Cooking The Books

Journal of Corporate Finance 17 (2011) 371390
Contents lists available at ScienceDirect
Journal of Corporate Finance

j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j c o r p f i n
Cooking the books: Recipes and costs of falsied nancial statements in China
Michael Firth a,, Oliver M. Rui b,1, Wenfeng Wu c,2
a b c
Department of Finance and Insurance, Lingnan University, 8 Castle Peak Road, Tuen Mun, NT, Hong Kong, China School of Accountancy, Chinese University of Hong Kong, Hong Kong Antai College of Economics and Management Shanghai Jiaotong University, Shanghai, China
a r t i c l e
i n f o
a b s t r a c t
We examine the causes and consequences of falsified financial statements in China. Using bivariate probit regression analysis, we find that firms with high debt and that plan to make equity issues are more likely to manipulate their earnings and thus have to restate their financial reports in subsequent years. We also find that corporate governance structures have an effect on the occurrence and detection of financial fraud. There are significant negative consequences to fraudulent financial statements. Restating firms suffer negative abnormal stock returns, increases in their cost of capital, wider bidask spreads, a greater frequency of modified audit opinions, and greater CEO turnover. We also find that firms located in highly developed regions suffer more severe consequences when they manipulate their accounts. 2010 Elsevier B.V. All rights reserved.
Available online 25 September 2010 JEL classication: G14 K22 M41
Keywords: Fraud Financial restatements Regional development Corporate governance Causes and consequences of restatements
1. Introduction High quality nancial information is a necessary condition for an efcient and vibrant stock market. However, trying to measure quality is a challenging task for researchers. The quality of nancial statements is often examined with reference to earnings management or earnings quality.3 However, the measurement of earnings management and earnings quality as done in accounting studies does not provide direct evidence that managers have manipulated earnings (Agrawal and Chadha, 2005). In contrast, a nancial restatement is often a direct admission by managers of false accounting and nancial misrepresentation. Restatements represent corrections to previously-issued nancial statements and these corrections usually occur because of accounting manipulations in prior years. Thus, the restatements are prima facie evidence of low quality nancial information disclosures in prior periods. We examine nancial restatements made by listed rms in China. In particular, we investigate the characteristics of rms that make restatements in order to understand why they occur. We then examine the consequences of restatements. While there are several research studies on restatements, they mainly use U.S. data and many of them relate to violations of accounting principles
Corresponding author. Tel.: + 852 2616 8950; fax: + 852 2462 1073. E-mail addresses: marth@ln.edu.hk (M. Firth), oliver@baf.msmail.cuhk.edu.hk (O.M. Rui), wfwu@sjtu.edu.cn (W. Wu). 1 Room 1010, Teaching Building at Chak Cheung Street, Faculty of Business Administration, The Chinese University of Hong Kong, Shatin, Hong Kong, China. Tel.: + 852 2609 7594; fax: + 852 2603 5114. 2 Antai College of Economics and Management, Shanghai Jiao Tong University, No. 535 Fahuazhen Road, Shanghai 200052, China. Tel.: + 86 21 52301194; fax: + 86 21 52301087. 3 Schipper and Vincent (2003) posit that earnings quality is multi-dimensional. Possible measures of earnings quality include persistence of earnings, value relevance of earnings, ability of earnings to predict future cash ows, and the conservative recognition of earnings. 0929-1199/$ see front matter 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.jcorpn.2010.09.002
372
M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390
(e.g., Anderson and Yohn, 2002; Hennes et al., 2008; Palmrose et al., 2004; Plumlee and Yohn, 2009).4 In contrast, we investigate falsied accounts rather than technical violations of accounting standards or unintentional errors of omission. The deliberate falsication of the accounts is a type of nancial fraud. Given the different historical, legal, and institutional backgrounds between China and the U.S. (Allen et al., 2005), we should not automatically impute or generalize the ndings from the latter to the former. Nevertheless, there are some similarities between the two countries, not least of which is China's willingness to adopt or modify the best governance practices from the developed countries, and this allows us to use the U.S. studies as a point of reference in our research. In addition to examining nancial statement misrepresentation in a different market setting, we also extend the literature in two other important ways. First, we use a two-stage procedure that examines the propensity to falsify the accounts and the reasons for its subsequent discovery and disclosure. This methodological advance is important as it allows us to gain a better understanding of the forces behind nancial restatements.5 Second, we recognize that there are substantial differences in economic and market development across the different regions of China and this can have a profound effect on our results. We therefore incorporate these regional differences into our research design.6 Restatements of the nancial accounts are the result of a multi-stage process. The rst stage is the decision to falsify the statements and therefore commit nancial fraud. Subsequent stages include the discovery of the false accounting and the reporting of it in a restatement. However, most prior studies use a basic probit or logit model of restatements and do not differentiate between the different stages leading to the revelation of nancial fraud (Dechow et al., 2010). One innovation of our study is that we use a two-stage process where we model the propensity to commit accounting fraud and, in the second stage, model the reporting of the fraud in a subsequent period. To identify the characteristics of, and the motivations behind, the propensity to falsify the nancial statements and its subsequent disclosure, we use the bivariate probit regression with partial observability technique. This approach allows us to overcome the problem of distinguishing between the motives to manipulate the nancial statements and the subsequent detection and reporting of the falsied accounts. Recent research has focused on institutional factors to explain earnings quality and differences in quality across countries (e.g., Bushman et al., 2004; Raonic et al., 2004). Institutional factors include the prevailing economic conditions, the governance of rms, institutional and regulatory frameworks, and the legal environment (e.g., investor protection and legal enforcement). One little understood characteristic of China's reforms is the very uneven distribution of economic and legal development across the country. Natural resources and human capital resources account for some of the regional differences in development but political connections with the country's leadership elite are also very important. Great variations in regional per capita income and education levels are one manifestation of the vast differences in development across the country. We argue that the differences in regional development have signicant effects on the incidence of, and consequences of, falsied accounts. In particular, we believe nancial statement quality is higher for those rms located in well-developed provinces and consequently investors rely heavily on nancial statements in their decision-making. This implies that the market will react strongly to the nancial restatements of rms located in well-developed regions but there will be a much more muted response in poorly developed regions as investors have already discounted the low quality of nancial reports. To test our beliefs, we explicitly account for market development using a set of indexes designed to capture differences in economic, political, legal, and institutional factors across regions. The advantage of conducting inter-regional studies within one country is that we can capture the effect of institutions on the quality of nancial information free of contamination due to country differences in reporting and disclosure rules, taxation, and bankruptcy laws. China has now emerged from the embryonic stage of capitalism and become a major economic power with signicant inward investment from other parts of the world. Portfolio investment from the U.S. and elsewhere is accelerating as international investors search for high growth markets with substantial domestic demand. One impediment to this growth, however, is investors' lack of knowledge about China's capital markets environment and the quality of nancial information. These concerns are real as high quality nancial information is a major factor in assuring the supply of inward international investment. Our study is therefore important as it yields key insights into the quality of nancial information by making an in-depth study of the causes and consequences of falsied accounts. We establish reasons that lead rms to manipulate their nancial statements and we examine whether there are safeguards or governance features that help reduce the accounting fraud from taking place. We also investigate the inuences that lead to the detection and disclosure of the false accounting. We provide a comprehensive examination of the consequences of restatements and this contrasts with prior research studies, which tend to focus on one or two consequences at a time (e.g., stock returns or executive turnover) while ignoring others. Examining market consequences is important as an absence of them may imply that investors ignore nancial information altogether. We identify 813 restatements in the period 2000 to 2005. Of these cases, 542 relate to the corrections of non-nancial information and we do not consider these any further in our study. The remaining 271 cases involve corrections to the income statement and the balance sheet. On average, about 3.7% of listed rms restate their nancial numbers each year. We nd evidence that rms are more likely to manipulate their nancial statements if they make equity issues, are highly levered, are controlled by the central government, and are located in less developed regions. This proclivity to restate is tempered, however, if a rm has a high percentage of directors
There are also studies on accounting errors and fraud uncovered in the SEC's enforcement actions in the U.S. (e.g., Karpoff et al., 2009). As we discuss later in Section 3.4, there are practical problems in implementing the bivariate probit model and these need to be overcome to achieve meaningful results. 6 As we will review and discuss later, there are several studies on nancial restatements in China. These studies do not use a two-stage process to explain restatements and do not control for a region's economic and market development. Furthermore, these other studies limit their examination of consequences of restatements to stock returns whereas we examine a large number of consequences. For the stock returns analysis, we examine the rst announcement of false accounting rather than its subsequent disclosure as a restatement in the annual report.
5 4
373
with nancial expertise. Given that false accounting has occurred, disclosure of it through the medium of restatements is more likely when the board is independent and the rm is non-government controlled and located in a more developed region. The restatements have important consequences for rms. First, there is a signicant negative stock return when the accounting fraud is announced. Second, the restating rm's cost of capital increases and its bidask spread widens. Third, rms that are forced to restate their accounts nd it harder to sell more shares after the restatement. Thus, capital markets extract a cost from rms that engage in false accounting. Fourth, there is increased uncertainty about the veracity of future nancial statements, which is manifested by an increase in the proportion of modied audit opinions. Fifth, top management turnover increases implying there are costs to managers as well as to the stockholders. In general, these consequences are much more pronounced for rms located in better-developed provinces. Investors expect high quality nancial statements in the better-developed provinces and so when rms are forced to restate their earnings the market punishes them severely. We organize the remainder of the paper as follows. The next section briey reviews China's economic reforms, regional development, and the rules regarding restatements. Section 3 discusses the research design and the sample. The characteristics of the restating rms are discussed in Section 4 and the consequences of restatements are described in Section 5. Conclusions are presented in the last section. 2. China's economic reforms and fraudulent nancial statements China began its transition from a centrally planned socialist system to a market-based economy in the late 1970s. The transition can be characterized as a gradualist approach with the reforms slowly unfolding (Chen et al., 2006a). In 1990 and 1991, China opened its two stock exchanges in Shanghai and Shenzhen, respectively, and there are now more than 1600 listed rms whose combined market capitalization is the second largest in the world. The central or local government and wholly state owned companies (SOEs) are often the major stockholders in many listed rms. A major challenge for the economic reforms has been the need to develop the legal and nancial infrastructure necessary for private ownership and stock market investment. To his end, China has copied, with modications, the best practices from the U.S. and other developed nations. For example, accounting standards have been introduced (the rst standard appeared in 1993) and strong progress is being made to convert them to international standards. In 1998/9, auditing rms disafliated from their former state owners and became independent. The rst auditing standards were introduced in 1995 and they have started to reect international norms and practices. The government has promoted the development of mutual funds and investment banks to provide nancial services and advice to individuals and corporations. China's regulator for securities markets and listed rms, the China Securities and Regulatory Commission (CSRC) is modeled, in part, on the SEC in the U.S. and the Securities and Futures Commission (SFC) in Hong Kong. The CSRC monitors listed rms and other stock market participants, which include banks, accountants, and auditors. The CSRC investigates companies and their nancial statements on a regular basis and in special cases when there are allegations of wrong-doing (Chen et al., 2005). Just like the SEC in the U.S., the CSRC has limited resources and only pursues those cases where it believes it has a strong chance of proving guilt. Despite the gradualist approach to the economic reforms, bottlenecks have appeared that impede progress. A major constraint is the lack of experienced and qualied personnel, a problem that pervades the legal, economic, nancial, business, and accounting/auditing sectors. Another problem area is the lack of good ethical behavior by business executives, which is due to ignorance and weak law enforcement that makes people think they can get away with fraud and other wrong-doing. While companies in the developed world have evolved systems of checks and balances and developed good corporate governance mechanisms in order to deter fraudulent nancial reporting,7 these systems are more rudimentary in China. In some cases, good governance systems exist on paper but they are not implemented in practice. In some other cases, the implementation is perfunctory and amounts to little more than paying lip service to the ideals of good governance. Financial restatements occur because of incompetence or the deliberate manipulation of nancial statements in prior periods. 2.1. Financial statement fraud The CSRC requires listed rms to restate the nancial statements when errors are detected. In about 67% of cases (542 out of 813 cases), the restatement consists of correcting information relating to ownership, top management, directors, and other nonnancial-fraud matters. However, we limit ourselves in this study to an investigation of false nancial reporting, where the income statement and balance sheets have been corrected for previous manipulations.8 Companies are required to disclose the restated nancial statements as a Material Events Special Alert. The Special Alert is sent to the rm's shareholders, the stock exchanges, and published in nancial newspapers and the company's website. Immediate disclosure is required. The restated nancial accounts have to be audited. The accounting errors may have been identied by a number of sources including the CSRC, the Chinese Institute of Certied Public Accountants (CICPA) via their monitoring and quality assurance reviews, the rm's auditors, or the rm itself. The Special Alerts rarely state which source identied the accounting error. The Alerts, however, do give details of the manipulated accounts, the money amounts,9 and dates. Appendix A summarizes the regulations covering restatements and
However, this has not prevented all fraud as recent scandals attest (e.g., Enron, Worldcom, Parmalat, Royal Ahold, and HIH). Our sample is restricted to deliberate distortions or falsications by managers. 9 Unfortunately, the level of disclosure of the money amounts involved (e.g., the distortion in net income) varies across cases. The lack of a consistent disclosure of money amounts means we cannot use this information in our cross-sectional tests.
8 7
374
Appendix B gives an example of a Special Alert issued by WeiDa Medical Applied Technology Co. Ltd. WeiDa used a fraudulent accounting treatment for shutdown losses and so reported a prot in 2002. The prot enabled WeiDa to come out of its Special Treatment designation. However, after the restatement, the prot became a loss and WeiDa was reclassied by the CSRC and the Shenzhen Stock Exchange as a Special Treatment rm. The Special Treatment (ST) moniker means the rm has two years of losses and risks being delisted if the losses continue for a third year. The ST designation is a warning to both the rm and to investors and is something rms want to avoid and something that may drive the managers to falsify the nancial statements. Financial restatements in China have also been studied by Zhou and Ma (2005), Li (2008) and Wu and Wang (2009). Zhou and Ma (2005) and Li (2008) examine all restatements including those of a minor nature that do not include falsied accounts. All three studies use a basic one-stage logit model. Between them, the studies nd that rm size, nancial leverage, ownership, losses, and auditor have associations with the likelihood of a restatement. Li (2008) reports a negative stock return of 2.26% that accompanied the announcement of the restatement. In contrast, Wu and Wang (2009) reported a non-signicant three day abnormal return of 0.48%; these returns were measured at the same time the annual earnings were announced and so it is difcult to disentangle the impact of the restatement from the concurrent earnings announcement for the latest year. None of these studies used a two-stage model and so they mix up the propensity to falsify the accounts stage and the detection and disclosure of fraud stage. Moreover, they limit their analysis of the consequences to stock returns and do not control for regional differences in economic and legal development. Our study therefore extends prior China-based research on restatements in terms of research method (two-stage model), control variables (e.g., the impact of regional development), and an extensive analysis of economic consequences.
2.2. Regional development While China's reforms have led to rapid national economic growth and a substantial increase in personal wealth, these gains are not evenly spread throughout the country (Demurger, 2001; Demurger et al., 2002; Tsui, 1996).10 In addition, the implementation of legal and nancial markets reforms have not been consistently applied throughout China.11 For example, the coastal regions and the major municipalities (e.g., Guangdong, Beijing, and Shanghai) have developed much faster than the western and inland provinces. In part, the regional disparities reect the preferences of China's top leaders (e.g., the leaders' favor the region they come from or where they built their career). We believe that the reasons for, and the consequences of, nancial restatements will depend, in part, on where the company is located. For example, the skill and experience levels of nancial executives and people's views of ethics will likely differ across regions depending on the degree of market development of the region. Most rms have a dominant investor and that investor or its representative is located in the same province as the rm. The dominant investors have a strong inuence on the quality of the rm's nancial statements. These investors' views on ethics, governance, and attitudes towards minority shareholders will vary depending on where they come from. Likewise, the judicial system, including, importantly, law and regulatory enforcement, varies a lot across regions and this will have an impact on a rm's incentives to manipulate the nancial statements. In particular, we expect that false accounting will be more likely for those rms located in provinces with low legal and economic development (Xia and Fang, 2005; Sun et al., 2005). Investors recognize that different levels of development may have an impact on the quality of nancial reporting and so they factor this into their decision-making. Assuming that investors believe the nancial reporting quality of a rm located in a poor developed province is low then they will not be surprised to see a restatement (compared to rms located in a well-developed province) and there will be fewer adverse consequences. It is therefore important that we account for market development in our analyses. To study the effect of regional development we make use of an index of market intermediaries and legal environment (including investor protection and protection of property rights). The index (MLEGAL) has a development score for each province and major municipality and is compiled by China's National Economic Research Institute (NERI) (Fan and Wang, 2003). We also use a comprehensive index (MINDEX) of regional development (also compiled by NERI), which captures the following aspects12: (1) the relations between government and markets, such as the role of markets in allocating resources and enterprise burden in addition to normal taxes; (2) the development of non-state business, such as the ratio of industrial output by the private sector to total industrial output; (3) development of product markets, such as regional trade barriers; (4) development of factor markets such as FDI and mobility of labor; (5) development of market intermediaries and the legal environment (e.g., the protection of property rights). The comprehensive index (MINDEX) gives similar rankings to the index based on legal and institutional factors (MLEGAL).
10 There is a strand of research that shows that differences in nancial practice and economic performance across nations depend on the legal and institutional underpinnings of the countries concerned (e.g., La Porta et al., 2000; Bushman et al., 2004; Raonic et al., 2004). These studies assume homogeneity within a country and while this is a valid assumption for most nations, it does not apply in China where regional differences are enormous. 11 Although there are national laws, national accounting standards, and national governance guidelines, the enforcement of them varies signicantly across the regions. 12 Demurger et al. (2002) also compile indices of regional development in China using data up to 1999. The Fan and Wang index is more up to date and more appropriate for our needs.
375
Table 1 Number of restatement samples during 20002005. We collect 271 restatement announcements made by China listed companies during the period 20002005 (rms in the nance industry are omitted). Panel A: the number of restatements by year and by stock exchange Year Shanghai Number 2000 2001 2002 2003 2004 2005 Total 23 30 47 24 18 8 150 Percentage 4.1 4.7 6.7 3.1 2.2 1.0 3.5 Shenzhen Number 0 18 38 21 24 20 121 Percentage 0.0 3.6 7.7 4.3 4.6 3.8 4.0 Total Number 23 48 85 45 42 28 271 Percentage 2.2 4.2 7.1 3.6 3.1 2.1 3.7
Panel B: the number of restatements by industry Industry name Agriculture Mining Food, beverage Textile/Apparel Timber, furniture Paper making, printing Petroleum, chemistry, plastics Electronics Metal, non-metal Machinery, equipment, instrument Medicine, biological product Other manufacturing industries Power, gas and water Construction Transportation IT Retail Real estate Social service Communication Conglomerate Total Industry code A B C0 C1 C2 C3 C4 C5 C6 C7 C8 C9 D E F G H J K L M Number of restatements 16 2 15 8 0 5 37 14 28 38 17 4 9 0 8 20 10 8 10 2 20 271 Percentage of restatements within an industry 8.5 1.9 4.6 2.5 0.0 3.7 4.6 6.3 4.3 3.4 3.7 4.1 2.9 0.0 2.7 4.4 1.9 2.3 4.5 3.0 4.3 3.7
We use the CSRC (Chinese Securities Regulation Commission) industry classication standard. As most of rms belong to the Manufacturing industry whose code begins with C, we use the rst two codes to classify these samples. Our sample does not include the nancial industry whose code begins with I. Panel C: number of restatements and development index by province Province Guangdong Zhejiang Fujian Jiangsu Shanghai Tianjin Shandong Beijing Liaoning Hainan Henan Chongching Anhui Sichuan Guangxi Hebei Hubei Jiangxi Hunan Jilin Yunnan Heilongjiang Neimengru Shanxi MINDEX score 9.02 8.27 8.09 7.58 7.44 6.70 6.63 6.37 6.04 5.99 5.91 5.84 5.78 5.60 5.35 5.26 5.08 4.99 4.96 4.93 4.92 4.62 4.45 4.42 MLEGAL score 9.42 6.57 6.58 5.80 8.89 6.60 4.96 10.83 5.47 5.87 4.70 3.52 4.56 4.12 4.40 4.96 4.52 4.06 3.42 5.02 4.11 5.36 4.63 5.39 Number of restatement cases 19 10 6 15 26 5 14 13 20 7 3 5 11 19 5 13 13 2 12 11 3 12 4 3 Ratio of restatements to listed rms (%) 2.3 2.4 2.6 3.5 3.2 3.6 3.5 3.0 6.6 5.8 1.8 3.2 5.2 5.2 4.2 7.5 3.8 1.6 5.3 5.6 2.6 6.5 4.1 2.5 (continued on next page)
376 Table 1 (continued)
Panel C: number of restatements and development index by province Province Shaanxi Gansu Guizhou Ningxia Xingjiang Qinghai Xizang MINDEX score 4.21 4.04 3.93 3.48 3.45 3.15 1.69 MLEGAL score 3.00 3.06 3.35 3.06 3.87 4.40 4.41 Number of restatement cases 5 5 2 2 3 0 3 Ratio of restatements to listed rms (%) 3.6 4.9 2.6 3.2 2.2 0.0 6.4
MINDEX is a comprehensive index to capture the regional market development from the following aspects: (1) the relations between government and markets, such as the role of markets in allocating resources and enterprises' burden in addition to normal taxes; (2) the development of non-state business, such as ratio of industrial output by the private sector to total industrial output; (3) development of product markets, such as regional trade barriers; (4) development of factor markets such as FDI and mobility of labor; (5) development of market intermediaries and the legal environment (such as the protection of property rights). MLEGAL is a sub-index of MINDEX, which represents the legal environment. The MINDEX and MLEGAL scores shown above are the average scores during period 19992002.
3. Research design 3.1. Sample We undertake a thorough examination of corporate and stock exchange announcements to identify nancial restatements in China in the years 2000 to 2005. These restatements are the result of deliberate manipulation of the nancial reports. We exclude rms in the nance and nancial services industries as they are subject to different regulations. The restatements and some governance data are hand collected, while the other governance data and stock price data are taken from the China Stock Market and Accounting Research (CSMAR) database. Table 1, Panel A, shows the number of restating rms in each year. For example, in 2004 there are 18 rms listed on the Shanghai Stock Exchange that restated their nancial statements; these 18 rms represent 2.2% of all listed rms in Shanghai (818 rms) in 2004. On average, about 3.7% of listed rms restate their nancial statements each year. Panel B shows the industry distribution of restatements. Quite clearly, the Agriculture industry has the highest incidence of restatements, while timber/ furniture and construction have no restatements at all. Panel C shows the regional distribution of rms that restate. For example, during 20002005, 19 rms located in Guangdong province restate their nancial statements and this represents 2.3% of all rm-years in that province. Guangdong is considered to be a relatively well-developed province as its market development scores (MINDEX and MLEGAL) are high. There is a relatively high incidence of restatements for rms located in Hebei, Liaoning, and Heilongjiang and a low incidence for rms located in Jiangxi. The correlation between the proportion of restatements in a province and the market development score (MLEGAL) of that province is 0.234, which is signicant at the 0.05 level. Thus, there is a higher incidence of restatements by rms located in less developed regions; this nding is consistent with our arguments outlined earlier. 3.2. Control rms We use control rms as a benchmark when evaluating the characteristics of, and consequences of, nancial restatements; Efendi et al. (2007) and Agrawal and Chadha (2005) also use a control rm approach in their study of restatements and accounting scandals in the U.S. The use of a control rm is important in our study as changes in government policy can affect time-series comparisons (thus, before and after comparisons need to be evaluated against a control rm). The construction of the control rm group is as follows. For each restating rm we chose a non-restating rm that is in the same two-digit industry code, has been listed for the same number of years on the same stock exchange, and is nearest in size (sales and total assets). There are no signicant mean and median differences between the sample and control rms as regards company size (sales and total assets) and age (years since incorporation and years since rst listing). 3.3. Why do rms' restate their nancial statements? In an attempt to nd out why some rms restate while others do not, we examine the nancial characteristics of restating companies that differentiate them from non-restating rms. Firms that restate had previously manipulated their earnings and in most cases this means reported earnings are greater than they should have been. Possible motives for manipulating earnings upwards are to avoid reporting three years of losses (rms are delisted if they report three years of consecutive losses), to issue equity capital more cheaply by showing higher protability,13 and to give condence to the investors and creditors of highly levered companies. Furthermore, CEO and top management compensation in Chinese rms is dependent on a rm's reported earnings (Firth et al., 2006a) and so managers may be tempted to fraudulently boost reported net income. We therefore construct variables to capture losses, equity
13 Firms have to achieve specic earnings thresholds to issue seasoned equity offerings. For example, return on equity (ROE) has to average 10% or more in the three years prior to an application to make a rights issue and to average 6% or more for a private or public placement.
377
issues, and leverage. LOSS is coded one (1) if the rm reports two years of consecutive losses in the year prior to the manipulation and the manipulation turned a loss into a prot in the following year. Here, the manipulation is made to ensure the company makes a prot and so avoid three years of losses, which would lead to de-listing. SEO is coded one (1) if a rm makes an equity issue in the year after the manipulation.14 LEV is the rm's leverage (debt/total assets) at the date of the nancial fraud. Manipulation is more difcult to do if the rm has good governance mechanisms in place and so we examine the internal and external governance features of the company. These governance features include some of those used in prior research (Agrawal and Chadha, 2005; Chen et al., 2006b; Firth et al., 2007; Park and Shin, 2004; Xie et al., 2003). BOARD is the number of directors on the board. Some critics argue that large boards do not function well and become less effective in constraining the wanton behavior of a CEO or executive chairman (who might be behind the manipulation) (Jensen, 1993). On the other hand, large boards make it more difcult for the CEO or chairperson to obtain unanimous consent for fraudulent, questionable, or controversial actions. We examine the proportion of outside non-executive directors (OUT%) on the board. These directors often represent the controlling shareholder or other major investor, or are independent directors.15 The non-executive directors' impact on earnings manipulation may be different from that in the West (Dahya and McConnell, 2005) although the exact form of the difference is an empirical matter. DUAL is coded one (1) if the CEO and chairman is the same person. If the CEO and chairperson is the same person they have a lot of power and they have more ability to manipulate earnings if they want to do so (Brickley et al., 1997; Efendi et al., 2007). CFO is coded one (1) if the chief nancial ofcer is a member of the board. If the CFO is on the board this gives them more power and inuence and non-nancial directors will be more willing to accept the advice of the CFO (either for or against manipulation). FINBACK% is the number of directors who have a nancial background (as evidenced by professional qualications in accounting, economics, or nance). These directors are more likely to understand the (potential) nancial fraud and they either can acquiesce (in which case the non-nancial directors will follow their advice) or object (which would lead to less manipulation). Thus, FINBACK% could have a positive or a negative relation with nancial restatements.16 While directors with nancial expertise should reduce inadvertent errors, this does not mean that deliberate manipulation will be reduced. Indeed, having nancial expertise enables directors and CFOs to undertake and perpetuate complex accounting fraud. In the international literature, many studies argue that there are differences in audit quality across audit rms and the Big4 auditors provide the best quality (Chung et al., 2009). In China, the international Big4 audit rms typically audit domestic listed rms through their local afliates and so their quality is not identical to that of the Big4 in the U.S. and other developed countries. In 2003, the CSRC identied and named 15 auditors that it believed had the highest quality. We code Big15 one (1) if the auditor is one of the CSRCdesignated auditors. The Big15 auditors may inhibit earnings manipulation and have a negative relation with accounting fraud. We include three ownership variables in our analysis. They are: the proportion of shares owned by the largest stockholder (TOP); a dummy variable (CENTRAL) set equal to one (1) if the largest stockholder is the central government; and a dummy variable (PRIVATE) set equal to one (1) if the largest stockholder is a private investor or a foreign rm. Chinese rms are characterized as having a dominant stockholder who owns substantially more shares than the second largest owner does (Chen et al., 2009a). The major stockholder is effectively the controlling stockholder as the other institutional investors are not very active or vocal. In determining ownership, we take care to trace the ultimate owner. A dominant owner (e.g., when TOP is high) has a lot of inuence over the rm. The dominant owner can be a force for good or for bad (e.g., committing nancial fraud). The bureaucrats who administer the central government's shareholdings are usually career civil servants who have little commercial acumen. Hence they exercise little oversight over a rm's managers. Furthermore, these bureaucrats are held responsible for achieving the government's economic and social goals and they may use the listed rm to meet these objectives (e.g., by expropriating wealth away from the listed rm). Financial reports may be falsied to hide these activities and to show better performance to the government. Based on the above reasoning, we expect that rms where the central government is the largest shareholder will have more nancial restatements. We use the t-test and Mann-Whitney Z-test to test for differences in means between the restatement group and the control group. In addition, we use a bivariate probit model with partial observability to identify the characteristics that help us differentiate between the restating rms and non-restating rms. The bivariate probit model is described in the next section.
3.4. Bivariate probit model One inherent problem with the basic regression approach is that it is possible that the non-restating rms have manipulated their nancial statements but they have not restated them. This can arise if the CSRC, the rm's directors, and the auditors have not identied the manipulation (or if the rm has identied the error, it may not want to disclose it). This issue represents an identication problem and it reduces the ability of the model to explain the restatement (Wang, 2004; Chen et al., 2006b) and makes it difcult to interpret the coefcients. To illustrate, an independent variable (e.g., a governance variable such as Big15) could have a negative effect on earnings manipulation and a positive effect in discovering it (and making a restatement). The
14 In sensitivity tests, we alternatively code SEO as one if a rm just satises the protability criterion for making a SEO. The results from using this alternative measure of SEO are qualitatively the same as the ones reported in this paper. 15 Since 2003 listed rms have been required to have boards where at least one-third of the directors are independent. Before this date there were few independent directors. Even after 2003 there are questions as to how independent the independent directors are, and whether they fully understand their duties and responsibilities. 16 In a different context, Guner et al. (2008) show that increasing nancial expertise on U.S. rms' boards does not always benet stockholders.
378
simple probit model does not catch this subtlety and the coefcient on the variable will be difcult to interpret. One approach to resolve the inherent problem of treating non-observed nancial reporting failure as no-nancial reporting failure is to make use of a bivariate probit model with partial observability; see Poirier (1980) and Haque and Haque (2008) for details of the technique. Here, we model detected nancial reporting failure as a function of the joint realizations of two latent variables (nancial reporting failure and restatement)17: Financial Reporting Failure: Fj = 1 if rm j has a nancial reporting failure. Otherwise, Fj = 0. Detected: Dj = 1 if rm j's nancial reporting failure is detected (i.e., earnings are restated). Otherwise Dj = 0. From this, we have the following reduced form equations: Fj = x1 j 1 + uj Dj = x2 j 2 + vj where x1j is the vector of variables that helps explain a rm's propensity to manipulate earnings and x2j is the vector of variables that helps explain why nancial reporting failure is detected. uj and vj are the disturbance terms. The interaction of Fj and Dj is denoted Zj. Thus: Zj = Fj 4 Dj : Then Zj = 1 indicates a restatement. The empirical model for Zj is P Zj = 1 = P Fj = 1 & Dj = 1 = P Fj = 1 P Dj = 1 j Fj = 1 = x1 j 1 ; x2 j 2 P Zj = 0 = P Fj = 0 or Dj = 0 = P Fj = 0 + P Fj = 1 P Dj = 0 j Fj = 1 = 1 x1 j 1 ; x2 j 2 : The log-likelihood function for the model is n h i h io L1 ; 2 = zj ln x1 j 1 ; x2 j 2 + 1zj ln 1 x1 j 1 ; x2 j 2 : Full identication of the model parameters requires that x1j and x2j do not contain exactly the same variables (Poirier, 1980; Greene, 1995). Note that a simple probit model, which is used in most prior restatement studies, is as follows: Zj = Fj = xj + j Pr Zj = 1 = Pr Fj = 1 = x1 j : If Dj is not always one, the coefcients in the simple probit model will differ from those in the bivariate probit model. Wang (2004) uses a similar bivariate probit with partial observability approach in her study of securities class action litigation in the U.S. Abowd and Farber (1982) and Chidambaran and Prabhala (2003) are others who have used this method in their studies on labor markets and stock option repricing. While the bivariate probit model is conceptually the best approach to use, there are practical difculties in implementing it. One issue is that there is no developed theory on what variables and what functional forms explain accounting fraud and its subsequent disclosure. In the absence of a formal theory, we use empiricism to develop an appropriate model. The peculiar nature of identication in partially observed bivariate probit models (Poirier, 1980) results in the models having poor convergence properties (see Farber, 1983; Heywood and Mohanty, 1993, 1994). The poor convergence is exacerbated when there are a large number of independent variables (Haque and Haque, 2008; Comola, 2009) and when the independent variables are correlated (Heywood and Mohanty, 1994). We face similar problems in our tests. To improve convergence properties, we explore a number of parsimonious models. The model (P(Fj = 1)) that provides the best t is: RESTATE = 0 + 1 OUT % + 2 CFO + 3 FINBACK + 4 CENTRAL + 5 BIG15 + 6 MLEGAL + 7 LEV + 8 LOSS + 9 SEO:
17
The following section draws heavily on Poirier (1980) and Wang (2004).
379
The conditional detection model (P(Dj = 1/Fj = 1)) includes OUT%, CFO, FINBACK, CENTRAL, BIG15, MLEGAL, and SEO. Regulators (CSRC and stock exchanges) may investigate rms making SEOs and their examinations may culminate in the rm having to restate their nancial statements. Several governance variables are included in the detection model to see if they are associated with the detection of fraud and the restatements of the accounts. 3.5. Consequences of restatements Financial restatements are likely to have negative consequences for rms and their top executives (Karpoff et al., 2008a,b). To measure economic consequences we examine changes in stock returns, cost of capital, capital raising exercises, bidask spreads, and the incidence of modied audit reports. In addition, we examine whether restating rms are more likely to change their CEO. We calculate these measures with respect to the control group of rms. To evaluate the impact on stockholder wealth we perform an event study in which we estimate the cumulative abnormal stock returns of the rms around the restatement. We identify the event day as the rst day that the public is informed about the restatement. The abnormal return for security i on event date t is ARi;t = Ri;t E Ri;t j It where ARi,t, Ri,t, and E(Ri, t|It) are the abnormal, actual, and expected returns for time period t, respectively. It is the information on which the expected return depends. We use the market-adjusted returns model, the matched-rm model, and the market model to calculate the expected returns. The three methods yield similar conclusions and so we just report the market-adjusted returns model results. We accumulate ARi,t to obtain cumulative abnormal returns (CARs), using various event windows ranging from 10 days before to 10 days after the event day. Easley and O'Hara (2004) show that information risks cannot be diversied away. Information risk refers to the likelihood that rm-specic information pertinent to the investor pricing decision is of poor quality. The restatements capture information risk in this study. Using U.S. data, Murphy et al. (2009) show that reputation losses from nancial fraud increase the cost of capital. To examine whether nancial restatements increase the cost of equity, we need to estimate the expected rate of return on equity. There is, however, a continuing debate on how to estimate the expected rate of return. The literature shows that reverseengineering valuation models are appropriate to obtain estimates of the expected rate of return on equity investment. These reverse-engineering valuation models include the residual income valuation model, the abnormal growth in earnings model, and the dividend capitalization model (e.g., Claus and Thomas, 2001; Hribar and Jenkins, 2004; Daske, 2006; Attig et al., 2008; Pastor et al., 2008; Chen et al., 2009b). However, all these models require estimating the future growth rate of a rm and for this purpose most studies use growth rate estimates provided by nancial analysts. The models used in prior studies are of the form: h i ROEjt rj 4bpsjt1 1 + gj pjt = bpsjt + rj gj where pjt is the stock price for rm j at the end of year t, bpsjt and bpsjt 1 are the book value of equity for rm j at the end of years t and t 1, respectively. ROEjt is the return on equity for rm j at the end of year t. rj is cost of equity and gj is the future growth of residual income. The empirical testing of these types of models makes use of analysts' earnings growth forecasts. However, this approach has been criticized as the forecasts may be biased and therefore do not accurately reect the market's expectations (Easton, 2006; Easton and Sommers, 2007; Easton, 2009; Hou et al., 2010). O'Hanlon and Steele (2000)18 transform the above model into the following regression model to estimate the cost of equity and future growth rate: epsjt pjt bpsjt = 0 + 1 + jt bpsjt1 bpsjt1
where epsjt is the earnings per share at the end of year t. This regression may be estimated for any group/portfolio of stocks to obtain estimates of the expected rate of return, r, and the expected growth rate, g, for the portfolio. Thus, there is no need for analysts' forecasts of growth. 0 is the estimated cost of equity for the portfolio and g = (r 1) / (1 + 1). Because there was a lack of analyst growth forecasts in China at the time of our study, and because of severe conicts of interest that bias analysts' forecasts (Firth et al., 2010), we adopt the O'Hanlon and Steele model to calculate the cost of capital. We compare the difference in 0 between the restatement group and the non-restatement group. To date, there have been few applications of the O'Hanlon and Steele model and our paper is the rst to use it on emerging markets data.
18 See Easton (2006) for a discussion of this model. He states that the O'Hanlon and Steele (2000) model is the most suitable model for estimating the cost of capital. The model does not use analysts' forecasts and so is free from the bias inherent in the forecasts.
380
In order to test whether the difference in the cost of capital before and after the restatement is signicant, we employ the following regression with the before and after samples of the restatement and control samples:
epsjt pjt bpsjt = 0 + rR Restate + rA After + rD RestateAfter + 1 + gR Restate + gA After + gD RestateAfter + jt bpsjt1 bpsjt1
where Restate is a dummy variable, which is equal to one if it is a restatement rm, otherwise equal to zero; After is a dummy variable, which is coded one if the sample is after the restatement, else coded zero. The estimated coefcient of rD is used to test whether there is a difference in the change of cost of capital between the restatement and control samples. The coefcient of rR is the difference in the cost of capital between the restatement and control samples before the restatement. The coefcient of rA is the difference in cost of capital before and after the restatement year for the control sample. As another measure of cost of capital, we use the realized market-adjusted rates of return. In particular, we use the realized market-adjusted return in the month prior to the restatement (where we exclude the month of the restatement) and the one month after the restatement. The change in the market-adjusted returns indicates the change in cost of capital. As a comparison, we calculate the change in cost of capital for the control group. We also examine changes in Tobin's Q as it has an inverse relation with the cost of capital (Daske et al., 2008). Glosten and Milgrom (1985) argue that when information asymmetries exist among investors, the less-informed investors are concerned they will systematically lose money when they trade with well-informed investors. To protect themselves against the potential losses from trading with more-informed investors, the less-informed investors will decrease the price at which they are willing to buy and increase the price at which they are willing to sell. This will result in higher bidask spreads and lower liquidity. Financial restatements can increase the adverse selection problem by increasing information asymmetries between the lessinformed and better-informed investors. We therefore test whether restatements are associated with wider bidask spreads. Another costly consequence of a restatement occurs if the likelihood of receiving a modied (i.e., qualied) audit opinion (MAO) increases. The auditor may believe audit risk increases after a restatement. We predicate this on the belief that a restatement signals that management is less competent and more dishonest than previously thought. Issuing a modied audit opinion is a rational response by an audit rm when they perceive the client has become more risky. We therefore examine whether MAOs increase after a nancial restatement. The consequences we outline above relate to potential losses to investors. However, we also examine the consequences for top management. In particular, we investigate whether CEO turnover increases in the year after restatement. Given that restatements are the result of manipulations of the accounts, we expect that the top manager is more likely to be replaced. In the U.S., Hennes et al. (2008), nd that outside director and top management turnover increases after restatements; further, the dismissed executives suffer reductions in pay and benets if and when they nd new jobs.19 Fich and Shivdasani (2007) nd that outside directors lose reputation if their rms engage in nancial fraud. 4. Characteristics of restating rms Table 2 compares the nancial and governance characteristics of restating and non-restating rms. Restating rms have a small percentage of directors with an accounting and nancial background (the mean is 26.5%) when compared to the control rm (the mean is 31.1%). The difference is signicant at the 0.01 level. Thus, a relative lack of nancial expertise in the boardroom appears to be a precursor to nancial restatements and thus accounting manipulation. The other boardroom variables are not important in differentiating between restating and non-restating rms. The Big15 auditors are less likely to have clients that restate their accounts.20 Restating rms are more likely to have an agency of the central government as the controlling shareholder. Consistent with our hypotheses on the motives for accounting manipulation we nd that restating rms are more likely to have made an SEO, have two years of losses (followed by a prot in the year of the nancial manipulation), and have higher leverage. However, only SEO is statistically signicant. Firms located in the less developed provinces are more likely to restate their nancial reports. The univariate tests detailed in Table 2 do not distinguish between the different stages of false accounting and the subsequent restatements. To examine the joint impacts of the different variables, and to model both the fraud and the disclosure of fraud, we turn to bivariate probit regression with partial observability. Table 3 shows the results.21 The P(Fj = 1) column represents the fraud model while the P(Dj = 1/Fj = 1) column represents the detection and reporting of the false accounting. Companies making SEOs and highly levered rms are associated with the propensity to commit fraud and this is consistent with the motives for false manipulation that we discussed earlier. These rms have incentives to report higher earnings and this may lead them to issue false nancial statements. LOSS has a positive sign as expected although it is not signicant at conventional levels. Firms that have many directors with a nancial background are less likely to be associated with nancial statement fraud. Earlier, in Section 3, we argued that nance-competent directors could have a positive relation with fraud (e.g., they have the ability to perpetrate complex nancial fraud) or a negative relation (e.g., they have high ethical standards and understand the harmful effects of nancial fraud). Our results show that nancially-savvy directors help to reduce accounting fraud. Listed rms
However, earlier evidence (e.g., Agrawal et al., 1999) found no evidence of increased CEO turnover. Similar results obtain if we use the Big10 or the international Big4 (Big4's local afliates) in place of the Big15. 21 As mentioned previously, there is no theory to guide us in the selection of variables and so we examine a number of corporate governance mechanisms and various functional forms of the model. Correlations among the variables render some models very unstable. The reported results provide the best t.
20 19
M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390 Table 2 Univariate comparisons of restatement and non-restatement rms. Variables Mean Restate BOARD OUT% DUAL CFO FINBACK% BIG15 TOP PRIVATE CENTRAL GROWTH LOSS SEO LEV MINDEX MLEGAL Variables BOARD OUT% DUAL CFO FINBACK% BIG15 TOP PRIVATE CENTRAL GROWTH LOSS SEO LEV MINDEX MLEGAL 2.251 0.778 0.116 0.216 0.265 0.159 0.419 0.246 0.164 0.265 0.052 0.164 0.506 6.305 5.925 Denition Control 2.228 0.762 0.147 0.220 0.311 0.237 0.440 0.211 0.103 1.300 0.030 0.089 0.465 6.825 6.448 Difference (P-value) 0.024 0.016 0.030 0.004 0.046 0.078 0.021 0.034 0.060 1.035 0.022 0.075 0.041 0.520 0.524 (0.283) (0.257) (0.337) (0.915) (0.009) (0.036) (0.207) (0.377) (0.056) (0.231) (0.242) (0.061) (0.16) (0.001) (0.011) Median Restate 2.197 0.778 0.000 0.000 0.267 0.000 0.399 0.000 0.000 0.132 0.000 0.000 0.473 6.030 5.180 Control 2.197 0.778 0.000 0.000 0.293 0.000 0.407 0.000 0.000 0.133 0.000 0.000 0.448 6.685 5.900
381
Difference (P-value) 0.000 0.000 0.000 0.000 0.026 0.000 0.008 0.000 0.000 0.001 0.000 0.000 0.026 0.655 0.720 (0.309) (0.235) (0.337) (0.732) (0.361) (0.036) (0.71) (0.377) (0.056) (0.990) (0.242) (0.061) (0.194) (0.028) (0.095)
The number of board directors The proportion of directors who are not members of the management team A dummy variable taking the value one if the chairman and CEO positions are held by the same person A dummy variable coded one if the CFO or general accountant is on the board The percentage of directors who have an accounting or nancial background. If a director has a professional certicate of Accountancy or Economy, we take her/him as having an accounting/nancial background A dummy variable coded one (1) if the auditor belongs to the 15 auditors that are designated by the CSRC as a good reputation auditor The proportion of shares owned by the largest stockholder A dummy variable that equals one if the ultimate controlling stockholder is a private or foreign investor, else it equals zero A dummy variable that equals one if the ultimate controller is the central government, else it equals zero The sales growth in the two years prior to the date of the restatement A dummy variable taking the value one if the rm had recorded a loss in each of the two years prior to the accounting manipulation and made a prot in the year of the manipulation A dummy variable taking the value one if the rm makes a SEO in the year after the accounting manipulation Debt to total assets The marketization index of the province where the rm is located The legal environment index of the province where the rm is located
that are controlled by the central government are more likely to have fraudulent nancial statements. This reects the lack of oversight exercised by the government bureaucrats and/or their efforts to falsify the accounts to show better performance. Firms located in highly developed provinces have fewer nancial frauds.22 This reects more rigorous law enforcement and perhaps a better appreciation of good ethics. In contrast, a weak market environment may fail to produce credible disciplinary mechanisms to ensure that rms and their managers act honestly. In the detection model, we nd that OUT%, CENTRAL, and MLEGAL are signicant. Firms with more independent boards are more likely to detect and disclose false accounting reports while state controlled rms are less likely to report fraud. False accounting is more likely to be reported by rms located in the more developed regions.
5. Consequences of restatements 5.1. Stock returns Table 4 shows the market-adjusted returns for the restating rms. We have sufcient returns data for 267 observations; in four cases, the shares are suspended from trading. There are signicant negative abnormal returns in the periods [ 10, 1], [ 5, 1], [0, 5], [ 5, 5], [0, 10], and [ 10, 10] where day 0 is the announcement date. Disclosure of a restatement results in a statistically signicant fall in stock price.23 The stock returns analysis indicates that investors regard nancial statements as being pricesensitive information as the announcement of the accounting fraud causes a decline in prices. Panel B shows that the returns are more negative for companies located in highly developed provinces (MINDEX and MLEGAL above the median) although the differences are not statistically signicant. The results are also shown in Fig. 1.
The reported results use MLEGAL for the development index. However, similar conclusions are drawn when we use MINDEX instead of MLEGAL. Morck et al. (2000) demonstrate that stock returns in China are not very sensitive to rm-specic news; instead, an individual rm's returns are strongly linked to market-wide movements. The fact that we nd a signicant negative abnormal stock return is therefore unusual and shows that investors treat accounting restatements seriously.
23 22
382
Table 3 The bivariate probit regression results of the characteristics of restating rms. P(Fj = 1) Estimate OUT% CFO FINBACK CENTRAL BIG15 MLEGAL LEV LOSS SEO Intercept Model summary Chi-square Prob N chi2 Test of rho = 0 Variables OUT% CFO FINBACK% CENTRAL BIG15 LOSS SEO LEV MLEGAL Denition The proportion of directors who are not members of the management team A dummy variable coded one if the CFO or general accountant is on the board The percentage of directors who have an accounting or nancial background. If a director has a professional certicate of Accountancy or Economy, we take her/him as having an accounting/nancial background A dummy variable that equals one if the ultimate controller is the central government, else it equals zero A dummy variable coded one (1) if the auditor belongs to the 15 auditors that are designated by the CSRC as a good reputation auditor A dummy variable taking the value one if the rm had recorded a loss in each of the two years prior to the accounting manipulation and made a prot in the year of the manipulation A dummy variable taking the value one if the rm makes a SEO in the year after the accounting manipulation Debt to total assets The legal environment index of the province where the rm is located 0.120 0.165 0.680 0.846 0.098 0.135 0.710 0.106 0.341 0.446 69.45 (0.001) Chi-square p-value 0.798 0.358 0.062 0.001 0.574 0.000 0.026 0.780 0.009 0.336 P(Dj = 1|Fj = 1) Estimate 5.559 0.223 1.990 2.849 1.502 1.369 p-value 0.096 0.587 0.213 0.004 0.105 0.001
6.271
0.048
19.97
Prob N chi2
(0.001)
Table 4 Market-adjusted abnormal return around restatement announcement. Panel A: market-adjusted CAR, around the restatement date (day = 0) Days ( 10,1) ( 5,1) ( 1,0) (0,+1) (0,+5) ( 5,+5) (0,+10) ( 10,+10) N 267 267 266 267 267 267 267 267 CAAR (%) 0.86 0.88 0.03 0.06 0.87 1.74 0.79 1.66 Z 1.817 2.625 0.134 0.303 2.350 3.506 1.608 2.419 T 1.558 2.570 0.123 0.404 2.618 3.513 1.939 2.356
Panel B: CAR ( 10,+10) classied by MINDEX and MLEGAL MINDEX b Median 0.54% N Median 1.85% Difference 1.31% T-value 0.47 MLEGAL b Median 1.26% N Median 2.21% Difference 0.95% T-value 1.09
, , indicates the 1%, 5%, and 10% signicance levels of the Z-statistics and t-statistics.
5.2. Cost of capital Table 5 shows the results for the cost of capital tests. In Panel A, we show the cost of capital before and after the restatement date for the restating rms and for the control rms (a control sample rm uses the same restatement date as the matched restatement rm). The cost of capital (0) is 3.94% before restatement and 5.89% after restatement.24 The corresponding costs of capital for the control group are 3.96% and 4.08%. Our estimates of cost of capital are plausible. While the estimated cost of capital is lower than many estimates for U.S. rms, there are good reasons for this. First, the interest rate on bank deposits has been xed at 1.71% or 1.98% throughout the period of our study and so our estimates of cost of equity capital are approximately one and a half to
24 Our estimation procedure uses a rm's stock price and this price may be inuenced by an announced rights issue (SEO). In light of this, we repeat all of our cost of capital analyses on the sample of restatement and control rms that do not have rights issues. The results are very close to those reported in Table 5. We thank the reviewer for alerting us to this question.
383
Fig. 1. Plots of CAR.
three times the interest rate. Second, people in China have very few investment opportunities unlike their counterparts in the West. Bank deposits and stock investment are the only two investments ordinary people can make (in very recent years real property has become an investment opportunity but only for the wealthy). The Chinese people have very high saving rates but very few investment alternatives in which to invest. This leads to a lower cost of capital than in Western countries. The Chinese government set the interest rate low (which leads to a lower cost of capital) to achieve high growth and high employment. The low cost of capital has led to a high growth rate in GDP and poor protability by Western standards (as the hurdle rate is low); this corroborates the ndings in Chen et al. (2006a). China's ability to impose low interest rates and cost of capital is facilitated by the non-convertibility of its currency, the renminbi (RMB). We nd that the cost of capital increases by 1.96% for the restating rms but it increases by just 0.11% for the control sample. The rate of increase (from 3.94% to 5.89%) is 50% and this is much higher than the percentage increase reported in the U.S. (Hribar and Jenkins, 2004). To test if the increase in cost of capital is signicant we run a regression model that distinguishes between the cost of capital before and after restatements and the results are shown in Panel B. The coefcient rD is positive and statistically signicant (rD = 0.018). Thus, one important consequence for a restating rm is that there is a signicant increase in its cost of capital. Panels C and D show that the increase in the cost of capital for restating rms is more prominent for those companies located in highly developed provinces (MINDEX and MLEGAL above the median). This result is consistent with investors believing that rms located in highly developed provinces have high quality nancial reports and so the revelation of false accounting in a restatement disclosure causes a major reassessment of management credibility and overall earnings quality. In contrast, a restatement by a rm in a poor developed province is less of a shock for investors. We also use realized market-adjusted returns to represent a rm's cost of capital. The results show a signicant decline in cost of capital from before to after the restatement (Table 5, Panel E). In contrast, there is no change in cost of capital for the control sample. Table 5, Panel F, shows that the increases in cost of capital are mainly for those rms located in more highly developed provinces. The results for the stock return estimates of cost of capital (Panels E and F) are similar to the estimates of cost of capital using the O'Hanlon and Steele (2000) method (Panels A and C). Thus, our conclusions are robust to the two different measures of cost of capital. Daske et al. (2008) argue that an increase in the cost of capital, ceteris paribus, should lead to a decrease in Tobin's Q. We therefore examine the changes in Tobin's Q from before the restatement to after. Tobin's Q is calculated as (total assets book value of equity + market value of equity)/total assets. We show the results in Table 5, Panel G. There is a signicant decline in Tobin's Q for the restatement rms, which implies an increase in the cost of capital. Panel H of Table 5 show that the declines in control sample-adjusted Tobin's Q are more apparent in highly developed regions. Another consequence of the disclosure of false accounting is that rms may nd it more difcult to sell new shares.25 This is partly the result of the increases in cost of capital discussed above. As a direct test of the ability to sell new shares, we examine the proportion of rms that make SEOs in the three years prior to restatement and the proportion that make SEOs in the three years after a restatement. The results are shown in Panel I of Table 5. Restating rms make more SEOs before the restatement than do the control rms although the difference is not statistically signicant (p = 0.364). After the restatement, the proportion of rms making SEOs falls dramatically for the restating rms. The reduction is more severe for the restating rms than for the control rms.26 The evidence indicates that restating rms nd it much more difcult to issue new shares in the aftermath of the restatement.
We thank the reviewer for suggesting this analysis. The proportion of control rms that make SEOs also falls. The reduction in SEOs over time reects capital controls imposed by the government (via the CSRC) in the later part of our sample period. This illustrates the importance of using control rms as benchmarks.
26
25
384
Table 5 Comparison of the cost of capital before and after the restatement. This table shows whether the cost of capital of restatement rms increases after the restatement. We estimate the cost of capital using the method of O'Hanlon and Steele (2000):
epsjt pjt bpsjt = 0 + 1 + jt bpsjt1 bpsjt1
where epsjt and bpsjt are the earnings per share and book value per share of rm j in year t, respectively. pjt is the closing price of rm j in year t. The estimated intercept term 0 = cost of capital (r), and the estimated parameter 1 = (r g)/(1 + g), where g is the expected growth rate. We conduct the above regressions with the restatement sample and control sample before and after the restatement, respectively. The results are shown in Panel A. Panel A: cost of capital before and after the restatement Restatement sample Before the restatement 3.94% (0.185) After the restatement 5.89% (0.115) Difference 1.96% Control sample Before the restatement 3.96% (0.161) After the restatement 4.08% (0.117) Difference 0.11%
The number in parenthesis is the adjusted R-square of the regression model. In order to test whether the difference in cost of capital before and after the restatement is signicant, we employ the following regression with the before and after samples of the restatement and control samples:
epsjt bpsjt1
= 0 + rR Restate + rA After + rD RestateAfter + 1 + gR Restate + gA After + gD RestateAfter
pjt bpsjt bpsjt1
+ jt
where Restate is a dummy variable, which is equal to one if it is a restatement rm, otherwise equal to zero; After is a dummy variable, which is coded one if the sample is after the restatement, else coded zero. The estimated coefcient of rD is used to test whether there is a difference in the change of cost of capital between the restatement and control samples. The coefcient of rR is the difference in cost of capital between the restatement and control samples before the restatement. The coefcient of rA is the difference in cost of capital before and after the restatement for the control sample. The results are shown in Panel B. Panel B: difference in cost of capital before and after restatement 0 0.040 (0.000) rR 0.000 (0.976) rA 0.001 (0.879) rD 0.018 (0.047) Adj-R2 0.185
The numbers in the parenthesis is the p-value of the t-statistic. Panel C: cost of capital before and after the restatement grouped by market development index MINDEX (MLEGAL) b Median Before the restatement MINDEX MLEGAL 3.48% (0.267) 4.05% (0.189) After the restatement 4.05% (0.264) 4.10% (0.232) Difference (%) 0.58 0.05 MINDEX (MLEGAL) N Median Before the restatement 4.87% (0.219) 4.54% (0.269) After the restatement 6.62% (0.099) 6.71% (0.103) Difference (%) 1.75 2.17
The number in parenthesis is the adjusted R-square of the model. In order to test whether there is any difference in terms of cost of capital for rms under different market development conditions, we employ the following regression:
epsjt bpsjt1
= 0 + rMindex After Mindex + 1 + gMindex After Mindex
pjt bpsjt bpsjt1
+ jt
where After is a dummy variable, which coded one if it is after the restatement, otherwise zero. Mindex is MINDEX, or MLEGAL, which are dened in Table 1. The estimated coefcient of rMindex indicates the effect of Mindex on the change in cost of capital after the restatement. GMindex captures the effect of MINDEX and MLEGAL on the change in expected growth rate after the restatement. The results are reported in Panel D. Panel D: the effect of market development index (MINDEX, MLEGAL) on the difference in the cost of capital from before to after the restatement 0 Coefcient MINDEX MLEGAL 0.038 0.037 p-value (0.000) (0.000) rMindex Coefcient 0.003 0.003 p-value (0.014) (0.006) 0.225 0.228 Adj-R2
Panel E: market-adjusted stock returns from one month before to one month after the restatement (%) Restatement sample Before the restatement 0.018 After the restatement 1.341 Difference 1.360 (0.031) Control sample Before the restatement 0.714 After the restatement 0.339 Difference 0.375 (0.600)
The number in parenthesis is the p-value of the t-statistic.
M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390 Table 5 (continued)
385
Panel F: the effect of market development index (MINDEX, MLEGAL) on the difference in the market-adjusted one-month stock return before and after the restatement MINDEX (MLEGAL) b Median Before the restatement MINDEX MLEGAL 0.475 0.646 After the restatement 1.514 1.475 Difference 1.038 (0.266) 0.829 (0.33) MINDEX (MLEGAL) N Median Before the restatement 0.483 0.683 After the restatement 1.180 1.208 Difference 1.663 (0.051) 1.891 (0.041)
Panel G: the Tobin's Q in the year before and after the restatement (%) Restatement sample Before the restatement 1.588 After the restatement 1.422 Difference 0.166 (0.001) Control sample Before the restatement 1.483 After the restatement 1.418 Difference 0.064 (0.139)
Panel H: the effect of market development index (MINDEX, MLEGAL) on the difference in the control sample-adjusted Tobin's Q in the year before and after the restatement MINDEX (MLEGAL) b Median Before the restatement MINDEX MLEGAL 0.054 0.027 After the restatement 0.009 0.020 Difference 0.045 (0.417) 0.006 (0.914) MINDEX (MLEGAL) N Median Before the restatement 0.068 0.095 After the restatement 0.005 0.006 Difference 0.063 (0.353) 0.101 (0.108) Difference between the two groups' difference 0.008 (0.85) 0.085 (0.039)
Panel I: the proportion of rms making SEOs before and after restatement Three years prior to the restatement Restatement rms Control rms Difference 0.328 0.286 0.042 (0.364) Three years after the restatement 0.097 0.133 0.035 (0.276) Difference 0.231 0.153 0.078 (0.091)
The number in parenthesis is the p-value of the t-statistic. Table 6 Comparison of the bidask spread before and after accounting restatements. To test whether accounting restatements affect the liquidity of the stock, we compare the bidask spread before and after the restatement for restatement rms and control rms, respectively. We calculate the average daily relative effective spread for one month prior to and one month post the restatement announcement for each rm. In Panel A, we provide the results of the bidask spread. We then divide the restatement sample into two groups based on the MINDEX or MLEGAL score. One group is rms with MINDEX (MLEGAL) scores below the median, while the other group is rms with MINDEX (MLEGAL) scores above the median. We report the bidask spread one month before and after the restatement for these two groups. The results are shown in Panel B. Panel A: the differences in bidask spreads before and after the restatement Before the restatement (%) Restatement rms Control rms Difference 0.0717 0.0889 0.0172 After the restatement (%) 0.0788 0.0837 0.0049 Difference (%) 0.0070 0.0052 0.0122
indicates the 1% signicance level of the t-statistic. Panel B: the comparison of the bidask spread before and after the restatement for the restatement rms grouped by MINDEX (or MLEGAL) Below the median Before the restatement (%) MINDEX MLEGAL 0.0812 0.0805 After the restatement (%) 0.0814 0.0759 Difference (%) Above the median Before the restatement (%) 0.0636 0.0645 After the restatement (%) 0.0708 0.0762 Difference (%) The difference between the two groups' difference (%) 0.0071 0.0162
0.0002 0.0045
0.0072 0.0117
and indicates the 5% and 10% signicance levels of the t-statistics.
5.3. Bidask spread We compute the average bidask spread for the month prior to the restatement date and compare it with the bidask spread in the one month after restatement.27 A comparison is then made with the control group. We show the results in Table 6. Panel A shows the bidask spread increases for the restatement rms but declines for the control group rms. The difference in the changes (0.0122%) is signicant at the 0.01 level. Our result contrasts with the U.S. experience where Palmrose et al. (2004) found no signicant evidence of restatements affecting bidask spreads. In Panel B we show that the increase in bidask spreads is
27 We follow Cai (2004) and others and calculate the spread as (Ask Bid)/(Ask + Bid)/2. Note that prices are quoted in fen and the minimum price movement is one fen (approximately $0.001U.S.). Thus, bidask spreads are small in comparison to those in the U.S.
386
Table 7 Comparison of modied (qualied) audit opinions before and after the restatement. As one of the consequences of restatements is an increase in the probability of a modied audit opinion, we compare the percentage of modied audit opinions one year before and one year after the restatement announcement for restatement rms and control rms, respectively. Panel A Year 1 Restatement rms (%) Control rms (%) Difference (%) 4.69 5.51 0.83 Year 0 9.59 3.31 6.29 Difference 1,0 4.91 2.21 7.11 p-value = 0.002
Panel B Year 1 MINDEX Above the median (%) Below the median (%) Difference (%) Above the median (%) Below the median (%) Difference (%) 5.04 1.75 3.29 2.50 4.43 1.93 Year 0 13.45 0.88 12.57 9.17 5.31 3.86 Difference 1,0 8.40 0.87 9.28 p-value = 0.001 6.67 0.89 5.78 p-value = 0.039
MLEGAL
stronger for rms that are located in provinces with high market development (i.e., those with MLEGAL scores above the median). The results show that rms that need to restate their nancial statements are viewed as becoming more risky and so the bidask spreads widen. 5.4. Modied (qualied) audit reports (MAOs) When a rm admits, via a nancial restatement, that its prior accounts are erroneous this implies the errors are a deliberate act by management and that the governance structures and monitoring mechanisms are unable to prevent such an act from occurring. This is likely to increase the audit risk as perceived by the external auditor. A rational response of the auditor to the increased audit risk is to increase the threshold for giving a clean opinion. Therefore we expect to see an increase in the proportion of modied audit opinions (MAOs) being given to restating rms after the restatement. To test this hypothesis, we compare the change in MAOs from before a restatement to after and compare this change to the control group. The results are shown in Table 7. In Panel A we see that there is an increase in MAOs for restating rms whereas the control group has a decrease. The difference is statistically signicant with a p-value of 0.002. In Panel B we nd that the relative increase in MAOs for restating rms is mainly conned to rms located in provinces with a high development score. 5.5. CEO turnover So far, we have considered the consequences of false nancial reporting for investors by examining stock returns, cost of capital, bidask spreads, and modied audit reports. However, there could also be consequences for top management. We therefore examine CEO turnover before and after the restatement (one year before and one year after28) and compare this to the control group.29 Table 8 shows the results. The turnover increases after a restatement (to 30.8%); thus, about 31% of CEOs lose their jobs within one year of the restatement. In comparison, the turnover rates decline for the control group. The difference in changes in turnover rates for restating and non-restating rms (10.5%) is statistically signicant. The evidence suggests that accounting restatements have costly consequences for CEOs. The CEOs are more likely to be dismissed after a restatement as they carry the responsibility for the false accounting. Our results are consistent with those reported in the U.S. (Hennes et al., 2008; Collins et al., 2009). Panel B shows that the increases in CEO turnover after a restatement do not depend on the level of market development where the rm is located. Unfortunately, we are unable to trace where the CEO goes after leaving the restating rm and are therefore unable to ascertain whether they obtain a worse (or similar, or better) position.30 Thus, we cannot carry out the type of analyses undertaken by Karpoff et al. (2008a) in the U.S. Furthermore, we do not have data on the other ofcers and directors of the rm and so we cannot examine the consequences of restatements on them.
28 A review of the public announcements that disclose the CEO replacement indicates that falsied nancial statements are a major reason for top executive dismissal. 29 In a robustness test, the control group is rened to match restatement rms and non-restatement rms based on return on assets. We do this because prior research (Firth et al., 2006b) shows that a rm's protability is an important factor in the executive turnover decision. The results from this additional test mirror those reported in Table 8 and so our ndings are robust to alternative specications of the control group. 30 Some CEOs may return to government jobs, parent SOEs, or move to foreign rms. Data on the top management jobs at these organizations are not publicly available.
387
Table 8 Comparison of the CEO turnover before and after the restatement. As one of the consequences of restatements is an increase in the probability of CEO turnover, we compare the percentage of CEO turnover one year before and one year after the restatement announcement for restatement rms and control rms, respectively. Panel A: the difference in CEO turnover before and after the restatement Before the restatement (%) Restatement rms Control rms Difference 28.5 26.8 1.7 After the restatement (%) 30.8 18.7 12.1 Difference (%) 2.3 8.2 10.5
indicates the 5% signicance level of the t-statistic. Panel B: comparison of CEO turnover for the restatement rms grouped by MINDEX and MLEGAL MINDEX or MLEGAL MINDEX/MLEGAL Below the median Before the restatement CEO turnover MINDEX MLEGAL 33.04 33.62 After the restatement CEO turnover 36.61 33.62 Difference MINDEX/MLEGAL Above the median Before the restatement CEO turnover 23.28 22.32 After the restatement CEO turnover 27.59 30.36 Difference The difference between two groups' differences
3.57 0.00
4.31 8.04
0.74 8.04
6. Conclusion The quality of a rm's published nancial statements is a major concern to investors, regulators, and other parties. However, the measurement of quality is contentious and a broad consensus on its denition remains elusive. Nevertheless, if a rm restates its nancial statements it represents an admission that its prior accounts are false. Financial restatements are extensively studied in the U.S. but there is relatively little literature on restatements in other countries. To help remedy this deciency our study examines restatements of nancial reports in China. Falsied nancial statements are common in China. We nd that rms are more likely to manipulate their nancial information when the rm issues new equity and when it is controlled by the central government. The manipulation allows a rm's nancial statements to look better than they should although it leads to a restatement in a subsequent year. The relations between restatements and governance variables are complex. For example, rms that have many directors with a nancial background are less likely to have restatements. The percentage of independent directors and the presence of a CFO on the board have no relation with fraudulent nancial statements. We nd no evidence that a major audit rm inhibits nancial fraud in listed rms. Restatements are more likely for rms located in less developed provinces. Detection and reporting of false accounting is more likely when the central government is not the major stockholder and when rms are located in more developed regions. There is some weak evidence that an independent board is more likely to insist on restatements. Financial restatements in China have economic consequences. Stock prices fall, cost of capital increases, access to capital markets declines, and bidask spreads widen. Restatements increase the risk perception of the rm as manifested in widening bidask spreads and an increase in modied audit opinions. Top management is not immune to the consequences of restatements as we show that the CEO is more likely to be replaced in the year after the restatement. Our results also show that the causes and consequences of restatements depend on where the rm is located. Although there are national laws, standards, and governance guidelines, the application and enforcement of them varies a lot. In particular, the application and enforcement are greater in more developed provinces. Thus, investors expect better nancial reporting quality in highly developed provinces and so restatements are a major shock that lead to negative consequences for rms (e.g., an increase in the cost of capital and widening of the bidask spread). In contrast, investors believe that rms located in poorly developed provinces have lower quality nancial statements and so restatements are less of a shock. While many previous studies have shown that the institutional and legal environment of a country have an impact on rm value and accounting quality, our study shows that there can also be differences inside a country. Thus, in large transition and emerging market economies, the progress of change can vary a lot across the different regions of a country and this will have an impact on the prevalence and consequences of falsied nancial reporting.
Acknowledgments We thank the editors and the reviewer for helpful comments and suggestions on earlier versions of the paper. We also thank Gordon Richardson and workshop participants at The Chinese University of Hong Kong, City University, Lingnan University, and The Journal of Corporate Finance Special Conference on Emerging Markets for helpful comments on the paper. Michael Firth and Oliver Rui acknowledge nancial support from a Hong Kong SAR Competitive Earmarked Research Grant (LU340307). Oliver Rui also acknowledges nancial support from the Chinese University of Hong Kong (projects 2070410, 6902327, and 6902456) and Wenfeng Wu acknowledges nancial support from the National Science Fund Committee of China (Project Number 70672074) and the Shanghai Pujiang Program.
388
Appendix A In order to regulate how listed companies disclose the correction of nancial information, to enhance their credibility and promptness and to protect the legal rights of investors, the China Securities Regulatory Commission issued Preparation Conventions of Information Disclosure by Companies Offering Securities to the Public No. 19 Correction of nancial information and related disclosures. These conventions apply to all publicly listed companies. Preparation Conventions of Information Disclosure by Companies Offering Securities to the Public No. 19 Correction of nancial information and related disclosures 1. In order to regulate how listed companies disclose the correction of nancial information, to enhance their credibility and promptness and to protect the legal rights of investors, the China Securities Regulatory Commission has issued the Preparation Conventions of Information Disclosure by Companies Offering Securities to the Public No. 19 in accordance with the Company Law of the People's Republic of China and the Securities Law of the People's Republic of China. 2. The Convention is applicable under the following circumstances: A company is ordered by the government authorities to correct errors included in any previous periodic reports As determined by the Board of Directors, a company volunteers to disclose errors included in any previous periodic reports, or Other events that are deemed warranted by the CSRC. 3. Companies that t the above criteria should release to the public a Material Events Special Alert to disclose the corrected nancial information. 4. The modied nancial statements should follow the related formats set by the China Securities Regulatory Commission. 5. After a previous set of annual nancial statements has been corrected, the modied annual nancial statements should be audited by a qualied CPA rm. 6. A special alert should contain the following: A detailed explanation of the modications from the board of directors or company management the effects of the modications on the nancial position and business operations of the company modied nancial statements which have been audited, notes to nancial statements associated with the correction, and the name of the CPA rm issuing the audit report. 7. Disclosures on modied nancial statements are required under three circumstances: If the nancial information of previous periods (including annual, semi-annual and quarterly) has been amended, a company should disclose, (i) amongst other sets of annual nancial statements under modication due to correction of the events, the modied annual nancial statements of the latest full accounting year; (ii) amongst other sets of interim nancial statements under modication due to correction of the events, the modied interim nancial statements of the latest interim period; If the correction is only made on interim nancial statements published in the current year, a company should disclose all interim nancial statements of the current year (quarterly and semi-annual) which have been modied due to correction of the events; or If the correction is made on interim nancial statements of the prior year although the annual nancial statements of the prior year have yet to been published, a company should disclose all of the interim nancial statements which have been modied due to correction of the events. 8. All amended items on the modied nancial statement should be highlighted in bold. 9. If a company modies the annual nancial information released three years or before, which has no effects on the nancial statements over the past three years, it is not required to reveal the amended nancial information thereafter. 10. The China Securities Regulatory Commission reserves the rights to interpret the terms and conditions of the convention.
Appendix B Stock Symbol: ST Weida Stock Code: 000603 Series Code: 2003-014 WeiDa Medical Applied Technology CO. LTD Board of Directors Restatement of 2002 Annual Report Our company and all the members of the board of directors guarantee the verity, correctness and completeness of this announcement. And we take joint responsibility for any possible false record, misleading statement and signicant omission in this announcement. Our company released the annual report of year 2002 and its abstract on April 24th, 2002. After an investigation by the Shenzhen Stock Exchange (SZE), it is found that our company's accounting policy for 1,122,300 yuan of shutdown losses is not appropriate. According to the requirement of the SZE, we restate the following data: Original: the main prot this year in the income statement and operation data abstract: net prot after deduction of nonrecurring gains and losses amounts to 66,779.52 yuan. Restatement: net prot after deduction of non-recurring gains and losses amounts to 1,055,520.48 yuan. Due to the adjustment of the net prot after deduction of non-recurring gains and losses, the related changes are as follows: Earnings per share after deduction of non-recurring gains and losses 0.00094 yuan Net asset yield rate after deduction of non-recurring gains and losses (diluted) -0.91%
389
Net asset yield rate after deduction of non-recurring gains and losses (weighted average) 1.26% Earnings per share after deduction of non-recurring gains and losses (diluted) 0.00094 yuan Earnings per share after deduction of non-recurring gains and losses (weighted average) 0.00094 yuan. After deduction of non-recurring gains and losses, the net prot of our company is a loss. According to the related regulation, the stock of our company will remain under the special treatment. Announcement here by WeiDa Medical Applied Technology CO. LTD Board of Directors June 11th, 2003 References
Abowd, J.M., Farber, H.S., 1982. Job queues and union status of workers. Ind. Labor Relat. Rev. 35, 354367. Agrawal, A., Chadha, S., 2005. Corporate governance and accounting scandals. J. Law Econ. 48, 371406. Agrawal, A., Jaffe, J.F., Karpoff, J.M., 1999. Management turnover and governance changes following the revelation of fraud. J. Law Econ. 42, 209341. Allen, F., Qian, M., Qian, J., 2005. Law, nance, and economic growth in China. J. Financ. Econ. 77, 57116. Anderson, K.L., Yohn, T.L., 2002. The Effect of 10K Restatements on Firm Value, Information Asymmetries, and Investors' Reliance on Earnings (September 2002). Available at SSRN: http://ssrn.com/abstract=332380 or doi:10.2139/ssrn.332380. Attig, N., Guedhami, O., Mishra, D., 2008. Multiple shareholders, control contests, and implied cost of equity. J. Corp. Finance 14, 721737. Brickley, J.A., Coles, J., Jarrell, G., 1997. Leadership structure: separating the CEO and chairman of the board. J. Corp. Finance 3, 189220. Bushman, R., Piotroski, J., Smith, A., 2004. What determines corporate transparency? J. Acc. Res. 42, 207252. Cai, J., 2004. Bidask spreads for trading Chinese stocks listed on domestic and international exchanges. Working paper. Shanghai Stock Exchange. Chen, G.M., Firth, M., Gao, D.N., Rui, O.M., 2005. Is China's securities regulatory agency a toothless tiger? Evidence from enforcement actions. J. Acc. Public Policy 24, 451488. Chen, G.M., Firth, M., Rui, O.M., 2006a. Have China's enterprise reforms led to improved efciency and protability? Emerg. Mark. Rev. 7, 82109. Chen, G.M., Firth, M., Gao, D.N., Rui, O.M., 2006b. Do ownership structure and governance mechanisms have an inuence on deterring corporate nancial fraud in a transitional economy? Regulatory enforcement actions in China. J. Corp. Finance 12, 424448. Chen, G.M., Firth, M., Xu, L., 2009a. Does the type of ownership control matter? Evidence from China's listed companies. J. Bank. Finance 33, 171181. Chen, K.C.W., Chen, Z., Wei, K.C.J., 2009b. Legal protection of investors, corporate governance, and the cost of equity capital. J. Corp. Finance 15, 273289. Chidambaran, N.K., Prabhala, N.R., 2003. Executive stock option repricing, internal governance mechanisms, and management turnover. J. Financ. Econ. 69, 153189. Chung, R., Firth, M., Kim, J.B., 2009. Big 4 conservatism around the world. Working paper. Concordia University. Claus, J., Thomas, J., 2001. Equity risk premium as low as three percent? Evidence from analysts' earnings forecasts for domestic and international stocks. J. Finance 56, 16291666. Collins, D., Masli, A., Reitenga, A.L., Sanchez, J.M., 2009. Earnings restatements, the SarbanesOxley Act and the disciplining of chief nancial ofcers. J. Acc. Auditing Finance 24, 134. Comola, M., 2009. The Network Structure of Informal Arrangements: Evidence from Rural Tanzania. Available at SSRN: http://llssrn.com.abstract=94603. Dahya, J., McConnell, J., 2005. Outside directors and corporate board decisions. J. Corp. Finance 11, 3760. Daske, H., 2006. Economic benets of adopting IFRS or U.S.-GAAP have the expected costs of equity capital really increased? J. Bus. Finance Acc. 33, 329373. Daske, H., Hail, L., Leuz, C., Verdi, R., 2008. Mandatory IFRS Reporting around the World: Early Evidence on the Economic Consequences. Working Paper. The University of Chicago. Dechow, P.M., Ge, W., Larson, C.R., Sloan, R.G., 2010. Predicting Material Accounting Misstatements. Working Paper. University of California, Berkeley. Demurger, S., 2001. Infrastructure development and economic growth: an explanation for regional disparities in China? J. Comp. Econ. 29, 95117. Demurger, S., Sachs, J.D., Woo, W.T., Bao, S.M., Chang, G., 2002. The relative contributions of location and preferential policies in China's regional development: being in the right place and having the right incentives. China Econ. Rev. 13, 444465. Easley, D., O'Hara, M., 2004. Information and the cost of capital. J. Finance 54, 15531582. Easton, P., 2006. Use of forecasts of earnings to estimate and compare cost of capital across regimes. J. Bus. Finance Acc. 33, 374394. Easton, P., 2009. Estimating the cost of capital implied by market prices and accounting data. Found. Trends Acc. 2, 241364. Easton, P., Sommers, G., 2007. Effects of analysts' optimism on estimates of the expected rate of return implied by earnings forecasts. J. Acc. Res. 45, 9831015. Efendi, J., Srivastava, A., Swanson, E.P., 2007. Why do corporate managers misstate nancial statements? The role of option compensation and other factors. J. Financ. Econ. 85, 667708. Fan, G., Wang, X.L., 2003. The Report on the Relative Process of Marketization of Each Region in China. The Economic Science Press. (in Chinese). Farber, H.S., 1983. The determination of the union status of workers. Econometrica 51, 737. Fich, E.M., Shivdasani, A., 2007. Financial fraud, director reputation, and shareholder wealth. J. Financ. Econ. 86, 306336. Firth, M., Fung, P., Rui, O.M., 2006a. Corporate governance and CEO compensation in China. J. Corp. Finance 12, 693714. Firth, M., Fung, P., Rui, O.M., 2006b. Firm performance, governance structure, and top management turnover in a transitional economy. J. Manage. Stud. 43, 12891330. Firth, M., Fung, P., Rui, O.M., 2007. Ownership, two-tier board structure, and the informativeness of earnings evidence from China. J. Acc. Public Policy 26, 463496. Firth, M., Lin, C., Liu, P., Xuan, Y., 2010. The Client is King: Do Mutual Fund Relationships Bias Analyst Recommendations? Working paper. Available at SSRN: http:// papers.ssrn.com/sol3/papers.cfm?abstract_id=1363773. Glosten, L., Milgrom, P., 1985. Bid, ask and transaction prices in a specialist market with heterogeneously informed traders. J. Financ. Econ. 14, 71100. Greene, W.H., 1995. LIMDEP: Users Manual, Version 7. Econometric Software Inc., Bellport, New York. Guner, A.B., Malmendier, U., Tate, G., 2008. Financial expertise of directors. J. Financ. Econ. 88, 323354. Haque, R., Haque, M., 2008. Bivariate probit models of labour market status. In: Haque, R., Haque, M. (Eds.), Gender, Ethnicity and Employment Non English Speaking Background Migrant Women in Australia. Springer-Verlag, Heidelberg, pp. 171201. Hennes, K., Leone, A., Miller, B., 2008. The importance of distinguishing errors from irregularities in restatement research: the case of restatements and CEO/CFO turnover. Acc. Rev. 83, 14871519. Heywood, J.S., Mohanty, M.S., 1993. Testing for state and local job queues. J. Labor Res. 14, 455467. Heywood, J.S., Mohanty, M.S., 1994. The role of employer and workplace size in the US federal sector job queue. Oxf. Bull. Econ. Stat. 56, 171188. Hou, K., van Dijk, M.A., Zhang, Y., 2010. The Implied Cost of Capital: A New Approach. Working Paper. Ohio State University. Hribar, P., Jenkins, N.T., 2004. The effect of accounting restatements on earnings revisions and the estimated cost of capital. Rev. Acc. Stud. 9, 337356. Jensen, M.C., 1993. The modern industrial revolution, exit and the failure of internal control systems. J. Finance 48, 831880.
390
Karpoff, J., Lee, D.S., Martin, G.S., 2008a. The consequences to managers for nancial misrepresentation. J. Financ. Econ. 88, 193215. Karpoff, J., Lee, D.S., Martin, G.S., 2008b. The cost to rms of cooking the books. J. Financ. Quant. Anal. 43, 581612. Karpoff, J., Lee, D.S., Martin, G.S., 2009. The Legal Penalties for Financial Misrepresentation. Working Paper. University of Washington. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 2000. Investor protection and corporate governance. J. Financ. Econ. 58, 327. Li, C.Q., 2008. The Study of the Quality of Accounting Disclosure by Listed Companies: A Perspective from Restatements of Annual Reports. Shanghai Stock Exchange. Available at: http://www.sse.com.cn/sseportal/webapp/datapresent/SSEDisquisitionAndPublicationAct (in Chinese). Morck, R., Yeung, B., Yu, W., 2000. The information content of stock markets: why do emerging stock markets have synchronous stock price movements? J. Financ. Econ. 58, 215260. Murphy, D.L., Shrieves, R.E., Tibbs, S.L., 2009. Understanding the penalties associated with corporate misconduct: an empirical examination of earnings and risk. J. Financ. Quant. Anal. 44, 5583. O'Hanlon, J., Steele, A., 2000. Estimating the equity risk premium using accounting fundamentals. J. Bus. Finance Acc. 27, 10511084. Palmrose, Z., Richardson, V., Scholz, S., 2004. Determinants of market reactions to earnings restatements. J. Acc. Econ. 37, 5990. Park, Y., Shin, H., 2004. Board composition and earnings management in Canada. J. Corp. Finance 10, 431457. Pastor, L., Sinha, M., Swaminathan, B., 2008. Estimating the intertemporal risk-return tradeoff using the implied cost of capital. J. Finance 6, 28592897. Plumlee, M., Yohn, T.L., 2009. Analysis of the Underlying Causes of Restatements. Working Paper. University of Utah. Poirier, D.J., 1980. Partial observability in bivariate probit models. J. Econometrics 12, 209217. Raonic, I., McLeay, S., Asimakopoulos, I., 2004. The timeliness of income recognition by European companies: an analysis of institutional and market complexity. J. Bus. Finance Acc. 31, 115148. Schipper, K., Vincent, L., 2003. Earnings quality. Acc. Horiz. 17, 97110. Sun, Z., Liu, F.W., Li, Z.Q., 2005. Market development, government inuence and corporate debt maturity structure. Econ. Res. J. 5, 5263 (in Chinese). Tsui, K.Y., 1996. Economic reform and interprovincial inequalities in China. J. Dev. Econ. 50, 353368. Wang, T.Y., 2004. Economic Determinants of Corporate Fraud Propensity and Detection. Working Paper. University of Maryland. Wu, M., Wang, X., 2009. The Quality of Financial Reporting in China. Working Paper. The University of Hong Kong. Xia, L.J., Fang, Y.Q., 2005. Government control, institutional environment and rm value: evidence from the Chinese securities market. Econ. Res. J. 5, 4150 (in Chinese). Xie, B., Davidson III, W., DaDalt, P., 2003. Earnings management and corporate governance: the role of the board and the audit committee. J. Corp. Finance 9, 295316. Zhou, C.S., Ma, G., 2005. The ownership structure and nancial restatements in Chinese listed companies. Financ. Res. 8292 (in Chinese).

Cooking The Books

Загружено:

Сведения о документе

Исходное описание:

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Cooking The Books

Загружено:

Авторское право:

Доступные форматы

Journal of Corporate Finance 17 (2011) 371390

Contents lists available at ScienceDirect

Journal of Corporate Finance

Available online 25 September 2010 JEL classication: G14 K22 M41

M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390

M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390

M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390

M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390

376 Table 1 (continued)

M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390

M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390

M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390

M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390

M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390

M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390

M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390

Fig. 1. Plots of CAR.

M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390

epsjt pjt bpsjt = 0 + 1 + jt bpsjt1 bpsjt1

= 0 + rR Restate + rA After + rD RestateAfter + 1 + gR Restate + gA After + gD RestateAfter

pjt bpsjt bpsjt1

= 0 + rMindex After Mindex + 1 + gMindex After Mindex

pjt bpsjt bpsjt1

The number in parenthesis is the p-value of the t-statistic.

M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390 Table 5 (continued)

and indicates the 5% and 10% signicance levels of the t-statistics.

M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390

M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390

M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390

M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390

M. Firth et al. / Journal of Corporate Finance 17 (2011) 371390

Вам также может понравиться