Abstract
CHINA JOURNAL OF ACCOUNTING STUDIES 2019, VOL. 7, NO. 2, 147–169 https://doi.org/10.1080/21697213.2019.1642604 ARTICLE Financial report similarity and the likelihood of administrative punishment:based on the empirical evidence of textual analysis a b Aimin Qian and Dapeng Zhu a b Business School, University of International Business and Economics, Beijing, China; School of Accounting, Shanghai Lixin University of Accounting and Finance, Shanghai, China ABSTRACT KEYWORDS Administrative punishment; This paper uses A-share non-financial listed companies in the similarity; textual analysis; Chinese Stock Exchange from 2008 to 2016 as a sample and information disclosure investigates the influence of financial reports similarity on the likelihood of administrative punishment. The empirical result shows that the greater similarity to the last MD&A, the higher probability of administrative punishments for fraud firms; and the greater similarity to the last Non-MD&A, the lower probability of administrative punishments for fraud firms. Namely, regulatory agencies pay more attention to the information content of MD&A and the stability and compliance of Non-MD&A respectively. Further analysis shows that state property rights can mitigate the positive relationship between MD&A similarity and the likelihood of being punished, and better textual readability can aggravate the negative relationship between Non-MD&A similarity and the likelihood of being punished. Finally, after considering the review and preview part of MD&A, the corporate governance and accounting policy part of Non-MD&A respectively, the conclusions still stand. This paper provides empirical evidence for governments to promote policies about regulatory enforcement and informa- tion disclosure. 1. Introduction According to CSMAR database statistics, China’s regulatory authorities disclosed a total of 4011 administrative punishments for fraud A-share listed companies during 2008 to 2017, with 1644 listed companies punished by the regulatory authorities for illegal information disclosure, delayed disclosure, unauthorised use of funds and so on. Administrative punishments disclosed by regulatory authorities show that the violations of listed companies not only occur in the current period of investigation, but also can be traced back to the previous periods. As an emerging market economy, China’s judicial system and capital market supervision are still in the stage of continuous improvement and development. There are many problems, such as limited law enforcement resources CONTACT Dapeng Zhu zhudapeng.happy@163.com School of Accounting, Shanghai Lixin University of Accounting and Finance, Shanghai, China Paper accepted by Kangtao Ye. © 2019 Accounting Society of China 148 A. QIAN AND D. ZHU and insufficient supervision (Jiang & Kim, 2015; Kato & Long, 2006). Therefore, regulators can only conduct sports supervision and selective enforcement according to different situations (Dai & Yang, 2006). Not all listed companies’ violations are discovered and punished in time. Therefore, the influencing factors that lead listed companies to become the subject of regulatory supervision and punishment are a common concern in the theoretical and practical circles. The high-quality information disclosure of listed companies can effectively alleviate information asymmetry, improve the efficiency of market resource allocation, and safe- guard the legitimate rights and interests of shareholders properly (Hu & Tan, 2013). The annual financial report is an important part of mandatory and regular information disclosure of listed companies and a main way for external information users to under- stand the current financial situations and predict future development trends. In recent years, non-financial information and textual information characteristics of financial reports have received increasing attention from both theoretical and practical circles. Prior literature shows that non-financial information disclosure can help investors better evaluate a company’s market value, reduce the cost of capital and analysts’ forecasting errors, and improve auditors’ audit quality (Dhaliwal, Li, Tsang, & Yang, 2011; Dhaliwal, Radhakrishnan, Tsang, & Yang, 2012), as well as influence a company’s investment efficiency (Cheng, Tan, & Liu, 2012). The tone, length, readability, and similarity of financial reports affect the information processing and decision making of external information users, specifically in stock returns, stock trading volume (Feldman et al., 2010; Lee, 2012; Miller, 2010; Price et al., 2011), and analysts forecasts (Lawrence, 2013; Loughran & Mcdonald, 2014; Reuven et al., 2011). The earliest literature on the financial report similarity in the field of accounting can be traced back to Brown and Tucker (2010). They use MD&A similarity to measure information content and find stock prices positively respond to MD&A modification. Hanley and Hoberg (2010) take IPO prospectuses as research texts and find lower similarity of IPO prospectuses; namely, a higher proportion of unique information that reflects the individual characteristics of a company is beneficial to the accuracy of stock pricing. Hao and Su (2014) draw the same conclusion based on the IPO prospectus of Chinese listed companies. Tetloc (2011) finds a greater similarity of 10-Ks leads to a decrease of stock return volatility and stock trading volume. Meng, Yang, and Lu (2017) find the lower similarity of MD&A can reduce future stock price crash risk. However, little literature researches into whether financial report similarity affects the likelihood of administrative punishment and whether regulators’ focus differentiates in the different chapters of financial reports. Overall, the contributions of this study are as follows: first, different from previous studies that research the determinants of corporate fraud and administrative punish- ment from the perspective of internal corporate governance and external regulatory environment (Fang, Huang, & Karpoffthe, 2016; Hu, Wang, & Xin, 2015; Khanna, Kim, & Lu, 2015; Lennox & Pittman, 2010; Lu & Hu, 2016; Teng, Xin, & Gu, 2016; Wang, Winton, & Yu, 2010), this study researches the likelihood of administrative punishment for fraud firms from the perspective of information disclosure, providing direct empirical evidence for the usefulness of a financial report. Second, textual analysis is based on computer and artificial intelligence programs to read accounting documents and quantify account- ing texts. This paper enriches the economic consequences of financial report textual CHINA JOURNAL OF ACCOUNTING STUDIES 149 characteristics by researching the influence of textual similarity on administrative pun- ishment for fraud firms and compensates for the research gap in the field regarding the relationship between textual analysis and enforcement actions (Hao & Su, 2014; Jiang, Ma, & Xiong, 2014; Meng et al., 2017; Xie & Lin, 2015). Third, an annual financial report of listed company includes many sections, such as ‘Management Discussion and Analysis’, ‘Changes in Shares and Information about Shareholders’, ‘Directors, Supervisors, Senior Management and Employees’, ‘Financial Report’. Our study not only focuses on the effect of MD&A similarity on the likelihood of administrative punishment, but also other sections. Finally, the conclusions of this paper help the capital market participants understand the mechanism between information disclosure and administrative punish- ment for fraud firms, provide valuable references for the regulatory agencies to for- mulate policies about information disclosure and ensure the prosperity of China’s capital market in the future. The remainder of this paper is structured as follows. Section 2 introduces the institu- tional background and theoretical analysis; Section 3 is the research design; Section 4 describes the empirical results; Section 5, further analysis; and Section 6 conclusions. 2. Institutional background and theoretical analysis 2.1. Institutional background Inorder to regulate listed companies’ annual financial report preparation and disclosure behaviour, the China Securities Regulatory Commission (CSRC) formulates the Contents and Formats of Information Disclosure for Public Offering Companies No. 2: Contents and Formats of Annual Financial Report (hereinafter referred to as ‘Guideline No. 2’) based on The Security Law of the People’s Republic of China, The Company Law of the People’s Republic of China and other relevant laws and provisions of the China Securities Regulatory Commission. ‘Guideline No. 2’ specifies some basic requirements for the preparation of the annual financial report, the content of each section, and the deadline for disclosure. Since 2001, the China Securities Regulatory Commission has revised ‘Guideline No. 2’ several times and there have been many changes in the content of each section of financial report. Among them, after the revision in 2012, the order and content of the sections of financial reports have been greatly adjusted compared with previous periods, and the section of ‘Internal Control’ was added. After the revision in 2014, the ‘Preference shares’ section was added. In 2016, the ‘Corporate Bonds’ section was further added. The changes of financial report sections from 2001-2016 are shown in Table1 ‘Management Discussion and Analysis’ is the review and analysis of business during the reporting period and the outlook and evaluation of future development trends. It is the essence of non-financial information in an annual financial report. China introduced this system in 2001 to help investors better understand a company’s operating results, financial situation and potential future changes. MD&A had been the main content of ‘Report of the Board of Directors’ section until 2015. According to ‘Guideline No. 2’, listed companies shall disclose information about operation and investment and analyse financial situations and operating results during the reporting period in the ‘Report of the Board of Directors’. Since 2015, MD&A has become an independent section, namely 150 A. QIAN AND D. ZHU Section IV, ‘Management Discussion and Analysis’. In 2016, it was renamed ‘Business Discussion and Analysis’. The contents of this section mainly include: Overview, Main Business Analysis, Main Business Composition, Asset and Liability Analysis, Core Competitiveness Analysis, Investment Analysis and Future Development Prospects. Specifically, Future Development Prospects include: industry development trends, devel- opment strategy and business plan, the use and source of funding and risk factor analysis. 2.2. Theoretical analysis Information disclosure of the financial report could alleviate information asymmetry between the internal and external company (Kryzanowski & Ying, 2013), and is the window for external information users to understand a company’s internal business decisions. Prior literature shows that the textual information disclosed in the financial report – such as research and development activities, social responsibility, and account- ing policies – has increasingly attracted the attention of institutional investors, small and medium investors, analysts and other stakeholders (Cormier & Magnan, 2014; Hope, 2015; Merkley, 2014; Xu & Tang, 2010). With the improvement and popularity of computers’ natural language processing ability, the textual characteristics of financial reports have a subtle and continuous influence on information acquisition, on the processing and interpretation of accounting information users, and also affect a com- pany’s business investment decisions, an investor’s capital market decisions and an analyst’s forecast behaviour (Hoberg & Phillips, 2010, 2017; Lee, 2012; Lin & Xie, 2017; Miller, 2010; Reuven et al., 2011). In recent years, although Chinese government regulatory agencies have continuously increased regulatory involvement and the degree of regulation and enforcement, the violations of listed companies are not still prohibited, and the means of violations are constantly diverse. High enforcement costs and limited enforcement resources result in regulators being able to execute randomly with a certain probability, rather than a thorough investigation of all listed companies’ violations. Since there are constraints on regulatory resources and enforcement costs, the financial information and textual infor- mation disclosed in the financial reports of listed companies become important for the allocation of enforcement resources by regulatory authorities. In China, law enforcement officials working in regulatory agencies, analysts and auditors, have financial and legal professional education backgrounds and strong financial information interpretation capabilities. The enforcement officials of regulatory authorities can directly make a preliminary judgement on whether the listed company has been involved in fraud behaviour by comparing the financial performance and textual information horizontally or vertically, thereby enabling further allocation of regulatory resources to conduct an investigation of potential target companies. At the same time, analysts’ behaviour and abnormal trading performance in the capital market are also important information sources that cause regulators to pay attention to the possibly fraudulent behaviour of listed companies. Rational investors and analysts can identify and respond to the abnormal characteristics of textual information disclosed by listed companies (Miller, 2010; Reuven et al., 2011; Lee, 2012; Lin and Xie, 2017). Therefore, regulators can indirectly make decisions regarding enforcement resource allocation based on analyst’s CHINA JOURNAL OF ACCOUNTING STUDIES 151 following and forecast, the stock price and trade volume of listed companies in the capital market. MD&A is the ‘soul’ of financial reports of listed companies. Listed companies mainly discuss and analyse accounting information in financial statements, major events that occur in the current period or will occur in the future, and try to provide useful information for decision-making to investors in a concise and understandable way. A large volume of the literature proves that MD&A can provide incremental information for investors and analysts to help predict future performance (Cole & Jones, 2004; Xue, Xiao, & Pan, 2010), improve analysts’ forecast accuracy, and reduce prediction diver- gence (Bryan, 1997; Clarkson, Kao, & Richardson, 1999). As the core content of textual information, ‘Guideline No. 2’ clearly formulates articles to regulate MD&A disclosure. For example, listed companies should interpret and analyse financial data in accordance with corporate business in the reporting period, not just repeat the contents of financial statements or use inane and stereotyped language. Therefore, the usefulness of deci- sion-making and information content are key elements for evaluating the quality of MD&A information disclosure. There are two main reasons for the high textual similarity of current MD&A compared with the previous period. The first reason is the company’s information disclosure behaviour. The high textual similarity of MD&A indicates that a listed company has a negative attitude when disclosing current MD&A information. The information disclosed is only copied from the previous year’s financial report, leading to a large quantity of information similar to the previous annual financial report. Listed companies with negative attitudes fail to provide objective analysis of actual business operations and the internal and external environment’s impact on financial information during the reporting period. In that situation, listed companies provide redundant information rather than incremental information, which cannot effectively alleviate the information asymmetry between the company and external information users. The above behaviour violates the basic disclosure requirements in accordance with ‘Guideline No. 2’. The second reason is the company’s actual business operations. The high similarity of MD&A may mean that few changes have occurred or will occur in a company’s current or future periods. There are no new major investment projects, no merger and acquisition plans, no recent developments in R&D, and a lack of sufficient interpretations. Under these circumstances, a listed company with high MD&A similarity has difficult in giving a good impression of corporate governance, reliable accounting information, and business performance to regulatory agencies, which increases the company’s risk of being inves- tigated and punished in the current period. Thus, based on the above analysis, we propose the first hypothesis. H1: The textual similarity of MD&A in a listed company’s financial reports is positively correlated with the likelihood of administrative punishment. Unlike MD&A’s information disclosure specifications, non-MD&A of listed company’s annual financial report are stereotypical and inertial, most of which does not change over time. Although the annual financial reports of listed companies have become longer and longer in recent years (Loughran & Mcdonald, 2014), the disclosure contents and formats of non-MD&A sections have been clearly defined by ‘Guideline No. 2’, such as, ‘Important 152 A. QIAN AND D. ZHU Statements and Contents’ and ‘Company Profiles’. Under this circumstance, the low similarity of non-MD&A may be due to the fact that a listed company does not strictly comply with the requirements of ‘Guideline No. 2’ and arbitrarily deletes or increases information with the motive of strategic information disclosure or attracting attention. In addition, for the sections ‘Changes in Shares and Information about Shareholders’, ‘Directors, Supervisors, Senior Management and Employees’ and ‘Corporate Governance’, unless mergers and acquisitions or transfer of control rights happen in the current period, which affects corporate governance structures and personnel structures strongly, the financial report will not change significantly compared with the previous period. The high similarity of non-MD&A means no major changes in the business strategy and governance structure during the reporting period. The prior beliefs of investors, analysts and regulators on the listed companies are further consolidated and confirmed and reduce the uncertainty of future value assessment (Bozanic & Thevenot, 2015). As a result, regulators are more concerned with the stability and conformity of non-MD&A. Based on the above theoretical analysis, we propose the second hypothesis. H2: The textual similarity of non-MD&A in a listed company’s financial report is nega- tively correlated with the likelihood of administrative punishment. 3. Research design 3.1. Data and sample This paper takes A-share listed companies in China’s Shanghai and Shenzhen Stock Exchanges from 2008 to 2016 as a research sample. The reason why the sample starts from 2008 is because the format of financial reports is quite different before 2008, resulting in a significant number of samples missing from the process of Python text processing (missing rate is about 50%). On the basis of the initial sample, this paper excludes some firms as follows: (1) all firms in the financial industry; (2) firms with missing textual similarity data, financial data and corporate governance data; and (3) firms with unreasonable data, such as insolvent firms, ST firms. Finally, we get 17,661 observations. The annual financial reports of listed companies are downloaded from www.cninfo. com.cn. Financial data, corporate governance data and administrative punishment data are mainly from the CSMAR database. Textual information data are obtained by proces- sing annual financial reports of listed companies through machine learning. The dis- closure contents and orders of MD&A have been constantly changing since 2008. Therefore, this paper mainly selects part of ‘Report of the Board of Directors’, ’Discussion and Analysis of Business Situation’, ‘Management Discussion and Analysis’ for text processing, in addition, we use Stata14.0 to analyse data. 3.2. Variable definitions Punish: listed companies’ annual financial reports are disclosed during 1 January to 30 April in the following year. In order to examine the causal relationship between the similarity of financial report information disclosure and the probability of administrative CHINA JOURNAL OF ACCOUNTING STUDIES 153 punishment, we use the data of financial reports similarity from 2008 to 2016 and the data of administrative punishment for fraud firms disclosed from 1 May to 31 December 2009–2017. We set a dummy variable. If the listed company is punished by regulatory agencies in the current period, the value is 1, otherwise 0. MD&A_Sim and Other_Sim: we use the financial report similarity between current and last period as a measure. Regarding the calculation of Chinese text similarity, first, we use Jieba to split Chinese words and delete Arabic numerals, punctuation marks, pictures and tables during splitting. Then, we use the cosine similarity calculation method (Latent Semantic Indexing) to calculate MD&A similarity and non-MD&A simi- larity respectively. In order to reduce the impact of financial characteristics, corporate governance and other factors on enforcement penalty, we include the following control variables based on prior literature: size, leverage, profitability, audit opinion, analyst following, the largest shareholder’s share proportion, the size of director board, duality. Finally, we include year dummy variable and industry dummy variable. Detailed explanations of the main variables are given in Table 2. 3.3. Empirical model Based on prior studies (Khanna et al., 2015; Lu & Hu, 2016), we estimate the following regression model to test our hypotheses. If the coefficient β of Sim is positive, it means that high MD&A similarity tends to increase the likelihood of administrative punishment, which is consistent with hypotheses 1. If the coefficient β of Sim is negative, it means that high non-MD&A similarity tends to decrease the likelihood of administrative punish- ment, which is consistent with hypotheses 2. ProbitðÞ Punish ¼ 1 ¼ β þ β Sim þ β Control Variables þ Year i;t i;tþ1 0 1 þ Industry þ ε (1) i;t 4. Empirical results 4.1. Descriptive statistics Descriptive statistics are shown in Table 3. In order to eliminate the influence of outliers, we winsorise all continuous variables at the 1% and 99% levels. As can be seen from Table 3, during the sample period, 13.6% of the samples are penalised by the regulatory authorities for violations. The mean value of MD&A textual similarity is 0.289, the mean value of non-MD&A textual similarity is 0.451, and the mean value of non-MD&A textual similarity is higher than that of MD&A textual similarity. From the statistical results of quantiles, there is a big difference in the financial report textual similarity among different listed companies. The descriptive statistics of the main control variables are almost consistent with previous studies. Figure 1 shows the trend of the mean value of the annual financial reports similarity from 2008 to 2016. As the lengths of annual financial reports of listed companies are getting longer, the textual similarity of annual financial reports generally fluctuates from 2008 to 2016. In 2016, MD&A similarity and non-MD&A similarity both reach the highest 154 A. QIAN AND D. ZHU Table 1. Annual financial report sections of listed companies from 2001 to 2016. 2001 2007 2012 2014 2016 SECTION I Important Important Important Important Important Statements and Statements and Statements, Statements, Statements, Contents Contents Contents and Contents and Contents and Definitions Definitions Definitions SECTION II Company Profile Company Profile Company Profile Company Profile Company Profile and Key Financial Results SECTION III Key Financial Key Financial Key Financial Key Financial Business Profile Results Results Results Results SECTION IV Changes in Shares Changes in Shares Report of the Report of the Management and Information and Information Board of Board of Discussion and about about Directors Directors Analysis Shareholders Shareholders SECTION V Directors, Directors, Significant Events Significant Events Significant Events Supervisors, Supervisors, Senior Senior Management Management and Employees and Employees SECTION VI Corporate Corporate Changes in Shares Changes in Shares Changes in Shares Governance Governance and Information and Information and Information about about about Shareholders Shareholders Shareholders Preference Shares Preference Shares SECTION VII Profile of Profile of Directors, Shareholders’ Shareholders’ Supervisors, Meeting Meeting Senior Management Employees and SECTION Report of the Report of the Corporate Directors, Directors, VIII Board of Board of Governance Supervisors, Supervisors, Directors Directors Senior Senior Management Management and Employees and Employees SECTION IX Report of the Report of the Internal Control Corporate Corporate Board of Board of Governance Governance Supervisors Supervisors SECTION X Significant Events Significant Events Financial Report Internal Control Corporate Bonds SECTION XI Financial Report Financial Report Documents Financial Report Financial Report Available for Reference SECTION XII Documents Documents Documents Documents Available for Available for Available for Available for Reference Reference Reference Reference point of 0.411 and 0.647. The mean value of MD&A similarity is always lower than that of non-MD&A similarity, and the fluctuation degree of MD&A similarity is significantly smaller than that of non-MD&A similarity. It is worth noting that the mean value of the financial report similarity of listed companies reached the lowest point in 2012. This may be related to the revision of ‘Guideline No. 2’ by the China Securities Regulatory Commission. After the revision, there were large adjustments in the section order and disclosure content of annual financial reports compared with the previous year. Since 2012, ‘Guideline No. 2’ has emphasised the language specifications of MD&A disclosure: ‘Language expression shall be plain, clear and easy to understand’, ‘Do not use inane and stereotypical languages’. Table 4 reports the correlation coefficients of main variables. As can be seen from Table 4, since annual financial reports and non-MD&A of annual financial reports significantly CHINA JOURNAL OF ACCOUNTING STUDIES 155 Table 2. Variable definitions. Variable Variable type sign Variable definition Dependent variable Punish Dummy variable, if listed companies are punished by regulatory agencies in the current period, the value is 1, otherwise 0. Independent MD&A_Sim MD&A similarity between current and last period. Variable Other_Sim Non-MD&A similarity between current and last period. Control variables Size The natural logarithm of company total assets at the end of the year. Lev Total debt/total assets. ROA The total return on assets = net profit/total assets. AO Audit opinion dummy variable, if it is an unqualified opinion, the value is 0, otherwise 1. Analyst The natural logarithm of analysts’ report. H1 The largest shareholder’s share proportion. Board_size The size of the board = the natural logarithm of the number of directors. Dual Dummy variable, if the head of director board and general manager is the same person, the value is 1, otherwise, 1. Industry Industry dummy variable, there are 21 industries and 20 industry dummy variables according to CSRC 2001 industry classification. Year Year dummy variable, there are 8 Year dummy variables from 2008 to 2016. Table 3. Summary statistics. Variables Mean Min P25 P50 P75 Max sd Punish 0.107 0.000 0.000 0.000 0.000 1.000 0.310 MD&A_Sim 0.289 0.030 0.164 0.259 0.386 0.843 0.166 Other_Sim 0.451 0.016 0.266 0.439 0.620 0.953 0.237 Size 21.955 19.302 21.043 21.799 22.692 25.782 1.270 Lev 0.448 0.048 0.276 0.446 0.617 0.906 0.215 ROA 0.037 −0.166 0.012 0.034 0.063 0.197 0.053 AO 0.035 0.000 0.000 0.000 0.000 1.000 0.185 Analyst 2.630 0.000 1.386 2.890 3.932 5.421 1.620 H1 0.353 0.003 0.231 0.333 0.458 0.900 0.153 Board_size 2.258 0.000 2.197 2.303 2.303 2.996 0.233 Dual 0.239 0.000 0.000 0.000 0.000 1.000 0.426 0.700 0.647 0.600 0.500 0.505 0.499 0.461 0.446 0.435 0.411 0.400 0.380 0.357 0.328 0.309 0.300 0.296 0.267 0.264 0.255 0.248 0.244 0.200 0.170 0.100 0.000 2008 2009 2010 2011 2012 2013 2014 2015 2016 othersim mdasim Figure 1. Financial report textual similarity from 2008 to 2016. 156 A. QIAN AND D. ZHU Table 4. Correlated coefficient matrix of main variables. Punish Report_Sim Other_Sim MD&A_Sim Size Lev ROA AO Analyst H1 Board_size Dual Punish 1 Report_Sim −0.011 1 Other_Sim −0.011 0.943 1 MD&A_Sim 0.016 0.514 0.458 1 Size −0.027 0.150 0.130 0.045 1 Lev 0.054 0.060 0.058 −0.004 0.465 1 ROA −0.101 −0.079 −0.075 −0.048 −0.038 −0.409 1 AO 0.131 0.024 0.022 0.016 −0.122 0.115 −0.176 1 Analyst −0.070 −0.031 −0.032 −0.006 0.408 −0.106 0.420 −0.179 1 H1 −0.064 −0.056 −0.046 −0.066 0.210 0.061 0.093 −0.097 0.124 1 Board_size −0.016 0.005 0.003 −0.029 0.257 0.153 −0.001 −0.014 0.131 0.019 1 Dual 0.023 −0.035 −0.033 0.006 −0.164 −0.146 0.063 −0.008 0.049 −0.052 −0.180 1 CHINA JOURNAL OF ACCOUNTING STUDIES 157 overlap, the correlation coefficient between the overall similarity of an annual financial report (Report_Sim) and non-MD&A similarity (Other_Sim) is 0.94. MD&A is short and takes up a small proportion of an annual financial report. However, as the most important part of annual financial report, MD&A has certain links with other information disclosed in annual financial reports, such as financial statement data and accounting policies. The correlation coefficient between MD&A similarity (MD&A_Sim) and the overall similarity of annual financial report (Report_Sim) is 0.51. The correlation coefficient between MD&A similarity (MD&A_Sim) and non-MD&A similarity (Other_Sim) is 0.45, which indicates that there are no serious collinearity problems. The correlation coefficient between the penalty for fraud firms (Punish) and MD&A similarity (MD&A_Sim) is positive, and the correlation coefficient between the penalty for fraud firms (Punish) and non-MD&A similarity (Other_Sim)is negative, which is consistent with the expectations of Hypotheses 1 and 2. 4.2. Empirical results In order to test the influence of financial report similarity on the likelihood of adminis- trative punishment, we use model (1) for the regression test. The results in column (1) of Table 5 show that the coefficient of MD&A_Sim is equal to 0.153, which is significantly negative at the 10% level (Z = 1.73). This shows that the higher the MD&A similarity between the current and the previous period and the lower the information content, the higher the likelihood of being investigated and punished by regulatory agencies in the current period. The results in column (2) of Table 5 show that the coefficient of Other_Sim is equal to −0.123, which is significantly negative at the 10% level (Z = −1.89). This shows that the higher the non-MD&A similarity between the current and previous period, the lower the likelihood of being investigated and punished by regulatory agencies in the current period. Column (3) of Table 5 shows the regression results of MD&A_Sim and Other_Sim in the model at the same time. The coefficient of MD&A_Sim is equal to 0.234, which is significantly positive at the 5% level of significance (Z = 2.51), and the coefficient of Other_Sim is equal to −0.182, which is significantly negative at the 1% level of significance. The above regression results verify Hypotheses 1 and 2, namely, regulators pay more attention to the information content and decision- making usefulness of MD&A disclosure, and the stability and compliance of non-MD&A disclosure. 4.3. Robustness tests First, consider alternative measures of the main variables. Since the year and industry of listed companies may have a certain correlation with the textual information disclosure characteristics of financial reports, in the robustness test, we use textual similarity adjusted by the year–industry mean value for measurement. The regression results in column (1) of Table 6 show that the coefficient of MD&A_Sim is equal to 0.225, which is significantly positive at the 5% level of significance (Z = 2.38); the coefficient of Other_Sim is equal to −0.185, which is significantly negative at the 1% level of signifi- cance (Z = −2.64). The results of the robustness test show that the main regression conclusions are still valid. 158 A. QIAN AND D. ZHU Table 5. Textual similarity and enforcement penalty. Variables (1) (2) (3) MD&A_Sim 0.153* 0.234** (1.73) (2.51) Other_Sim −0.123* −0.182*** (−1.89) (−2.63) Size −0.046*** −0.044*** −0.043*** (−2.99) (−2.84) (−2.79) Lev 0.452*** 0.454*** 0.457*** (5.61) (5.63) (5.67) ROA −1.228*** −1.242*** −1.235*** (−4.14) (−4.19) (−4.16) AO 0.598*** 0.600*** 0.600*** (10.05) (10.07) (10.07) Analyst −0.028*** −0.029*** −0.030*** (−2.67) (−2.80) (−2.85) H1 −0.466*** −0.478*** −0.473*** (−4.96) (−5.09) (−5.04) Board_size 0.050 0.051 0.053 (0.85) (0.86) (0.90) Dual 0.068** 0.067** 0.065** (2.21) (2.18) (2.12) Constant −0.471 −0.429 −0.476 (−1.45) (−1.32) (−1.46) Year FE Yes Yes Yes Ind FE Yes Yes Yes N 17,661 17,661 17,661 0.058 0.058 0.059 Pseudo-R This table reports the results of financial report similarity on enforcement penalty. From columns 1 to 3, the independent variable is MD&A_Sim, Other_Sim, both MD&A_Sim and Other_Sim. All the variables are defined in Table 2. T-statistics used to signal the robustness of standard errors clustered at the firm level. ***, **, and *, indicating statistical significance at 1%, 5%, and 10% levels respectively. Second, we test the impact of overall financial report similarity on the likelihood of administrative punishment for fraudulent companies. The regression results in column (2) of Table 6 show that the coefficient of Report_Sim is equal to −0.167, which is significantly positive at the 1% level of significance (Z = −2.60). In column (3) of Table 6, MD&A similarity is added to the regression model. The coefficient of MD&A_Sim is equal to 0.297, which is significantly positive at the 1% level of significance (Z = 3.08), and the coefficient of Report_Sim is equal to −0.258, which is significantly negative at the 1% level of significance (Z = −3.63). The annual financial report is composed of MD&A and non-MD&A. Non-MD&A takes up a very high proportion of annual financial report. Therefore, the impact of overall annual financial report similarity on the administrative punishment is the same as that of non-MD&A. Third, we solve endogenous problems. There may have a problem of reverse causal relationship. Listed companies maybepunishedbyregulatoryagenciesfor violations in the previous periods. Due to the ‘deterrent effect’ of administrative punishment, listed companies have a motivation to adjust the information disclo- sure of financial reports and improve information disclosure quality in the current period (Fisch, 2009; Li, 2007). Therefore, we further consider whether listed com- panies have been punished by the regulatory authorities in the period t–1and t–2, CHINA JOURNAL OF ACCOUNTING STUDIES 159 Table 6. Robustness tests. Variables (1) (2) (3) (4) (5) MD&A_Sim 0.225** 0.297*** 0.275*** 0.237** (2.38) (3.08) (2.72) (2.54) Other_Sim −0.185*** −0.187** −0.182*** (−2.64) (−2.49) (−2.62) Report_Sim −0.167*** −0.258*** (−2.60) (−3.63) Size −0.043*** −0.043*** −0.041*** −0.056*** −0.044*** (−2.79) (−2.73) (−2.63) (−3.30) (−2.81) Lev 0.457*** 0.454*** 0.457*** 0.501*** 0.459*** (5.67) (5.63) (5.67) (5.61) (5.70) ROA −1.235*** −1.240*** −1.229*** −1.299*** −1.228*** (−4.17) (−4.18) (−4.14) (−3.95) (−4.14) AO 0.600*** 0.601*** 0.602*** 0.565*** 0.602*** (10.07) (10.09) (10.10) (8.13) (10.09) Analyst −0.030*** −0.030*** −0.031*** −0.018 −0.030*** (−2.85) (−2.88) (−2.97) (−1.56) (−2.85) H1 −0.473*** −0.483*** −0.480*** −0.480*** −0.472*** (−5.14) (−5.11) (−4.71) (−5.02) (−5.04) Board_size 0.053 0.052 0.056 0.063 0.055 (0.90) (0.88) (0.94) (0.97) (0.93) Dual 0.065** 0.067** 0.064** 0.043 0.065** (2.12) (2.16) (2.08) (1.29) (2.12) Agent_admin 0.097 (0.65) Agent_turnover 0.021 (0.71) Constant −0.504 −0.443 −0.511 −0.311 −0.501 (−1.55) (−1.37) (−1.57) (−0.87) (−1.53) Year FE Yes Yes Yes Yes Yes Ind FE Yes Yes Yes Yes Yes N 17,661 17,661 17,661 15,722 17,661 Pseudo-R 0.058 0.058 0.059 0.054 0.059 This table reports the results of robustness tests. All the variables are defined in Table 2. T-statistics used to signal the robustness of standard errors clustered at the firm level. ***, **, and *, indicating statistical significance at 1%, 5%, and 10% levels respectively. and exclude the listed companies punished by regulatory agencies in the period of t–1and t–2. Column (4) of Table 6 shows that the regression results of MD&A_Sim and Other_Sim are still consistent with the main regression conclusions after removing the sample of listed companies that have been penalised for violations in the past two periods. Finally, we exclude an alternative explanation of corporate governance quality. Corporate governance is an important factor affecting listed companies’ informa- tion disclosure behaviours and violations of laws and regulations (Cai & Wu, 2007; Lu, Zhu, & Hu, 2012;Wang etal., 2018). Financial report similarity may be an indication of some missing variables of corporate governance. For the above problem, we further control some confounding variables of corporate governance in the robustness test by adding agency costs as control variables. Referring to the study of Ang, Cole, and Lin (2000), agency costs are measured by the rate of asset turnover and administrative expenses. Column (5) of Table 6 shows that the regression results of MD&A_Sim and Other_Sim are consistent with the main regres- sion conclusions. 160 A. QIAN AND D. ZHU 5. Further analysis 5.1. The impact of ownership on financial report similarity and the likelihood of administrative punishment The information content of MD&A is the focus of regulators. As an emerging and transitional market economy, the difference of ownership causes significant differences between China’s state-owned enterprises and private enterprises in business objectives, resource acquisition, and regulatory supervision (Correia, 2014; Fan, Rui, & Zhao, 2008; Shleifer & Vishny, 1994). State-owned enterprises, regarded as the eldest sons of the People’s Republic of China, enjoy more advantages and privileges in the product market and factor market than private enterprises, and have stronger monopoly power in the market (Brandt & Li, 2003; Fang, 2007). Private enterprises are faced with unfair treat- ment and discrimination in terms of resource acquisition and policy support. Specifically, private enterprises often fail to raise funds through formal channels, which leads to losing opportunities to grow and develop because of insufficient funds or higher financing costs. Compared with state-owned enterprises, private enterprises are more motivated to disclose information out of financing needs and resource acquisition. According to the capital market trading motive hypothesis, the greater the information disclosure of a listed company, the greater the degree of transparency, which helps to alleviate information asymmetry between external shareholders and internal manage- ment and avoids being regarded as a defective product in the ‘second-hand car’ market, thereby reducing financing costs and increasing stock prices (Dhaliwal et al., 2011; Tetloc, 2011; Wang, Yu, & An, 2014). Private listed companies hope to disclose more information about a company’s current operating conditions and future development prospects through MD&A information disclosure, improve corporate transparency and reduce external financing costs. Therefore, in the process of enforcement, regulatory agencies pay more attention to the quality of MD&A information disclosure of private listed companies. MD&A information disclosure directly affects the efficiency of resource allocation in capital markets and non-public markets. For non-MD&A, regulators are mainly concerned with the compliance and stability of disclosures, especially for state-owned listed companies. On the one hand, state-owned enterprises take on some political burdens of maintaining economic growth, social stability, and employment. State-owned listed companies should strictly abide by the central government’s policy and requirements when appointing senior executives and accounting firms, disclosing the information of social responsibilities (Zeng & Chen, 2006; Zhou, 2007). If state-owned enterprises violate ‘Guideline No. 2’ by adjusting annual financial report disclosure without authorisation, that would be not in line with the needs of catering to ‘political correctness’ for state-owned enterprises. At the same time, if there is a big change in non-MD&A of state-owned enterprises in the current period, regulators will pay more attention and investigate whether corporate govern- ance structure, business strategy, has undergone major changes, and whether these changes have negative social spillover effects on employee employment and social stability. On the other hand, compared with state-owned enterprises, private enterprises have weaker resilience in the face of macroeconomic fluctuations and policy uncertain- ties. In order to cope with the crisis, private enterprises may adopt mergers and CHINA JOURNAL OF ACCOUNTING STUDIES 161 acquisitions to achieve a brand-new governance structure (Kato & Long, 2006; Zhou, 2007) or adjust the cost structure by means of salary adjustment (Ma & Zhang, 2013). The low similarity of non-MD&A has a warning effect of operational risks. Moreover, in the process of decision-making, private enterprises are less subject to policy guidance and constraints than state-owned enterprises, resulting in more aggressive business strategies (Liang & Yu, 2014). Above all, non-MD&A similarity of state-owned and private listed companies is the focus of regulatory authorities. In order to test the effect of the nature of property rights on the relationship between (non)MD&A similarity and administrative punishment, we add an interaction of owner- ship and similarity in model (1). The regression results in column (1) of Table 7 show that the interaction coefficient of MD&A similarity and ownership is −0.366 at the 5% level of significance, which indicates that the state property right weakens the relationship between MD&A similarity and the likelihood of administrative punishment. Regulators are more concerned with MD&A information content disclosed by private listed compa- nies than state-owned listed companies. The regression results in column (2) of Table 7 show that the interaction coefficient of non-MD&A similarity and ownership is −0.171, Table 7. The impact of ownership and textual readability. (1) (2) (3) (4) Variables Ownership Ownership Readability Readability MD&A_Sim 0.250** 0.216* (2.34) (1.78) Other_Sim −0.049 0.011 (−0.59) (0.12) SOE/Read −0.209*** −0.205*** 0.010 0.010 (−6.60) (−6.49) (0.25) (0.25) MD&A_Sim*SOE/Read −0.366* −0.162 (−1.83) (−0.89) Other_Sim*SOE/Read −0.171 −0.283** (−1.23) (−2.15) Size −0.026* −0.025 −0.049** −0.048** (−1.65) (−1.55) (−2.49) (−2.42) Lev 0.490*** 0.490*** 0.452*** 0.453*** (6.08) (6.08) (5.58) (5.59) ROA −1.300*** −1.310*** −1.231*** −1.242*** (−4.38) (−4.42) (−4.15) (−4.19) AO 0.589*** 0.591*** 0.599*** 0.600*** (9.90) (9.94) (10.05) (10.07) Analyst −0.040*** −0.041*** −0.028*** −0.029*** (−3.84) (−3.93) (−2.68) (−2.77) H1 −0.373*** −0.382*** −0.466*** −0.473*** (−3.96) (−4.06) (−4.97) (−5.03) Board_size 0.095 0.094 0.050 0.051 (1.53) (1.51) (0.85) (0.85) Dual 0.031 0.032 0.068** 0.067** (1.00) (1.03) (2.21) (2.16) Constant −0.869*** −0.897*** −0.383 −0.412 (−2.61) (−2.69) (−0.94) (−1.01) Year FE Yes Yes Yes Yes Ind FE Yes Yes Yes Yes N 17,661 17,661 17,661 17,661 Pseudo-R 0.062 0.062 0.058 0.059 This table reports the impact of ownership and textual readability on the relationship between (non) MD&A and enforcement penalty respectively. All the variables are defined in Table 2.T-statistics used to signal the robustness of standard errors clustered at the firm level. ***, **, and *, indicating statistical significance at 1%, 5%, and 10% levels respectively. 162 A. QIAN AND D. ZHU but is not significant. There is no significant difference among state-owned companies and private companies in the relationship between non-MD&A similarity and the like- lihood of enforcement penalty. Regardless of whether it is a state-owned or private listed company, the compliance and stability of financial report disclosure is an impor- tant factor when officials make decisions of enforcement penalty. 5.2. The impact of readability on financial report similarity and the likelihood of administrative punishment Klare (1963)defines readability as the ease of reading comprehension due to writing style. In accounting and finance research, readability is defined as the ability of an individual investor or analyst to absorb value-related information from financial disclo- sures. When there is a large number of professional terms or obscure statements in the financial report, it increases the difficulty for external information users to correctly read and understand financial reports, which leads to a lag in the investors’ response to financial report information disclosure (You & Zhang, 2009) and increases predictive divergence and uncertainty in analysts’ earnings forecasts (Lawrence, 2013). With the improvement of regulatory system and the intensifying of law enforcement, textual readability has even become an alternative way for company executives to conceal violations and confuse external information users with redundant disclosure. Li (2008) finds that when the company’s current profit is low, the readability of the financial report is worse. Managers make it impossible for external information users to make the right investment decisions based on the information disclosed in financial reports through lengthy and complicated language. Lo et al. (2017) find that accrued earnings management is significantly positively correlated with information complexity. Earnings management behaviour is covered by manipulating information complexity, making textual information disclosure an auxiliary tool for executives to gain personal gains. Due to the different emphasis of regulators on MD&A and non-MD&A information disclosure, the moderating effect of financial report readability also differentiates between MD&A similarity and non-MD&A similarity. Compared with non-MD&A, MD&A is shorter and more refined with more information content, which is an important information source for external investors and analysts to make evaluation and invest- ment decisions. Therefore, external information users will spend more time and energy analysing MD&A. Moreover, when the low readability impedes regulatory enforcement and judgement, regulators can also learn from investor’s investment decisions and analysts’ reports. Different from the MD&A section, the longer non-MD&A with lower information content draws less attention. In this case, textual readability becomes an important factor affecting the judgement of information users. Because information users have limited time and energy, more readable information could help information users understand information and improve information processing efficiency, which increases the influence of non-MD&A similarity on administrative punishment. Overall, textual readability has more significant influence on the relationship between non- MD&A similarity and the likelihood of administrative punishment than MD&A similarity. Regarding the measure of readability, the most commonly used is the Fog Index (Li, 2008). Since the Fog Index is a metric based on the spelling specifications of English words, whether it is applicable to Chinese remains to be verified. We employ the CHINA JOURNAL OF ACCOUNTING STUDIES 163 method of Loughran and Mcdonald (2014) by using the file size of an annual financial report as a measure of textual readability. And we standardise the measure by dividing a company’s total assets to overcome the impact of the company’s size on the size of financial report. We construct a dummy variable (Read) as a readability indicator by dividing the whole sample into two groups based on the median of standardised financial report size of each industry in each year. If the report size is below the median value, the value is equal to 1, which means financial report is more readable; otherwise, the value is equal to 0, which means financial report is less readable. The regression results in column (3) of Table 7 show that the interaction coefficient of MD&A similarity and text readability is 0.162, which is not significant. Textual readability has no signifi- cant effect on the relationship between MD&A similarity and the likelihood of adminis- trative punishment. The regression results in column (4) of Table 7 show that the interaction coefficient of non-MD&A similarity and textual readability is −0.283, which is significant at the 5% level. This shows that financial report readability aggravates the impact of non-MD&A similarity on the likelihood of administrative punishment for fraudulent companies. The above results indirectly support the findings of Li (2008) and Lo et al. (2017), explaining why textual readability becomes an alternative way for listed companies to conceal negative information and manipulation behaviour. 5.3. (Non)MD&A similarity in segments and administrative punishment In further analysis, we process the information of MD&A and non-MD&A. MD&A consists of two parts: Review and Preview. The Review part mainly includes business operations, profit composition, asset and liability situations and investment situations. Executives often explain the reasons for the changes in the current financial statement from the perspective of business operations, and disclose more about the major investment activities and business adjustments, which are related for users to make decisions in the current period. The Preview part focuses on the company’s future development strategy, next year’s business plan, and the risks the company may face. Compared with the Review part, the Preview part provides more forward-looking information to help external information users reduce uncertainty about the company’s future evaluation and it provides more incremental information related to decision-making (Meng et al., 2017; Xue et al., 2010). Therefore, the similarity of the Preview part is more powerful than that of the Review part for administrative punishment. Non-MD&A includes a wide range of content. ‘Changes in Shares and Shareholders’, ‘Directors, Supervisors, Senior Management and Employees’ and ‘Corporate Governance’ mainly disclose information about corporate governance. Prior literature finds that employee liquidity, executive personalities, board structure and shareholder structure are significantly related to corporate frauds (Aobdia, 2018; Beasley, 1996; Chen et al. 2005; Dechow et al., 1996; Gao et al., 2018; Gu & Liu, 2013). If the similarity of the corporate governance part decreases in the current period, it is more likely to attract the greater attention of regulators to conduct a deep investigation on the reasons of corporate governance restructuring. The footnotes of financial statements in non- MD&A primarily reflect the accounting policies and accounting estimates used by listed companies. Accounting policies have both mandatory and selective characteristics, and accounting policy changes have become a tool for earnings management (Zhang & Lu, 164 A. QIAN AND D. ZHU 2011). Even if the accounting policy changes of listed companies have been approved by the board of directors and auditors, the performance of the capital market is not good (Xie et al., 2017). It is easy to cause regulators and investors to doubt whether the company has fraudulent behavour. In order to test the influence of MD&A segment similarity on the probability of administrative punishment for listed companies, we regress model (1). The results in columns (1) and (2) of Table 8 show that the regression coefficient of Rev_Sim is equal to 0.020, but not significant; the regression coefficient of Pre_Sim is equal to 0.169, which is significantly positive at the 5% level (Z = 2.35). This shows that the information content of the MD&A Preview part has more significant impact than the Review part on whether the listed company is investigated and punished by regulatory authorities in the current period. This is consistent with the conclusion of Meng et al. (2017), which studies the relationship between MD&A similarity and the risk of stock price crash. The results in columns (3) and (4) of Table 8 show that the segment of corporate governance similarity in non-MD&A is significantly negatively correlated with the possibility of administrative Table 8. (Non)MD&A similarity in segments. (1) (2) (3) (4) Variables Review Preview Corporate governance Accounting policy Rev_Sim 0.020 (0.24) Pre_Sim 0.169** (2.35) Gov_Sim −0.186** (−2.25) Acc_Sim −0.130** (−2.12) Size −0.043*** −0.043*** −0.043*** −0.044*** (−2.69) (−2.73) (−2.71) (−2.79) Lev 0.469*** 0.479*** 0.469*** 0.470*** (5.63) (5.75) (5.67) (5.68) ROA −1.280*** −1.278*** −1.209*** −1.184*** (−4.15) (−4.14) (−3.99) (−3.91) AO 0.639*** 0.639*** 0.613*** 0.613*** (10.15) (10.17) (10.13) (10.05) Analyst −0.027** −0.027** −0.027*** −0.029*** (−2.53) (−2.55) (−2.60) (−2.77) H1 −0.479*** −0.472*** −0.500*** −0.469*** (−5.00) (−4.93) (−5.21) (−4.87) Board_size 0.079 0.080 0.045 0.086 (1.29) (1.30) (0.73) (1.39) Dual 0.058* 0.056* 0.064** 0.076** (1.83) (1.78) (2.02) (2.40) Constant −0.523 −0.558* −0.435 −0.522 (−1.58) (−1.68) (−1.32) (−1.57) Year FE Yes Yes Yes Yes Ind FE Yes Yes Yes Yes N 16,804 16,804 16,840 16,840 Pseudo-R 0.058 0.059 0.059 0.059 This table reports the impact of ownership and textual readability on the relationship between (non)MD&A and enforcement penalty respectively. All the variables are defined in Table 2. T-statistics used to signal the robustness of standard errors clustered at the firm level. ***, **, and *, indicating statistical significance at 1%, 5%, and 10% levels respectively. Sample reduction is because machine learning cannot accurately recognise the text of ‘Review’, ’Preview’, ‘Corporate Governance’, ‘Accounting Policy’. CHINA JOURNAL OF ACCOUNTING STUDIES 165 punishment (β = −0.186, Z = −2.25). The relationship between the segment of account- ing policy similarity in non-MD&A and the possibility of administrative punishment is also significantly negatively correlated at the level of 5% significance (β = −0.130, Z = −2.12). The above regression results further support Hypotheses 1 and 2. Regulators are more concerned about MD&A’s information content and usefulness in decision-making, especially with regard to the forward-looking information in MD&A, while regulators emphasise the stability and compliance of non-MD&A, especially for the stability and compliance of corporate governance and accounting policies. 6. Conclusion Information disclosure in financial reports and its economic consequences have always been a core issue in accounting and finance research. The popularity of computer technology and artificial intelligence allows us to further analyse accounting texts and study related accounting and financial issues. This paper studies the economic conse- quences of financial report similarity from the perspective of regulatory enforcement for fraud-listed companies. The conclusions show that a greater similarity between the listed company’s current and previous MD&A leads to a greater probability of being investigated and punished by the regulatory authorities in the current period, while a lower similarity between current and previous non-MD&A leads to a greater probability of being investigated and punished by the regulatory authorities in the current period. Regulators have a different emphasis on listed companies’ MD&A and non-MD&A information when conducting enforcement measures. More specifically, regulators place a greater emphasis on the information content of MD&A and the stability and compliance of non-MD&A. Further analysis finds that owing to the low dependence on external financing and resource acquisition, regulators lower the regulatory require- ments for state-owned listed companies’ MD&A disclosure. State-owned ownership significantly weakens the relationship between MD&A similarity and the likelihood of administrative punishment. Since non-MD&A texts are longer and less important than MD&A texts, information users are not willing to spend more time on non-MD&A texts. Better textual readability exacerbates the impact of non-MD&A similarity on the like- lihood of administrative punishment, while readability has no significant impact on the relationship between MD&A similarity and the likelihood of administrative punishment. Finally, this paper examines the impact of MD&A’s ‘Review’ and ‘Preview’, non-MD&A’s ‘Corporate Governance’ and ‘Accounting Policy’ similarity on the likelihood of adminis- trative punishment respectively. The conclusions of this paper still stand. The conclusions of this paper have important practical significance. First, the conclu- sions of this paper prove the usefulness of financial report textual information from the perspective of regulatory enforcement. Regulators and other external information users should be fully aware of the importance of listed companies’ textual information disclosure, and use computer natural language processing to fully exploit more valuable accounting information and economic facts behind the texts to reduce information asymmetry and protect the legitimate interests of investors. Second, more information disclosure in financial reports is helpful for external information users to understand the company’s operating situations and improve the information asymmetry between com- panies and external information users. As the listed companies’ annual financial reports 166 A. QIAN AND D. ZHU are getting longer and longer, each section of financial report can be differentiates with regard to in disclosure information and textual characteristics. Owing to the limited cognitive ability and energy, information users cannot generalise the information in different sections of financial reports. Only by fully grasping the key disclosed informa- tion and textual features of different sections can information users reduce information processing costs and improve the information usefulness for decision-making. Finally, in the process of financial report textual processing, we find that some listed companies fail to comply with ‘Guideline No. 2’ when completing annual financial reports, such as the use of traditional Chinese characters, and inconsistency in section titles. Chinese regulatory authorities should improve the requirements for information disclosure of listed companies, clearly define the disclosure contents and specifications for financial reports and other disclosure documents, further guide and regulate the information disclosure behaviour of listed companies, and promote the stable and healthy develop- ment of the capital market in China. Acknowledgments We are grateful for numerous helpful comments from Yanli Chen and Yu Xin at the 2018 CJAS conference, two anonymous reviewers, and Kangtao Ye (Associate editor). Disclosure statement No potential conflict of interest was reported by the authors. References Aobdia, D. (2018). Employee mobility, noncompete agreements, product-market competition, and company disclosure. Review of Accounting Studies, 23, 296–346. Ang, J.S., Cole, R.A., & Lin, J.W. (2000). Agency costs and ownership structure. Journal of Finance, 55 (1), 81–106. Beasley, M. S. (1996). An empirical analysis of the relation between the board of director composi- tion and financial statement fraud. The Accounting Review, 71(4), 443–465. Bozanic, Z., & Thevenot, M. (2015). Qualitative disclosure and changes in sell-side financial analysts’ information environment. Contemporary Accounting Research, 32(4), 1595–1616. Brandt, L., & Li, H. (2003). Bank discrimination in transition economics: Ideology, information, or incentives? Journal of Comparative Economics, 31(3), 387–413. Brown, S., & Tucker, J.W. (2010). Large-sample evidence on firms’ year-over-year MD&A modifica- tions. Journal of Accounting Research, 49(2), 309–346. Bryan, S.P. (1997). Incremental information content of required disclosures contained in manage- ment discussion and analysis. The Accounting Review, 72(2), 285–301. Cai, Z., Wu, S. (2007). An empirical study on predicting illegal disclosure of listed companies in China from the perspective of finance, market and governance. Management Review, 19(1), 25– 33. (in Chinese). Chen, G., Lin, H, & Wang, L. (2005). Corporate Governance, Reputation Mechanism and the Behavior of Listed Firms in Committing in Fraud. Nankai Business Review, 8(6), 35–40. (in Chinese). Cheng, X.S., Tan, Y.C., & Liu, J.M. (2012). The non-financial information, the external financing, and the investment efficiency: A study based on the constraint of external systems. Management World, 7, 137–150. (In Chinese). CHINA JOURNAL OF ACCOUNTING STUDIES 167 Clarkson, P.M., Kao, J.L., & Richardson, G.D. (1999). Evidence that management discussion and analysis (MD&A) is a part of a firm’s overall disclosure package. Contemporary Accounting Research, 16(1), 111–134. Cole, C.J., & Jones, C.J. (2004). The usefulness of MD&A disclosures in the retail industry. Journal of Accounting, Auditing and Finance, 19(4), 361–388. Cormier, D., & Magnan, M. (2014). The impact of social responsibility disclosure and governance on financial analysts’ information environment. Corporate governance, 14(4), 467–484. Correia, M.M. (2014). Political connections and SEC enforcement. Journal of Accounting & Economics, 57(2–3), 241–262. Dai, Z.Y., & Yang, X.W. (2006). Indirect law-enforcement cost, indirect harm and selective law- enforcement. Economic Research Journal, 9,94–102. (In Chinese). Dechow, P., Sloan, R., & Sweeney, A. (1996). Causes and consequences of earnings manipulation: An analysis of firms subject to enforcement actions by the SEC. Contemporary Accounting Research, 13,1–36. Dhaliwal, D.S., Li, O. Z., Tsang, A., & Yang, Y.G. (2011). Voluntary non-financial disclosure and the cost of equity capital: The initiation of corporate social responsibility reporting. The Accounting Review, 86(1), 59–100. Dhaliwal, D.S., Radhakrishnan, S., Tsang, A., & Yang, Y.G. (2012). Non-financial disclosure and analyst forecast accuracy: international evidence on corporate social responsibility disclosure. The Accounting Review, 87(3), 723–759. Fan, J., Rui, O., & Zhao, M. (2008). Public governance and corporate finance: Evidence from corruption cases. Journal of Comparative Economics, 36(3), 343–364. Fang, J.X. (2007). Ownership, institutional environment and capital allocation. Economic Research Journal, 12,82–92. (In Chinese). Fang, V.W., Huang, A.H., & Karpoffthe, J.M. (2016). Short selling and earnings management: A controlled experiment. Journal of Finance, 71(3), 1251–1294. Feldman, R., Govindaraj, S., Livnat, J., Segal, B. (2010). Management’s Tone change, Post earnings announcement drift and accruals. Review of Accounting Studies, 15(4), 915–953. Franco, G., Hope, O., Vyas, D., & Zhou, Y. (2015). Analyst report readability. Contemporary Accounting Research, 32(1), 76–104. Gao, H., Zhang, H., & Zhang, J. (2018). Employee turnover likelihood and earnings management: Evidence from the inevitable disclosure doctrine. Review of Accounting Studies, 23, 1424–1470. Hanley, K. W., Hoberg, G. (2010). The information content of IPO prospectuses. Review of Financial Studies, 23(7), 2821–2864. Hao, X.C., & Su, Z.X. (2014). Can primary risk disclosure mitigate IPO Underpricing? Evidence from text analysis. Journal of Finance and Economics, 5,42–53. (In Chinese). Hoberg, G., & Phillips, G. (2010). Product market synergies and competition in mergers and acquisitions: A text-based analysis. Review of Financial Studies, 23(10), 3773-3811. Hoberg, G., & Phillips, G. (2017). Conglomerate industry choice and product language. Management Science, 64(8), 1–21. Hu, X., Wang, X., & Xin, B. (2015). Insider trading: Does being a neighbor of the securities and exchange commission matter? Managerial and Decision Economics, 38(1), 209–247. Hu, Y.M., & Tan, Y.C. (2013). The non-fnancial information disclosure: Literature Review and future research prospects. Accounting Research, 3,20–26. (In Chinese). Jiang, F., & Kim, K.A. (2015). Corporate governance in China: A modem perspective. Journal of Corporate Finance, 32(3), 190–217. Jiang, Y.H., Ma, C.Q., & Xiong, X.X. (2014). The GEM listed corporations inertia disclosure, informa- tion similarity and asset pricing–empirical test based on improved fama - French model. Chinese Journal of Management Science, 8,56–63. (In Chinese). Kato, T., & Long, C. (2006). Executive turnover and firm performance in China. American Economic Review, 96(2), 363–367. Khanna, V., Kim, E.H., & Lu, Y. (2015). CEO connectedness and corporate fraud. Journal of Finance, 70(3), 1203–1252. Klare, G. R. (1963). The measurement of readability, Iowa State University Press. 168 A. QIAN AND D. ZHU Kryzanowski, L., & Ying, Z. (2013). Financial restatements and Sarbanes Oxley: Impact on Canadian firm governance and management turnover. Journal of Corporate Finance, 21(5), 87–105. Lawrence, A. (2013). Individual investors and financial disclosure. Journal of Accounting & Economics, 56(1), 130–147. Lee, Y. (2012). The effect of quarterly report readability on information efficiency of stock prices. Contemporary Accounting Research, 29(4), 1137–1170. Lennox, C., & Pittman, J.A. (2010). Big five audits and accounting fraud. Contemporary Accounting Research, 27(1), 209–247. Li, W. (2007). The economic consequence of government regulation on accounting information disclosure: Empirical findings from Chinese Security Market. Accounting Research, 8,13–21. (in Chinese). Li, F. (2008). Annual report readability, current earnings, and earnings persistence. Journal of Accounting and Economics, 45(2–3), 221–247. Liang, Q., & Yu, F. (2014). Financial crisis, state ownership and capital investment. Economic Research Journal, 49(4), 47–61. (in Chinese). Lin, L., & Xie, D. (2017). Do analysts recommendation revisions use management tones? A textual analysis based on earnings communication conferences. Management World, 11, 125–145. (in Chinese). Lo, K., Ramos, F., & Rogo, R. (2017). Earnings management and annual report readability. Journal of Accounting and Economics, 63(1), 1–25. Loughran, T., & Mcdonald, B. (2014). Measuring readability in financial disclosures. Journal of Finance, 69(4), 1643–1671. Lu, Y., & Hu, J.Y. (2016). The impact of hometown connectedness between the CEO and board of directors on corporate fraud. Nankai Business Review, 2,52–62. (In Chinese). Lu, Y., Zhu, Y.J., & Hu, X.Y. (2012). Institutional shareholding and corporate fraud: Evidence from China. Nankai Business Review, 1,13–23. (In Chinese). Ma, Y., & Zhang, Z. (2013). A study on financial crisis shocks, managerial earnings incentive and cost and expense stickiness. Nankai Business Review, 16 (6), 70–80. (in Chinese). Meng, Q.B., Yang, J.H., & Lu, B. (2017). The informative content of management discussion and analysis and stock price crash risk——based on text vectorization method. China Industrial Economics, 12, 132–150. (In Chinese). Merkley, K.J. (2014). Narrative disclosure and earnings performance: Evidence from R&D disclo- sures. The Accounting Review, 89(2), 725–757. Miller, B.P. (2010). The effects of reporting complexity on small and large investor trading. The Accounting Review, 85(6), 2107–2143. Price, S.M., Doran, J.S., Peterson, D.R., & Bliss, B.A. (2011). Earnings conference calls and stock returns: The incremental informativeness of textual tone. Journal of Banking & Finance, 36(4), 992–1011. Reuven, L., Li, F., & Kenneth, M. (2011). The effect of annual report readability on analyst following and the properties of their earnings forecasts. The Accounting Review, 86(3), 1087–1115. Shleifer, A., & Vishny, R.W. (1994). Politicians and firms. Quarterly Journal of Economics, 109(4), 995– Teng, F., Xin, Y., & Gu, X.L. (2016). Product market competitiveness and corporate fraud. Accounting Research, 9,32–40. (In Chinese). Tetloc, P.C. (2011). All the news that’s fit to reprint: Do investors react to stale information? Review of Financial Studies, 24(5), 1481–1512. Wang, T., Winton, A., & Yu, X. (2010). Corporate fraud and business conditions: Evidence from IPOs. Journal of Finance, 65(6), 2255–2292. Wang, Y.Y., Yu, L.S., & An, R. (2014). Does non-financial information improve information environ- ment? Evidence from disclosure of corporate social responsibility. Journal of Financial Research, 8, 178–191. (In Chinese). Wang, K., Wang, H.,Li, D., & Dai, X. (2018). Complexity of annual report and management self- interest: Empirical evidence from Chinese listed firms. Management World, (12), 120–132. (in Chinese). CHINA JOURNAL OF ACCOUNTING STUDIES 169 Xie, D.R., & Lin, L. (2015). Do management tones help to forecast firms’ future performance: a textual analysis based on annual earnings communication conferences of listed companies in China. Accounting Research, (2), 220–227. (In Chinese). Xie, D., Liao, K., & Z,D. (2017). Controlling Shareholder's Share pledge and implicit choice of accounting policy for R&D. Accounting Research, (3), 30–38. (in Chinese). Xu, X., & Tang, Q. (2010). Analyst Following and Corporate Earnings Management: Evidence from China. Journal of Financial Research, (7), 124-139. (in Chinese). Xue, S., Xiao, Z.Z., & Pan, M.L. (2010). Does management discussion and analysis provide useful information? An empirical exploration based on losing listed companies. Management World, 5, 130–140. (In Chinese). You, H., & Zhang, X. (2009). Financial reporting complexity and investor underreaction to 10-K information. Review of Accounting Studies, 14(4), 559–586. Zeng, Q.S., & Chen, X.Y. (2006). State stockholder, excessive employment and labor cost. Economic Research Journal, 5,74–86. (In Chinese). Zhang, H., & Lu, C. (2011). Stock option awards and accounting policy choices: Analysis based on asset impairment accounting. Journal of Finance and Economics, 37(7), 60–70. (in Chinese). Zhou, L.A. (2007). Governing China’s local officials: An analysis of promotion tournament model. Economic Research Journal, 7,36–50. (In Chinese).
Journal
China Journal of Accounting Studies
– Taylor & Francis
Published: Apr 3, 2019
Keywords: Administrative punishment; similarity; textual analysis; information disclosure
