financial indicators as input variables to establish a decision tree, neural network and Bias belief network model. Two of the overall mean variance hypothesis testing and Logistic regression equations are used by Lou Quan [9] to study the characteristics of financial reporting fraud companies. Chen Guoxin [10] uses China's listed company data to establish the 4 indicatorsInternational Journal of Security and Its ApplicationsVol.8, No.4 (2014)Copyright ⓒ 2014 SERSC 39of fraud. Cheng Liang [11] established the one-way ANOVA recognition model based onfraudulent financial information fraud detection and the overall accuracy was86%.MaoDaowei, Zhu Min [12]selected 136 samples and used analytical review methods toestablish a one-way ANOVA model, multivariate, discriminate model, linear probabilitymodel and Logistic regression model. They used the models to forecast and the Logisticregression model prediction accuracy was up to 80%. Wang Ya and Yuan Quan [13] selected35 samples and 35 control samples, chose five indicator variables to establish a Logisticregression model to quantify the financial reporting fraud identification, the model’s accuracyis up to 84%.Based on principal component analysis and logistic regression methods, we select theappropriate indicators from each respect of the enterprise such as the financial situation andthe basic aspects to build recognition model which has a higher accuracy compared withordinary single-factor model and logistic regression model. In this paper, the second and thirdpart will explain the basic idea of Logistic regression method and the establishment ofideological principal component logistic regression model in detail. A simple logisticregression model and principal component-logistic regression model will be created in thefourth part. We will use the two models to predict and compare the effect and make aconclusion.2. The Basic Idea of Logistic Regression MethodLogistic regression model, a multiple-variable analysis category, was proposed by J.Berkson in 1944 [14]. It is a common method of sociology, biostatistics, clinical,quantity psychology, econometrics, marketing and other statistical empirical analysis. Itis most widely used in medicine, and then gradually expands the scope of application.logistic regression is mainly used to predict the dependent variable and the relationshipbetween a set of discrete explanatory variables. The most commonly used is binarylogistic regression and the value of the variable contains only two categories, forexample: good or ba
การแปล กรุณารอสักครู่..
