کلاهبرداری های مالیاتی توسط شرکت ها و حسابرسی بهینه

Tax fraud is an issue of increasing importance in China. One particularly significant fraud involves excessive claims for the rebate of VAT on exported goods. This fraud has two interesting features. First, it requires the collusion of an intermediary to supply the false documentation that supports a rebate application. Second, the punishment schedule is convex—with capital punishment used in major fraud cases. These features ensure that the payoff function of a firm engaging in fraud is strictly concave in the level of fraud. This gives a well-defined optimization without the need to appeal to risk aversion. We show that the existence of fraud does not affect the real output decision of the firm nor the tax policy of the government. Audit resources can be used to detect firms engaged in fraud as well as the intermediaries who supply false documents. Under reasonable assumptions it is shown that resources should be focused on detecting firms and not intermediaries. Finally, if the government must take action on fraud a convex punishment scheme is shown to be optimal.

In many countries there is growing public concern over issues of tax evasion and fraud in economic activities. Evidence indicates that tax evasion is especially severe in developing and transition economies (Schneider & Enste, 2000). The figures reported by Jia (2002) show that marketization in China has been accompanied by the development of a hidden economy. The economic literature has focused largely upon the evasion of income taxation. In contrast, corporate tax fraud is an important issue in China and a particular concern of its tax system. Tax fraud is a form of economic crime and the source of a potentially serious loss of revenue to the Chinese government. The Chinese tax system relies on a form of voluntary compliance. Under Chinese law, all taxpayers are required to assist the tax authority by reporting honestly their income and profits and paying a level of tax based on their reported incomes over time. The system is actively enforced by the State Administration of Taxation (SAT) and the Chinese courts which can target investigations and impose substantial penalties for evasion and fraud. In the 2005 tax year the SAT raised revenue of 3086.6 billion Yuan, or 16.93% of GDP. On the other hand, Jia (2002) estimated the revenue lost through evasion and fraud at 444.6 billion Yuan, which was about 35.11% of the total tax revenue of 1266.6 billion Yuan in 2000. In this paper, we analyze tax fraud in claims for Value Added Tax (VAT) rebates on exported products. This fraud has some especially interesting features from the perspective of tax theory. Producers can claim a VAT rebate on the value of exports, so overstating the value of exports generates an unwarranted rebate. To commit such a fraud one requires a set of supporting documents: a false VAT invoice, a customs manifest, and a foreign exchange settlement memo. A producer committing fraud therefore requires the assistance of another party to supply the documentation. This interaction between the parties is the first interesting aspect of the model. For the VAT fraud the government has a very clear punishment scale with potentially extreme penalties, rising to capital punishment for a sufficient level of fraud. This explicit penalty scheme is the second interesting aspect since it is clearly convex, rather than being the linear fine used in most models of tax evasion. Many theoretical and empirical studies have addressed the economics of tax evasion. The standard approach was introduced by Allingham and Sandmo (1972) in the context of personal income taxation. They assumed that fines were based on the amount of income hidden by a taxpayer and addressed the determinants of individual income tax compliance. This basic model has been extended in a number of directions which are surveyed in Myles (1995) and Slemrod and Yitzhaki (2000). There is rather less literature on tax evasion by firms. Firms can evade taxation either by misreporting sales or profit, or by making false declarations about input use. It is possible that all these methods may be required simultaneously to disguise evasion if the information gathering process of the revenue service is sufficiently thorough. The tax evasion decision for competitive firms has been analyzed by Virmani (1989), Yamada (1990) and Cremer and Gahvari (1993). Tax evasion by monopolistic firms has been studied by Marrelli (1984) and evasion by oligopolistic firms by Marrelli and Martina (1988). Various forms of non-compliance with VAT have been reviewed in Keen and Smith (2006): the authors list under-reported sales, failure to register (“ghost” businesses), misclassification of commodities (as tax-exempt or subject to lower rates) as common for VAT and retail sales tax evasion, as well as forms that are distinct for VAT evasion, such as false claims for credit or refunds, credit claimed on purchases that are not creditable, and bogus traders, or firms set up solely to generate invoices used to recover VAT. Keen and Mintz (2004) show how an optimal VAT schedule can be constructed to balance the tax revenues and collection costs (including both administrative and compliance costs). Das-Gupta and Gang (2003) examine the impact of a particular VAT enforcement technique, namely, matching of purchase and sales invoices, on the enforcement and efficiency of VAT. They find that sufficiently intensive audit can induce truthful reporting but may at the same time distort the purchase–sale and the input–output decisions of the firm. Fedeli and Forte (1999) model a VAT evasion decision as an outcome of bargaining between the firms involved in the VAT chain. In particular, their model predicts that for recovering the lost revenues an increase in the probability of audit is preferable to an increase in the penalty for evasion. Aizenman and Jinjarak (2008) investigate how different political and structural economic factors affect the collection efficiency of VAT in a cross-sectional sample of countries, but notably not including China. In these papers a traditional, consumption-based form of VAT is assumed. The unusual feature of VAT in China is that it is production-based and capital goods are included in the VAT base (Ahmad, Singh, & Lockwood, 2004). Furthermore, China employs a separate export rebate system, with the rebate rates independent of rates paid on creditable inputs. Whalley and Wang (2007) analyze the welfare implications of these two features of the VAT system in China using a monetary trade model and argue that the switch to the traditional form of VAT in China will not be welfare-improving. They emphasize the importance of export rebates in their analysis, but do not address the issue of evasion. A similar study is undertaken in Lin (2008). Fisman and Wei (2001) present an empirical case study of tax evasion in China, with the focus on China’s imports from Hong Kong and the evasion of tariffs plus VAT. They show that more revenue is lost for products subject to higher tax rates; furthermore, they find strong evidence of evasion by mislabelling, or misclassification of higher tax products as lower tax products. Our paper contributes to this literature by presenting a theoretical analysis of fraud involving a VAT rebate for export sales. The analysis of tax evasion by firms faces a difficulty when combining profit maximization with a linear punishment schedule: expected profit is linear in the amount of evasion, so the decision to evade is all or nothing. Marrelli (1984) and Marrelli and Martina (1988) resolve this problem by assuming the owner of the firm is risk averse. A similar approach is undertaken in de Mello (2009). This is acceptable in the context of an owner–manager but does not fit convincingly with wider ownership of the firm’s equity. Stöwhase and Traxler (2005) assume instead that there is a cost of evading which is convex in the amount of evasion but do not identify the source of this cost. In contrast, the model we study is necessarily non-linear for reasons entirely explicable for the form of tax fraud we analyze. First, the announced fraud penalty structure is explicitly convex. We detail this in a brief summary of the Chinese VAT system in Section 2. Second, the act of fraud we consider involves the participation of an additional party who must supply false documentation to support the fraud. Section 3 shows how the cost of securing false documentation provides an explanation of the evasion cost used by Stöwhase and Traxler (2005). Section 4 demonstrates that this structure of costs and punishments implies the value of fraud undertaken by the firm is independent of the tax rate. This produces a separation of the government choice problem studied in Section 5 into two distinct components. The first component determines the optimal tax rate, and this rate is unaffected by the existence of fraud. Fraud does not, therefore, imply that the tax rate is distorted from its efficient value. The auditing and punishment policy is a consequence of the second component. If the marginal cost of auditing firms and document suppliers is the same then resources should be focused upon auditing firms, not document suppliers. It is shown in Section 6 that the same conclusions hold under an alternative specification of the punishment structure. Section 7 demonstrates that if there is a limit to the level of punishment that can be exacted for a relatively small amount of fraud then convexity of punishment emerges endogenously as the optimal structure of punishments. Section 8 provides the conclusions and some technical details are presented in Appendix A.

The paper has provided an analysis of VAT fraud in the context of the Chinese rebate for exported goods. The interesting features of this situation are the convex punishment schedules which are defined in the relevant tax law and the fact that an additional party has to be involved in the supply of the documentation required to support the fraud. The analysis shows that there is a well-defined choice problem even for a firm which maximizes profit. This is due to the convexity of punishments in the level of fraud. It is therefore unnecessary to impose risk aversion upon the preferences of the firm. The decision process of the firm separates into two parts. The real decisions of the firm – on input use and the level of output – are unaffected by the possibility of engaging in fraud. Two options for the punishment schedules were considered: based on the extent of the fraud and based on the value of fraud. In the first case the extent of fraud increases as the tax rate increases. In the second case the value of the fraud remains constant as the tax rate is changed. The interesting finding is that in both cases the optimal tax rate is unaffected by the existence of fraud. Hence, there is no incentive for the government to reduce the tax rate to reduce fraud or of raising it to cover for lost revenue. Fraud therefore does not imply a deviation of the tax rate from its efficient value. The choice between focusing audit policy upon the firm or upon the supplier of false documentation depends primarily upon the cost of monitoring. If the marginal costs of monitoring the firm and the document supplier are equal, the policy is most likely to involve audits focused upon the firm. The analysis showed that when the punishment function is concave the firm has an incentive to commit the maximum possible fraud unless it is monitored with certainty. Therefore, a government with limited resources to monitor compliance will employ a convex punishment in order to reduce fraud. This result provides support for the structure of Chinese criminal law and shows that auditing and convex punishments can be determined endogenously in this model of crime. We must note that the convex punishment structure for tax fraud in China should not, perhaps, be viewed as a unique phenomenon. Shikalova (2004) analyzed payroll tax evasion by firms in Russia, under a slightly different set of assumptions. Model calibration to Russian data showed that the shape of the punishment function for the payroll tax evasion is convex in the amount of labor income in “black cash”.