CHINA JOURNAL OF ACCOUNTING STUDIES 2018, VOL. 6, NO. 1, 1–23 https://doi.org/10.1080/21697213.2018.1480005 The economic consequences of ﬁnancial fraud: evidence from the product market in China* a a b Qingquan Xin , Jing Zhou and Fang Hu a b Economics and Business Administration, Chongqing University, China; The Department of Accounting Finance and Economics, Griﬃth University, Australia ABSTRACT KEYWORDS economic consequences; This article measures the spillover eﬀect of the reputation loss ﬁnancial fraud; product induced by ﬁnancial fraud on the product market. Using 294 market; regulatory penalties regulatory penalties cases between 2004 and 2012, we ﬁnd that, compared with the control ﬁrms, ﬁrms engaging in ﬁnancial fraud exhibit a decline in sales revenue by 11.9–17.1% and a decrease in their gross proﬁt margin on sales by 2.4–2.8% in the three years after punishment. Furthermore, we ﬁnd that the sales revenue from the top ﬁve large customers falls 43.9–55.1% in the post- punishment period, while sales revenue from small customers does not decline signiﬁcantly. Overall, our analyses suggest that the damaged reputation as a result of ﬁnancial fraud has a major impact on the product market. Our ﬁndings help understand the trust mechanism in the Chinese product market. 1. Introduction Financial fraud is a phenomenon that exists throughout the world. It may be evidenced by a violation of information disclosure obligations that may result in reputational penalties to companies and managers in the capital market and labour market, including the decline in share prices (Armour, Mayer, & Polo, 2017; Karpoﬀ, Lee, & Martin, 2008), the increase in ﬁnancing costs (Graham, Li, & Qiu, 2008; Hribar & Jenkins, 2004; Kravet & Shevlin, 2010), and the loss of jobs for directors and executives (Agrawal & Cooper, 2017; Fich & Shivdasani, 2007). This paper investigates whether and how a company’s reputa- tional loss as a result of ﬁnancial fraud spreads to the product market in China. While a ﬁrm’s ﬁnancial fraud appears not to directly involve the channel partners (customers and suppliers) in the product market, whether and how a ﬁrm’s fraud aﬀects these partners in the supply chain is an interesting topic. The economic literature points out that commodity trading contracts are accompa- nied by a series of explicit and implicit commitments, such as suppliers’ commitments to product quality as well as maintenance in the future. These commitments often depend on the other channel parties’ reputation for self-protection due to the lack of explicit law protection (or the cost of using the law mechanism is too high) (Klein & Leﬄer, 1981; Shapiro, 1983). Although ﬁnancial fraud is not directly aﬃliated with customers and CONTACT Qingquan Xin email@example.com *Paper accepted by Kangtao Ye. © 2018 Accounting Society of China 2 XIN ET AL. suppliers, it may lead customers and suppliers to re-evaluate fraud ﬁrms’ abilities and incentives to enforce the product market contracts, thereby aﬀecting their subsequent trading behaviour. First, serving as a signal of dishonesty, ﬁnancial fraud makes customers and suppliers cast doubt on a ﬁrm’s commitments in the product market, which will weaken the incentives for customers and suppliers to sign contracts with the company. Second, ﬁnancial fraud directly aﬀects a ﬁrm’s ﬁnancing abilities and ﬁnancing costs, as well as the adjustments of corporate governance (such as the departure of executives). This leads to great diﬃculties and uncertainties in a company’s production and operation activities. Thus, it is impossible for fraud ﬁrms to fulﬁl their existing commitments (or future commitments) (Cornell & Shapiro, 1987). Given these uncertainties, customers and sup- pliers are concerned about whether to continue the contracts with fraud ﬁrms. In addition, the stakeholders including customers and suppliers use ﬁnancial reports and rely on the quality of accounting information to monitor and control trading contracts in the product market (FASB, 1978; Raman & Shahrur, 2008). For example, ﬁrms engaging in ﬁnancial fraud may mislead customers and suppliers through misstating the ﬁrms’ performance, or they may falsify the purchase–sale contracts using distorted accounting performance. After regulatory penalties are announced and imposed on fraud ﬁrms, customers and suppliers ﬁnd that the operation of these ﬁrms is much worse than it is supposed to be. Given the foregoing reasons, it can reasonably be inferred that fraud ﬁrms will suﬀer a loss in the product market following the exposure of ﬁnancial fraud. This paper contributes to understanding the trust mechanism in the context of the Chinese product market. A large amount of the literature points out that, although China has become the world’s second largest economy, its institutions in the market, such as the legal enforcement mechanism, are still relatively weak, and commercial transactions are more often sustained through political or social networks (Allen, Qian, & Qian, 2005; Hung, Wong, & Zhang, 2015). If the parties to transactions in the product market have established trust primarily through social networks, commodity trading can be con- ducted eﬃciently through the exchange of information within the social trust circle. The counterparties’ damaged reputation due to ﬁnancial fraud should have less inﬂuence on trading within the circle. On the contrary, China has experienced market-oriented reforms over the past 40 years. The trading contract in the product market to some extent exhibits the characteristic of an arm’s-length transaction, that is, the public information plays a more important role in solving issues of information asymmetry and trust. As information becomes more transparent, the ﬁrms engaging in fraud are subject to more public scrutiny. That will have more reputational eﬀects on the main- tenance of contract. Examining the economic consequences of ﬁnancial fraud in the product market provides a reﬂection on the trust mechanism in the Chinese product market. The sample consists of 294 listed manufacturing companies involved in ﬁnancial fraud cases during 2004–2012. First, we examine ﬁrms’ performance in the product market following a punishment announcement using two proxies: sales revenue and Financial fraud in this paper refers to the manipulation of ﬁgures in ﬁnancial statements, the missing disclosure of major information, misleading disclosure, or the postponement of disclosure. We identify ﬁnancial fraud based on the regulatory authorities’ enforcement actions. The detailed deﬁnition of fraud is presented below. CHINA JOURNAL OF ACCOUNTING STUDIES 3 gross proﬁt margin. Compared with the control ﬁrms (those non-fraud ﬁrms by match- ing), we ﬁnd that fraud ﬁrms’ sales revenue declines by 11.9–17.1% in the three years after punishment and the gross proﬁt margin decreases by 2.4–2.8%, suggesting that ﬁnancial fraud has a substantial adverse eﬀect on the product market performance. We then analyse and examine the reputational penalties incurred by ﬁnancial fraud from the perspective of customer concentration. After a ﬁrm’s fraud is announced in the market, would the large customers or small customers be more or less willing to continue trading with the company? In theory, the trade between large customers and ﬁrms is more likely to arise from a relationship based on transactions. Williamson (1985) has demonstrated that more frequent transaction and more relationship-speciﬁc investments lead large customers to have a stronger incentive for reputation. Therefore, the large customers have strong incentives to keep clear of fraud ﬁrms with damaged reputations. On the other hand, large customers have a (bilaterally) locked relationship with a penalised company because of the relationship-speciﬁc investments. It will result in higher costs if they refuse to trade with the penalised company in the future. How the customer concentration changes following the regulatory punishment for fraud ﬁrms is, thus, an empirical issue. We manually collect and analyse the data on the ratio of sales revenue from the top ﬁve large customers to ﬁrms’ total revenue disclosed in the annual reports. We ﬁnd that compared with the control ﬁrms, in the three-year window surrounding the announce- ment of the regulatory penalties, fraud ﬁrms’ customer concentration drops by 3.2–3.5% after the punishment and the decline rate exceeds 10.6% relative to the mean of the total sample. Further, we decompose sales revenue into two parts: revenue from the top ﬁve customers (large customers) and revenue from the remaining customers (small customers), we ﬁnd that the sales revenue from the top ﬁve customers falls by 43.9– 55.1% after the punishment, which is very signiﬁcant economically. On the contrary, sales from small customers do not suﬀer a signiﬁcant decline following regulatory penalties. This evidence indicates that large customers are more likely to reduce the commercial transactions with the penalised fraud ﬁrms. These customers have less pressure from relationship-speciﬁc investment, and are not afraid of losing the transac- tional relationship with fraud ﬁrms. The reason may be that the manufacturing industry has been generally dominated by the buyer’s market over recent years so that large customers in this industry have more negotiation power in commercial transactions. We further examine the changes in supplier concentration. If manufacturing industry is generally dominated by the buyer’s market, then the reputational eﬀects on suppliers should be small. The suppliers will face higher losses in speciﬁc investment projects and have more diﬃculties in ﬁnding new large customers when they terminate the com- mercial transactions with the penalised fraud ﬁrms. Thus, the change in supplier con- centration for fraud ﬁrms should be not as much as the change in customer concentration following the regulatory punishment. Our data show that compared with the control ﬁrms, fraud ﬁrms’ supplier concentration (the proportion of purchase from the top ﬁve suppliers in the total purchase amount) changes little surrounding the punishment announcement. However, we ﬁnd that the trade credit oﬀered by suppliers to fraud ﬁrms declines in the post-punishment period. We also ﬁnd that the severity of regulatory penalties increases these consequences to ﬁnancial fraud ﬁrms in the product market. Speciﬁcally, the decline in sales 4 XIN ET AL. revenue, gross proﬁt margin, and customer concentration are more pronounced when regulatory penalties are more severe. We perform some robustness tests, including using cash sales in replacement of total sales revenue, controlling for ﬁnancing ability or debt level, controlling for management turnover, using dynamic models to test the parallel trend assumption, and re-selecting the control sample. The above results are all robust. Our study contributes to the literature that examines the economic consequences of corporate fraud in China. Recent literature mainly studies the impact of ﬁnancial fraud on the capital market or the labour market. Yet, little research has examined and measured consequences of a fraudulent ﬁnancial statement on the product market. Karpoﬀ et al. (2008) suggest that shareholder loss caused by ﬁnancial fraud is due to investors’ anticipation of losing customers and sales revenue in the product market. To our knowledge, only Johnson, Xie, and Yi (2014) provide empirical evidence on the reputational penalties of ﬁnancial fraud in the context of the American product market. The academic community still lacks theoretical analyses and empirical evidence from an emerging economy. Our examination of the reputational losses spilling over to the product market will shed some lights on understanding the reputational eﬀects and trust mechanism in an emerging market. Furthermore, existing research shows that the characteristics of customers and supplies will aﬀect the quality as well as managerial behaviour in the preparation of ﬁnancial statements, but their ﬁndings are inconsistent. Bowen, DuCharme, and Shores (1995) ﬁnd that ﬁrms are willing to choose long-run income-increasing accounting methods to earn a good impression of customers and suppliers with respect to their ﬁnancial performance. Under this framework, Raman and Shahrur (2008) examine the impact of customer–supplier relationship-speciﬁc investments on earnings management. They ﬁnd that the supplier industry's intensity of relationship- speciﬁc investments is positively related to the magnitude of earnings management, and the relationship between the customer industry’sintensity of speciﬁc investments and earnings management largely depends on ﬁrms’ bargaining power. However, Hui, Klasa, and Yeung (2012) suggest that, when a ﬁrm’s customers or suppliers have greater bargaining power, the ﬁrm’s accounting conservatism is higher. Wang and Liu (2014) and Fang and Zhang (2016) conducted this research in China. This paper provides more evidence by investigating how the product market responds to cor- porate fraud. In particular, we ﬁnd that large customers are more likely to reduce the size of transactions with fraudulent ﬁrms than are small customers, which is diﬀerent from prior studies. Last but not least, this paper extends the empirical literature on China’s securities market regulation and legal enforcement. Based on a sample of listed companies penalised by the China Securities Regulatory Commission, Chen et al. (2006) and Zhang and Ma (2005) examine the determinants of corporate fraud; Chen, Firth, Gao, and Rui (2005) and Liebman and Milhaupt (2008) investigate the economic conse- quences of regulatory penalties to listed ﬁrms; Xin, Hunag, and Yi (2013) study the phenomenon of independent directors suﬀering regulatory punishment in ﬁnancial misrepresentation cases. This paper focuses on the product market consequences of fraud ﬁrms due to reputational eﬀects, supplementing and extending existing research on China’s securities supervision. CHINA JOURNAL OF ACCOUNTING STUDIES 5 2. Literature review Assuming reputation serves as a mechanism for providing incentives and abilities to enforce contracts, a damaged reputation will increase the parties’ concerns over the transactions involved as they result in adverse consequences to ﬁrms and management. Firms engaging in fraud explicitly violate the laws and regulations about information disclosure of listed ﬁrms, which is a breach of legitimacy. It will lead information users to doubt ﬁrms’ commitments on contracts and reputation, thus aﬀecting engagement in transactions. Prior studies have provided evidence that accounting fraud in American companies results in 27% shareholder loss (Karpoﬀ et al., 2008), and the equity costs will increase after the fraud has been detected (Hribar & Jenkins, 2004; Kravet & Shevlin, 2010). In terms of debt ﬁnancing contracts, Graham et al. (2008) ﬁnd that ﬁrms’ bank loans have signiﬁcant higher spreads, shorter maturities, higher likelihood of being secured, and more covenant restrictions after earnings restatement. Chen, Chang, and Lo (2013) suggest that in the post-restatement period, ﬁrms depend more on debt ﬁnancing and less on equity ﬁnancing. Using class action lawsuits as a proxy for fraud, Johnson et al. (2014) ﬁnd that customers impose signiﬁcant reputational penalties on fraud ﬁrms, including a higher likelihood of a break-up in the business with fraud ﬁrms, a signiﬁcant drop in sales revenue from large customers, and a decline in the operating performance as a result of increased selling costs. The management of ﬁnancial fraud ﬁrms has also received reputational penalties. For example, Fich and Shivdasani (2007) ﬁnd that the independent directors experience a higher likelihood of losing jobs than other non-fraud ﬁrms when they are facing share- holder class lawsuits on account of ﬁnancial fraud. Agrawal and Cooper (2017) show that after the restatement announcement, CEOs and CFOs face, respectively, a 14% and 10% greater probability of being replaced in restating ﬁrms than in control ﬁrms. Moreover, some studies point out that ﬁnancial scandals even have an adverse eﬀect on employ- ees’ job hunting. Regarding the corporate fraud and regulatory penalties of Chinese listed companies, some studies have provided evidence. Chen et al. (2005) suggest that over the ﬁve days surrounding the regulatory punishment announcement, ﬁrms’ stock prices fall by 1–2%. They also suggest that after the regulatory punishment, ﬁrms experience an increased likelihood of auditor turnover, a higher incidence of modiﬁed audit opinions, a greater rate of CEO departure, and wider bid-ask spreads, which indicates that the China Securities Regulatory Commission is not a toothless tiger. Liebman and Milhaupt (2008) carefully examine the eﬀectiveness of public condemnation from stock exchanges. They ﬁnd that public condemnation has a signiﬁcantly adverse impact on fraudulent ﬁrms’ stock prices. In addition, according to the results of the interview investigation, public condemnation seems to adversely aﬀect the companies’ future business operations (such as bank loans). Xin et al. (2013) examine the phenomenon that independent directors are subject to regulatory sanction for ﬁrms’ ﬁnancial mis- representation. They show that the penalties on independent directors are signiﬁcantly lower than those of non-independent directors. They also ﬁnd that the board seats held by independent directors experience a signiﬁcant decline following the regulatory See Harvard Business Review, 2016 (9), ‘The Scandal Eﬀect’ (in Chinese). 6 XIN ET AL. punishment, probably as the result of independent directors’ intentions to leave high- risk listed companies. Moreover, other studies ﬁnd that the role of the regulatory penalties of the China Securities Regulatory Commission is relatively limited. For exam- ple, Song, Li, and Ji (2011) investigate the eﬀect of regulatory penalties arising from listed companies’ management forecast violations, and they ﬁnd that regulatory penal- ties do not lower the probability of the subsequent management forecast violations. Based on the Chinese market environment, Hung et al. (2015) divide corporate scandals into three categories: political scandals, mixed scandals and market scandals, and then examine the value eﬀects of diﬀerent categories of scandals. They suggest that political ties and market credibility are both valuable to ﬁrms. They argue that, if commercial transac- tions rely more on political networks and rely less on market credibility, then the destruc- tion of ﬁrms’ political ties will have a much more signiﬁcantly adverse eﬀect on stock returns than the destruction of market credibility. In their research, corporate ﬁnancial fraud is classiﬁed as market scandals. They ﬁnd that market scandals lead to a loss of about 13% in the two years surrounding the scandal event, but the scandals signalling the destruction of political ties result in a decline of around 40%, which indicates that political ties are more valuable to contracting than market credibility. We conclude that the product market is one of the main sources for creating ﬁrm’s value. Examining the economic consequences of ﬁnancial fraud on the product market will have more insights into reputational eﬀects in the Chinese business environment. 3. Sample selection and descriptive statistics This paper selects all the A-share listed manufacturing companies with enforcement actions for ﬁnancial fraud from 2004–2012 as the initial sample. The main reason why the sample period is limited to 2004–2012 is that we need to examine the changes in transactions between the ﬁrms and their customers in the three years surrounding the punishment, and we can only obtain the data on customer concentration for 2001–2015. To ensure ﬁrms have data within the three-year window, the sample period is narrowed to 2004–2012. To exclude the variation across diﬀerent industries, we select the sample from the manufacturing industry that is the most characterised by product market competition. Corporate ﬁnancial fraud often refers to the manipulation of ﬁgures in ﬁnancial state- ments and the omitting, misleading, or postponing disclosure of important information. On 9 January 2003, the China Securities Regulatory Commission promulgated its ‘Several Provisions of the Supreme People’s Court on Trial of Civil Compensation Cases Triggered by False Statements in the Securities Market’,which deﬁnes corporate ﬁnancial fraud as, ‘making false statements or misleading statements that are contrary to the facts of major events, or making major omissions in the disclosure of information or disclosing information unfairly’ (Article 17). In accordance with the classiﬁcation criteria of ‘Chinese listed compa- nies’ regulations violation database’ from the China Security Market and Accounting Research (CSMAR), we identify seven violation types in the database as ﬁnancial fraud: ‘ﬁctitious proﬁts’, ‘virtual assets’, ‘false statements’, ‘signiﬁcant omissions’, ‘untrue disclo- sure’, ‘general accounting mishandling’ and ‘postponed disclosure’ (Xin, Huang, & Yi, 2013). By processing the violation data from CSMAR, we obtain a total of 294 penalised cases in 282 listed manufacturing companies engaging in ﬁnancial fraud during 2004–2012. Table 1, Panel A reports the sample distribution by year. CHINA JOURNAL OF ACCOUNTING STUDIES 7 Table 1. Descriptive statistics of manufacturing listed companies subject to regulatory penalties for fraud. Panel A: Sample distribution by year N 18 23 14 16 13 33 37 55 85 294 Year 2004 2005 2006 2007 2008 2009 2010 2011 2012 Total Panel B: Punishment types and institutions Punishment types Ordering Public Public Warning & Total correction criticism condemnation monetary ﬁnes Shanghai 12 9 17 0 38 Shenzhen 45 49 30 2 126 CSRC 112 2 0 26 140 CMF 1 0 0 4 5 Multiple institutions (13) (0) (0) (2) (15) Total 157 60 47 30 294 In response to corporate fraud, the regulatory authorities such as the China Securities Regulatory Commission (CSRC), the Shanghai and Shenzhen Stock Exchanges, and the China Ministry of Finance (CMF) may impose various administrative sanctions on ﬁrms, including ordering correction, public criticism, public condemnation, oﬃcial warning, and monetary ﬁnes. Ordering correction is mainly aimed at companies with minor infraction, requiring illegal companies to rectify within a time limit. The CSRC, CMF and stock exchanges may all issue notices ordering correction for violation with a slight nature. For the more severe violation, the CSRC imposes either an oﬃcial warning or monetary ﬁnes on the company, both of which are formal administrative penalties. In general, oﬃcial warnings and monetary ﬁnes are used simultaneously for certain violations but, in some cases, separately. In terms of corporate ﬁnancial fraud, the monetary ﬁnes for ﬁrms are generally between 100,000 and 600,000 RMB. Since the stock exchanges are the most liquid and active bodies in the Chinese securities market and have an information advantage in securities regulation, the Security Law entitles the stock exchanges to rights of disciplinary actions, over- seeing ﬁrms who violate information disclosure in the practice of securities reg- ulation. Speciﬁcally, the Exchanges usually implement the two punishment methods – public criticism and public condemnation. Although public condemna- tion is more severe than public criticism, both apply to ﬁrms with minor violation. Investors cannot ﬁle lawsuits against the oﬀenders according to either of these two penalty reports. In addition, according to the Accounting Law, the CMF has the authority to supervise and inspect ﬁrms’ accounting misconduct and it can make ordering correction or monetary ﬁnes in accordance with the seriousness of the case. Table 1, Panel B presents the descriptive statistics of punishment types and regulatory authorities. According to the ‘Several Provisions of the Supreme People’s Court on Trial of Civil Compensation Cases Triggered by False Statements in the Securities Market’, issued on 9 January 2003, the precondition for investors to ﬁle civil lawsuits for ﬁnancial fraud is that the company has already been imposed on administrative penalties by CSRC, CMF and their dispatched agencies. Ordering correction, public criticism and public condemnation cannot be used as a prerequisite for investor prosecution. 8 XIN ET AL. In order to empirically examine the product market changes for fraud ﬁrms surrounding the year of punishment, we need to have a control sample for bench- marking. We select the control sample based on ﬁrms’ ﬁnancial characteristics prior to the year of punishment for the 294 ﬁnancial fraud cases. Speciﬁcally, we screen manufacturing companies that have never been subject to regulatory penalties (non- fraud ﬁrms) during the sample period and construct a control sample including all ﬁrm-year observations (3344 in total). We then pool the sample of 3344 control samples and the sample of 294 fraud ﬁrms, resulting in a pooled sample of 3638 ﬁrm-year observations. Then we analyse the determinants of ﬁnancial fraud, using the model as follows: FRAUD ; t ¼ α þ α SIZE ; t 1 þ α LEV ; t 1 þ α BTM ; t 1 þ α ROA ; t 1 i 0 2 i 3 i 4 i 5 i þα EP ; t 1 þ α LISTAGEi; t 1 þ YEARFE þ ε ; t (1) 6 i 7 i In model (1), the dependent variable is FRAUD , , which is an indicator variable i t for 294 fraud ﬁrms. Independent variables are for ﬁnancial characteristics such as the total assets and debt-to-assets ratio. The deﬁnition of each variable is sum- marised in Table 2. By regressing model (1), we can get the propensity score for each observation. Then we can match each fraud ﬁrm with the non-fraud ﬁrms based on the closest propensity score. As a result, we obtain 294 pairs of fraud and control ﬁrms. The propensity scores of each pair do not exceed 0.02, indicat- ing a good match. Table 3, panel A reports the regression results of propensity score matching estima- tion using model (1). It shows that the ﬁnancial fraud ﬁrms have a smaller company size (SIZE), a higher debt-to-assets ratio (LEV), and a lower performance level (ROA). Panel B represents the results of validity test of the propensity score matching. There are no signiﬁcant diﬀerences in ﬁnancial characteristics between the two samples, indicating that our control sample is suitable. We employ the data of three years before and after punishment (6-year window) to examine the economic consequences of ﬁnancial fraud in the product market. Based on 294 pairs of fraud and non-fraud ﬁrms matched on propensity scores, we adopt the following model: Y ; t ¼ α þ α FRAUD þ α POST þ α FRAUD POST þ CONTROLS ; t 1 i 0 1 i 2 3 i i þYEARFE þ PAIRFE þ ε ; t (2) In model (2), the dependent variable is Y , , which represents ﬁrms’ observed i t performance in the product market, including sales revenue, gross proﬁtmargin, and customer concentration. In particular, if fraud ﬁrms are involved in earnings manipulation, wehavecorrected the ﬁnancial statement as much as possible according to the enforcement actions, and adjusted the dependent variable Y , . i t FRAUD is an indicator variable which is equal to one for fraud ﬁrms and zero for control ﬁrms. POST is an indicator variable, taking the value of one for the post- punishment period (year t+1, t+2 and t+3), and zero for the pre-punishment period (year t–3, t–2and t–1). In order to eliminate the eﬀect of shock in the As a robustness test, we delete the seven cases of earnings manipulation fraud and re-estimate the model. The results of our analyses remain qualitatively similar. CHINA JOURNAL OF ACCOUNTING STUDIES 9 Table 2. Deﬁnition of variables. Variable Deﬁnition LN(SALES) Natural logarithm of ﬁrms’ total sales revenue. GROSS_MARGIN Gross proﬁt margin, equal to (sales revenue – sales costs)/sales revenue. CUST_CONCEN Customer concentration, equal to the proportion of sales revenue from the top ﬁve customers to ﬁrm’s total sales revenue. FRAUD Indicator variable equal to one for the ﬁnancial fraud ﬁrms, and zero for the control ﬁrms. POST Indicator variable equal to one for the post-punishment period (year t+1, t+2 and t+3), and equal to zero for the pre-punishment period (year t–3, t–2 and t–1). SIZE Natural logarithm of ﬁrms’ total assets. LEV The debt-to-assets ratio, equal to the total liability divided by total assets. ROA Return on assets. BTM Book-to-market ratio, equal to the proportion of net assets per share to stock price per share. EP Earnings-to-price ratio, equal to the proportion of earnings to market value. LISTAGE Natural logarithm of the number of years the ﬁrm has been listed. TOP5 The sum of the proportion of shares held by the top ﬁve largest shareholders. INDEP The proportion of independent directors on the board. LN(SALES_BIG) Natural logarithm of sales revenue from the top ﬁve large customers. LN(SALES_SMALL) Natural logarithm of sales revenue from the remaining customers. SUPPLIER Equal to the ratio of purchase amount from the top ﬁve suppliers to the total purchase CONCENTRATION amount. TRADE CREDIT The ratio of the changes in (notes payable + accounts payable – prepayments) to the total assets. PUN_MILD Indicator variable equal to one for the punishment type of ordering correction and zero otherwise. PUN_SEVERITY Indicator variable equal to one for the punishment types of public criticism, public and zero otherwise. condemnation, oﬃcial warning, or monetary ﬁnes LN(CASH_SALES) Natural logarithm of cash sales which is the item in cash ﬂow statement “cash from selling commodities or oﬀering labour”. FIN_CASH Debt ﬁnancing ability, equal to the proportion of diﬀerence in the item of “borrowing” and the item of “cash paid for debt” in cash ﬂow statement to the total assets. TURNOVER Indicator variable equal to one if the chairman or CEO leaves the oﬃce during the event window [t+0, t+3], and his or her age is less than 60, and zero otherwise. –1 BEFORE Indicator variable for the ﬁrst year before punishment (year t–1) and zero otherwise. AFTER° Indicator variable for the year of punishment (year t+0) and zero otherwise. AFTER Indicator variable for the ﬁrst year after punishment (year t+1) and zero otherwise. 2+ Indicator variable for the second (or third) year (year t+2, t+3) after punishment and zero AFTER otherwise. distribution of punishment year, we remove the observations in the year of punishment (year t+0). We are interested in the interaction term of FRAUD and POST.If the coeﬃcient on the interaction term is negative and statistically signiﬁcant, it indicates that ﬁnancial fraud ﬁrms experience a worse performance in the product market after punishment than the control ﬁrms. In model (2), we include a series of control variables that may aﬀect ﬁrms’ performance in the product market, such as ﬁrm Assuming that ﬁrm A is subject to regulatory penalties for ﬁnancial fraud in the year 2008. According to the matching results of PSM, the control sample of ‘ﬁrm A-2008’ maybe ‘ﬁrm B-2010’ (the punishment year of the fraud ﬁrm is not the same as the matching year of the control ﬁrm). Thus, for ﬁrm A, POST is equal to one for the year of 2009–2011, and equal to zero for the year of 2005–2007. For ﬁrm B, POST is equal to one for the year of 2011–2013, and equal to zero for the year of 2007–2009. This is because the event window of the treating sample is not the same as that of control sample, which may have an eﬀect on results. The purpose of our paper is to ﬁnd a control sample with the closest probabilities of being subject to regulatory penalties for each treating sample, therefore we do not require the punishment year of the treating sample to be equivalent to the matching year of the control sample. In addition, we control for year-ﬁxed eﬀects in model (2), which can eﬀectively eliminate the impact of diﬀerences in years between the two samples on product market performance. As a robustness test, for each fraud ﬁrm, we identify a control ﬁrm in the same punishment year with the closest propensity score. This procedure leads to a new control sample. Then we re-conduct the empirical regression and the results remain the same (see Panel E, Table 9 for the results). 10 XIN ET AL. Table 3. Regression of PSM and matching results. Panel A: Regression of PSM Dependent variable: FRAUD SIZE –0.308*** [–4.26] LEV 1.110*** [3.13] ROA –3.134** [–2.38] BTM –0.041 [–0.39] EP –1.978*** [–3.59] LISTAGE –0.156 [–1.03] CONSTANT 3.644*** [2.59] YEAR FE YES Pseudo R 0.114 N 3638 Panel B: The diﬀerences in characteristics between the fraud and control ﬁrms Fraud ﬁrms Control ﬁrms Mean (The treating sample, N=294) (The control sample, N=294) diﬀerence test SIZE 21.290 21.363 –0.073 LEV 0.541 0.533 0.008 ROA –0.001 0.006 –0.007 BTM 0.750 0.756 –0.006 EP –0.039 –0.037 –0.002 LISTAGE 2.256 2.278 –0.022 Notes: Panel A reports the regression results of propensity score matching estimation using the model (1). Regression coeﬃcients are in the table and T values are in brackets. Panel B reports the mean diﬀerences in variables between fraud ﬁrms (the treating sample) and control ﬁrms (the control sample). All continuous variables are winsorised at the *, **, *** indicate two-sided signiﬁcance levels of 10%, 5%, and 1%, respectively. 1st and 99th percentiles. size, debt-to-assets ratio, accounting performance, book-to-market ratio, and so on. Additionally, we control for year-ﬁxed eﬀects as well as matching-pair-ﬁxed eﬀects. The descriptive statistics for the main variables used in model (2) is presented in Table 4. Table 4. Descriptive statistics of major research variables. Variable N Mean Sd P25 P50 P75 LN(SALES) 3501 20.861 1.442 20.052 20.891 21.766 GROSS_MARGIN 3501 0.215 0.160 0.112 0.186 0.284 CUST_CONCEN 3516 0.303 0.216 0.146 0.247 0.412 SIZE 3476 21.312 1.106 20.605 21.253 21.955 LEV 3476 0.558 0.470 0.351 0.514 0.639 ROA 3476 0.014 0.103 0.006 0.024 0.054 BTM 3461 0.750 0.871 0.300 0.540 1.010 EP 3461 0.004 0.192 0.007 0.027 0.062 LISTAGE 3516 2.231 0.556 1.946 2.303 2.639 TOP5 3478 0.507 0.145 0.404 0.507 0.611 INDEP 3439 0.344 0.078 0.333 0.333 0.375 Notes: See Table 2 for the detailed deﬁnition of variables. All continuous variables are winsorised at the 1st and 99th percentiles. CHINA JOURNAL OF ACCOUNTING STUDIES 11 4. Product market performance This paper investigates the product market performance for fraud ﬁrms, measured by sales revenue and gross proﬁt margin. Sales revenue is the main source of ﬁrms’ value creation, and gross proﬁt margin is an important indicator to assess ﬁrms’ competitive- ness in the product market. Prior literature, examining how fraud aﬀects the shareholder wealth, holds the view that investors’ anticipation of losing customers and sales revenue in the product market is the main reason for a decline in stock prices after ﬁnancial fraud has been detected (e.g. Karpoﬀ et al., 2008). However, as far as we know, there is no empirical evidence that directly examines the impact of ﬁnancial fraud on sales revenue and gross proﬁt margin in the Chinese market. Figures 1(a) and 1(b) plot the changes in sales revenue and gross proﬁt margin surrounding the punishment period for fraud and control ﬁrms, respectively. The ﬁgures show that the natural logarithm of sales revenue (LN(SALES)) and gross proﬁt margin (GROSS_MARGIN) show a constant trend in the two samples for the pre-punishment period (year t–3, t–2 and t–1), suggesting our study meets the requirement of a parallel 21.50 21.20 20.90 20.60 20.30 20.00 T- 3 T - 2 T- 1 T + 0 T + 1 T + 2 T + 3 Fraud firms Control firms Figure 1. (a) Change trend in (LN(SALES)). 0.23 0.22 0.21 0.20 0.19 0.18 0.17 T- 3 T - 2 T- 1 T + 0 T + 1 T + 2 T + 3 Fraud firms Control firms Figure 1. (b) Change trend in (GROSS_MARGIN). 12 XIN ET AL. trend using the diﬀerence-in-diﬀerences approach. In addition, the change trend in the two samples shows signiﬁcant diﬀerences in the post-punishment period (year t+1, t+2 and t+3), suggesting regulatory penalties on fraud have an adverse impact on the product market. Table 5, Panel A reports the univariate analyses of sales revenue and gross proﬁt margin for the two samples. For sales revenue, the mean (median) of natural logarithm of sales revenue (LN(SALES)) for ﬁnancial fraud ﬁrms in the post-punishment period is 0.282 (0.361) higher than that in pre-punishment period, and the diﬀerences in means Table 5. The product market consequences of ﬁnancial fraud: sales revenue and gross proﬁt margin. Panel A: Univariate analysis Sales revenue (LN(SALES)) Gross proﬁt margin (GROSS_MARGIN) POST=0 POST =1 Diﬀerence POST =0 POST =1 Diﬀerence Fraud ﬁrms 20.574 20.856 0.282*** 0.219 0.214 –0.005 (FRAUD=1) (20.563) (20.924) (0.361)*** (0.193) (0.181) (–0.012)* N=881 N=866 N=881 N=866 Control ﬁrms 20.769 21.250 0.481*** 0.203 0.222 0.019* (FRAUD=0) (20.792) (21.241) (0.449)*** (0.179) (0.190) (0.011)* N=881 N=873 N=881 N=873 Diﬀerence-in-diﬀerences –0.199*** –0.024*** (–0.088) (–0.023)*** Panel B: Multiple regression results Dependent variable: LN(SALES) Dependent variable: GROSS_MARGIN (1) (2) (3) (4) FRAUD –0.139* –0.086 0.019** 0.017** [–1.689] [–1.211] [2.045] [1.974] POST –0.464*** –0.332*** –0.004 –0.009 [–5.021] [–4.057] [–0.366] [–0.816] FRAUD*POST –0.187** –0.127* –0.028*** –0.024** [–2.563] [–1.962] [–2.759] [–2.528] SIZE –0.028*** [–4.441] LEV 0.392* –0.007 [1.790] [–0.342] ROA 2.686*** [6.818] BTM 0.561*** –0.025*** [10.676] [–3.999] EP –0.443** 0.100*** [–2.564] [5.737] LISTAGE 0.246*** –0.011 [2.680] [–0.888] TOP5 0.582* 0.140*** [1.899] [3.567] INDEP –1.754*** 0.150* [–3.224] [1.740] CONSTANT 21.397*** 19.534*** 0.263*** 0.835*** [81.041] [54.598] [9.542] [6.295] YEAR FE YES YES YES YES PAIR FE YES YES YES YES 0.511 0.593 0.328 0.380 adj. R N 3410 3410 3410 3410 Notes: The ﬁgures in panel A are the sample mean, and the sample median is in brackets. The T-test is used for testing the diﬀerences in mean and the Ranksum test is used for testing the diﬀerences in median. The ﬁgures in panel B are regression coeﬃcients and T values are in brackets. The T-value is based on standard errors adjusted for ﬁrm-level clustering. See Table 2 for deﬁnition of variables. All continuous variables are winsorised at the 1st and 99th *, **, *** percentiles. indicate two-sided signiﬁcance levels of 10%, 5%, and 1%, respectively. CHINA JOURNAL OF ACCOUNTING STUDIES 13 and medians are statistically signiﬁcant (both at the 0.01 level). These results indicate that the absolute level of sales revenue in fraud ﬁrms does not decline. This is likely to be confounded by the rapid economic growth and the inﬂation of prices. For control ﬁrms, the mean (median) of LN(SALES) is 0.481 (0.449) higher than that in pre-punish- ment period, and the diﬀerence is more than that of fraud ﬁrms. The diﬀerence-in- diﬀerences test shows that the diﬀerence of the change in the mean sales revenue between the fraud ﬁrms and control ﬁrms is –0.199 and is signiﬁcant at the 0.1 level, indicating that, compared with the control ﬁrms, ﬁnancial fraud ﬁrms experience a signiﬁcant decline in sales revenue following the regulatory punishment. In terms of gross proﬁt margin (GROSS MARGIN), its mean (median) drops from 21.9% (19.3%) in the pre-punishment period to 21.4% (18.1%) in the post-punishment period. The mean and median of gross proﬁt margin for the control ﬁrms increase signiﬁcantly following the punishment, respectively. The diﬀerence-in-diﬀerences test shows that, compared with the control ﬁrms, the mean and median of gross proﬁt margin for ﬁnancial fraud ﬁrms decrease by 2.4% and 2.3% respectively, both statistically signiﬁcant at the 0.01 level. Overall, the results of Table 5, panel A suggest that compared with the control ﬁrms, both sales revenue and gross proﬁt margin decline following regulatory punishment, indicating that fraud ﬁrms suﬀer a certain reputational loss in the product market. Table 5, Panel B presents the multiple regression results for product market perfor- mance. Consistent with our hypothesis, the coeﬃcients on the interaction term FRAUD*POST are all signiﬁcantly negative. These results indicate that, compared with the control sample, fraud ﬁrms have statistically signiﬁcantly less sales revenue and gross proﬁt margin following the regulatory punishment. Speciﬁcally, as reported in columns 1–2, the coeﬃcients on FRAUD*POST are –0.187 and –0.127, respectively, which suggests that compared with the control ﬁrms, the sales revenue of fraud ﬁrms decline by 11.9–17.1% in the three years after punishment. In columns 3–4, the coeﬃcients on FRAUD*POST are –0.028 and –0.024 respectively, indicating that fraud ﬁrms’ gross proﬁt margin falls by 2.4–2.8%. Relative to the mean of gross proﬁt margin (0.215), the decreasing rate of gross proﬁt margin is more than 11%. Overall, the evidence in Table 5 indicates that ﬁrms engaging in fraud have reputational eﬀects, suﬀering substantial losses in the product market. 5. Customer concentration One of the key issues that this paper focuses on is whether large customers are likely to terminate their contracting with fraud ﬁrms. As mentioned in the introduction, large customers have higher demand for the reputation of parties in the supply chain. The exposure of corporate misconduct behaviour may have a greater impact on large customers. On the contrary, large customers may have alternative mechanisms to maintain trust with the channel parties, or they may face higher switch costs due to previous relationship-speciﬁc investments after the termination of contracting with fraud ﬁrms. So far, we have no theoretical predictions on this issue. Figure 2 illustrates the change in the mean of customer concentration (CUST_CONCEN) surrounding the punishment period. As illustrated in the ﬁgure, fraud ﬁrms have a higher customer concentration than control ﬁrms. Speciﬁcally, in the pre- 14 XIN ET AL. 0.35 0.32 0.29 0.26 0.23 0.20 T- 3 T - 2 T- 1 T + 0 T + 1 T + 2 T + 3 Fraud firms Control firms Figure 2. Change trend in (CUST_CONCEN). punishment period (years t–3, t–2 and t–1), the diﬀerence in customer concentration between the two samples remains stable. But in the post-punishment period, the diﬀerence gradually decreases. This indicates that customer concentration meets the assumption of parallel trend using the diﬀerence-in-diﬀerences approach. And it also suggests that ﬁnancial fraud has an eﬀect on the contracting relationship between large customers and fraud ﬁrms, speciﬁcally, large customers rely more on the reputation of contracting parties. Table 6, panel A reports the univariate analysis of customer concentration. For fraud ﬁrms, the mean and median of the ratio of sales revenue of the top ﬁve customers to total sales revenue both decline 1.8% following the regulatory punishment. Compared with the control ﬁrms, its mean and median decrease by 3.3% and 2.7%, respectively, both statistically signiﬁcant at the 0.01 level. Panel B reports the multiple regression results. It shows that the coeﬃcients on FRAUD*POST are –0.035 and –0.032, respectively, which are both signiﬁcant at the 0.05 level. Given the mean of customer concentration is -0.303, compared with the control ﬁrms customer concentration of fraud ﬁrms declines by 10.6–11.6% after fraud. This evidence suggests that large customers are more likely to reduce the transactions with fraud ﬁrms than small customers when the regulatory penalties are imposed on fraud ﬁrms. Based on the results in Tables 5 and 6, we infer that the decline in sales revenue for fraud ﬁrms may be mainly due to the decreasing transaction amount of large customers. To further examine this issue, we distinguish between the changes in sales revenue from large customers and from small customers surrounding the punishment announcement. Columns 3–4of Table 6, panel B presents the regression results of the sales revenue from large customers. They show that the coeﬃcients on FRAUD*POST are –0.801 and –0.578, which are signiﬁcant at the 0.01 and 0.05 levels, respectively. From the perspective of economic signiﬁcance, sales revenue from large customersdeclinesby43.9–55.1% after punishment, which is a signiﬁcant decline. In contrast, as reported in columns 5–6of Table 6, panel B, sales revenue from small customersdoesnot exhibitasigniﬁcant decline in the post-punishment period, as the coeﬃcients on FRAUD*POST are negative but not signiﬁcant. These ﬁndings CHINA JOURNAL OF ACCOUNTING STUDIES 15 Table 6.. The product market consequences of ﬁnancial fraud: customer concentration. Panel A: Univariate analysis POST=0 POST=1 Diﬀerences N Mean Median N Mean Median Mean Median Fraud ﬁrms 882 0.334 0.275 876 0.316 0.257 –0.018* –0.018* (FRAUD=1) Control ﬁrms 882 0.274 0.224 876 0.290 0.233 0.015 0.009 (FRAUD=0) Diﬀerence-in-diﬀerences –0.033** –0.027*** Panel B: Multiple regression results CUST_CONCEN LN(SALES_BIG) LN(SALES_SMALL) (1) (2) (3) (4) (5) (6) FRAUD 0.064*** 0.057*** 0.422** 0.408** –0.332*** –0.251*** [5.076] [4.642] [2.262] [2.433] [–3.197] [–3.804] POST 0.034* 0.024 –0.613* 0.139 –0.730*** –0.184** [1.950] [1.419] [–1.838] [0.496] [–5.631] [–2.319] FRAUD*POST –0.035** –0.032** –0.801*** –0.578** –0.117 0.003 [–2.478] [–2.232] [–2.827] [–2.017] [–1.183] [0.037] SIZE –0.045*** 1.129*** 1.231*** [–5.891] [8.685] [24.266] LEV –0.069*** –1.347** 0.002 [–3.410] [–2.545] [0.018] ROA 0.033 2.447** 1.310*** [0.457] [2.001] [2.860] BTM –0.010 –0.008 0.025 [–1.129] [–0.059] [0.488] EP –0.022 –0.421 0.000 [–0.654] [–0.683] [0.001] LISTAGE 0.023 –0.322* –0.272*** [1.528] [–1.963] [–3.347] TOP5 0.107** –0.301 –0.394 [2.084] [–0.413] [–1.430] INDEP –0.011 –1.002 –0.394 [–0.122] [–0.581] [–0.800] CONSTANT 0.303** 1.233*** 19.417*** –4.005 20.864*** –5.642*** [2.545] [6.088] [22.974] [–1.561] [32.753] [–4.236] YEAR FE YES YES YES YES YES YES PAIR FE YES YES YES YES YES YES adj. R 0.309 0.342 0.181 0.257 0.470 0.727 N 3422 3422 3410 3410 3410 3410 Notes: In panel A, the T-test is used for testing the diﬀerences in mean and the Ranksum test is used for testing the diﬀerences in median. The dependent variables in panel B are CUST_CONCEN, LN(SALES_BIG) and LN(SALES_SMALL). LN(SALES_BIG) is the natural logarithm of sales revenue from the top ﬁve large customers, and LN(SALES_SMALL)is the natural logarithm of sales revenue from the remaining customers. See Table 2 for the deﬁnition of other variables. The ﬁgures in panel B are regression coeﬃcients and T values are in brackets. The T-value is based on standard errors adjusted for ﬁrm-level clustering. All continuous variables are winsorised at the 1st and 99th percentiles. *, **, *** indicate two-sided signiﬁcance levels of 10%, 5%, and 1%, respectively. demonstrate that the main reason for a declineinsales revenuefor fraud ﬁrmsisthe termination of contracting with large customers. Why are large customers willing to terminate the transactions with fraud ﬁrms and yet continue relationship-speciﬁc investments? We argue that this may be explained by the dominance of the buyer’s market in the current manufacturing industry. In the buyers’ dominated market, customers (buyers), especially large customers, have 16 XIN ET AL. advantages over the transaction negotiation. This may reduce large customers’ dependence on speciﬁc investments and may lower the costs of searching for new contracts. In contrast, it is more diﬃcult and costly for the suppliers to ﬁnd new customers when buyers dominate the market. Suppliers bear more costs if terminat- ing transactions with fraud ﬁrms on speciﬁc investments. Therefore, we expect that the reputation of a ﬁrm has less inﬂuence on supplier concentration following the regulatory punishment, but suppliers become stricter with contract terms and pay- ments due to risk consideration. Column 1 of Table 7 presents the regression results on supplier concentration measured by the ratio of purchase amount from the top ﬁve suppliers to the total purchase amount. It shows that the coeﬃcient on FRAUD*POST is positive but not signiﬁcant, indicating supplier concentration exhibits no signiﬁcant change following the punishment on fraud. Column 2 of Table 7 shows the regression results on the trade credit oﬀered by suppliers. As reported in column 2, the coeﬃcient on FRAUD*POST is signiﬁcantly negative, suggesting that suppliers provide less trade Table 7. The eﬀects of ﬁnancial fraud on suppliers. (1) (2) SUPPLIER CONCENTRATION TRADE CREDIT FRAUD 0.030** 0.003 [2.106] [1.078] POST 0.017 0.008** [1.039] [2.269] FRAUD*POST 0.008 –0.007* [0.443] [–1.750] SIZE –0.025*** 0.000 [–2.951] [0.240] LEV 0.002 –0.009 [0.074] [–1.520] ROA 0.098 0.035 [1.300] [1.328] BTM 0.020** –0.001 [2.006] [–0.195] EP –0.044 0.009 [–1.311] [0.698] LISTAGE –0.042** –0.003 [–2.408] [–0.962] TOP5 –0.050 0.031*** [–0.917] [2.627] INDEP 0.067 0.069*** [0.623] [2.655] CONSTANT 0.880*** –0.008 [4.845] [–0.214] YEAR FE YES YES PAIR FE YES YES adj. R2 0.276 0.027 N 3422 3422 Notes: The dependent variable in column (1) is SUPPLIER CONCENTRATION, which is equal to the ratio of purchase amount from the top ﬁve suppliers to the total purchase amount. The dependent variable in column (2) is the TRADE CREDIT provided by the suppliers, which is equal to the change in (notes payable + accounts payable – prepayments) divided by the total assets. Figures in the table are regression coeﬃcients and T values are in brackets. The T-value is based on standard errors adjusted for ﬁrm-level clustering. All continuous variables are winsorised at the 1st and 99th levels of 10%, 5%, and 1%, respectively. percentiles. *, **, *** indicate two-sided signiﬁcance CHINA JOURNAL OF ACCOUNTING STUDIES 17 credit to fraud ﬁrms after regulatory penalties. These ﬁndings are consistent with our hypotheses. 6. Severity of regulatory penalties and product market consequences If the regulatory penalties to the ﬁnancial fraud ﬁrms are more severe, it means that ﬁrms have engaged in more serious fraud. We expect that the reputational loss increases with the worsened reputation.Totestthisprediction, we create two indicator variables PUN_MILD and PUN_ SEVERITY according to the severity of punishment types. PUN_MILD is equal to one for the mild punishment - ordering correction and zero otherwise. PUN_SEVERITY is equal to one for the more severe punishment – public criticism, public condemnation, oﬃcial warning, or monetary ﬁnes – and zero otherwise. We re-run our regression model by replacing FRAUD in model (2) with PUN_MILD and PUN_ SEVERITY. Table 8 reports the regression results. In regression on sales revenue and customer concentration, the coeﬃcients on the interaction term of PUN_MILD and POST are negative but insigniﬁcant. Table 8. Regulatory penalties and product market performance. (1) (2) (3) LN(SALES) GROSS_MARGIN CUST_CONCEN PUN_MILD 0.112 0.026** 0.053*** [1.250] [2.373] [3.319] PUN_SEVERITY –0.324*** 0.006 0.063*** [–3.108] [0.437] [3.195] POST –0.236*** –0.004 0.025 [–2.857] [–0.359] [1.404] PUN_MILD*POST –0.083 –0.021** –0.025 [–1.187] [–2.008] [–1.618] PUN_SEVERITY*POST –0.201** –0.029** –0.041* [–2.029] [–2.018] [–1.906] SIZE –0.029*** –0.045*** [–4.505] [–5.854] LEV 0.380* –0.007 –0.069*** [1.679] [–0.374] [–3.413] ROA 2.695*** 0.033 [6.934] [0.465] BTM 0.555*** –0.025*** –0.010 [10.761] [–3.925] [–1.156] EP –0.485*** 0.098*** –0.022 [–2.843] [5.632] [–0.666] LISTAGE 0.262*** –0.010 0.023 [2.837] [–0.797] [1.523] TOP5 0.551* 0.139*** 0.108** [1.802] [3.550] [2.095] INDEP –1.877*** 0.143* –0.010 [–3.394] [1.664] [–0.108] CONSTANT 19.688*** 0.862*** 1.230*** [50.167] [6.303] [5.981] YEAR F.E. YES YES YES PAIR F.E. YES YES YES adj. R 0.600 0.381 0.342 N 3410 3410 3422 Notes: Figures in the table are regression coeﬃcients and T values are in brackets. The T-value is based on standard errors adjusted for ﬁrm-level clustering. All continuous variables are winsorised at the 1st and 99th percentiles. *, **, *** indicate two-sided signiﬁcance levels of 10%, 5%, and 1%, respectively. 18 XIN ET AL. However, the coeﬃcients on the interaction term of PUN_SEVERITY and POST are all signiﬁcantly negative for all three regressions. These ﬁndings are consistent with the conjecture that ﬁrms receiving more severe penalties suﬀer more reputa- tional loss in the product market. 7. Robustness tests 7.1. Alternative measures There is an alternative explanation that the decline in sales revenue after the regulatory penalties may be caused by accounting conservatism rather than the reduction of sales contracts. To exclude this possible explanation, we use a measure other than accounting numbers – the natural logarithm of cash sales (LN(SALES_CASH)) as a proxy for customer contracting and re-do our tests. As cash sales are wholly based on a cash basis, this proxy is less likely to be manipulated by accountants and can reﬂect the eﬀects on actual sales activities. As reported in columns 1–2of Table 9, panel A, the coeﬃcients on FRAUD*POST are both signiﬁcantly negative, which is consistent with the previous ﬁndings. 7.2. Control for debt ﬁnancing ability The reputation losses the fraud ﬁrms suﬀer in the product market might be triggered by the lack of ﬁnancing channels or high ﬁnancing costs for fraud ﬁrms. With ﬁnancial constraints, ﬁrms have diﬃculties sustaining normal production and operation, so decreasing product market performance such as sales revenue and gross proﬁt margin. The economic con- sequences of regulatory penalties are the secondary consequences of the penalties imposed by the capital market (and/or labour market), rather than being initiated by reputational penalties for ﬁnancial fraud ﬁrms imposed by the product market. We believe that capital, labour and product markets interact with each other. Previous literature argues that fraud aﬀects stock prices as fraud has an impact on product sales, which in turn aﬀects the stock prices (Karpoﬀ et al., 2008). Following this logic, we argue that fraud aﬀects ﬁrms’ ﬁnancing frictions in the capital market, which in turn aﬀects the product market perfor- mance. While it is hard to disentangle the eﬀects of product market participants’ incentives and the eﬀects of ﬁnancing friction, we attempt to do the following tests to exclude the alternative explanations. First, in the multiple regressions of Tables 5 and 6, we have already controlled for ﬁrms’ ﬁnancial constraints such as debt-to-assets ratio, accounting perfor- mance, book-to-market ratio, and earnings-to-price ratio, which have already largely absorbed the eﬀect of ﬁnancial frictions. In addition, as a robustness test, we control for the ratio of net debt ﬁnancing to total assets (FIN_CASH). The regression results are presented in Table 9, panel B. These results are consistent with the previous ﬁndings. Managers may defer the revenue recognition in the face of fraud risk. In untabulated ﬁndings, we ﬁnd that compared with the control ﬁrms, FIN_CASH of fraud ﬁrms experiences a signiﬁcant decline, indicating fraud may have an impact on ﬁrms’ debt ﬁnancing ability. CHINA JOURNAL OF ACCOUNTING STUDIES 19 Table 9. Robustness tests. (1) (2) Panel A: A proxy for sales revenue: cash sales (LN(SALES_CASH)) FRAUD –0.131 –0.073 [–1.502] [–1.523] POST –0.553*** –0.039 [–4.794] [–0.622] FRAUD*POST –0.253*** –0.115* [–2.907] [–1.777] CONTROLS NO YES YEAR FE YES YES PAIR FE YES YES 0.489 0.776 adj. R N 3410 3410 LN(SALES) GROSS_MARGIN CUST_CONCEN Panel B: Controlling for debt ﬁnancing ability FRAUD –0.095 0.017* 0.057*** [–1.339] [1.926] [4.641] POST –0.329*** –0.009 0.024 [–4.028] [–0.806] [1.422] FRAUD*POST –0.117* –0.024** –0.032** [–1.809] [–2.543] [–2.214] FIN_CASH 0.839*** 0.051* 0.007 [3.911] [1.746] [0.138] CONTROLS YES YES YES YEAR FE YES YES YES PAIR FE YES YES YES adj. R 0.594 0.381 0.342 N 3410 3410 3422 LN(SALES) GROSS_MARGIN CUST_CONCEN Panel C: Controlling for the eﬀect of management turnover FRAUD –0.091 0.017** 0.057*** [–1.286] [2.023] [4.537] POST –0.378*** –0.002 0.013 [–4.475] [–0.198] [0.771] FRAUD*POST –0.133** –0.023** –0.034** [–2.050] [–2.399] [–2.309] TURNOVER 0.046 –0.000 0.006 [0.430] [–0.027] [0.324] TURNOVER*POST 0.058 –0.009 0.014 [0.965] [–1.103] [1.086] CONTROLS YES YES YES YEAR FE YES YES YES PAIR FE YES YES YES adj. R 0.593 0.380 0.342 N 3410 3410 3422 LN(SALES) GROSS_MARGIN CUST_CONCEN Panel D: Dynamic Approach FRAUD –0.080 0.017* 0.056*** [–1.100] [1.924] [4.230] –1 BEFORE –0.182*** –0.021*** –0.000 [–4.516] [–2.805] [–0.035] –1 FRAUD* BEFORE –0.020 –0.001 0.002 [–0.463] [–0.142] [0.170] AFTER –0.251*** –0.014 0.010 [–3.993] [–1.475] [0.733] FRAUD* AFTER –0.114** –0.020** –0.019 [–2.198] [–2.192] [–1.313] AFTER –0.395*** –0.012 0.022 [–4.577] [–0.955] [1.295] FRAUD* AFTER –0.125* –0.028** –0.031** (Continued) 20 XIN ET AL. Table 9. (Continued). LN(SALES) GROSS_MARGIN CUST_CONCEN [–1.794] [–2.480] [–2.001] 2+ AFTER –0.549*** –0.025* 0.027 [–5.022] [–1.716] [1.266] 2+ FRAUD* AFTER –0.153** –0.025** –0.032* [–1.989] [–2.387] [–1.871] CONTROLS YES YES YES YEAR FE YES YES YES PAIR FE YES YES YES adj. R 0.604 0.389 0.349 N 3986 3986 4002 LN(SALES) GROSS_MARGIN CUST_CONCEN Panel E: Re-selecting the control sample FRAUD 0.057 0.025 0.008 0.015* 0.051*** 0.057*** [0.727] [0.359] [0.872] [1.785] [3.819] [4.259] POST 0.066 0.096 0.016* 0.017** 0.009 0.008 [1.154] [1.623] [1.892] [2.044] [0.632] [0.554] FRAUD*POST –0.198*** –0.113* –0.018* –0.016* –0.029** –0.030** [–2.746] [–1.803] [–1.842] [–1.736] [–2.083] [–2.119] CONTROLS NO YES NO YES NO YES YEAR FE YES YES YES YES YES YES PAIR FE YES YES YES YES YES YES adj. R 0.551 0.612 0.350 0.399 0.311 0.334 N 3408 3408 3408 3408 3408 3408 Notes: The dependent variable LN(SALES_CASH) in panel A is the natural logarithm of cash sales which is the item in cash ﬂow statement ‘cash from selling commodities or oﬀering labour’. In panel B, we control for ﬁrm’s debt ﬁnancing ability (FIN_CASH), which is equal to the proportion of diﬀerence in the item of ‘borrowing’ and the item of ‘cash paid for debt’ in cash ﬂow statement to the total assets. In panel C, we control for the eﬀect of chairman or CEO –1 turnover (TURNOVER, TURNOVER*POST) in the event window [0, 3] on product market. In panel D, BEFORE , AFTER°, 2+ AFTER1 and AFTER are indicator variables for the ﬁrst year before punishment (year t–1), the year of punishment (year t+0), the ﬁrst year after punishment (year t+1), the second (or third) year after punishment (year t+2, t+3), respectively. In panel E, we require the punishment year for fraud ﬁrms is the same as the matching year for the control ﬁrms. See Table 2 for the deﬁnition of variables. The ﬁgures in the table are regression coeﬃcients and T values are in brackets. The T-value is based on standard errors adjusted for ﬁrm-level clustering. All continuous variables are winsorised at the 1st and 99th percentiles. *, **, *** indicate two-sided signiﬁcance levels of 10%, 5%, and 1%, respectively. 7.3. The eﬀect of management turnover Agrawal and Cooper (2017) point out that there is a greater management turnover for fraud ﬁrms after suﬀering fraud. This will give more uncertainty to ﬁrms in operation and manage- ment, and aﬀect product market performance. In addition, if personal connections of manage- ment play an important role in maintaining the commercial transactions with large customers, then thedepartureofmanagerswillincreasethe likelihood of large customers terminating transactions with the ﬁrms. To exclude the eﬀect of management turnover, we add an indicator variable, TURNOVER, and its interaction with POST (TURNOVER*POST) in our regression model. TURNOVER is equal to one if the chairman or CEO leaves the ﬁrm during the event window [t+0, t+3], and his or her age is less than 60, and zero otherwise. The data show that the means of TURNOVER for fraud ﬁrms and control ﬁrms are 74.14% and 62.92%, respectively. The diﬀerence between these two samples is signiﬁcant at the 0.01 level (T-value=2.95), which indicates that the top management for fraud ﬁrmsismorelikelytoleave oﬃce after regulatory penalties. As reported in Table 9,panel C, thecoeﬃcients on FRAUD*POST are all signiﬁcantly Chairman or CEO must retire at the age of 60. It excludes retirement as a reason for departure. CHINA JOURNAL OF ACCOUNTING STUDIES 21 negative after controlling for the eﬀect of management turnover, indicating our ﬁndings cannot be caused by top management turnover. 7.4. Dynamic approach We employ a diﬀerence-in-diﬀerences approach to study the eﬀect of ﬁnancial fraud on product market. The assumption for adopting this method is that the treating sample and the control sample must satisfy the parallel trend assumption. In previous analyses, we demonstrate our sample meets the requirement of parallel trend by plotting the mean trend changes in sales revenue, gross proﬁt margin and customer concentration. In order to corroborate our results, we further perform a dynamic analysis. We set a total of four –1 0 2+ –1 indicator variables, BEFORE , AFTER , AFTER1and AFTER .Speciﬁcally, BEFORE is equal to one for the ﬁrst year before the punishment (year t–1) and zero otherwise, AFTER is equal to one for the year of punishment (year t+0) and zero otherwise, AFTER1 is equal to one for 2+ the ﬁrst year after punishment (year t+1) and zero otherwise, AFTER is equal to one for the second (or third) year after punishment (year t+2, t+3) and zero otherwise. We replace –1 the post-punishment indicator (POST) in model (2) with these four indicators BEFORE , 0 2+ AFTER , AFTER1and AFTER .If the ﬁndings in this paper are correct, we should observe the –1 coeﬃcient on FRAUD*BEFORE is insigniﬁcant, and the coeﬃcients on FRAUD*AFTER1and 2+ FRAUD*AFTER are signiﬁcantly negative. Table 9, panel D reports the corresponding regression results, and these results are consistent with our expectations. 7.5. Re-selecting the control sample In the selection of the control sample, we do not require the year of punishment for ﬁnancial fraud ﬁrms to be the same as the matching year of the control sample. As a robustness test, we adopt another approach to select the control ﬁrms. Speciﬁcally, based on the year of punishment for each ﬁnancial fraud ﬁrm, we identify a non-fraud ﬁrm in the same year with the closest propensity score as the control sample, then re-do the main tests. Table 9,Panel E presents the regression results. It shows that the ﬁndings remain the same. 8. Conclusions Using 294 cases of manufacturing listed companies subject to ﬁnancial fraud in China during 2004–2012, this paper is the ﬁrst study that has systematically examined the economic consequences of ﬁnancial fraud on the Chinese product market. We docu- ment that compared with the control ﬁrms, ﬁnancial fraud ﬁrms’ sales revenue and gross proﬁt margin exhibit a signiﬁcant decline following the punishment. Moreover, fraud ﬁrms are more likely to lose sales revenue from large customers, while the sales revenue from small customers does not change signiﬁcantly surrounding the punishment period. Overall, the ﬁnancial misconduct behaviour of listed companies in the capital market leads to substantial reputational losses in the product market. Since we include the year of punishment in the dynamic analysis, thus for the regression of fraud on sales revenue (LN(SALES)) and gross proﬁt margin (GROSS_MARGIN), the number of observations changes from 3410 in the previous tables to 3422. For the regression on customer concentration, the number of observations changes from 3422 in the previous tables to 4002. 22 XIN ET AL. Our ﬁndings contribute to the understanding of the product market in the Chinese context. If the contracting parties in the product market rely heavily on non-market mechanisms such as relationships, then the parties’ reputation is unlikely to have a substantial eﬀect on the trust mechanism of the relationship-based transactions. In contrast, if the contracting and transactions in the product market are mainly deter- mined by contracting parties’ market performance and reputation, that is, transac- tions are commenced through the fair, transparent and competitive market participation, then ﬁrms’ misconduct behaviour should have a serious impact on the maintenance of commitments in transactions. As shown in our research results, ﬁnancial fraud ﬁrms’ sales revenue from large customers falls almost by half after being penalised. This implies that, in the process of market-oriented reform in China, market forces play a more leading role in resource allocation decisions in the product market. Acknowledgements We appreciate the helpful comments and suggestions from reviewers and editors. Qingquan Xin acknowledges the ﬁnancial support from the National Natural Science Foundation of China (Grant No.71272087) and the Fundamental Research Funds for the Central Universities (Grant No. CDJSK100209). Funding This work was supported by the National Natural Science Foundation of China ; Fundamental Research Funds for the Central Universities [CDJSK100209]; References Agrawal, A., & Cooper, T. (2017). Corporate governance consequences of accounting scandals: Evidence from top management, CFO and auditor turnover. Quarterly Journal of Finance, 7(1), 1– Allen, F., Qian, J., & Qian, M. (2005). Law, ﬁnance, and economic growth in China. Journal of Financial Economics, 77(1), 57–116. Armour, J., Mayer, C., & Polo, A. (2017). Regulatory sanctions and reputational damage in ﬁnancial markets. Journal of Financial and Quantitative Analysis, 52(4), 1429–1448. Bowen, R.M., DuCharme, L., & Shores, D. (1995). Stakeholders’ implicit claims and accounting method choice. Journal of Accounting and Economics, 20(3), 255–295. Chen, X., Cheng, Q., & Lo, A.K. (2013). Accounting restatements and external ﬁnancing choices. Contemporary Accounting Research, 30(2), 750–779. Chen, G., Firth, M., Gao, D.N., & Rui, O.M. (2005). Is China’s securities regulatory agency a toothless tiger? Evidence from enforcement actions. Journal of Accounting and Public Policy, 24(6), 451– Chen, G., Firth, M., Gao, D.N., & Rui, O.M. (2006). Ownership structure, corporate governance, and fraud: Evidence from China. Journal of Corporate Finance, 12(3), 424–448. Cornell, B., & Shapiro, A.C. (1987). Corporate stakeholders and corporate ﬁnance. Financial Management, 16(1), 5–14. Fang, H.X., & Zhang, Y. (2016). Supplier/customer relationship transaction, earnings management and the auditor’s decision-making behavior. Accounting Research,(1),79–86 (in Chinese). CHINA JOURNAL OF ACCOUNTING STUDIES 23 Financial Accounting Standards Board (FASB), 1978. Statement of Financial Accounting Concepts No. 1. Objectives of ﬁnancial reporting by business enterprises. Norwalk, CT. Fich, E.M., & Shivdasani, A. (2007). Financial fraud, director reputation, and shareholder wealth. Journal of Financial Economics, 86(2), 306–336. Graham, J.R., Li, S., & Qiu, J. (2008). Corporate misreporting and bank loan contracting. Journal of Financial Economics, 89(1), 44–61. Hribar, P., & Jenkins, N.T. (2004). The eﬀect of accounting restatements on earnings revisions and the estimated cost of capital. Review of Accounting Studies, 9(2), 337–356. Hui, K.W., Klasa, S., & Yeung, P.E. (2012). Corporate suppliers and customers and accounting conservatism. Journal of Accounting and Economics, 53(1), 115–135. Hung, M., Wong, T.J., & Zhang, F. (2015). The value of political ties versus market credibility: Evidence from corporate scandals in China. Contemporary Accounting Research, 32(4), 1641– Johnson, W.C., Xie, W., & Yi, S. (2014). Corporate fraud and the value of reputations in the product market. Journal of Corporate Finance, 25(2), 16–39. Karpoﬀ, J.M., Lee, D.S., & Martin, G.S. (2008). The cost to ﬁrms of cooking the books. Journal of Financial & Quantitative Analysis, 43(3), 581–611. Klein, B., & Leﬄer, K.B. (1981). The role of market forces in assuring contractual performance. Journal of Political Economy, 89(4), 615–641. Kravet, T., & Shevlin, T. (2010). Accounting restatements and information risk. Review of Accounting Studies, 15(2), 264–294. Liebman, B.L., & Milhaupt, C.J. (2008). Reputational sanctions in china’s securities market. Columbia Law Review, 108(4), 929–983. Raman, K., & Shahrur, H. (2008). Relationship-speciﬁc investments and earnings management: evidence on corporate suppliers and customers. Accounting Review, 83(4), 1041–1081. Shapiro, C. (1983). Premiums for high quality products as returns to reputations. Quarterly Journal of Economics, 98(4), 659–679. Song, Y.L., Li, Z.W., & Ji, X.W. (2011). On the regulatory eﬀects of the Chinese securities regulation in view of the management forecasts frauds. Journal of Financial Research,(6), 136–149 (in Chinese). Wang, X.Y., & Liu, F. (2014). Customers bargaining power and suppliers’ accounting conservatism. China Accounting Review,(Z1), 389–404 (in Chinese). Williamson, O.E. (1985). The economic institutions of capitalism. London: Simon and Schuster. Xin, Q.Q., Huang, M.L., & Yi, H.R. (2013). The listed-company’s fraud in statement and the super- vision and penalty of independent directors: An analysis based on the perspective of the individual independent director. Management World,(5), 131–143, 175, 188 (in Chinese). Zhang, J., & Ma, G. (2005). Law, corporate governance and corporate fraud. Management World, (10), 113–123 (in Chinese).
China Journal of Accounting Studies
– Taylor & Francis
Published: Jan 2, 2018
Keywords: economic consequences; financial fraud; product market; regulatory penalties