Government decentralisation and corporate fraud: Evidence from listed state-owned enterprises in China

Hang Liu; Xiaorong Li

doi:10.1080/21697213.2015.1100090

Government decentralisation and corporate fraud: Evidence from listed state-owned enterprises in China

Liu, Hang; Li, Xiaorong 2015-10-02 00:00:00 China Journal of Accounting Studies, 2015 Vol. 3, No. 4, 320–347, http://dx.doi.org/10.1080/21697213.2015.1100090 Government decentralisation and corporate fraud: Evidence from listed state-owned enterprises in China a,b c Hang Liu * and Xiaorong Li a b School of Accountancy, Dongbei University of Finance and Economics, Dalian, China; China Internal Control Research Center, Dalian, China; School of Public Finance / China’s Public Finance Development Synergetic Innovation Center, Central University of Finance and Economics, Beijing, China This paper examines the economic consequences of government decentralisation from the perspective of corporate fraud. Theoretically, government decentralisation reduces the political costs of state intervention and hence decreases the probability of state- owned enterprises (SOEs) to engage in fraud. It also aggravates the agency costs (the costs of managerial self-dealing), thereby increasing the probability of SOEs to com- mit fraud. Using pyramidal layers as a proxy of government decentralisation for SOEs, empirical results show that government decentralisation signiﬁcantly lowers the probability of SOEs to commit fraud. Further categorisation of types of fraud shows that government decentralisation primarily deters disclosure-related fraud and market transaction-related fraud. Finally, the effect of decentralisation on corporate fraud is more pronounced for SOEs in which government intervention is more likely. Keywords: agency costs; fraud; government decentralisation; political costs; pyramidal layers 1. Introduction Since the reform and opening up of China, multiple reforms have been conducted by the Chinese government upon SOEs to improve their operational efﬁciency. In retro- spect, although speciﬁc measures vary, the key has always been decentralisation. Since the 1990s, the Chinese government has proposed a reform plan for SOEs to construct the modern enterprise system. During this reform period, a large number of state-owned assets have been decoupled from government agencies to form new companies. Although the ultimate controllers of these new companies are within the sphere of gov- ernment, actual operational power has been gradually decentralised to specialised state- owned asset management companies or to the SOEs themselves through pyramidal ownership structures. Government intervention in ﬁrms is greatly decreased by this measure (Fan, Wong, & Zhang, 2013). This paper tries to explore the effect of govern- ment decentralisation, in the form of pyramid structures, upon the probability of SOEs to commit fraud, beginning in the 1990s. The economic consequences of the above-mentioned reform are considered controversial in the current literature. Qian (1996) argues, by theoretical analysis, that government decentralisation reduces the political costs of government intervention but *Corresponding author. Email: liuhang@dufe.edu.cn Paper accepted by Donghua Chen. © 2015 Accounting Society of China China Journal of Accounting Studies 321 increases agency costs (the costs of managerial self-dealing). This argument is empirically conﬁrmed by Fan et al. (2013). Adopting the proxy of pyramidal layers to measure the degree of decentralisation, the latter ﬁnd that political and agency costs jointly determine the degree of decentralisation, and government decentralisation improves ﬁrm output and employment efﬁciency in SOEs, demonstrating that the reduction in political costs is greater than the increase in agency costs. The same ﬁnd- ing can be observed in Wang and Xiao (2009) from a ﬁrm value point of view, in Cheng, Xia, and Yu (2008) from an investment efﬁciency viewpoint and in Liu and Li (2012) from a corporate tax burden perspective. Nonetheless, Zhong, Ran, and Wen (2010) discover that opportunistic behaviour of managers bred by government decen- tralisation can invalidate the investment decisions of SOEs. Therefore, agency costs cannot be neglected. Based on the data of the Chinese A-share listed SOEs in 2004–2010 and follow- ing the literature (Cheng et al., 2008; Fan et al., 2013; Liu & Li, 2012; Wang & Xiao, 2009; Zhong et al., 2010), this paper adopts the number of pyramidal layers between the ultimate controller and the listed ﬁrm as a proxy of the degree of government decentralisation over SOEs. The paper examines empirically this proxy’s effect on corporate fraud. The results show that pyramidal layers are signiﬁcantly negatively correlated with the probability of corporate fraud, demonstrating that gov- ernment decentralisation signiﬁcantly lowers the probability of SOEs to commit fraud. A detailed categorisation of fraud types, as provided in the fraud database of CSMAR, indicates that fraud reduction occurs via two mechanisms, information disclo- sure and stock market transactions; fraud reduction is not signiﬁcant in capital-related fraud or other types of fraud. A further classiﬁcation by severity of fraud leads to the conclusion that there is no signiﬁcant difference among fraud activities with various levels of severity affected by government decentralisation. Finally, the effect of govern- ment decentralisation upon corporate fraud exists only in SOEs that are more likely to be inﬂuenced by the government because a more severe government intervention is associated with a greater reduction in political costs caused by decentralisation. Thus, the conclusion that government decentralisation decreases the probability of corporate fraud is strengthened. The implications of this paper are the following: (1) Theoretically, government decentralisation reduces the political costs of SOEs but increases the agency costs (Qian, 1996). Both effects are empirically sup- ported (Cheng et al., 2008; Fan et al., 2013; Liu & Li, 2012; Wang & Xiao, 2009; Zhong et al., 2010). In this paper, we ﬁnd that the pyramidal structure of SOEs formed by government decentralisation signiﬁcantly reduces the probabil- ity of SOEs to commit fraud, demonstrating that from the perspective of corpo- rate fraud, the reduction in political costs resulting from decentralisation is signiﬁcant. This result expands the existing literature and supports the decen- tralisation reform of government over SOEs from the viewpoint of accounting information quality. (2) Where government intervenes heavily in SOEs, investors need to exercise more caution because those SOEs tend to be more likely to commit fraud and cause enormous loss to investors (Chen, Firth, Gao, & Rui, 2005; Dechow, Sloan, & Sweeney, 1996; Hung, Wong, & Zhang, in press). 322 Liu and Li The remainder of this paper is structured as follows: Section 2 presents an institutional background and literature review. Section 3 comprises theoretical analysis and hypothesis. Section 4 introduces the research design and lists descriptive statistics. Section 5 discusses empirical results. Section 6 concludes. 2. Institutional background and literature review 2.1. Institutional background: Government decentralisation and pyramidal structures of SOEs Prior to the reform and opening up in 1978, all SOEs in China were directly under the control of either the central or a local government. Managers in SOEs have a very lim- ited autonomy in such operational activities as employment, production and distribu- tion, with all decisions ﬂowing from government agencies. SOEs were actually only attachments to government agencies and were extremely inefﬁcient because of low incentive levels for managers and strong political objectives (Fan et al., 2013). After the reform and opening up, to enhance the vitality of SOEs, SOEs were empowered to manage independently to be responsible for their own proﬁts or losses, and to reserve a certain proportion of proﬁt within ﬁrms. These measures incentivise managers in the short run and activate state-owned assets (Groves et al., 1994). How- ever, a lack of ﬁrm operational information and professional operations capacity restrains government from effectively controlling managers’ opportunistic behaviours, such as concealing proﬁts or transferring assets to satisfy private interests (Qian, 1996). Moreover, during this period, state-owned assets continued to be directed in the former Soviet mode of a central planning system and were directly under the control of the government, resulting in severe interventions in ﬁrms by the government (Qian, 1996). To alter this situation, the operation of state-owned assets was transformed, from the 1990s, from the planning system to the market system. During this transformation state-owned assets were decoupled from government agencies (Fan et al., 2013). Con- cerning listed SOEs, the means of government decentralisation are largely the follow- ing: ﬁrst, the government injects capital into newly established listed SOEs and holds the shares of these companies through a state-asset management agency (e.g. State- owned Assets Supervision and Administration Commission of the State Council, SASAC). Alternatively, the government can indirectly control the listed SOEs through pyramidal structures. In this case, there are often intermediate companies between gov- ernment agencies (i.e. the ultimate controllers) and the listed SOEs. The intermediate companies are usually unlisted parent SOEs (Fan et al., 2013). In addition, the number of intermediate companies varies in different pyramidal structures. Appendix 1 lists a typical pyramidal structure of SOEs, in which SASAC, a government agency, 100% controls the China South Industry Group Corporation, which in turn 100% controls the China Changan Automobile Group, which controls the Chongqing Changan Automo- bile Corporation (stock code: 000625) with a shareholding ratio of 47.51%. As tabled, this pyramidal ownership structure distances ﬁrms and government, relieving ﬁrms of direct government control. A greater distance from the top decreases the likelihood of government intervention in the ﬁrms (Aghion & Tirole, 1997; Fan et al., 2013). Thus, the pyramidal structure of SOEs stems from government decentralisation reform and lessens direct government intervention imposed upon SOEs. China Journal of Accounting Studies 323 2.2. Literature review Prior literature explores the economic consequences of government decentralisation theoretically and empirically. Qian (1996) argues, by theoretical analysis, that decentralisation reform of SOEs by the Chinese government confronts two major issues: the political costs of government intervention and agency costs. Speciﬁcally, on the one hand, government decentralisa- tion reduces government intervention in ﬁrms, thereby reducing ﬁrms’ political costs and improving their efﬁciency; on the other hand, government decentralisation empow- ers managers with more autonomy, thereby increasing agency costs and harming ﬁrms’ efﬁciency as the higher information costs resulting from decentralisation make supervis- ing more difﬁcult. Therefore, Qian believes that an effective SOE reform strategy lies in the establishment of an effective corporate governance system as part of the govern- ment decentralisation reform to reduce ﬁrms’ political costs and inhibit agency costs. Qian’s theoretical analysis is conﬁrmed empirically by Fan et al. (2013), who study the factors and economic consequences of government decentralisation using the exam- ple of government decentralisation in the form of pyramidal structures in China since the 1990s. Adopting pyramidal layers to depict the degree of decentralisation, the authors ascertain that political and agency costs jointly determine the degree of decen- tralisation and that government decentralisation improves the professionalism of SOE managers, production efﬁciency, and the operational performance of employees, demon- strating that the reduction in political costs is greater than the increase in agency costs on average. Later studies demonstrate the positive economic consequences from government decentralisation. Cheng et al. (2008) ﬁnd that it reduces excessive investment in SOEs. Wang and Xiao (2009) observe that a higher degree of government decentralisation over SOEs is associated with a greater valuation of ﬁrms. Liu and Li (2012) show that SOEs with a higher degree of government decentralisation tend to have a lower policy burden (i.e. a lower corporate tax burden). With regard to agency costs, Zhong et al. (2010) demonstrate that government decen- tralisation at a low level can reduces political costs and constrain ﬁrms’ over-investment of free cash ﬂow. However, a further increase in government decentralisation will induce a signiﬁcant rise of agency costs, resulting in an inverse U-shaped relationship between the degree of government decentralisation and ﬁrms’ over-investment of free cash ﬂow. Quan, Wu, and Wen (2010) show that government decentralisation is signiﬁcantly positively correlated with excess cash compensation to SOE managers by utilising the pyramidal layers as one ingredient of SOE managers’ power. All the evidence points to the profound effect upon ﬁrms’ economic activities of agency costs derived from government decentralisation. Based on the above literature, this paper examines the effect of government decen- tralisation in the form of pyramidal structures upon the probability of SOEs to commit fraud. 3. Theoretical analysis and research hypothesis Aghion and Tirole (1997) argue that a trustworthy decentralisation can be achieved by a pyramidal structure without ownership transfer. Compared with a government promise not to interfere, the pyramidal structure is far more reliable because its long intermedi- ate chains and sophistication require more costs for government to interfere with the 324 Liu and Li daily operation of ﬁrms. Moreover, there will be distortion and selective deviation of information transmission via intermediate chains within the structure, making it more difﬁcult to interfere. Therefore, a pyramidal structure can effectively reduce administra- tive intervention of government in SOEs and thereby decrease political costs (Fan et al., 2013; Qian, 1996). However, it is not costless to form the structure. On the one hand, although opera- tional management of SOEs is decentralised, government remains the owner of SOEs. The agent conﬂict between principal (government) and agent (managers) motivates managers to pursue opportunistic behaviour; on the other hand, within the structure, the challenge to gather information and the complexity of the organisation deter govern- ment from effective supervision. Thus, agency costs are increased by the decentralisa- tion (Qian, 1996). Therefore, the pyramidal structure brought by government decentralisation reform not only reduces government intervention in SOEs but also increases agency costs. This paper conducts its theoretical analysis from these two perspectives. 3.1. Analysis from the perspective of government intervention Low ﬁrm-speciﬁc ﬁnancial transparency has always been attributed largely to govern- ment intervention (Bushman, Piotroski, & Smith, 2004; Bushman & Piotroski, 2006). International evidence demonstrates that there is a relatively low level of ﬁnancial transparency for ﬁrms within a nation in which the political economy is characterised by high state ownership of ﬁrms, high state ownership of banks, and a high risk of state expropriation (Bushman et al., 2004). There is also a low level of accounting con- servatism for ﬁrms in a nation with greater risks of being appropriated by government or a larger share of SOEs (Bushman & Piotroski, 2006). For China, there are at least the following three procedures through which government intervention will affect the information quality of listed SOEs. During listing, government intervention damages the information quality of ﬁrms and boosts their fraud probability (Aharony, Lee, & Wong, 2000). The quantity of listed ﬁrms is stringently regulated by a security supervision agency; therefore, for a ﬁrm to be listed requires a certain performance threshold and for it to pass the censor- ship of the supervision agency, consequently causing the listing qualiﬁcation to become a scarce resource (Huang, Cheng, Li, & Wei, 2014). Additionally, an SOE listing ignites tremendous beneﬁts, including post-listing ﬁnancing for ﬁrms, enhanced reputa- tion with the government, and accumulation of ofﬁcials’ political capital (Hung, Wong, & Zhang, 2012). In striving for the qualiﬁcation, SOEs will be ﬁnancially packaged by government (Aharony et al., 2000) and receive proﬁts through related-party transactions by government (Aharony, Wang, & Yuan, 2010), leading to a severely distorted quality of information about SOEs. Information quality will be further distorted during the daily operation of SOEs as follows. (1) In the scenario of government intervention, the operational goal of SOEs is not to maximise the interests of shareholders but rather to shoulder partial policy bur- dens such as more taxes (Wu, 2009), more employment (Zeng & Chen, 2006), and more over-investment (Bai & Lian, 2013), all to the detriment of SOEs’ performance. To cover the bad news of deteriorating performance, preserve governmental reputation and obtain reﬁnancing qualiﬁcation, another round of ﬁnancial packaging will be pursued (Piotroski & Wong, 2012; Piotroski, Wong, & Zhang, 2015), biasing the information quality of SOEs and culminating in corporate fraud (Gao & Song, 2007; China Journal of Accounting Studies 325 Zhu & Lee, 2008). (2) A lack of many external monitoring mechanisms for SOEs also contributes to low information quality, speciﬁcally embodied in the following points: ﬁrst, as stated previously, policy burdens drag down SOEs’ performance, with the gov- ernment frequently paying the loss (Piotroski & Wong, 2012). The resulting soft budget constraint lessens the restraining effect of liability upon SOEs and lowers creditors’ demand for high-quality ﬁnancial information (Chen, Chen, Lobo, & Wang, 2010; Rao & Jiang, 2011). Second, government intervention deprives SOEs of the demand for high-quality external auditing; consequently, SOEs often hire local small-scale account- ing ﬁrms of low quality (Wang, Wong, & Xia, 2008), reducing the role of auditors in restraining corporate fraud (Lennox & Pittman, 2010) and practically facilitating SOEs to commit fraud. (3) Dai, Pang, and Liu (2011) discover that government intervention also debilitates the role of media supervision over corporate fraud. All the evidence supports the conclusion that the information quality of SOEs is damaged and that the role of supervision mechanisms in constraining corporate fraud is limited by govern- ment intervention. After the occurrence or exposure of SOEs’ fraud, the very existence of government intervention weakens the degree of punishment for SOEs, decreasing SOEs’ cost of violating rules and practically contributing to SOEs’ fraud. Speciﬁcally, Chen, Jiang, Liang, and Wang (2011) prove that there is an issue of selective law enforcement in Chinese supervision agencies. Because SOEs shoulder partial political costs, SOEs are penalised more leniently after fraud exposure. Moreover, Chen, Li, Rui, and Xia (2009) and Firth, Rui, and Wu (2011) show that the umbrella of government background lav- ishes privileges upon SOEs from supervision or judiciary agencies. Speciﬁcally, when facing possible civil litigation, investors believe that trial results are more favourable to SOEs; thus, stock prices of SOEs decrease less (Chen et al., 2009). SOEs can also have better results in appeals (Firth et al., 2011). Succinctly, extant government intervention increases the fraud probability of SOEs. Therefore, with the deepening of decentralisation reform, the degree of government intervention in SOEs will be smaller, decreasing the probability of fraud. On this point, there is a signiﬁcantly negative correlation between the degree of government decentral- isation and the probability of corporate fraud. We term this effect the ‘government intervention hypothesis’. 3.2. Analysis from the perspective of agency costs Managers’ self-interested behaviour has always been an important determinant of cor- porate fraud (Dechow, Ge, & Schrand, 2010). Prior literature proves that from multiple perspectives. One perspective is to examine the inﬂuence of managers’ incentives upon corporate fraud. Although managers’ compensation incentives are considered as an important mechanism to relieve the agency problem (Jensen & Meckling, 1976), they also embody the agent issue and induce managers’ self-interested behaviour (Bebchuk, Fried, & Walker, 2002). Burns and Kedia (2006), Efendi, Srivastava, and Swanson (2007), Harris and Bromiley (2007), and Hong, Hu, and Guo (2012) all discover that managers’ compensation incentives signiﬁcantly increase the probability of fraud in ﬁrms. Bergstresser and Philippon (2006) observe that equity incentives stimulate earn- ings management of ﬁrms because a high level of incentives enables managers to amass private beneﬁts by such methods as insider trading, publishing false information and adjusting earnings to manipulate stock prices. Erickson, Hanlon, and Maydew 326 Liu and Li (2006) and Armstrong, Jagolinzer, and Larcker (2010) do not ﬁnd relevant evidence, but Armstrong, Larcker, Ormazabal, and Taylor (2013) conﬁrm that a stronger correla- tion between managers’ wealth and company risk is associated with a higher probabil- ity to commit fraud. The conclusion is consistent with managers’ self-interested behaviour resulting in a higher probability of corporate fraud. The other angle is to consider the relationship between corporate governance and the probability of corporate fraud. The potential theoretical hypothesis is that a sound corporate governance system can effectively inhibit managers’ agency problem, thereby reducing corporate fraud. For instance, Beasley (1996) and Chen, Firth, Gao, and Rui (2006) ﬁnd that a higher proportion of external directors, particularly with a ﬁnance background (Agrawal & Chadha, 2005), can reduce ﬁrms’ probability to commit fraud. Dechow et al. (1996) show that when the CEO is simultaneously chairman of the board, there is a higher probability of corporate fraud. Lennox and Pittman (2010) observe that the employment of the international Big Five public accounting ﬁrms, rep- resenting high audit quality, signiﬁcantly reduced corporate fraud. Furthermore, Liu, Luo, Zhang, and Chen (2013) conclude that although sound internal control can reduce fraud, executive power centralisation weakens the supervisory role of internal control over managers’ irregularities. To summarise, the evidence consistently indicates that managers’ opportunistic behaviours can increase the probability of corporate fraud. After decentralisation reform, the government is unable to conduct face-to-face supervision of managers and the complexity of a pyramidal structure weakens the governmental supervision effect (Qian, 1996), all possibly contributing to the increasing agency costs of SOEs and thus breeding a higher fraud probability. From this standpoint, there is a signiﬁcantly posi- tive correlation between the degree of government decentralisation and corporate fraud probability. We term this effect the ‘agency costs hypothesis’. Theoretically, there are two distinct hypotheses concerning the inﬂuence of govern- ment decentralisation upon corporate fraud. Which is the dominant one? This paper tries to answer this question empirically. 4. Research design and descriptive statistics 4.1. Empirical model and variables deﬁnition To investigate the inﬂuence of government decentralisation on corporate fraud, the fol- lowing empirical model is deﬁned: ProbðÞ FRAUD ¼ 1¼ a þ a LAYER þ a LNMCAP þ a ROA þ a LEV it 0 1 i;t 2 i;t 3 i;t 4 i;t þa MB þ a BIG4 þ a MANHOLD þ a INDEP þ 5 i;t 6 i;t 7 i;t 8 it a DUALITY þ a LNBSIZE þ a HBSHARE þ 9 i;t 10 i;t 11 i;t (1) a MEETING þ a LOSS þ a ST þ a GOV SHARE 12 i;t 13 i;t 14 i;t 15 i;t þa GOV INTERV þ a TOP1 þ a RPT 16 i;t 17 i;t 18 i;t P P þa ORECTA þ Industry þ Year þ e 19 i;t i;t In Model (1), FRAUD is the dependent variable, indicating whether ﬁrm i is engaged in fraud in year t. To depict this variable, the method commonly found in investigation of Chinese companies is adopted, i.e. downloading all the listed companies’ fraud information from the fraud database in China Stock Market & Accounting Research (CSMAR). We can observe in which year the listed ﬁrms engaged in fraud under the China Journal of Accounting Studies 327 title of ‘the year of violation’ in the database. When a listed ﬁrm commits fraud in the current year, FRAUD equals 1, and otherwise 0. The ﬁrst independent variable of Model (1) is LAYER, describing the degree of gov- ernment decentralisation over SOEs. Following Fan et al. (2013), Cheng et al. (2008), Liu and Li (2012), Wang and Xiao (2009), and Zhong et al. (2010), the number of pyramidal layers from the ultimate controller to the listed SOE is utilised to measure the degree of governmental decentralisation. Speciﬁcally, when the ultimate controller directly controls the listed company, the pyramidal layer is 1. If there is an intermediate controller between them, the pyramidal layer is 2, and so on. To control other factors affecting corporate fraud, following Chen et al. (2006) and Chen, Chen, Li, and Ni (2013), these control variables are included in the model: LNMCAP is the natural logarithm of a ﬁrm’s market capitalisation; the higher political cost of a large-size ﬁrm lowers the probability of fraud. ROA, denoting the ﬁrm’s prof- itability, is equal to net income divided by total assets. Firms with better performance have weaker motivation to manipulate earnings and make misstatements; thus, they are less likely to commit fraud. LEV, ﬁrm leverage, is equal to the total liabilities divided by the total assets. Highly leveraged ﬁrms with a higher bankruptcy cost are more likely to be involved in fraud. However, they will also be strictly supervised by credi- tors, thereby lowering the probability of fraud. MB, market to book ratio, is equal to the total market value of equity divided by the book value of equity. To obtain continu- ous ﬁnancing, the high-growth ﬁrm may manipulate earnings through fraud activities. BIG4 indicates whether the ﬁrm has hired an international big four auditing ﬁrm. If yes, BIG4 equals 1, otherwise 0. Lennox and Pittman (2010) ﬁnd that ‘big ﬁve’ audit- ing can effectively reduce corporate fraud probability; therefore, we control for this variable. Next, we control for a series of variables related to corporate governance. MANHOLD represents the percentage of management shareholding; INDEP is the pro- portion of independent directors. DUALITY indicates whether the chairman of the board and CEO are the same person (1 = ‘yes’,0 = ‘no’). LNBSIZE, the board size, is equal to the natural logarithm of the number of directors. HBSHARE indicates whether the ﬁrm issues H-shares or B-shares simultaneously (1 = ‘yes’,0 = ‘no’). MEETING is equal to the natural logarithm of the number of board meetings. We also control for two variables closely related to the ﬁrm’s ﬁnancial reporting motivation. In general, a stronger motivation of the ﬁrm to manipulate ﬁnancial reports is associated with a higher probability of fraud. The two variables are LOSS, denoting whether there is a loss for the ﬁrm in the previous period (1 = ‘yes’,0 = ‘no’), and ST, indicating whether the ﬁrm is specially treated (1 = ‘yes’,0 = ‘no’). Companies with a previous loss and ST have a greater motivation to commit fraud. To control other possible mechanisms of government intervention besides pyramidal layers inﬂuencing corporate fraud, two control variables are incorporated: GOV_SHARE, standing for the proportion of state- owned shares of the company, and GOV_INTERV, denoting the degree of regional gov- ernment intervention, measured by the index of ‘reducing government intervention over the enterprise’ from Fan, Wang, and Zhu (2011). In addition, this paper adopts pyrami- dal layers as a proxy of the degree of government decentralisation over SOEs. How- ever, the literature indicates that the pyramidal ownership structure itself may lead to a series of agency problems, reducing the degree of ﬁrms’ information asymmetry, which may have a signiﬁcant effect on corporate fraud (Fan & Wong, 2002; Leuz, Nanda, & Wysocki, 2003). Thus, we include two variables to control for the agency problem related to the pyramidal structure: RPT, representing the degree of a ﬁrm’s related-party transactions, is equal to the total amount of the ﬁrm’s related-party transactions between 328 Liu and Li the listed ﬁrm and the parent ﬁrm divided by the total assets of the listed ﬁrm. ORECTA denotes the degree of tunnelling by the controlling shareholder and is equal to inter-corporate loans by the controlling shareholder in the other receivables divided by the total assets (Jiang, Lee, & Yue, 2010). We also include the percentage of shares held by the largest shareholder (TOP1) to control for the possible effect of ownership concentration on corporate fraud. Finally, we control for year (YEAR) and industry (INDUSTRY) ﬁxed effects. 4.2. Sample and data We select all A-share listed SOEs from 2004 to 2010 as the initial sample. The sample interval begins in 2004 because the CSMAR database has collected the ownership structure ﬁgures of listed ﬁrms in China from their annual reports since 2004, through which the core variable LAYER is manually gathered. The interval ends in 2010 because the CSMAR database identiﬁes corporate fraud retrospectively as it emerges. More recent years may be less complete in data. This research uses only SOEs as the sample ﬁrms because this paper studies the economic consequences of government decentralisation, which targets only SOEs. Excluding the observations of the ﬁnancial industry and with missing data from the initial sample, our ﬁnal sample includes 6,111 observations representing 1,147 unique ﬁrms. Concerning data sources, data on pyramidal layers come from the authors’ manual collection, the data of inter-corporate loans by the controlling shareholder in the other receivables from the WIND database, and other data from the CSMAR database. 4.3. Descriptive statistics Descriptive statistics of variables are presented in Table 1. To eliminate the effect of outliers, we winsorise all continuous variables at the 1% and 99% levels. Panel A of Table 1 shows that the average of FRAUD is 0.097, meaning that 9.7% of the sample companies are engaged in fraud. The proportion is higher than 7.4% in the descriptive statistics of Chen et al. (2013). This is the result of corporate fraud being detected later; thus, in conjunction with the continuous renewal of the fraud database, fraud not previ- ously reported in the company may also be observed. Panel B of Table 1 lists the dis- tribution of LAYER. The results show that only 6.1% of the pyramidal layer equals 1 and that most of the pyramidal layer is 2. This demonstrates that in this sample, gov- ernment decentralisation is rather common. Panel C of Table 1 indicates that LAYER changes annually. The results show that the annual average LAYER has a steadily rising annual trend, consistent with the reality of SOE reform continuously deepening, which, to some extent, indicates that it is appropriate to adopt pyramidal layers as a proxy for decentralised government. To investigate government decentralisation affecting corporate fraud, the sample is ﬁrst divided into four parts according to the pyramidal layers, with the number of the layers, respectively equal to 1, 2, 3, and greater than or equal to 4. The probability of fraud in each of the four parts is reported in Figure 1(a). Figure 1(a) shows that corpo- rate pyramidal layers for 1 and 2 entail a very close fraud probability. When the num- ber of pyramidal layers rises to 3, the probability of corporate fraud decreases to 9%. When the number is greater than or equal to 4, the probability drops to 7.6%, with a decreased amount of approximately 23% (1 – 7.6% / 9.9%) compared with companies China Journal of Accounting Studies 329 Table 1. Descriptive statistics. Panel A: Firm characteristics Variables N Mean STD Q1 Median Q3 FRAUD 6,111 0.097 0.296 0.000 0.000 0.000 FRAUD_DISCLOSE 6,111 0.080 0.272 0.000 0.000 0.000 FRAUD_MARKET 6,111 0.015 0.122 0.000 0.000 0.000 FRAUD_CAPITAL 6,111 0.024 0.153 0.000 0.000 0.000 FRAUD_OTHER 6,111 0.058 0.235 0.000 0.000 0.000 LAYER 6,111 2.452 0.887 2.000 2.000 3.000 LNMCAP 6,111 21.869 1.213 20.986 21.729 22.562 ROA 6,111 0.027 0.068 0.010 0.029 0.056 LEV 6,111 0.521 0.205 0.380 0.526 0.652 MB 6,111 3.226 3.069 1.447 2.312 4.005 BIG4 6,111 0.095 0.294 0.000 0.000 0.000 MANHOLD 6,111 0.001 0.009 0.000 0.000 0.000 INDEP 6,111 0.353 0.050 0.333 0.333 0.364 DUALITY 6,111 0.099 0.299 0.000 0.000 0.000 LNBSIZE 6,111 2.247 0.205 2.197 2.197 2.398 HBSHARE 6,111 0.111 0.314 0.000 0.000 0.000 MEETING 6,111 2.055 0.377 1.792 2.079 2.303 LOSS 6,111 0.113 0.317 0.000 0.000 0.000 ST 6,111 0.056 0.230 0.000 0.000 0.000 GOV_SHARE 6,111 0.348 0.233 0.146 0.377 0.533 GOV_INTERV 6,111 6.459 3.126 4.330 6.780 8.820 TOP1 6,111 0.408 0.159 0.282 0.403 0.528 RPT 6,111 0.124 0.234 0.000 0.025 0.152 ORECTA 6,111 0.002 0.013 0.000 0.000 0.000 Panel B: Distribution of pyramidal layers LAYER No. of OBS =1 373 6.10 =2 3,499 57.26 =3 1,607 26.30 ≥4 632 10.34 Panel C: Pyramidal layers by year Year Mean Year Mean 2004 2.276 2008 2.489 2005 2.342 2009 2.600 2006 2.374 2010 2.671 2007 2.439 FRAUD is an indicator that equals 1 if a listed ﬁrm commits fraud in the current year, otherwise 0; FRAUD_- DISCLOSE is an indicator that equals 1 if a listed ﬁrm commits fraud in the current year and the fraud is related to information disclosure (fabricating proﬁt, ﬁctitious assets, false statements, delayed disclosure, major omissions, untrue disclosure, dishonest listings, and general accounting misconduct), otherwise 0; FRAUD_MARKET is an indicator that equals 1 if a listed ﬁrm commits fraud in the current year and the fraud is related to market transactions (insider trading, illegal stock purchases, and manipulating stock prices), other- wise 0; FRAUD_CAPITAL is an indicator that equals 1 if a listed ﬁrm commits fraud in the current year and the fraud is related to corporate capital (illegal investments, unauthorised changes in capital use, the occupa- tion of listing ﬁrms’ assets by the controlling shareholders, and illegal guarantees), otherwise 0; FRAUD_- OTHER is an indicator that equals 1 if a listed ﬁrm commits fraud in the current year and the fraud is of other types, otherwise 0; LAYER is the proxy of the degree of government decentralisation over SOEs, deﬁned as the number of pyramidal layers from the ultimate controller to the listed SOE; LNMCAP is the natural log- arithm of a ﬁrm’s market capitalization; ROA is deﬁned as the net income divided by the total assets; LEV is deﬁned as the total liabilities divided by the total assets; MB is deﬁned as the total market value of equity divided by the book value of equity; BIG4 is an indicator that equals 1 if a ﬁrm has hired an international ‘big four’ auditing ﬁrm, otherwise 0; MANHOLD is the percentage of management shareholding; INDEP is the percentage of independent directors; DUALITY is an indicator that equals 1 if a ﬁrm’s chairman of the board and its CEO is the same person, otherwise 0; LNBSIZE is the natural logarithm of the number of direc- tors; HBSHARE is an indicator that equals 1 if a ﬁrm issues H-shares or B-shares simultaneously, otherwise 0; MEETING is the natural logarithm of the number of board meetings; LOSS is an indicator that equals 1 if there is a loss for the ﬁrm in the previous year, otherwise 0; ST is an indicator that equals 1 if a ﬁrm is trea- ted special, otherwise 0; GOV_SHARE is the percentage of state-owned shares; GOV_INTERV is the degree of regional government intervention, deﬁned as the index of ‘reducing government intervention over the enter- prise’ from Fan et al. (2011); TOP1 is the percentage of shares held by the largest shareholder; RPT is deﬁned as the total amount of a ﬁrm’s related-party transactions between the listed ﬁrm and the parent ﬁrm divided by the total assets of the listed ﬁrm; ORECTA is deﬁned as the inter-corporate loans by the controlling share- holder in the other receivables divided by the total assets. 330 Liu and Li (a) 12% 10.43% 9.92% 10% 9.02% 7.59% 8% 6% 4% 2% 0% Layer=1 Layer=2 Layer=3 Layer>=4 (b) 7% 5.88% 6% 4.97% 4.40% 5% 3.56% 3.44% 4% 3% 2.35% 2% 1% 0% Layer unchanged Layer increased Layer decreased Fraud in the previous year but no fraud in this year No fraud in the previous year but fraud in this year Figure 1. Graphical evidence. (a) Pyramidal layers and corporate fraud probability. (b) Changes of pyramidal layers and changes of corporate fraud probability. when the number of layers is equal to 1. According to the results of the univariate test, the fraud probability of ﬁrms with three layers is signiﬁcantly lower than that of the ﬁrms with layers fewer than 3 (p value is 0.063). However, the fraud probability of the ﬁrms with three, or more than three, pyramidal layers is not signiﬁcantly different (p value is 0.140). Furthermore, we report the changes in corporate fraud probability brought about by corporate pyramidal layers changes. In our sample, the pyramidal layers of 4,259 observations have no changes compared with the previous year; however, those of 523 observations have increased, and those of 255 observations decreased. The rele- vant changes in the fraud probability are collected and shown in Figure 1(b). Figure 1(b) shows that for ﬁrms with no change in pyramidal layers compared with the previous year, their probability of having no fraud the previous year but having fraud this year is 3.56%. Their probability of having fraud the previous year but no fraud this year is 4.40%. A univariate test indicates that the difference is signiﬁcant at the 5% level (p value is 0.026). Firms whose pyramidal layers increased over the previous year have a probability of having no fraud the previous year but having fraud this year of 4.97%, higher than the 3.44% probability of fraud the previous year but no fraud this year. The univariate test notes that the difference is signiﬁcant at the 12% level (p value is 0.114). For ﬁrms with pyramidal layers decreasing com- pared with the previous year, their probability of having no fraud the previous year but fraud this year is 2.35%, lower than the probability of 5.88% of fraud the previ- ous year but no fraud this year. The univariate test indicates that the difference is signiﬁcant at the 5% level (p value is 0.025). China Journal of Accounting Studies 331 Figure 1 illustrates in a preliminary manner the ‘government intervention hypothe- sis’, namely that decentralised government can reduce the probability of corporate fraud. However, considering that there are many other factors also affecting corporate fraud, we must be cautious about the results. Untabulated results show that the Pearson and Spearman correlation coefﬁcients of LAYER and FRAUD are signiﬁcantly negative at the 5% level, which is consistent with the conclusion of Figure 1. In addition, the absolute value of most of the correlation coefﬁcients is below 0.5, indicating that there is no serious multicollinearity problem for the empirical model. 5. Empirical results 5.1. Government decentralisation and corporate fraud In this section, we employ multivariate regression analyses. Because FRAUD is a dummy variable, we use the probit regression model. The empirical results are listed in Table 2. The regression results in Table 2 are divided into four parts, each discussed sepa- rately. The results of the ﬁrst column show that the coefﬁcient on LAYER is signiﬁ- cantly negative at the 5% level, indicating that with the increase of government decentralisation degree over SOEs, the probability of corporate fraud lowers signiﬁ- cantly, consistent with the results of descriptive statistics and the theoretical expectation of the ‘government intervention hypothesis’. The right-hand side of the ﬁrst column shows the marginal effect of each variable. The marginal effect of LAYER is –0.013, indicating that for every one-unit increase of the pyramidal layer, the probability of cor- porate fraud would decrease by 1.3%; in the results of Column (2), we set up a dummy variable DUMLAYER according to whether LAYER is greater than 2; if yes, DUM- LAYER equals 1, otherwise 0. The beneﬁt of binary handling LAYER is that it can pre- vent the results of Column (1) from being caused by the fraud probability of an extremely large LAYER. The results of Column (2) show that the coefﬁcient on DUM- LAYER is signiﬁcantly negative at the 1% level, further supporting the results in Col- umn (1). Additionally, the marginal effect of DUMLAYER is –0.025, indicating that the average corporate fraud probability of the sample with LAYER greater than 2 is 2.5% less than that with LAYER less than or equal to 2. Taken together, the results of Columns (1) and (2) indicate that government decen- tralisation can reduce the probability of SOEs to engage in fraud. However, might there remain a controversy over whether the pyramidal layers actually are limited to repre- senting the degree of government decentralisation over SOEs? A possible method of alleviating this concern is to examine whether the pyramidal layers also affect the prob- ability of non-state-owned enterprises (NSOEs) to commit fraud. An afﬁrmative con- clusion will invalidate our empirical results. The results using a NSOEs sample are listed in Column (3), with the coefﬁcient on LAYER not signiﬁcant. With the use of DUMLAYER as the independent variable, the results of Column (4) report that the coef- ﬁcient on DUMLAYER remains insigniﬁcant. This suggests that the association between pyramidal layers and corporate fraud does not exist in NSOEs. Among the regression results of control variables, the coefﬁcient on LNMCAP is signiﬁcantly negative, indicating that the fraud probability of large-size ﬁrms is low. The coefﬁcient on ROA is signiﬁcantly negative, indicating that high-proﬁtability ﬁrms are unlikely to engage in fraud. Among the variables of corporate governance, the 332 Liu and Li Table 2. The effect of government decentralisation on corporate fraud. (1) Marginal (2) Marginal (3) (4) Variables SOEs effects SOEs effects NSOEs NSOEs LAYER −0.097** −0.013 −0.004 (–2.43) (–0.12) DUMLAYER −0.197*** −0.025 0.026 (–2.92) (0.36) LNMCAP −0.208*** −0.028 −0.210*** −0.028 −0.143*** −0.145*** (–4.14) (–4.14) (–2.75) (–2.77) ROA −1.017** −0.136 −1.037** −0.138 −1.408*** −1.404*** (–2.40) (–2.44) (–4.60) (–4.59) LEV 0.377** 0.050 0.373** 0.050 −0.114 −0.118 (2.04) (2.03) (–0.90) (–0.93) MB 0.011 0.001 0.011 0.002 0.017** 0.017** (1.09) (1.11) (2.18) (2.20) BIG4 −0.240 −0.028 −0.238 −0.027 −0.125 −0.129 (–1.13) (–1.13) (–0.54) (–0.56) MANHOLD −3.283 −0.438 −3.298 −0.440 −0.295 −0.273 (–1.31) (–1.32) (–1.19) (–1.11) INDEP 0.278 0.037 0.256 0.034 −0.391 −0.376 (0.44) (0.41) (–0.58) (–0.55) DUALITY 0.099 0.014 0.101 0.014 0.092 0.093 (0.96) (0.99) (1.21) (1.22) LNBSIZE 0.274 0.037 0.273 0.036 0.183 0.187 (1.41) (1.40) (0.93) (0.94) HBSHARE −0.066 −0.008 −0.075 −0.010 −0.437** −0.441** (–0.40) (–0.46) (–1.99) (–2.02) MEETING 0.147* 0.020 0.149* 0.020 0.221** 0.218** (1.76) (1.79) (2.57) (2.54) LOSS 0.180** 0.027 0.183** 0.027 0.199** 0.197** (2.49) (2.53) (2.34) (2.32) ST 0.338** 0.056 0.333** 0.055 0.526*** 0.526*** (2.56) (2.53) (3.77) (3.77) GOV_SHARE 0.300* 0.040 0.307* 0.041 −0.047 −0.055 (1.68) (1.72) (–0.14) (–0.16) GOV_INTERV −0.002 −0.000 −0.002 −0.000 −0.007 −0.007 (–0.20) (–0.13) (–0.65) (–0.62) TOP1 −0.738*** −0.099 −0.748*** −0.100 −0.681** −0.683** (–2.73) (–2.77) (–2.56) (–2.57) RPT −0.103 −0.014 −0.101 −0.013 −0.192 −0.197 (–0.81) (–0.81) (–1.16) (–1.19) ORECTA 6.898*** 0.921 6.832*** 0.911 7.279*** 7.256*** (4.09) (4.05) (2.71) (2.70) CONSTANT 2.665** 2.542** 1.542 1.550 (2.38) (2.23) (1.21) (1.22) YEAR YES YES YES YES INDUSTRY YES YES YES YES N 6,111 6,111 3,861 3,861 Pseudo R 0.108 0.109 0.097 0.098 The dependent variable is FRAUD. DUMLAYER is an indicator that equals 1 if LAYER is greater than 2, otherwise 0. Refer to Table 1 for the deﬁnitions of the other variables. ***, ** and * denote signiﬁcance at the 1, 5 and 10% levels, respectively. Z-statistics computed with robust standard errors clustered at the ﬁrm level are reported in parentheses. The sample in Columns (1) and (2) is SOEs. The sample in Columns (3) and (4) is NSOEs. China Journal of Accounting Studies 333 coefﬁcient on the number of board meetings (MEETING) is signiﬁcantly positive, consistent with the conclusion of Chen et al. (2006). The coefﬁcients on the two vari- ables LOSS and ST are signiﬁcantly positive, meaning that the ﬁrms with a loss or the risk of delisting have stronger motivation to participate in fraud. The coefﬁcient on GOV_SHARE is positive, suggesting that the ﬁrms with more state-owned shares are more likely to engage in fraud. However, even after controlling for the variable of the proportion of state shares, the key variable LAYER remains signiﬁcant. This suggests that although the government owns the same percentage of shares among different SOEs, its ability to intervene in ﬁrms remains affected by the length of the control chain. The coefﬁcient on TOP1 is signiﬁcantly negative, implying that a higher own- ership concentration is associated with a lower probability of corporate fraud. The coef- ﬁcient on ORECTA is signiﬁcantly positive at the 1% level, suggesting that more serious controlling shareholder tunnelling activity is associated with a higher probability of fraud in the listed companies, consistent with Fan and Wong (2002) and Leuz et al. (2003). In other words, to hide tunnelling, ﬁrms tend to manipulate earnings and cause accounting irregularities. 5.2. Robustness tests To increase the credibility of the conclusion, the following robustness tests are conducted. First, in the empirical models of this paper, there may be some uncontrolled vari- ables affecting both pyramidal layers and corporate fraud. In other words, this model may have omitted variables. To mitigate the effects of missing variables on the conclu- sion, we include the ﬁrm ﬁxed effects in the model. This method can effectively control factors that do not change with time trends but may also affect corporate fraud and pyramidal layers. Theoretical and empirical studies of Fan et al. (2013) and Xia and Chen (2007) ﬁnd that regional institution environment and the characteristics of indus- try strategy are important factors affecting government control mode and the degree of decentralisation and that they change little over time. Therefore, controlling the ﬁrm ﬁxed effect can largely eliminate the effect of those omitted variables on the conclusion. Speciﬁcally, the results including ﬁrm ﬁxed effects are shown in Table 3. Table 3 shows, after controlling for the ﬁrm ﬁxed effects, that many previously sig- niﬁcant control variables become no longer signiﬁcant, perhaps the result of a minimal annual change in variables. However, the coefﬁcient on LAYER remains signiﬁcantly negative at the 5% level, indicating that even controlling for those variables with no ﬂuctuations over time, the conclusion of the paper remains. Therefore, the problem of missing variables will not seriously inﬂuence our conclusion. Second, Zhong et al. (2010) observe that there is an inverse U-shaped relationship between the pyramidal layers and over-investment of free cash ﬂow in SOEs, implying that decentralisation can reduce the over-investment of free cash ﬂow to a certain extent but, after reaching that certain extent, the agency costs increase signiﬁcantly, causing inefﬁcient investment. Does such a nonlinear relationship also exist in our paper? To con- ﬁrm this, we include LAYER and its square, LAYER , in the model. The unreported results indicate that the coefﬁcient on LAYER is insigniﬁcant. Therefore, there is no evidence of a nonlinear relationship between government decentralisation and corporate fraud. Third, the transfer of corporate control rights is likely to make listed ﬁrms’ location in the pyramidal ownership structure change, leading to the number of pyramidal layers between the controlling shareholder and the listed ﬁrm changing. Nonetheless, this 334 Liu and Li Table 3. Controlling for ﬁrm ﬁxed effects. Variables Coefﬁcients Z values LAYER −0.275** (–2.47) LNMCAP 0.115 (0.66) ROA −1.736* (–1.95) LEV −0.348 (–0.62) MB −0.019 (–0.85) BIG4 0.089 (0.17) MANHOLD −4.190 (–0.78) INDEP 0.031 (0.02) DUALITY −0.040 (–0.17) LNBSIZE 0.912* (1.78) MEETING 0.179 (1.01) LOSS 0.308** (2.17) GOV_SHARE −0.014 (–0.03) GOV_INTERV 0.059 (1.37) TOP1 −2.335** (–2.46) RPT −0.012 (–0.05) ORECTA 11.618*** (3.03) CONSTANT −2.969 (–0.82) YEAR YES FIRM YES N 1,533 Pseudo R 0.228 The dependent variable is FRAUD. Refer to Table 1 for the deﬁnitions of the variables. ***, ** and * denote signiﬁcance at the 1, 5 and 10% levels, respectively. Z-statistics computed with robust standard errors clus- tered at the ﬁrm level are reported in parentheses. Sample size is reduced because the observations whose val- ues of FRAUD are constant during the sample period are automatically deleted due to perfect collinearity after including ﬁrm ﬁxed effects into the regression model. change does not directly result from government decentralisation. If the transferred con- trol of the ﬁrm also affects corporate fraud, then the relationship between pyramidal layers and corporate fraud is likely to be caused by the transfer of control rights. To eliminate the inﬂuence of the alternative explanation, we obtained the list of listed ﬁrms whose control rights have been shifted by equity trading in 2005–2010, from the CSMAR database. We then eliminated these listed ﬁrms from our sample and reran model (1). Unreported results show that the coefﬁcient on LAYER remains negative and signiﬁcant at the 5% level (the coefﬁcient is –0.117, and the z value is –2.30). There- fore, the transfer of corporate control rights does not affect our conclusion. Fourth, in the process of forming a pyramidal structure, there may be a divergence between the ultimate controlling shareholder’s voting rights and cash ﬂow rights over the listed ﬁrm of the pyramidal structure, possibly leading to an agency problem for the con- trolling shareholder, thereby affecting the corporate information environment (Claessens, Djankov, Fan, & Lang, 2002; Fan & Wong, 2002). Hence, pyramidal layers having inﬂuence on corporate fraud is likely due to the inﬂuence of the divergence between the controlling shareholder’s voting rights and cash ﬂow rights. To eliminate this alternative explanation, we include two variables in the model (1): CASHR, representing the cash ﬂow rights of the corporate controlling shareholder; and WEDGE, the difference between the voting rights and cash ﬂow rights of the controlling shareholder. After rerunning model (1) including these two variables, unreported results show that the coefﬁcient on LAYER is signiﬁcantly negative (the coefﬁcient is –0.120, and z value is –2.72). The China Journal of Accounting Studies 335 coefﬁcients on CASHR and WEDGE are not signiﬁcant (the coefﬁcient on CASHR is – 0.669, the z value is –1.34; the coefﬁcient on WEDGE is 0.003, the z value is 0.000). These results are consistent with Fan et al. (2013); because the Chinese government has strict rules for the transfer of state-owned shares, the divergence between the controlling shareholders’ voting rights and cash ﬂow rights in the pyramidal structure of SOEs is very small. This leads to the low information content of WEDGE. 5.3. Government decentralisation and corporate fraud: Categorising fraud types Categorising fraud types and examining how government decentralisation reduces speciﬁc types of fraud has policy implications. According to Chen et al. (2013), corporate fraud is classiﬁed into four types: (1) FRAUD_DISCLOSE, related to information disclosure (fabricating proﬁt, ﬁctitious assets, false statements, delayed dis- closure, major omissions, untrue disclosure, dishonest listings, and general accounting misconduct); (2) FRAUD_MARKET, connected with market transactions (insider trad- ing, illegal stock purchases, and manipulating stock prices); (3) FRAUD_CAPITAL, concerned with corporate capital (illegal investments, unauthorised changes in capital use, the occupation of listing ﬁrms’ assets by the controlling shareholders, and illegal guarantees); and (4) FRAUD_OTHER, referring to other types. Table 1 lists the descrip- tive statistics of these four types of fraud, with FRAUD_DISCLOSE the most and FRAUD_MARKET the least likely because of the larger crackdown and heavier punish- ment by supervisory agencies. Theoretically, how does government decentralisation affect different types of fraud? For information disclosure fraud, the current literature suggests that government intervention signiﬁcantly affects the enterprise information environment and causes low-quality accounting information for ﬁrms (Piotroski & Wong, 2012). This stems pri- marily from the decrease in the supply of and demand for high-quality accounting information for SOEs from government intervention (Piotroski & Wong, 2012). The demand side, embodied in government intervention, reduces the contract function of accounting information, for instance, by the direct appointment of CEOs to SOEs by government and by private and political information channelling between government and managers (Fan et al., 2007). Moreover, the soft budget constraint of SOEs lowers the demand of creditors for high-quality ﬁnancial statements based on debt contracts. The supply side often allows SOEs to pursue political rather than operational aims. This distortion motivates SOEs to conceal bad news from the public and damages the information environment (Piotroski et al., 2015). Therefore, it is expected that govern- ment decentralisation signiﬁcantly reduces corporate fraud in information disclosure. Second, some types of market transactions fraud (e.g. insider trading, illegal stock purchases, and stock price manipulation) are closely related to information disclosure. Speciﬁcally, abundant literature proves that managers can manipulate stock prices and proﬁt from doing so by managing information disclosure. For example, Cheng and Lo (2006) discover that managers disclose bad news to reduce stock prices before purchas- ing stocks. Cheng, Luo, and Yue (2013) further observe that managers also achieve the goal of affecting stock price by manipulating the precision of management forecasts. Bergstresser and Philippon (2006) show that managers affect stock price through earn- ings management. Government decentralisation reduces the probability of information disclosure fraud; therefore, it is believed that with a higher quality of information dis- closed due to less government intervention, market transaction fraud will happen at a signiﬁcantly lower rate. 336 Liu and Li Finally, we argue that government decentralisation has a limited effect upon corpo- rate fraud related to capital. Speciﬁcally, Jian and Wong (2010) conﬁrm that controlling shareholders of SOEs support rather than expropriate listed companies through related- Table 4. Government decentralisation on corporate fraud: Categorising fraud types. (1) (2) (3) (4) Variables FRAUD_DISCLOSE FRAUD_MARKET FRAUD_CAPITAL FRAUD_OTHER LAYER −0.096** −0.144** 0.030 −0.073 (–2.13) (–2.41) (0.43) (–1.53) LNMCAP −0.216*** −0.075 −0.281*** −0.265*** (–3.72) (–1.29) (–3.13) (–4.58) ROA −1.477*** −0.207 −1.671*** −0.434 (–3.34) (–0.27) (–3.05) (–0.84) LEV 0.256 0.346 0.431 0.373* (1.31) (1.21) (1.64) (1.67) MB −0.003 0.038*** −0.008 0.007 (–0.25) (2.87) (–0.63) (0.61) BIG4 −0.141 −0.384 −4.398*** −0.170 (–0.57) (–1.23) (–9.37) (–0.61) MANHOLD −4.719 0.222 1.207 −5.198 (–1.40) (0.07) (0.31) (–1.54) INDEP 0.449 −1.183 0.366 0.508 (0.66) (–1.17) (0.33) (0.63) DUALITY 0.107 0.223* 0.037 0.009 (0.94) (1.69) (0.25) (0.07) LNBSIZE 0.268 0.377 0.005 0.452* (1.22) (1.45) (0.02) (1.90) HBSHARE −0.307* 0.264 0.120 −0.174 (–1.66) (1.17) (0.37) (–0.81) MEETING 0.135 0.247* 0.007 0.080 (1.49) (1.92) (0.05) (0.84) LOSS 0.193*** 0.007 0.050 0.026 (2.60) (0.05) (0.45) (0.31) ST 0.373*** −0.459 0.350* 0.201 (2.69) (–1.48) (1.89) (1.30) GOV_SHARE 0.203 0.203 0.132 0.454** (1.07) (0.66) (0.43) (2.10) GOV_INTERV −0.007 −0.007 −0.028 0.014 (–0.58) (–0.46) (–1.50) (0.92) TOP1 −0.613** −1.063** 0.128 −0.442 (–2.12) (–2.17) (0.29) (–1.43) RPT −0.120 0.093 −0.204 −0.057 (–0.89) (0.41) (–1.03) (–0.41) ORECTA 6.516*** 1.862 8.690*** 4.742** (3.74) (0.45) (4.31) (2.32) CONSTANT 2.755** −0.682 3.590* 2.861** (2.12) (–0.55) (1.74) (2.19) YEAR YES YES YES YES INDUSTRY YES YES YES YES N 6,111 6,111 6,111 6,111 Pseudo R 0.117 0.155 0.181 0.106 The dependent variable in each column is FRAUD_DISCLOSE, FRAUD_MARKET, FRAUD_CAPITAL, and FRAUD_OTHER, respectively. Refer to Table 1 for the deﬁnitions of the variables. ***, ** and * denote sig- niﬁcance at the 1, 5 and 10% levels, respectively. Z-statistics computed with robust standard errors clustered at the ﬁrm level are reported in parentheses. China Journal of Accounting Studies 337 party transactions, a phenomenon not found in NSOEs. Similarly, Jiang et al. (2010) ﬁnd that the occupation of listed company capital by controlling shareholders primarily occurs in NSOEs. This phenomenon occurs because the latter’s governmental back- ground deprives them of capital, which frequently affects NSOEs (Allen, Qian, & Qian, 2005). Consequently, NSOEs are degraded as an ‘Automatic Teller Machine’ for their parent companies or provide guarantees for their parent companies to make loans. Therefore, SOEs are less likely to transgress on rules, resulting in a limited effect of decentralisation on funds related to capital. To test the above theoretical expectation, Table 4 lists the regression results of the effects of government decentralisation upon various types of fraud. When dependent variables are FRAUD_DISCLOSE and FRAUD_MARKET, the coefﬁcients on LAYER are both signiﬁcantly negative at the 5% level. When dependent variables are FRAUD_CAPITAL and FRAUD_OTHER, the coefﬁcients on LAYER are both not signiﬁcant. This proves that government decentralisation decreases information disclosure-related and market transactions-related fraud but has no signiﬁcant effect on capital-related fraud. 5.4. Government decentralisation and corporate fraud: Categorising the level of fraud severity The level of fraud severity directly relates to the economic consequences of fraud. Fol- lowing Chen, Jiang, Liang, and Wang (2011), the backward method is utilised to infer the level of fraud severity according to its consequences. Speciﬁcally, penalties for list- ing ﬁrms’ fraud by supervisory agencies are used to judge the level of severity. First, for the type of penalty, if the ﬁrm suffers no penalty or the penalty is limited only to a condemnation, fraud is not severe. If it contains ﬁnes, warnings, or criticisms, fraud is severe. Second, for ﬁnes, if the penalty is without ﬁnes, fraud is not severe; otherwise, it is severe. Finally, Chen, Jiang, Liang, and Wang (2011) note that there is a selective enforcement issue for supervisory agencies, namely that the penalties are not directly related to the level of severity. Hence, fraud types are further classiﬁed to discern the level of severity. Speciﬁcally, when violation type is fabricating proﬁt, false assets, delayed disclosure, illegal stock purchase, or other frauds, it is not severe; otherwise, it is severe (Chen, Jiang, Liang, and Wang 2011). Because government decentralisation may eliminate the umbrella of government for ﬁrms (Chen, Jiang, Liang, and Wang, 2011), so potentially exposing the ﬁrms to penal- ties, it is not clearly predicted by theory – for this part of the analysis – when ﬁrms might consider fraud with both severe and lenient penalties. Certainly, a higher severity of fraud induces more adverse economic consequences; thus, ﬁrms may strive not to commit fraud with severe penalties. Therefore, theoretically, it is ambiguous whether government decentralisation inhibits more severe fraud or less severe fraud. Thus, an empirical test is attempted for the analysis. In the test, the sample is divided into two, a sample with more severe fraud and no fraud and a sample with less severe fraud and no fraud, allowing effective discernment of which type of fraud is affected by govern- ment decentralisation. Relevant empirical results are presented in Table 5, demonstrat- ing that for the categorisation based on type of penalty, decentralisation largely decreases less-severe fraud, with the Column (2) LAYER being signiﬁcantly negative. For the categorisation based on the amount of ﬁnes, decentralisation largely decreases more-severe fraud, with the Column (3) LAYER being signiﬁcantly negative. Finally, for the categorisation based on the type of fraud, decentralisation largely decreases 338 Liu and Li Table 5. Government decentralisation on corporate fraud: Categorising the level of fraud severity. Type of penalty Amount of ﬁne Type of corporate fraud (1) (2) (3) (4) (5) (6) Variables More severe Less severe More severe Less severe More severe Less severe LAYER −0.096 −0.087* −0.159** −0.061 −0.073 −0.133*** (–1.56) (–1.94) (–2.11) (–1.45) (–1.54) (–2.85) LNMCAP −0.154* −0.222*** −0.038 −0.264*** −0.231*** −0.157*** (–1.91) (–4.25) (–0.51) (–4.92) (–3.55) (–2.74) ROA −1.382** −0.819* −1.990*** −0.504 −1.566*** 0.667 (–2.46) (–1.69) (–3.72) (–1.04) (–3.40) (1.02) LEV 0.024 0.441** 0.167 0.389* 0.311 0.403 (0.10) (2.13) (0.65) (1.88) (1.51) (1.50) MB 0.003 0.011 0.025* −0.000 0.002 0.022 (0.18) (1.04) (1.87) (–0.00) (0.17) (1.59) BIG4 −0.042 −0.284 0.059 −0.359 −0.334 −0.084 (–0.20) (–1.09) (0.31) (–1.24) (–0.96) (–0.42) MANHOLD −99.306 −1.867 −257.421 −2.094 −2.904 −3.877 (–1.12) (–0.81) (–1.24) (–0.90) (–0.97) (–1.03) INDEP 0.437 0.142 1.300 −0.138 0.552 −0.038 (0.46) (0.21) (1.24) (–0.21) (0.79) (–0.04) DUALITY −0.069 0.178 0.030 0.128 0.037 0.194 (–0.45) (1.59) (0.20) (1.15) (0.32) (1.39) LNBSIZE 0.111 0.344* 0.428 0.207 0.226 0.393 (0.34) (1.70) (1.21) (1.05) (0.99) (1.55) HBSHARE −0.251 −0.008 −0.112 −0.054 −0.186 0.093 (–1.35) (–0.04) (–0.59) (–0.29) (–0.92) (0.46) MEETING 0.086 0.169* 0.215 0.125 0.193** 0.031 (0.66) (1.87) (1.53) (1.43) (2.01) (0.28) LOSS 0.195* 0.175** 0.117 0.209*** 0.190** 0.120 (1.78) (2.24) (1.01) (2.72) (2.49) (1.08) ST 0.628*** 0.120 0.605*** 0.142 0.308** 0.298 (3.75) (0.74) (3.50) (0.99) (2.10) (1.59) China Journal of Accounting Studies 339 GOV_SHARE −0.245 0.465** −0.018 0.352* 0.234 0.329 (–0.97) (2.33) (–0.06) (1.82) (1.18) (1.26) GOV_INTERV −0.013 0.004 −0.010 0.001 −0.011 0.015 (–0.83) (0.31) (–0.59) (0.08) (–0.80) (1.01) TOP1 −0.551 −0.692** −0.927* −0.576** −0.608** −0.739* (–1.31) (–2.35) (–1.95) (–2.00) (–1.96) (–1.89) RPT −0.382* −0.025 −0.409 −0.012 −0.011 −0.248 (–1.74) (–0.18) (–1.22) (–0.10) (–0.07) (–1.24) ORECTA 5.321** 7.315*** 7.995*** 5.707*** 6.757*** 4.825* (2.27) (3.86) (3.28) (3.14) (3.76) (1.79) CONSTANT 0.709 2.713** −2.707 4.084*** 2.928** 0.906 (0.37) (2.39) (–1.57) (3.41) (1.98) (0.81) YEAR YES YES YES YES YES YES INDUSTRY YES YES YES YES YES YES N 5,693 5,934 5,672 5,955 5,924 5,703 Pseudo R 0.152 0.114 0.178 0.108 0.122 0.102 The dependent variable is FRAUD. Refer to Table 1 for the deﬁnitions of the variables. ***, ** and * denote signiﬁcance at the 1, 5 and 10% levels, respectively. Z-statistics com- puted with robust standard errors clustered at the ﬁrm level are reported in parentheses. 340 Liu and Li Table 6. Government decentralisation on corporate fraud: The effects of government intervention. Regional government intervention index Regional ﬁnancial deﬁcit degree Regional registered urban unemployment rate (1) (2) (3) (4) (5) (6) High government Low government High government Low government High government Low government Variables intervention intervention intervention intervention intervention intervention LAYER −0.121** −0.073 −0.133** −0.069 −0.139** −0.054 (–2.30) (–1.20) (–2.43) (–1.11) (–2.57) (–0.94) LNMCAP −0.129** −0.302*** −0.122* −0.305*** −0.170*** −0.235*** (–2.07) (–3.92) (–1.96) (–3.75) (–2.69) (–2.84) ROA −1.261** −0.953 −1.337*** −0.747 −0.380 −1.770*** (–2.43) (–1.32) (–2.69) (–0.94) (–0.70) (–2.72) LEV 0.288 0.514* 0.146 0.612** 0.442* 0.432 (1.25) (1.81) (0.63) (1.98) (1.83) (1.51) MB −0.001 0.033** 0.009 0.014 0.017 −0.008 (–0.10) (1.97) (0.79) (0.82) (1.40) (–0.48) BIG4 0.058 −0.417* 0.014 −0.476** 0.069 −1.144*** (0.17) (–1.84) (0.04) (–2.01) (0.27) (–4.00) MANHOLD −8.414 −0.106 −5.561 −0.999 −4.622 −2.469 (–1.47) (–0.04) (–0.75) (–0.38) (–0.66) (–0.93) INDEP 0.227 0.336 0.012 0.190 −0.154 0.597 (0.30) (0.32) (0.02) (0.16) (–0.19) (0.60) DUALITY 0.071 0.182 0.166 −0.045 0.198 −0.046 (0.56) (1.23) (1.31) (–0.26) (1.50) (–0.29) LNBSIZE 0.093 0.695** 0.198 0.299 0.152 0.362 (0.42) (2.06) (0.86) (0.89) (0.60) (1.27) HBSHARE −0.343 0.042 −0.044 −0.051 −0.229 0.125 (–1.00) (0.23) (–0.14) (–0.26) (–0.99) (0.56) MEETING 0.100 0.263** 0.223** 0.044 0.200* 0.092 (0.93) (2.07) (2.03) (0.34) (1.86) (0.70) LOSS 0.071 0.388*** 0.086 0.372*** 0.082 0.340*** (0.73) (3.76) (0.90) (3.37) (0.85) (3.22) ST 0.318** 0.497** 0.301** 0.518** 0.252 0.373* (2.19) (1.99) (1.99) (2.06) (1.53) (1.77) GOV_SHARE 0.532** 0.078 0.498** 0.147 0.146 0.572** (2.43) (0.29) (2.19) (0.52) (0.57) (2.37) China Journal of Accounting Studies 341 GOV_INTERV 0.034* −0.004 0.021 0.001 −0.006 0.005 (1.67) (–0.14) (1.16) (0.03) (–0.39) (0.24) TOP1 −1.246*** −0.051 −1.266*** −0.264 −0.968*** −0.640 (–3.60) (–0.13) (–3.57) (–0.60) (–2.77) (–1.57) RPT −0.191 0.042 −0.107 −0.087 −0.013 −0.192 (–1.25) (0.20) (–0.72) (–0.37) (–0.08) (–0.94) ORECTA 9.543*** 0.425 9.691*** 1.629 6.955*** 7.144*** (4.90) (0.10) (4.73) (0.47) (3.07) (2.90) CONSTANT 1.558 3.021* 1.073 4.629** 2.140 2.909 (1.11) (1.82) (0.75) (2.52) (1.58) (1.56) YEAR YES YES YES YES YES YES INDUSTRY YES YES YES YES YES YES N 3,180 2,931 3,150 2,961 3,340 2,771 Pseudo R 0.096 0.160 0.087 0.172 0.101 0.168 The dependent variable is FRAUD. Refer to Table 1 for the deﬁnitions of the variables. ***, ** and * denote signiﬁcance at the 1, 5 and 10% levels, respectively. Z-statistics com- puted with robust standard errors clustered at the ﬁrm level are reported in parentheses. 342 Liu and Li less-severe fraud, with the Column (6) LAYER being signiﬁcantly negative. Therefore, overall, there is no signiﬁcant disparity among fraud by various levels of severity affected by government decentralisation. The conclusion is most likely formed for two reasons: ﬁrst, as stated previously, under the decentralisation background, there is no obvious preference for SOEs between avoiding severe and not-severe commitment of fraud; second, there is a certain bias in the judgement of level of severity in the paper. 5.5. Government decentralisation and corporate fraud: The effects of government intervention The previous empirical results indicate that the relationship between government decen- tralisation and corporate fraud ﬁts well with the theoretical expectation of the ‘govern- ment intervention hypothesis’. To further support the hypothesis, sub-sample regression is conducted based on the degree of intervention of government over SOEs. Theoreti- cally, a stronger government intervention is associated with a greater decrease in the political costs caused by government decentralisation and thus a greater inﬂuence of government decentralisation upon corporate fraud. To conﬁrm this inference, following Fan et al. (2013), three variables are selected to measure the motive for government intervention: the ‘reducing government intervention over ﬁrms’ index from Fan et al. (2011), regional ﬁnancial deﬁcit degree and regional registered urban unemployment rate. When the level of government intervention is greater than the annual sample med- ian, it is included in the high government intervention group; otherwise, it is included in the low government intervention group. The results are reported in Table 6. The results in Table 6 imply that the coefﬁcients on LAYER are signiﬁcant within only the high government intervention group. Speciﬁcally, in Columns (1), (3) and (5), the coefﬁcients on LAYER are all signiﬁcantly negative at the 5% level. Nevertheless, they are all insigniﬁcant in the subsamples of low government intervention. Thus, the results in Table 6 further conﬁrm the ‘government intervention hypothesis’. 6. Conclusion Understanding the economic consequences of the deepening SOE reforms is critically important. From the perspective of the pyramidal structure of listed SOEs, this paper examines the effect of the degree of government decentralisation upon the probability of SOEs to commit fraud. Theoretically, government decentralisation reduces the politi- cal costs of government intervention, but simultaneously induces agency costs, both of which are vital ingredients for corporate fraud according to the literature. Therefore, government decentralisation might both decrease (‘government intervention hypothe- sis’) and increase corporate fraud probability (‘agency costs hypothesis’). Using the data of A-share listed SOEs in China in 2004–2010, this paper observes empirically that government decentralisation over SOEs signiﬁcantly reduces their fraud probability. Therefore, the relationship between government decentralisation and corpo- rate fraud ﬁts well with the ‘government intervention hypothesis’. Further results pro- ven by the categorisation of fraud types are that government decentralisation largely decreases fraud related to information disclosure and market transactions. However, there is no signiﬁcant difference among fraud activities with various levels of severity affected by government decentralisation. Finally, the effect of government decentralisa- tion upon corporate fraud largely occurs in SOEs in which government intervention is more likely, further conﬁrming the ‘government intervention hypothesis’. China Journal of Accounting Studies 343 The conclusions of this paper provide new empirical evidence for the study of the economic consequences of government decentralisation and the study of factors underlying corporate fraud. The conclusions also support SOE reforms by government decentralisation from the perspective of accounting information quality, thereby embodying important policy implications. The limitation of this paper lies in the exclusion of analysing types of government decentralisation other than the pyramidal structure formed since the 1990s; this necessi- tates future consideration. Acknowledgements The authors appreciate the helpful comments from two anonymous reviewers, Donghua Chen (Associate editor), Pauline Weetman (Language Editor), Liansheng Wu (Joint-editor), and Jason Zezhong Xiao (Joint-editor). Disclosure statement No potential conﬂict of interest was reported by the authors. Funding Hang Liu acknowledges ﬁnancial support from the National Natural Science Foundation of China (grant no. 71402017) and the Program for Liaoning Excellent Talents in University (grant no. WJQ2014035). Xiaorong Li acknowledges ﬁnancial support from the National Natural Science Foundation of China (grant no. 71503283), the Humanities and Social Science Research Project of the Ministry of Education in China (grant no. 14YJC630069), the Social Science Research Project of Beijing (grant no. 15JGC173), the Program for Innovation Research in Central Univer- sity of Finance and Economics, and Zhongcai-Pengyuan Local Finance Investment and Funding Research Institute. Notes 1. The current literature on the Chinese government decentralisation largely refers to the decentralisation in the economic ﬁeld since 1980s after the reform and opening up, whose core is ﬁnance decentralisation from central to local government. Government decentralisa- tion on SOEs is also part of it and is exposed to many studies (Groves, Hong, McMillan, & Naughton, 1994, 1995; Qian, 1996). This paper focuses on the economic consequences of government decentralisation on SOEs. 2. Refer to Fan et al. (2013) for a more detailed description of the formation of state-owned pyramidal structures. 3. The fraud database in CSMAR collects the information related to corporate fraud from the announcements released by the listed ﬁrms committing fraud, the coverage from media des- ignated by the China Securities Regulatory Commission (CSRC), and the announcements released by regulators since 1994. The information of this database includes items such as the time period of fraud, the type of fraud, the type of punishment. 4. In China, if a listed ﬁrm has two consecutive annual losses, market regulators will assign special treatment (ST) status to it. ST ﬁrms face various trading and ﬁnancial restrictions. In addition, if they make losses for one more year, trading will be suspended; if they still make losses in the fourth year, they will be delisted. 5. The index is only revealed by the year 2009, and our sample period is 2004–2010. Therefore, the index of 2010 uses data from 2009 as a replacement. In addition, the index is an inverse index. In other words, a smaller GOV_INTERV indicates a higher degree of government intervention in the company location. 344 Liu and Li 6. Note that the existing study largely concludes that the pyramidal structure will enhance corporate information opacity and that the following empirical results of our paper show that the pyramidal structure will reduce the probability of SOEs to engage in fraud. Thus, theoretically, this explanation does not affect the conclusion of our paper. 7. The listed ﬁrms provide the ownership structure ﬁgures in their annual reports. Through the ownership structure ﬁgures, we can manually identify the pyramidal layers between the con- trolling shareholders and listed ﬁrms. The CSMAR database collects all the ownership struc- ture ﬁgures, so we can download these ﬁgures from CSMAR and manually calculate each ﬁrm’s pyramidal layer. 8. As far as we know, three papers investigate the evolution of NSOEs’ pyramidal structures in China (Chen, Jin, & Liu, 2011; Li, Xin, & Yu, 2008; Liu, Zheng, & Zhu, 2010). The conclusions of these papers generally indicate that NSOEs’ pyramidal structures are more prevalent in regions with worse institutions. Although determinants of pyramidal structures of SOEs and NSOEs are different according to current literature (Fan et al., 2013), we still cannot remove the concern that some unobservable factors affect the usage of pyramidal structures both in SOEs and NSOEs. If these unobservable factors impact on corporate fraud, then the conclusion of our paper cannot be interpreted as government decentralisa- tion. 9. We expound primarily on Column (1)’s regression results of the control variables because the results are our main concern. 10. What requires an explanation is that the coefﬁcient on GOV_INTERV is not signiﬁcant, which is inconsistent with the theory. A possible reason is that the existing variable is not a good measure of the degree of government intervention. Thus, we perform the exploration in two ways: ﬁrst, we set up a dummy variable DUMGI according to the size of GOV_IN- TERV; when ﬁrms’ GOV_INTERV is less than or equal to the annual sample median, it equals 1, otherwise 0. Unreported results show that when the regression model does not contain LAYER, the coefﬁcient on DUMGI is signiﬁcantly positive at the marginal level of 10% (z value = 1.62). In the model containing LAYER, the z value of DUMGI’s coefﬁcient is changed to 1.41. Second, we use the political connection (PC) as another measure of government intervention. According to Fan, Wong, and Zhang (2007), when the Chairman of the board or CEO of the SOE was or is a government ofﬁcial, PC equals 1, otherwise 0. Unreported results show that when not containing LAYER in the regression model, the coef- ﬁcient on PC is signiﬁcantly positive at the 10% level (z value = 1.69). When the regression model contains LAYER, the z value of PC’s coefﬁcient decreases to 1.52. In addition, when we change the dependent variable to FRAUD_DISCLOSE (the most important and common category of corporate fraud), the regression results also conﬁrm our above conclusion, reﬂected by the coefﬁcients on DUMGI and PC being signiﬁcantly positive at the 5% and 10% level, respectively. 11. In our sample, the average of WEDGE is only 4%. References Aghion, P., & Tirole, J. (1997). Formal and real authority in organizations. Journal of Political Economy, 105,1–29. Agrawal, A., & Chadha, S. (2005). Corporate governance and accounting scandals. Journal of Law and Economics, 48, 371–406. Aharony, J., Lee, C.-W. J., & Wong, T. J. (2000). Financial packaging of IPO ﬁrms in China. Journal of Accounting Research, 38, 103–126. Aharony, J., Wang, J., & Yuan, H. (2010). Tunneling as an incentive for earnings management during the IPO process in China. Journal of Accounting and Public Policy, 29,1–26. Allen, F., Qian, J., & Qian, M. (2005). Law, ﬁnance, and economic growth in China. Journal of Financial Economics, 77,57–116. Armstrong, C. S., Jagolinzer, A. D., & Larcker, D. F. (2010). Chief executive ofﬁcer equity incentives and accounting irregularities. Journal of Accounting Research, 48, 225–271. Armstrong, C. S., Larcker, D. F., Ormazabal, G., & Taylor, D. J. (2013). The relation between equity incentives and misreporting: The role of risk-taking incentives. Journal of Financial Economics, 109, 327–350. China Journal of Accounting Studies 345 Bai, J., & Lian, L. (2013). Why do state-owned enterprises over-invest? Government intervention or managerial entrenchment. China Journal of Accounting Studies, 1, 236–259. Beasley, M. S. (1996). An empirical analysis of the relation between the board of director com- position and ﬁnancial statement fraud. The Accounting Review, 71, 443–465. Bebchuk, L. A., Fried, J. M., & Walker, D. I. (2002). Managerial power and rent extraction in the design of executive compensation. The University of Chicago Law Review, 69, 751–846. Bergstresser, D., & Philippon, T. (2006). CEO incentives and earnings management. Journal of Financial Economics, 80,511–529. Burns, N., & Kedia, S. (2006). The impact of performance-based compensation on misreporting. Journal of Financial Economics, 79,35–67. Bushman, R. M., & Piotroski, J. D. (2006). Financial reporting incentives for conservative accounting: The inﬂuence of legal and political institutions. Journal of Accounting and Economics, 42, 107–148. Bushman, R. M., Piotroski, J. D., & Smith, A. J. (2004). What determines corporate trans- parency? Journal of Accounting Research, 42, 207–252. Chen, D., Chen, Y., Li, O. Z., & Ni, C. (2013). Foreign residency rights and corporate frauds. Working Paper. Nanjing University, China, National University of Singapore, Singapore, and Renmin University of China, China. Retrieved from http://ssrn.com/abstract=2310426 Chen, H., Chen, J. Z., Lobo, G. J., & Wang, Y. (2010). Association between borrower and lender state ownership and accounting conservatism. Journal of Accounting Research, 48, 973–1014. Chen, G., Firth, M., Gao, D., & Rui, O. M. (2005). Is China’s securities regulatory agency a toothless tiger? Evidence from enforcement actions. Journal of Accounting and Public Policy, 24, 451–488. Chen, G., Firth, M., Gao, D., & Rui, O. M. (2006). Ownership structure, corporate governance, and fraud: Evidence from China. Journal of Corporate Finance, 12, 424–448. Chen, D., Jiang, D., Liang, S., & Wang, F. (2011). Selective enforcement of regulation. China Journal of Accounting Research, 4,9–27. Chen, D., Jin, X., & Liu, X. (2011). Government quality, social capital and endogenous determi- nants of pyramid structure of family business. China Industrial Economics, (7), 129–139. (In Chinese.) Chen, X., Li, M., Rui, O. M., & Xia, L. (2009). Judiciary independence and the enforcement of investor protection laws: Market responses to the “1/ 15” notice of the Supreme People’s Court of China. China Economic Quarterly, (9), 1–28. (In Chinese.) Cheng, Q. & Lo, K. (2006). Insider trading and voluntary disclosures. Journal of Accounting Research, 44, 815–848. Cheng, Q., Luo, T., & Yue, H. (2013). Managerial incentives and management forecast precision. The Accounting Review, 88, 1575–1602. Cheng, Z., Xia, X., & Yu, M. (2008). Government intervention, pyramidal structures, and invest- ments of local state-owned listed ﬁrms. Management World, (9), 37–47. (In Chinese.) Claessens, S., Djankov, S., Fan, J. P. H., & Lang, L. H. P. (2002). Disentangling the incentive and entrenchment effects of large shareholdings. Journal of Finance, 57, 2741–2771. Dai, Y., Pan, Y., & Liu, S. (2011). Media supervision, government intervention, and corporate governance: Evidence from the perspective of Chinese listed ﬁrms’ ﬁnancial restatements. World Economy,(11), 121–144. (In Chinese.) Dechow, P., Ge, W., & Schrand, C. (2010). Understanding earnings quality: A review of the proxies, their determinants and their consequences. Journal of Accounting and Economics, 50, 344–401. Dechow, P. M., Sloan, R. G., & Sweeney, A. P. (1996). Cases and consequences of earnings manipulations: An analysis of ﬁrms subject to enforcement actions by the SEC. Contempo- rary Accounting Research, 13,1–36. Efendi, J., Srivastava, A., & Swanson, E. P. (2007). Why do corporate managers misstate ﬁnan- cial statements? The role of option compensation and other factors. Journal of Financial Eco- nomics, 85, 667–708. Erickson, M., Hanlon, M., & Maydew, E. L. (2006). Is there a link between executive equity incentives and accounting fraud? Journal of Accounting Research, 44,113–143. Fan, G., X. Wang, & H. Zhu (2011). NBEI index of marketization of China’s provinces 2011 report. Economics Science Press. (In Chinese.) 346 Liu and Li Fan, J. P. H. & Wong, T. J. (2002). Corporate ownership structure and the informativeness of accounting earnings in East Asia. Journal of Accounting and Economics, 33, 401–425. Fan, J. P. H., Wong, T. J., & Zhang, T. (2007). Politically connected CEOs, corporate gover- nance, and post-IPO performance of China’s newly partially privatized Firms. Journal of Financial Economics, 84, 330–357. Fan, J. P. H., Wong, T. J., & Zhang, T. (2013). Institutions and organizational structure: The case of state-owned corporate pyramids. Journal of Law, Economics, and Organization, 29, 1253– Firth, M., Rui, O. M., & Wu, W. (2011). The effects of political connections and state ownership on corporate litigation in China. Journal of Law & Economics, 54, 573–607. Gao, L. & Song, S. (2007). Corporate governance and corporate transparency. Journal of Financial Research,(11), 28–44. (In Chinese.) Groves, T., Hong, Y., McMillan, J., & Naughton, B. (1994). Autonomy and incentives in Chinese state enterprises. Quarterly Journal of Economics, 109, 183–209. Groves, T., Hong, Y., McMillan, J., & Naughton, B. (1995). China’s evolving managerial labor market. Journal of Political Economy, 103, 873–892. Harris, J., & Bromiley, P. (2007). Incentives to cheat: The inﬂuence of executive compensation and ﬁrm performance on ﬁnancial misrepresentation. Organization Science, 18, 350–367. Hong, H., Hu, X., & Guo, C. (2012). Research on the identiﬁcation of listed companies’ ﬁnancial reporting fraud: Based on GONG theory. Accounting Research,(8), 84–90. (In Chinese.) Huang, Q., Cheng, M., Li, W., & Wei, M. (2014). Listing approach, political favours and earn- ings quality: Evidence from Chinese family ﬁrms. China Journal of Accounting Studies, 2, 13–36. Hung, M., Wong, T. J., & Zhang, T. (2012). Political considerations in the decision of Chinese SOEs to list in Hong Kong. Journal of Accounting and Economics, 53, 435–449. Hung, M., Wong, T. J., & Zhang, F. (In press). The value of political ties versus market credibil- ity: Evidence from corporate scandals in China. Contemporary Accounting Research. Jensen, M. C., & Meckling, W. H. (1976). Theory of the ﬁrm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics, 3, 305–360. Jian, M. & Wong, T. J. (2010). Propping through related party transactions. Review of Accounting Studies, 15,70–105. Jiang, G., Lee, C. M. C., & Yue, H. (2010). Tunneling through intercorporate loans: The China experience. Journal of Financial Economics, 98,1–20. Lennox, C., & Pittman, J. A. (2010). Big ﬁve audits and accounting fraud. Contemporary Accounting Research, 27, 209–247. Leuz, C., Nanda, D., & Wysocki, P. D. (2003). Earnings management and investor protection: An international comparison. Journal of Financial Economics, 69, 505–527. Li, Z., Xin, X., & Yu, X. (2008). Financial development, debt ﬁnancing constraint and the pyra- midal structure. Management World, (1), 123–135. (In Chinese.) Liu, H., & Li, X. (2012). Pyramidal structure, tax burden, and ﬁrm value: Evidence from local state-owned enterprises. Management World,(8), 91–105. (In Chinese.) Liu, Q., Luo, L., Zhang, Y., & Chen, H. (2013). Concentration of managerial power, internal con- trol, and accounting information quality. Nankai Business Review, (16), 15–23. (In Chinese.) Liu, Q., Zheng, Y., & Zhu, Y. (2010). The evolution and consequence of Chinese pyramids. Working Paper. Peking University, China, Sun Yat-Sen University, China, and Morgan Stanley. Piotroski, J. D., & Wong, T. J. (2012). Institutions and information environment of Chinese listed ﬁrms. In J.P.H. Fan & R. Morck (Eds.), Capitalizing China, (pp. 201–242). University of Chicago Press. Piotroski, J. D., Wong, T. J., & Zhang, T. (2015). Political incentives to suppress negative infor- mation: Evidence from Chinese listed ﬁrms. Journal of Accounting Research, 53, 405–459. Qian, Y. (1996). Enterprise reform in China: Agency problems and political control. Economics of Transition, 4, 427–447. Quan, X., Wu, S., & Wen, F. (2010). Managerial power, private income, and compensation rigging: Evidence from state-owned listed enterprises in China. Economic Research Journal, (11), 73–87. (In Chinese.) Rao, P., & Jiang, G. (2011). Monetary policy ﬂuctuations, bank credit, and accounting conser- vatism. Journal of Financial Research,(3), 51–71. (In Chinese.) China Journal of Accounting Studies 347 Wang, Q., Wong, T. J., & Xia, L. (2008). State ownership, the institutional environment, and auditor choice: Evidence from China. Journal of Accounting and Economics, 46,112–134. Wang, K., & Xiao, X. (2009). Ultimate government control structures and ﬁrm value: Evidence from Chinese listed companies. China Journal of Accounting Research, 2, 101–122. Wu, L. (2009). State ownership, preferential tax, and corporate tax burdens. Economic Research Journal,(10), 109–120. (In Chinese.) Xia, L., & Chen, X. (2007). Marketization, SOE reform strategy, and endogenously determined corporate governance structure. Economic Research Journal,(7), 82–95. (In Chinese.) Zeng, Q., & Chen, X. (2006). State stockholder, excessive employment, and labor cost. Economic Research Journal,(5), 74–86. (In Chinese.) Zhong, H., Ran, M., & Wen, S. (2010). Government intervention, internal control, and corporate investment. Management World,(7), 98–109. (In Chinese.) Zhu, C., & Lee, C.-W. J. (2008). State ownership and accounting conservatism. Accounting Research,(5), 38–45. (In Chinese.) Appendix 1. An example of pyramidal structure of SOE in China – The ownership structure of Chongqing Changan Automobile Corporation (stock code: 000625) at the 2010 ﬁscal year end SASAC 100% China South Industry Group Corporation 100% China Changan Automobile Group 45.71% Chongqing Changan Automobile Corporation Source: The 2010 annual report of Chongqing Changan Automobile Corporation http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png China Journal of Accounting Studies Taylor & Francis http://www.deepdyve.com/lp/taylor-francis/government-decentralisation-and-corporate-fraud-evidence-from-listed-iXWA5dOrLE

Loading next page...

References (65)

L. Bebchuk, Jesse Fried, David Walker (2002)
Managerial Power and Rent Extraction in the Design of Executive Compensation
Corporate Governance: Compensation of Executive & Directors eJournal
(2013)
Concentration of managerial power, internal control, and accounting information quality
Natasha Burns, Simi Kedia (2004)
The Impact of Performance-Based Compensation on Misreporting
Corporate Finance: Governance
K. Wang, Xing Xiao (2009)
Ultimate Government Control Structures and Firm Value:Evidence from Chinese Listed Companies
China journal of accounting research, 2
Joseph Fan, T. Wong, Tianyu Zhang (2007)
Politically connected CEOs, corporate governance, and Post-IPO performance of China's newly partially privatized firms
Journal of Financial Economics, 84
P. Aghion, Jean Tirole (2007)
You have printed the following article : Formal and Real Authority in Organizations
Robert Bushman, Joseph Piotroski, Abbie Smith (2003)
What Determines Corporate Transparency?
Economic Growth
Ming Jian, T. Wong (2010)
Propping through related party transactions
Review of Accounting Studies, 15
Zhicheng Liang (2001)
Evidence from China
(2012)
Research on the identification of listed companies' financial reporting fraud: Based on GONG theory
Q. Cheng, Ting Luo, Heng Yue (2013)
Managerial Incentives and Management Forecast Precision
S&P Global Market Intelligence Research Paper Series
Jared Harris, P. Bromiley (2006)
Incentives to Cheat: The Influence of Executive Compensation and Firm Performance on Financial Misrepresentation
SEIN Social Impacts of Business eJournal
(2010)
Managerial power, private income, and compensation rigging: Evidence from state-owned listed enterprises in China
Joseph Fan, T. Wong, Tianyu Zhang (2013)
Institutions and Organizational Structure: The Case of State-Owned Corporate Pyramids
Journal of Law Economics & Organization, 29
(2011)
NBEI index of marketization of China’s provinces 2011 report
(2010)
Government intervention, internal control, and corporate investment
C. Xinyuan, L. Mochou, X. Lijun (2010)
Judiciary Independence and The Enforcement of Investor Protection Laws: Market Responses to the "1/15" Notice of the Supreme People's Court of China
China Economic Quarterly
C. Armstrong, Alan Jagolinzer, D. Larcker (2009)
Chief Executive Officer Equity Incentives and Accounting Irregularities
CGN: Executive Compensation (Topic)
(1997)
Formal and real authority in organizations
, 105
(2008)
Government intervention, pyramidal structures, and investments of local state-owned listed firms
Qian Wang, T. Wong, Lijun Xia (2008)
State ownership, the institutional environment, and auditor choice: Evidence from China
Journal of Accounting and Economics, 46
L. Xin (2011)
Government Quality,Social Capital and Endogenous Determinants of Pyramid Structure of Family Business
China Industrial Economy
Joseph Piotroski, T. Wong (2012)
Institutions and Information Environment of Chinese Listed Firms
C. Armstrong, D. Larcker, G. Ormazabal, Daniel Taylor (2012)
The Relation between Equity Incentives and Misreporting: The Role of Risk-Taking Incentives
Managerial Accounting eJournal
Robert Bushman, Joseph Piotroski (2005)
Financial Reporting Incentives for Conservative Accounting: The Influence of Legal and Political Institutions
International Finance eJournal
Gong-meng Chen, M. Firth, Daniel Gao, Oliver Rui (2006)
Ownership structure, corporate governance, and fraud: Evidence from China
Journal of Corporate Finance, 12
Joseph Piotroski, T. Wong, Tianyu Zhang (2015)
Political Incentives to Suppress Negative Information: Evidence from Chinese Listed Firms
Journal of Accounting Research, 53
Q. Cheng, Kin Lo (2006)
Insider Trading and Voluntary Disclosures
Corporate Finance: Governance
(2008)
Financial development, debt financing constraint and the pyramidal structure
Donghua Chen, Yinying Chen, Oliver Li, Chenkai Ni (2013)
Foreign Residency Rights and Corporate Frauds
Corporate Governance & Accounting eJournal
Mingyi Hung, T. Wong, Tianyu Zhang (2012)
Political Considerations in the Decision of Chinese SOEs to List in Hong Kong
Journal of Accounting and Economics, 53
J. Aharony, J. Aharony, Chi-Wen Lee, T. Wong (2000)
Financial Packaging of IPO Firms in China
Corporate
Patricia Dechow, Weili Ge, Catherine Schrand (2010)
Understanding Earnings Quality: A Review of the Proxies, Their Determinants and Their Consequences
Journal of Accounting and Economics, 50
(2009)
State ownership, preferential tax, and corporate tax burdens
Y. Qian (1996)
Enterprise reform in China: agency problems and political control
Economics of Transition, 4
(2011)
Monetary policy fluctuations, bank credit, and accounting conservatism
(2010)
The evolution and consequence of Chinese pyramids
(2012)
Pyramidal structure, tax burden, and firm value: Evidence from local state-owned enterprises
C. Lennox, Jeffrey Pittman (2010)
Big Five Audits and Accounting Fraud
Contemporary Accounting Research, 27
Guohua Jiang, Charles Lee, Heng Yue (2010)
Tunneling through intercorporate loans: The China experience☆
Journal of Financial Economics, 98
Jap Efendi, Anup Srivastava, E. Swanson (2007)
Why Do Corporate Managers Misstate Financial Statements? the Role of Option Compensation and Other Factors
Corporate Finance: Governance
Franklin Allen, Jun Qian, Meijun Qian (2004)
Law, Finance, and Economic Growth in China
Law & Economics
Joseph Fan, T. Wong (2000)
Corporate Ownership Structure and the Informativeness of Accounting Earnings in East Asia
AFA 2001 New Orleans Meetings (Archive)
M. Beasley (1998)
An Empirical Analysis of the Relation between Board of Director Composition and Financial Statement Fraud
Mingyi Hung, T. Wong, Fang Zhang (2015)
The value of political ties versus market credibility: : Evidence from corporate scandals in China
Contemporary Accounting Research, 32
Hanwen Chen, Jeff Chen, Gerald Lobo, Yanyan Wang (2010)
Association Between Borrower and Lender State Ownership and Accounting Conservatism
CGN: Disclosure & Accounting Decisions (Topic)
J. Aharony, Jiwei Wang, Hongqi Yuan (2010)
Tunneling as an incentive for earnings management during the IPO process in China
Journal of Accounting and Public Policy, 29
(2007)
Marketization, SOE reform strategy, and endogenously determined corporate governance structure
S. Claessens, Simeon Djankov, Joseph Fan, Larry Lang (2002)
Disentangling the Incentive and Entrenchment Effects of Large Shareholdings
Journal of Finance, 57
Anup Agrawal, Sahiba Chadha (2003)
Corporate Governance and Accounting Scandals
Social Science Research Network
Merle Erickson, Michelle Hanlon, Edward Maydew (2006)
Is There a Link between Executive Equity Incentives and Accounting Fraud?
Corporate Finance: Governance
Daniel Bergstresser, Thomas Philippon (2004)
CEO Incentives and Earnings Management
Corporate Governance & Accounting eJournal
Michael Jensen, W. Meckling (1976)
Theory of the Firm
(2006)
State stockholder, excessive employment, and labor cost
Joseph Fan, T. Wong, Tianyu Zhang (2014)
Politically Connected CEOs, Corporate Governance, and the Post‐IPO Performance of China's Partially Privatized Firms
Corporate Finance: Capital Structure & Payout Policies eJournal
R. Kalra (2004)
Earnings Management and Investor Protection: An International Comparison
Cfa Digest, 34
Theodore Groves, Yongmiao Hong, J. Mcmillan, Barry Naughton (1994)
Autonomy and Incentives in Chinese State Enterprises
Quarterly Journal of Economics, 109
(2011)
Media supervision, government intervention, and corporate governance: Evidence from the perspective of Chinese listed firms’ financial restatements
Jun Bai, Lishuai Lian (2013)
Why do state-owned enterprises over-invest? Government intervention or managerial entrenchment
China Journal of Accounting Studies, 1
M. Firth, Gong-meng Chen, N. Gao, Oliver Rui (2005)
Is China's Securities Regulatory Agency a Toothless Tiger? Evidence from Enforcement Actions
Corporate Law: Securities Law
Qiongyu Huang, M. Cheng, Wenjing Li, Minghai Wei (2014)
Listing approach, political favours and earnings quality: Evidence from Chinese family firms
China Journal of Accounting Studies, 2
Donghua Chen, D. Jiang, Shangkun Liang, Fangping Wang (2011)
Selective enforcement of regulation
China journal of accounting research, 4
Patricia Dechow, Richard Sloan, A. Sweeney (1996)
Causes and Consequences of Earnings Manipulation: An Analysis of Firms Subject to Enforcement Actions by the SEC*
Contemporary Accounting Research, 13
M. Firth, Oliver Rui, Wenfeng Wu (2011)
The Effects of Political Connections and State Ownership on Corporate Litigation in China
The Journal of Law and Economics, 54
Theodore Groves, Yongmiao Hong, J. Mcmillan, Barry Naughton, Yongmiao Hong, Research Support, M. Aoki, T. Hoshi, J. Litwack, M. Machina, Y. Qian, V. Ramey
University of California, San Diego ___________________________________________ Department of Economics China's Evolving Managerial Labor Market China's Evolving Managerial Labor Market

Publisher: Taylor & Francis
Copyright: © 2015 Accounting Society of China
ISSN: 2169-7221
eISSN: 2169-7213
DOI: 10.1080/21697213.2015.1100090
Publisher site: See Article on Publisher Site

Abstract

China Journal of Accounting Studies, 2015 Vol. 3, No. 4, 320–347, http://dx.doi.org/10.1080/21697213.2015.1100090 Government decentralisation and corporate fraud: Evidence from listed state-owned enterprises in China a,b c Hang Liu * and Xiaorong Li a b School of Accountancy, Dongbei University of Finance and Economics, Dalian, China; China Internal Control Research Center, Dalian, China; School of Public Finance / China’s Public Finance Development Synergetic Innovation Center, Central University of Finance and Economics, Beijing, China This paper examines the economic consequences of government decentralisation from the perspective of corporate fraud. Theoretically, government decentralisation reduces the political costs of state intervention and hence decreases the probability of state- owned enterprises (SOEs) to engage in fraud. It also aggravates the agency costs (the costs of managerial self-dealing), thereby increasing the probability of SOEs to com- mit fraud. Using pyramidal layers as a proxy of government decentralisation for SOEs, empirical results show that government decentralisation signiﬁcantly lowers the probability of SOEs to commit fraud. Further categorisation of types of fraud shows that government decentralisation primarily deters disclosure-related fraud and market transaction-related fraud. Finally, the effect of decentralisation on corporate fraud is more pronounced for SOEs in which government intervention is more likely. Keywords: agency costs; fraud; government decentralisation; political costs; pyramidal layers 1. Introduction Since the reform and opening up of China, multiple reforms have been conducted by the Chinese government upon SOEs to improve their operational efﬁciency. In retro- spect, although speciﬁc measures vary, the key has always been decentralisation. Since the 1990s, the Chinese government has proposed a reform plan for SOEs to construct the modern enterprise system. During this reform period, a large number of state-owned assets have been decoupled from government agencies to form new companies. Although the ultimate controllers of these new companies are within the sphere of gov- ernment, actual operational power has been gradually decentralised to specialised state- owned asset management companies or to the SOEs themselves through pyramidal ownership structures. Government intervention in ﬁrms is greatly decreased by this measure (Fan, Wong, & Zhang, 2013). This paper tries to explore the effect of govern- ment decentralisation, in the form of pyramid structures, upon the probability of SOEs to commit fraud, beginning in the 1990s. The economic consequences of the above-mentioned reform are considered controversial in the current literature. Qian (1996) argues, by theoretical analysis, that government decentralisation reduces the political costs of government intervention but *Corresponding author. Email: liuhang@dufe.edu.cn Paper accepted by Donghua Chen. © 2015 Accounting Society of China China Journal of Accounting Studies 321 increases agency costs (the costs of managerial self-dealing). This argument is empirically conﬁrmed by Fan et al. (2013). Adopting the proxy of pyramidal layers to measure the degree of decentralisation, the latter ﬁnd that political and agency costs jointly determine the degree of decentralisation, and government decentralisation improves ﬁrm output and employment efﬁciency in SOEs, demonstrating that the reduction in political costs is greater than the increase in agency costs. The same ﬁnd- ing can be observed in Wang and Xiao (2009) from a ﬁrm value point of view, in Cheng, Xia, and Yu (2008) from an investment efﬁciency viewpoint and in Liu and Li (2012) from a corporate tax burden perspective. Nonetheless, Zhong, Ran, and Wen (2010) discover that opportunistic behaviour of managers bred by government decen- tralisation can invalidate the investment decisions of SOEs. Therefore, agency costs cannot be neglected. Based on the data of the Chinese A-share listed SOEs in 2004–2010 and follow- ing the literature (Cheng et al., 2008; Fan et al., 2013; Liu & Li, 2012; Wang & Xiao, 2009; Zhong et al., 2010), this paper adopts the number of pyramidal layers between the ultimate controller and the listed ﬁrm as a proxy of the degree of government decentralisation over SOEs. The paper examines empirically this proxy’s effect on corporate fraud. The results show that pyramidal layers are signiﬁcantly negatively correlated with the probability of corporate fraud, demonstrating that gov- ernment decentralisation signiﬁcantly lowers the probability of SOEs to commit fraud. A detailed categorisation of fraud types, as provided in the fraud database of CSMAR, indicates that fraud reduction occurs via two mechanisms, information disclo- sure and stock market transactions; fraud reduction is not signiﬁcant in capital-related fraud or other types of fraud. A further classiﬁcation by severity of fraud leads to the conclusion that there is no signiﬁcant difference among fraud activities with various levels of severity affected by government decentralisation. Finally, the effect of govern- ment decentralisation upon corporate fraud exists only in SOEs that are more likely to be inﬂuenced by the government because a more severe government intervention is associated with a greater reduction in political costs caused by decentralisation. Thus, the conclusion that government decentralisation decreases the probability of corporate fraud is strengthened. The implications of this paper are the following: (1) Theoretically, government decentralisation reduces the political costs of SOEs but increases the agency costs (Qian, 1996). Both effects are empirically sup- ported (Cheng et al., 2008; Fan et al., 2013; Liu & Li, 2012; Wang & Xiao, 2009; Zhong et al., 2010). In this paper, we ﬁnd that the pyramidal structure of SOEs formed by government decentralisation signiﬁcantly reduces the probabil- ity of SOEs to commit fraud, demonstrating that from the perspective of corpo- rate fraud, the reduction in political costs resulting from decentralisation is signiﬁcant. This result expands the existing literature and supports the decen- tralisation reform of government over SOEs from the viewpoint of accounting information quality. (2) Where government intervenes heavily in SOEs, investors need to exercise more caution because those SOEs tend to be more likely to commit fraud and cause enormous loss to investors (Chen, Firth, Gao, & Rui, 2005; Dechow, Sloan, & Sweeney, 1996; Hung, Wong, & Zhang, in press). 322 Liu and Li The remainder of this paper is structured as follows: Section 2 presents an institutional background and literature review. Section 3 comprises theoretical analysis and hypothesis. Section 4 introduces the research design and lists descriptive statistics. Section 5 discusses empirical results. Section 6 concludes. 2. Institutional background and literature review 2.1. Institutional background: Government decentralisation and pyramidal structures of SOEs Prior to the reform and opening up in 1978, all SOEs in China were directly under the control of either the central or a local government. Managers in SOEs have a very lim- ited autonomy in such operational activities as employment, production and distribu- tion, with all decisions ﬂowing from government agencies. SOEs were actually only attachments to government agencies and were extremely inefﬁcient because of low incentive levels for managers and strong political objectives (Fan et al., 2013). After the reform and opening up, to enhance the vitality of SOEs, SOEs were empowered to manage independently to be responsible for their own proﬁts or losses, and to reserve a certain proportion of proﬁt within ﬁrms. These measures incentivise managers in the short run and activate state-owned assets (Groves et al., 1994). How- ever, a lack of ﬁrm operational information and professional operations capacity restrains government from effectively controlling managers’ opportunistic behaviours, such as concealing proﬁts or transferring assets to satisfy private interests (Qian, 1996). Moreover, during this period, state-owned assets continued to be directed in the former Soviet mode of a central planning system and were directly under the control of the government, resulting in severe interventions in ﬁrms by the government (Qian, 1996). To alter this situation, the operation of state-owned assets was transformed, from the 1990s, from the planning system to the market system. During this transformation state-owned assets were decoupled from government agencies (Fan et al., 2013). Con- cerning listed SOEs, the means of government decentralisation are largely the follow- ing: ﬁrst, the government injects capital into newly established listed SOEs and holds the shares of these companies through a state-asset management agency (e.g. State- owned Assets Supervision and Administration Commission of the State Council, SASAC). Alternatively, the government can indirectly control the listed SOEs through pyramidal structures. In this case, there are often intermediate companies between gov- ernment agencies (i.e. the ultimate controllers) and the listed SOEs. The intermediate companies are usually unlisted parent SOEs (Fan et al., 2013). In addition, the number of intermediate companies varies in different pyramidal structures. Appendix 1 lists a typical pyramidal structure of SOEs, in which SASAC, a government agency, 100% controls the China South Industry Group Corporation, which in turn 100% controls the China Changan Automobile Group, which controls the Chongqing Changan Automo- bile Corporation (stock code: 000625) with a shareholding ratio of 47.51%. As tabled, this pyramidal ownership structure distances ﬁrms and government, relieving ﬁrms of direct government control. A greater distance from the top decreases the likelihood of government intervention in the ﬁrms (Aghion & Tirole, 1997; Fan et al., 2013). Thus, the pyramidal structure of SOEs stems from government decentralisation reform and lessens direct government intervention imposed upon SOEs. China Journal of Accounting Studies 323 2.2. Literature review Prior literature explores the economic consequences of government decentralisation theoretically and empirically. Qian (1996) argues, by theoretical analysis, that decentralisation reform of SOEs by the Chinese government confronts two major issues: the political costs of government intervention and agency costs. Speciﬁcally, on the one hand, government decentralisa- tion reduces government intervention in ﬁrms, thereby reducing ﬁrms’ political costs and improving their efﬁciency; on the other hand, government decentralisation empow- ers managers with more autonomy, thereby increasing agency costs and harming ﬁrms’ efﬁciency as the higher information costs resulting from decentralisation make supervis- ing more difﬁcult. Therefore, Qian believes that an effective SOE reform strategy lies in the establishment of an effective corporate governance system as part of the govern- ment decentralisation reform to reduce ﬁrms’ political costs and inhibit agency costs. Qian’s theoretical analysis is conﬁrmed empirically by Fan et al. (2013), who study the factors and economic consequences of government decentralisation using the exam- ple of government decentralisation in the form of pyramidal structures in China since the 1990s. Adopting pyramidal layers to depict the degree of decentralisation, the authors ascertain that political and agency costs jointly determine the degree of decen- tralisation and that government decentralisation improves the professionalism of SOE managers, production efﬁciency, and the operational performance of employees, demon- strating that the reduction in political costs is greater than the increase in agency costs on average. Later studies demonstrate the positive economic consequences from government decentralisation. Cheng et al. (2008) ﬁnd that it reduces excessive investment in SOEs. Wang and Xiao (2009) observe that a higher degree of government decentralisation over SOEs is associated with a greater valuation of ﬁrms. Liu and Li (2012) show that SOEs with a higher degree of government decentralisation tend to have a lower policy burden (i.e. a lower corporate tax burden). With regard to agency costs, Zhong et al. (2010) demonstrate that government decen- tralisation at a low level can reduces political costs and constrain ﬁrms’ over-investment of free cash ﬂow. However, a further increase in government decentralisation will induce a signiﬁcant rise of agency costs, resulting in an inverse U-shaped relationship between the degree of government decentralisation and ﬁrms’ over-investment of free cash ﬂow. Quan, Wu, and Wen (2010) show that government decentralisation is signiﬁcantly positively correlated with excess cash compensation to SOE managers by utilising the pyramidal layers as one ingredient of SOE managers’ power. All the evidence points to the profound effect upon ﬁrms’ economic activities of agency costs derived from government decentralisation. Based on the above literature, this paper examines the effect of government decen- tralisation in the form of pyramidal structures upon the probability of SOEs to commit fraud. 3. Theoretical analysis and research hypothesis Aghion and Tirole (1997) argue that a trustworthy decentralisation can be achieved by a pyramidal structure without ownership transfer. Compared with a government promise not to interfere, the pyramidal structure is far more reliable because its long intermedi- ate chains and sophistication require more costs for government to interfere with the 324 Liu and Li daily operation of ﬁrms. Moreover, there will be distortion and selective deviation of information transmission via intermediate chains within the structure, making it more difﬁcult to interfere. Therefore, a pyramidal structure can effectively reduce administra- tive intervention of government in SOEs and thereby decrease political costs (Fan et al., 2013; Qian, 1996). However, it is not costless to form the structure. On the one hand, although opera- tional management of SOEs is decentralised, government remains the owner of SOEs. The agent conﬂict between principal (government) and agent (managers) motivates managers to pursue opportunistic behaviour; on the other hand, within the structure, the challenge to gather information and the complexity of the organisation deter govern- ment from effective supervision. Thus, agency costs are increased by the decentralisa- tion (Qian, 1996). Therefore, the pyramidal structure brought by government decentralisation reform not only reduces government intervention in SOEs but also increases agency costs. This paper conducts its theoretical analysis from these two perspectives. 3.1. Analysis from the perspective of government intervention Low ﬁrm-speciﬁc ﬁnancial transparency has always been attributed largely to govern- ment intervention (Bushman, Piotroski, & Smith, 2004; Bushman & Piotroski, 2006). International evidence demonstrates that there is a relatively low level of ﬁnancial transparency for ﬁrms within a nation in which the political economy is characterised by high state ownership of ﬁrms, high state ownership of banks, and a high risk of state expropriation (Bushman et al., 2004). There is also a low level of accounting con- servatism for ﬁrms in a nation with greater risks of being appropriated by government or a larger share of SOEs (Bushman & Piotroski, 2006). For China, there are at least the following three procedures through which government intervention will affect the information quality of listed SOEs. During listing, government intervention damages the information quality of ﬁrms and boosts their fraud probability (Aharony, Lee, & Wong, 2000). The quantity of listed ﬁrms is stringently regulated by a security supervision agency; therefore, for a ﬁrm to be listed requires a certain performance threshold and for it to pass the censor- ship of the supervision agency, consequently causing the listing qualiﬁcation to become a scarce resource (Huang, Cheng, Li, & Wei, 2014). Additionally, an SOE listing ignites tremendous beneﬁts, including post-listing ﬁnancing for ﬁrms, enhanced reputa- tion with the government, and accumulation of ofﬁcials’ political capital (Hung, Wong, & Zhang, 2012). In striving for the qualiﬁcation, SOEs will be ﬁnancially packaged by government (Aharony et al., 2000) and receive proﬁts through related-party transactions by government (Aharony, Wang, & Yuan, 2010), leading to a severely distorted quality of information about SOEs. Information quality will be further distorted during the daily operation of SOEs as follows. (1) In the scenario of government intervention, the operational goal of SOEs is not to maximise the interests of shareholders but rather to shoulder partial policy bur- dens such as more taxes (Wu, 2009), more employment (Zeng & Chen, 2006), and more over-investment (Bai & Lian, 2013), all to the detriment of SOEs’ performance. To cover the bad news of deteriorating performance, preserve governmental reputation and obtain reﬁnancing qualiﬁcation, another round of ﬁnancial packaging will be pursued (Piotroski & Wong, 2012; Piotroski, Wong, & Zhang, 2015), biasing the information quality of SOEs and culminating in corporate fraud (Gao & Song, 2007; China Journal of Accounting Studies 325 Zhu & Lee, 2008). (2) A lack of many external monitoring mechanisms for SOEs also contributes to low information quality, speciﬁcally embodied in the following points: ﬁrst, as stated previously, policy burdens drag down SOEs’ performance, with the gov- ernment frequently paying the loss (Piotroski & Wong, 2012). The resulting soft budget constraint lessens the restraining effect of liability upon SOEs and lowers creditors’ demand for high-quality ﬁnancial information (Chen, Chen, Lobo, & Wang, 2010; Rao & Jiang, 2011). Second, government intervention deprives SOEs of the demand for high-quality external auditing; consequently, SOEs often hire local small-scale account- ing ﬁrms of low quality (Wang, Wong, & Xia, 2008), reducing the role of auditors in restraining corporate fraud (Lennox & Pittman, 2010) and practically facilitating SOEs to commit fraud. (3) Dai, Pang, and Liu (2011) discover that government intervention also debilitates the role of media supervision over corporate fraud. All the evidence supports the conclusion that the information quality of SOEs is damaged and that the role of supervision mechanisms in constraining corporate fraud is limited by govern- ment intervention. After the occurrence or exposure of SOEs’ fraud, the very existence of government intervention weakens the degree of punishment for SOEs, decreasing SOEs’ cost of violating rules and practically contributing to SOEs’ fraud. Speciﬁcally, Chen, Jiang, Liang, and Wang (2011) prove that there is an issue of selective law enforcement in Chinese supervision agencies. Because SOEs shoulder partial political costs, SOEs are penalised more leniently after fraud exposure. Moreover, Chen, Li, Rui, and Xia (2009) and Firth, Rui, and Wu (2011) show that the umbrella of government background lav- ishes privileges upon SOEs from supervision or judiciary agencies. Speciﬁcally, when facing possible civil litigation, investors believe that trial results are more favourable to SOEs; thus, stock prices of SOEs decrease less (Chen et al., 2009). SOEs can also have better results in appeals (Firth et al., 2011). Succinctly, extant government intervention increases the fraud probability of SOEs. Therefore, with the deepening of decentralisation reform, the degree of government intervention in SOEs will be smaller, decreasing the probability of fraud. On this point, there is a signiﬁcantly negative correlation between the degree of government decentral- isation and the probability of corporate fraud. We term this effect the ‘government intervention hypothesis’. 3.2. Analysis from the perspective of agency costs Managers’ self-interested behaviour has always been an important determinant of cor- porate fraud (Dechow, Ge, & Schrand, 2010). Prior literature proves that from multiple perspectives. One perspective is to examine the inﬂuence of managers’ incentives upon corporate fraud. Although managers’ compensation incentives are considered as an important mechanism to relieve the agency problem (Jensen & Meckling, 1976), they also embody the agent issue and induce managers’ self-interested behaviour (Bebchuk, Fried, & Walker, 2002). Burns and Kedia (2006), Efendi, Srivastava, and Swanson (2007), Harris and Bromiley (2007), and Hong, Hu, and Guo (2012) all discover that managers’ compensation incentives signiﬁcantly increase the probability of fraud in ﬁrms. Bergstresser and Philippon (2006) observe that equity incentives stimulate earn- ings management of ﬁrms because a high level of incentives enables managers to amass private beneﬁts by such methods as insider trading, publishing false information and adjusting earnings to manipulate stock prices. Erickson, Hanlon, and Maydew 326 Liu and Li (2006) and Armstrong, Jagolinzer, and Larcker (2010) do not ﬁnd relevant evidence, but Armstrong, Larcker, Ormazabal, and Taylor (2013) conﬁrm that a stronger correla- tion between managers’ wealth and company risk is associated with a higher probabil- ity to commit fraud. The conclusion is consistent with managers’ self-interested behaviour resulting in a higher probability of corporate fraud. The other angle is to consider the relationship between corporate governance and the probability of corporate fraud. The potential theoretical hypothesis is that a sound corporate governance system can effectively inhibit managers’ agency problem, thereby reducing corporate fraud. For instance, Beasley (1996) and Chen, Firth, Gao, and Rui (2006) ﬁnd that a higher proportion of external directors, particularly with a ﬁnance background (Agrawal & Chadha, 2005), can reduce ﬁrms’ probability to commit fraud. Dechow et al. (1996) show that when the CEO is simultaneously chairman of the board, there is a higher probability of corporate fraud. Lennox and Pittman (2010) observe that the employment of the international Big Five public accounting ﬁrms, rep- resenting high audit quality, signiﬁcantly reduced corporate fraud. Furthermore, Liu, Luo, Zhang, and Chen (2013) conclude that although sound internal control can reduce fraud, executive power centralisation weakens the supervisory role of internal control over managers’ irregularities. To summarise, the evidence consistently indicates that managers’ opportunistic behaviours can increase the probability of corporate fraud. After decentralisation reform, the government is unable to conduct face-to-face supervision of managers and the complexity of a pyramidal structure weakens the governmental supervision effect (Qian, 1996), all possibly contributing to the increasing agency costs of SOEs and thus breeding a higher fraud probability. From this standpoint, there is a signiﬁcantly posi- tive correlation between the degree of government decentralisation and corporate fraud probability. We term this effect the ‘agency costs hypothesis’. Theoretically, there are two distinct hypotheses concerning the inﬂuence of govern- ment decentralisation upon corporate fraud. Which is the dominant one? This paper tries to answer this question empirically. 4. Research design and descriptive statistics 4.1. Empirical model and variables deﬁnition To investigate the inﬂuence of government decentralisation on corporate fraud, the fol- lowing empirical model is deﬁned: ProbðÞ FRAUD ¼ 1¼ a þ a LAYER þ a LNMCAP þ a ROA þ a LEV it 0 1 i;t 2 i;t 3 i;t 4 i;t þa MB þ a BIG4 þ a MANHOLD þ a INDEP þ 5 i;t 6 i;t 7 i;t 8 it a DUALITY þ a LNBSIZE þ a HBSHARE þ 9 i;t 10 i;t 11 i;t (1) a MEETING þ a LOSS þ a ST þ a GOV SHARE 12 i;t 13 i;t 14 i;t 15 i;t þa GOV INTERV þ a TOP1 þ a RPT 16 i;t 17 i;t 18 i;t P P þa ORECTA þ Industry þ Year þ e 19 i;t i;t In Model (1), FRAUD is the dependent variable, indicating whether ﬁrm i is engaged in fraud in year t. To depict this variable, the method commonly found in investigation of Chinese companies is adopted, i.e. downloading all the listed companies’ fraud information from the fraud database in China Stock Market & Accounting Research (CSMAR). We can observe in which year the listed ﬁrms engaged in fraud under the China Journal of Accounting Studies 327 title of ‘the year of violation’ in the database. When a listed ﬁrm commits fraud in the current year, FRAUD equals 1, and otherwise 0. The ﬁrst independent variable of Model (1) is LAYER, describing the degree of gov- ernment decentralisation over SOEs. Following Fan et al. (2013), Cheng et al. (2008), Liu and Li (2012), Wang and Xiao (2009), and Zhong et al. (2010), the number of pyramidal layers from the ultimate controller to the listed SOE is utilised to measure the degree of governmental decentralisation. Speciﬁcally, when the ultimate controller directly controls the listed company, the pyramidal layer is 1. If there is an intermediate controller between them, the pyramidal layer is 2, and so on. To control other factors affecting corporate fraud, following Chen et al. (2006) and Chen, Chen, Li, and Ni (2013), these control variables are included in the model: LNMCAP is the natural logarithm of a ﬁrm’s market capitalisation; the higher political cost of a large-size ﬁrm lowers the probability of fraud. ROA, denoting the ﬁrm’s prof- itability, is equal to net income divided by total assets. Firms with better performance have weaker motivation to manipulate earnings and make misstatements; thus, they are less likely to commit fraud. LEV, ﬁrm leverage, is equal to the total liabilities divided by the total assets. Highly leveraged ﬁrms with a higher bankruptcy cost are more likely to be involved in fraud. However, they will also be strictly supervised by credi- tors, thereby lowering the probability of fraud. MB, market to book ratio, is equal to the total market value of equity divided by the book value of equity. To obtain continu- ous ﬁnancing, the high-growth ﬁrm may manipulate earnings through fraud activities. BIG4 indicates whether the ﬁrm has hired an international big four auditing ﬁrm. If yes, BIG4 equals 1, otherwise 0. Lennox and Pittman (2010) ﬁnd that ‘big ﬁve’ audit- ing can effectively reduce corporate fraud probability; therefore, we control for this variable. Next, we control for a series of variables related to corporate governance. MANHOLD represents the percentage of management shareholding; INDEP is the pro- portion of independent directors. DUALITY indicates whether the chairman of the board and CEO are the same person (1 = ‘yes’,0 = ‘no’). LNBSIZE, the board size, is equal to the natural logarithm of the number of directors. HBSHARE indicates whether the ﬁrm issues H-shares or B-shares simultaneously (1 = ‘yes’,0 = ‘no’). MEETING is equal to the natural logarithm of the number of board meetings. We also control for two variables closely related to the ﬁrm’s ﬁnancial reporting motivation. In general, a stronger motivation of the ﬁrm to manipulate ﬁnancial reports is associated with a higher probability of fraud. The two variables are LOSS, denoting whether there is a loss for the ﬁrm in the previous period (1 = ‘yes’,0 = ‘no’), and ST, indicating whether the ﬁrm is specially treated (1 = ‘yes’,0 = ‘no’). Companies with a previous loss and ST have a greater motivation to commit fraud. To control other possible mechanisms of government intervention besides pyramidal layers inﬂuencing corporate fraud, two control variables are incorporated: GOV_SHARE, standing for the proportion of state- owned shares of the company, and GOV_INTERV, denoting the degree of regional gov- ernment intervention, measured by the index of ‘reducing government intervention over the enterprise’ from Fan, Wang, and Zhu (2011). In addition, this paper adopts pyrami- dal layers as a proxy of the degree of government decentralisation over SOEs. How- ever, the literature indicates that the pyramidal ownership structure itself may lead to a series of agency problems, reducing the degree of ﬁrms’ information asymmetry, which may have a signiﬁcant effect on corporate fraud (Fan & Wong, 2002; Leuz, Nanda, & Wysocki, 2003). Thus, we include two variables to control for the agency problem related to the pyramidal structure: RPT, representing the degree of a ﬁrm’s related-party transactions, is equal to the total amount of the ﬁrm’s related-party transactions between 328 Liu and Li the listed ﬁrm and the parent ﬁrm divided by the total assets of the listed ﬁrm. ORECTA denotes the degree of tunnelling by the controlling shareholder and is equal to inter-corporate loans by the controlling shareholder in the other receivables divided by the total assets (Jiang, Lee, & Yue, 2010). We also include the percentage of shares held by the largest shareholder (TOP1) to control for the possible effect of ownership concentration on corporate fraud. Finally, we control for year (YEAR) and industry (INDUSTRY) ﬁxed effects. 4.2. Sample and data We select all A-share listed SOEs from 2004 to 2010 as the initial sample. The sample interval begins in 2004 because the CSMAR database has collected the ownership structure ﬁgures of listed ﬁrms in China from their annual reports since 2004, through which the core variable LAYER is manually gathered. The interval ends in 2010 because the CSMAR database identiﬁes corporate fraud retrospectively as it emerges. More recent years may be less complete in data. This research uses only SOEs as the sample ﬁrms because this paper studies the economic consequences of government decentralisation, which targets only SOEs. Excluding the observations of the ﬁnancial industry and with missing data from the initial sample, our ﬁnal sample includes 6,111 observations representing 1,147 unique ﬁrms. Concerning data sources, data on pyramidal layers come from the authors’ manual collection, the data of inter-corporate loans by the controlling shareholder in the other receivables from the WIND database, and other data from the CSMAR database. 4.3. Descriptive statistics Descriptive statistics of variables are presented in Table 1. To eliminate the effect of outliers, we winsorise all continuous variables at the 1% and 99% levels. Panel A of Table 1 shows that the average of FRAUD is 0.097, meaning that 9.7% of the sample companies are engaged in fraud. The proportion is higher than 7.4% in the descriptive statistics of Chen et al. (2013). This is the result of corporate fraud being detected later; thus, in conjunction with the continuous renewal of the fraud database, fraud not previ- ously reported in the company may also be observed. Panel B of Table 1 lists the dis- tribution of LAYER. The results show that only 6.1% of the pyramidal layer equals 1 and that most of the pyramidal layer is 2. This demonstrates that in this sample, gov- ernment decentralisation is rather common. Panel C of Table 1 indicates that LAYER changes annually. The results show that the annual average LAYER has a steadily rising annual trend, consistent with the reality of SOE reform continuously deepening, which, to some extent, indicates that it is appropriate to adopt pyramidal layers as a proxy for decentralised government. To investigate government decentralisation affecting corporate fraud, the sample is ﬁrst divided into four parts according to the pyramidal layers, with the number of the layers, respectively equal to 1, 2, 3, and greater than or equal to 4. The probability of fraud in each of the four parts is reported in Figure 1(a). Figure 1(a) shows that corpo- rate pyramidal layers for 1 and 2 entail a very close fraud probability. When the num- ber of pyramidal layers rises to 3, the probability of corporate fraud decreases to 9%. When the number is greater than or equal to 4, the probability drops to 7.6%, with a decreased amount of approximately 23% (1 – 7.6% / 9.9%) compared with companies China Journal of Accounting Studies 329 Table 1. Descriptive statistics. Panel A: Firm characteristics Variables N Mean STD Q1 Median Q3 FRAUD 6,111 0.097 0.296 0.000 0.000 0.000 FRAUD_DISCLOSE 6,111 0.080 0.272 0.000 0.000 0.000 FRAUD_MARKET 6,111 0.015 0.122 0.000 0.000 0.000 FRAUD_CAPITAL 6,111 0.024 0.153 0.000 0.000 0.000 FRAUD_OTHER 6,111 0.058 0.235 0.000 0.000 0.000 LAYER 6,111 2.452 0.887 2.000 2.000 3.000 LNMCAP 6,111 21.869 1.213 20.986 21.729 22.562 ROA 6,111 0.027 0.068 0.010 0.029 0.056 LEV 6,111 0.521 0.205 0.380 0.526 0.652 MB 6,111 3.226 3.069 1.447 2.312 4.005 BIG4 6,111 0.095 0.294 0.000 0.000 0.000 MANHOLD 6,111 0.001 0.009 0.000 0.000 0.000 INDEP 6,111 0.353 0.050 0.333 0.333 0.364 DUALITY 6,111 0.099 0.299 0.000 0.000 0.000 LNBSIZE 6,111 2.247 0.205 2.197 2.197 2.398 HBSHARE 6,111 0.111 0.314 0.000 0.000 0.000 MEETING 6,111 2.055 0.377 1.792 2.079 2.303 LOSS 6,111 0.113 0.317 0.000 0.000 0.000 ST 6,111 0.056 0.230 0.000 0.000 0.000 GOV_SHARE 6,111 0.348 0.233 0.146 0.377 0.533 GOV_INTERV 6,111 6.459 3.126 4.330 6.780 8.820 TOP1 6,111 0.408 0.159 0.282 0.403 0.528 RPT 6,111 0.124 0.234 0.000 0.025 0.152 ORECTA 6,111 0.002 0.013 0.000 0.000 0.000 Panel B: Distribution of pyramidal layers LAYER No. of OBS =1 373 6.10 =2 3,499 57.26 =3 1,607 26.30 ≥4 632 10.34 Panel C: Pyramidal layers by year Year Mean Year Mean 2004 2.276 2008 2.489 2005 2.342 2009 2.600 2006 2.374 2010 2.671 2007 2.439 FRAUD is an indicator that equals 1 if a listed ﬁrm commits fraud in the current year, otherwise 0; FRAUD_- DISCLOSE is an indicator that equals 1 if a listed ﬁrm commits fraud in the current year and the fraud is related to information disclosure (fabricating proﬁt, ﬁctitious assets, false statements, delayed disclosure, major omissions, untrue disclosure, dishonest listings, and general accounting misconduct), otherwise 0; FRAUD_MARKET is an indicator that equals 1 if a listed ﬁrm commits fraud in the current year and the fraud is related to market transactions (insider trading, illegal stock purchases, and manipulating stock prices), other- wise 0; FRAUD_CAPITAL is an indicator that equals 1 if a listed ﬁrm commits fraud in the current year and the fraud is related to corporate capital (illegal investments, unauthorised changes in capital use, the occupa- tion of listing ﬁrms’ assets by the controlling shareholders, and illegal guarantees), otherwise 0; FRAUD_- OTHER is an indicator that equals 1 if a listed ﬁrm commits fraud in the current year and the fraud is of other types, otherwise 0; LAYER is the proxy of the degree of government decentralisation over SOEs, deﬁned as the number of pyramidal layers from the ultimate controller to the listed SOE; LNMCAP is the natural log- arithm of a ﬁrm’s market capitalization; ROA is deﬁned as the net income divided by the total assets; LEV is deﬁned as the total liabilities divided by the total assets; MB is deﬁned as the total market value of equity divided by the book value of equity; BIG4 is an indicator that equals 1 if a ﬁrm has hired an international ‘big four’ auditing ﬁrm, otherwise 0; MANHOLD is the percentage of management shareholding; INDEP is the percentage of independent directors; DUALITY is an indicator that equals 1 if a ﬁrm’s chairman of the board and its CEO is the same person, otherwise 0; LNBSIZE is the natural logarithm of the number of direc- tors; HBSHARE is an indicator that equals 1 if a ﬁrm issues H-shares or B-shares simultaneously, otherwise 0; MEETING is the natural logarithm of the number of board meetings; LOSS is an indicator that equals 1 if there is a loss for the ﬁrm in the previous year, otherwise 0; ST is an indicator that equals 1 if a ﬁrm is trea- ted special, otherwise 0; GOV_SHARE is the percentage of state-owned shares; GOV_INTERV is the degree of regional government intervention, deﬁned as the index of ‘reducing government intervention over the enter- prise’ from Fan et al. (2011); TOP1 is the percentage of shares held by the largest shareholder; RPT is deﬁned as the total amount of a ﬁrm’s related-party transactions between the listed ﬁrm and the parent ﬁrm divided by the total assets of the listed ﬁrm; ORECTA is deﬁned as the inter-corporate loans by the controlling share- holder in the other receivables divided by the total assets. 330 Liu and Li (a) 12% 10.43% 9.92% 10% 9.02% 7.59% 8% 6% 4% 2% 0% Layer=1 Layer=2 Layer=3 Layer>=4 (b) 7% 5.88% 6% 4.97% 4.40% 5% 3.56% 3.44% 4% 3% 2.35% 2% 1% 0% Layer unchanged Layer increased Layer decreased Fraud in the previous year but no fraud in this year No fraud in the previous year but fraud in this year Figure 1. Graphical evidence. (a) Pyramidal layers and corporate fraud probability. (b) Changes of pyramidal layers and changes of corporate fraud probability. when the number of layers is equal to 1. According to the results of the univariate test, the fraud probability of ﬁrms with three layers is signiﬁcantly lower than that of the ﬁrms with layers fewer than 3 (p value is 0.063). However, the fraud probability of the ﬁrms with three, or more than three, pyramidal layers is not signiﬁcantly different (p value is 0.140). Furthermore, we report the changes in corporate fraud probability brought about by corporate pyramidal layers changes. In our sample, the pyramidal layers of 4,259 observations have no changes compared with the previous year; however, those of 523 observations have increased, and those of 255 observations decreased. The rele- vant changes in the fraud probability are collected and shown in Figure 1(b). Figure 1(b) shows that for ﬁrms with no change in pyramidal layers compared with the previous year, their probability of having no fraud the previous year but having fraud this year is 3.56%. Their probability of having fraud the previous year but no fraud this year is 4.40%. A univariate test indicates that the difference is signiﬁcant at the 5% level (p value is 0.026). Firms whose pyramidal layers increased over the previous year have a probability of having no fraud the previous year but having fraud this year of 4.97%, higher than the 3.44% probability of fraud the previous year but no fraud this year. The univariate test notes that the difference is signiﬁcant at the 12% level (p value is 0.114). For ﬁrms with pyramidal layers decreasing com- pared with the previous year, their probability of having no fraud the previous year but fraud this year is 2.35%, lower than the probability of 5.88% of fraud the previ- ous year but no fraud this year. The univariate test indicates that the difference is signiﬁcant at the 5% level (p value is 0.025). China Journal of Accounting Studies 331 Figure 1 illustrates in a preliminary manner the ‘government intervention hypothe- sis’, namely that decentralised government can reduce the probability of corporate fraud. However, considering that there are many other factors also affecting corporate fraud, we must be cautious about the results. Untabulated results show that the Pearson and Spearman correlation coefﬁcients of LAYER and FRAUD are signiﬁcantly negative at the 5% level, which is consistent with the conclusion of Figure 1. In addition, the absolute value of most of the correlation coefﬁcients is below 0.5, indicating that there is no serious multicollinearity problem for the empirical model. 5. Empirical results 5.1. Government decentralisation and corporate fraud In this section, we employ multivariate regression analyses. Because FRAUD is a dummy variable, we use the probit regression model. The empirical results are listed in Table 2. The regression results in Table 2 are divided into four parts, each discussed sepa- rately. The results of the ﬁrst column show that the coefﬁcient on LAYER is signiﬁ- cantly negative at the 5% level, indicating that with the increase of government decentralisation degree over SOEs, the probability of corporate fraud lowers signiﬁ- cantly, consistent with the results of descriptive statistics and the theoretical expectation of the ‘government intervention hypothesis’. The right-hand side of the ﬁrst column shows the marginal effect of each variable. The marginal effect of LAYER is –0.013, indicating that for every one-unit increase of the pyramidal layer, the probability of cor- porate fraud would decrease by 1.3%; in the results of Column (2), we set up a dummy variable DUMLAYER according to whether LAYER is greater than 2; if yes, DUM- LAYER equals 1, otherwise 0. The beneﬁt of binary handling LAYER is that it can pre- vent the results of Column (1) from being caused by the fraud probability of an extremely large LAYER. The results of Column (2) show that the coefﬁcient on DUM- LAYER is signiﬁcantly negative at the 1% level, further supporting the results in Col- umn (1). Additionally, the marginal effect of DUMLAYER is –0.025, indicating that the average corporate fraud probability of the sample with LAYER greater than 2 is 2.5% less than that with LAYER less than or equal to 2. Taken together, the results of Columns (1) and (2) indicate that government decen- tralisation can reduce the probability of SOEs to engage in fraud. However, might there remain a controversy over whether the pyramidal layers actually are limited to repre- senting the degree of government decentralisation over SOEs? A possible method of alleviating this concern is to examine whether the pyramidal layers also affect the prob- ability of non-state-owned enterprises (NSOEs) to commit fraud. An afﬁrmative con- clusion will invalidate our empirical results. The results using a NSOEs sample are listed in Column (3), with the coefﬁcient on LAYER not signiﬁcant. With the use of DUMLAYER as the independent variable, the results of Column (4) report that the coef- ﬁcient on DUMLAYER remains insigniﬁcant. This suggests that the association between pyramidal layers and corporate fraud does not exist in NSOEs. Among the regression results of control variables, the coefﬁcient on LNMCAP is signiﬁcantly negative, indicating that the fraud probability of large-size ﬁrms is low. The coefﬁcient on ROA is signiﬁcantly negative, indicating that high-proﬁtability ﬁrms are unlikely to engage in fraud. Among the variables of corporate governance, the 332 Liu and Li Table 2. The effect of government decentralisation on corporate fraud. (1) Marginal (2) Marginal (3) (4) Variables SOEs effects SOEs effects NSOEs NSOEs LAYER −0.097** −0.013 −0.004 (–2.43) (–0.12) DUMLAYER −0.197*** −0.025 0.026 (–2.92) (0.36) LNMCAP −0.208*** −0.028 −0.210*** −0.028 −0.143*** −0.145*** (–4.14) (–4.14) (–2.75) (–2.77) ROA −1.017** −0.136 −1.037** −0.138 −1.408*** −1.404*** (–2.40) (–2.44) (–4.60) (–4.59) LEV 0.377** 0.050 0.373** 0.050 −0.114 −0.118 (2.04) (2.03) (–0.90) (–0.93) MB 0.011 0.001 0.011 0.002 0.017** 0.017** (1.09) (1.11) (2.18) (2.20) BIG4 −0.240 −0.028 −0.238 −0.027 −0.125 −0.129 (–1.13) (–1.13) (–0.54) (–0.56) MANHOLD −3.283 −0.438 −3.298 −0.440 −0.295 −0.273 (–1.31) (–1.32) (–1.19) (–1.11) INDEP 0.278 0.037 0.256 0.034 −0.391 −0.376 (0.44) (0.41) (–0.58) (–0.55) DUALITY 0.099 0.014 0.101 0.014 0.092 0.093 (0.96) (0.99) (1.21) (1.22) LNBSIZE 0.274 0.037 0.273 0.036 0.183 0.187 (1.41) (1.40) (0.93) (0.94) HBSHARE −0.066 −0.008 −0.075 −0.010 −0.437** −0.441** (–0.40) (–0.46) (–1.99) (–2.02) MEETING 0.147* 0.020 0.149* 0.020 0.221** 0.218** (1.76) (1.79) (2.57) (2.54) LOSS 0.180** 0.027 0.183** 0.027 0.199** 0.197** (2.49) (2.53) (2.34) (2.32) ST 0.338** 0.056 0.333** 0.055 0.526*** 0.526*** (2.56) (2.53) (3.77) (3.77) GOV_SHARE 0.300* 0.040 0.307* 0.041 −0.047 −0.055 (1.68) (1.72) (–0.14) (–0.16) GOV_INTERV −0.002 −0.000 −0.002 −0.000 −0.007 −0.007 (–0.20) (–0.13) (–0.65) (–0.62) TOP1 −0.738*** −0.099 −0.748*** −0.100 −0.681** −0.683** (–2.73) (–2.77) (–2.56) (–2.57) RPT −0.103 −0.014 −0.101 −0.013 −0.192 −0.197 (–0.81) (–0.81) (–1.16) (–1.19) ORECTA 6.898*** 0.921 6.832*** 0.911 7.279*** 7.256*** (4.09) (4.05) (2.71) (2.70) CONSTANT 2.665** 2.542** 1.542 1.550 (2.38) (2.23) (1.21) (1.22) YEAR YES YES YES YES INDUSTRY YES YES YES YES N 6,111 6,111 3,861 3,861 Pseudo R 0.108 0.109 0.097 0.098 The dependent variable is FRAUD. DUMLAYER is an indicator that equals 1 if LAYER is greater than 2, otherwise 0. Refer to Table 1 for the deﬁnitions of the other variables. ***, ** and * denote signiﬁcance at the 1, 5 and 10% levels, respectively. Z-statistics computed with robust standard errors clustered at the ﬁrm level are reported in parentheses. The sample in Columns (1) and (2) is SOEs. The sample in Columns (3) and (4) is NSOEs. China Journal of Accounting Studies 333 coefﬁcient on the number of board meetings (MEETING) is signiﬁcantly positive, consistent with the conclusion of Chen et al. (2006). The coefﬁcients on the two vari- ables LOSS and ST are signiﬁcantly positive, meaning that the ﬁrms with a loss or the risk of delisting have stronger motivation to participate in fraud. The coefﬁcient on GOV_SHARE is positive, suggesting that the ﬁrms with more state-owned shares are more likely to engage in fraud. However, even after controlling for the variable of the proportion of state shares, the key variable LAYER remains signiﬁcant. This suggests that although the government owns the same percentage of shares among different SOEs, its ability to intervene in ﬁrms remains affected by the length of the control chain. The coefﬁcient on TOP1 is signiﬁcantly negative, implying that a higher own- ership concentration is associated with a lower probability of corporate fraud. The coef- ﬁcient on ORECTA is signiﬁcantly positive at the 1% level, suggesting that more serious controlling shareholder tunnelling activity is associated with a higher probability of fraud in the listed companies, consistent with Fan and Wong (2002) and Leuz et al. (2003). In other words, to hide tunnelling, ﬁrms tend to manipulate earnings and cause accounting irregularities. 5.2. Robustness tests To increase the credibility of the conclusion, the following robustness tests are conducted. First, in the empirical models of this paper, there may be some uncontrolled vari- ables affecting both pyramidal layers and corporate fraud. In other words, this model may have omitted variables. To mitigate the effects of missing variables on the conclu- sion, we include the ﬁrm ﬁxed effects in the model. This method can effectively control factors that do not change with time trends but may also affect corporate fraud and pyramidal layers. Theoretical and empirical studies of Fan et al. (2013) and Xia and Chen (2007) ﬁnd that regional institution environment and the characteristics of indus- try strategy are important factors affecting government control mode and the degree of decentralisation and that they change little over time. Therefore, controlling the ﬁrm ﬁxed effect can largely eliminate the effect of those omitted variables on the conclusion. Speciﬁcally, the results including ﬁrm ﬁxed effects are shown in Table 3. Table 3 shows, after controlling for the ﬁrm ﬁxed effects, that many previously sig- niﬁcant control variables become no longer signiﬁcant, perhaps the result of a minimal annual change in variables. However, the coefﬁcient on LAYER remains signiﬁcantly negative at the 5% level, indicating that even controlling for those variables with no ﬂuctuations over time, the conclusion of the paper remains. Therefore, the problem of missing variables will not seriously inﬂuence our conclusion. Second, Zhong et al. (2010) observe that there is an inverse U-shaped relationship between the pyramidal layers and over-investment of free cash ﬂow in SOEs, implying that decentralisation can reduce the over-investment of free cash ﬂow to a certain extent but, after reaching that certain extent, the agency costs increase signiﬁcantly, causing inefﬁcient investment. Does such a nonlinear relationship also exist in our paper? To con- ﬁrm this, we include LAYER and its square, LAYER , in the model. The unreported results indicate that the coefﬁcient on LAYER is insigniﬁcant. Therefore, there is no evidence of a nonlinear relationship between government decentralisation and corporate fraud. Third, the transfer of corporate control rights is likely to make listed ﬁrms’ location in the pyramidal ownership structure change, leading to the number of pyramidal layers between the controlling shareholder and the listed ﬁrm changing. Nonetheless, this 334 Liu and Li Table 3. Controlling for ﬁrm ﬁxed effects. Variables Coefﬁcients Z values LAYER −0.275** (–2.47) LNMCAP 0.115 (0.66) ROA −1.736* (–1.95) LEV −0.348 (–0.62) MB −0.019 (–0.85) BIG4 0.089 (0.17) MANHOLD −4.190 (–0.78) INDEP 0.031 (0.02) DUALITY −0.040 (–0.17) LNBSIZE 0.912* (1.78) MEETING 0.179 (1.01) LOSS 0.308** (2.17) GOV_SHARE −0.014 (–0.03) GOV_INTERV 0.059 (1.37) TOP1 −2.335** (–2.46) RPT −0.012 (–0.05) ORECTA 11.618*** (3.03) CONSTANT −2.969 (–0.82) YEAR YES FIRM YES N 1,533 Pseudo R 0.228 The dependent variable is FRAUD. Refer to Table 1 for the deﬁnitions of the variables. ***, ** and * denote signiﬁcance at the 1, 5 and 10% levels, respectively. Z-statistics computed with robust standard errors clus- tered at the ﬁrm level are reported in parentheses. Sample size is reduced because the observations whose val- ues of FRAUD are constant during the sample period are automatically deleted due to perfect collinearity after including ﬁrm ﬁxed effects into the regression model. change does not directly result from government decentralisation. If the transferred con- trol of the ﬁrm also affects corporate fraud, then the relationship between pyramidal layers and corporate fraud is likely to be caused by the transfer of control rights. To eliminate the inﬂuence of the alternative explanation, we obtained the list of listed ﬁrms whose control rights have been shifted by equity trading in 2005–2010, from the CSMAR database. We then eliminated these listed ﬁrms from our sample and reran model (1). Unreported results show that the coefﬁcient on LAYER remains negative and signiﬁcant at the 5% level (the coefﬁcient is –0.117, and the z value is –2.30). There- fore, the transfer of corporate control rights does not affect our conclusion. Fourth, in the process of forming a pyramidal structure, there may be a divergence between the ultimate controlling shareholder’s voting rights and cash ﬂow rights over the listed ﬁrm of the pyramidal structure, possibly leading to an agency problem for the con- trolling shareholder, thereby affecting the corporate information environment (Claessens, Djankov, Fan, & Lang, 2002; Fan & Wong, 2002). Hence, pyramidal layers having inﬂuence on corporate fraud is likely due to the inﬂuence of the divergence between the controlling shareholder’s voting rights and cash ﬂow rights. To eliminate this alternative explanation, we include two variables in the model (1): CASHR, representing the cash ﬂow rights of the corporate controlling shareholder; and WEDGE, the difference between the voting rights and cash ﬂow rights of the controlling shareholder. After rerunning model (1) including these two variables, unreported results show that the coefﬁcient on LAYER is signiﬁcantly negative (the coefﬁcient is –0.120, and z value is –2.72). The China Journal of Accounting Studies 335 coefﬁcients on CASHR and WEDGE are not signiﬁcant (the coefﬁcient on CASHR is – 0.669, the z value is –1.34; the coefﬁcient on WEDGE is 0.003, the z value is 0.000). These results are consistent with Fan et al. (2013); because the Chinese government has strict rules for the transfer of state-owned shares, the divergence between the controlling shareholders’ voting rights and cash ﬂow rights in the pyramidal structure of SOEs is very small. This leads to the low information content of WEDGE. 5.3. Government decentralisation and corporate fraud: Categorising fraud types Categorising fraud types and examining how government decentralisation reduces speciﬁc types of fraud has policy implications. According to Chen et al. (2013), corporate fraud is classiﬁed into four types: (1) FRAUD_DISCLOSE, related to information disclosure (fabricating proﬁt, ﬁctitious assets, false statements, delayed dis- closure, major omissions, untrue disclosure, dishonest listings, and general accounting misconduct); (2) FRAUD_MARKET, connected with market transactions (insider trad- ing, illegal stock purchases, and manipulating stock prices); (3) FRAUD_CAPITAL, concerned with corporate capital (illegal investments, unauthorised changes in capital use, the occupation of listing ﬁrms’ assets by the controlling shareholders, and illegal guarantees); and (4) FRAUD_OTHER, referring to other types. Table 1 lists the descrip- tive statistics of these four types of fraud, with FRAUD_DISCLOSE the most and FRAUD_MARKET the least likely because of the larger crackdown and heavier punish- ment by supervisory agencies. Theoretically, how does government decentralisation affect different types of fraud? For information disclosure fraud, the current literature suggests that government intervention signiﬁcantly affects the enterprise information environment and causes low-quality accounting information for ﬁrms (Piotroski & Wong, 2012). This stems pri- marily from the decrease in the supply of and demand for high-quality accounting information for SOEs from government intervention (Piotroski & Wong, 2012). The demand side, embodied in government intervention, reduces the contract function of accounting information, for instance, by the direct appointment of CEOs to SOEs by government and by private and political information channelling between government and managers (Fan et al., 2007). Moreover, the soft budget constraint of SOEs lowers the demand of creditors for high-quality ﬁnancial statements based on debt contracts. The supply side often allows SOEs to pursue political rather than operational aims. This distortion motivates SOEs to conceal bad news from the public and damages the information environment (Piotroski et al., 2015). Therefore, it is expected that govern- ment decentralisation signiﬁcantly reduces corporate fraud in information disclosure. Second, some types of market transactions fraud (e.g. insider trading, illegal stock purchases, and stock price manipulation) are closely related to information disclosure. Speciﬁcally, abundant literature proves that managers can manipulate stock prices and proﬁt from doing so by managing information disclosure. For example, Cheng and Lo (2006) discover that managers disclose bad news to reduce stock prices before purchas- ing stocks. Cheng, Luo, and Yue (2013) further observe that managers also achieve the goal of affecting stock price by manipulating the precision of management forecasts. Bergstresser and Philippon (2006) show that managers affect stock price through earn- ings management. Government decentralisation reduces the probability of information disclosure fraud; therefore, it is believed that with a higher quality of information dis- closed due to less government intervention, market transaction fraud will happen at a signiﬁcantly lower rate. 336 Liu and Li Finally, we argue that government decentralisation has a limited effect upon corpo- rate fraud related to capital. Speciﬁcally, Jian and Wong (2010) conﬁrm that controlling shareholders of SOEs support rather than expropriate listed companies through related- Table 4. Government decentralisation on corporate fraud: Categorising fraud types. (1) (2) (3) (4) Variables FRAUD_DISCLOSE FRAUD_MARKET FRAUD_CAPITAL FRAUD_OTHER LAYER −0.096** −0.144** 0.030 −0.073 (–2.13) (–2.41) (0.43) (–1.53) LNMCAP −0.216*** −0.075 −0.281*** −0.265*** (–3.72) (–1.29) (–3.13) (–4.58) ROA −1.477*** −0.207 −1.671*** −0.434 (–3.34) (–0.27) (–3.05) (–0.84) LEV 0.256 0.346 0.431 0.373* (1.31) (1.21) (1.64) (1.67) MB −0.003 0.038*** −0.008 0.007 (–0.25) (2.87) (–0.63) (0.61) BIG4 −0.141 −0.384 −4.398*** −0.170 (–0.57) (–1.23) (–9.37) (–0.61) MANHOLD −4.719 0.222 1.207 −5.198 (–1.40) (0.07) (0.31) (–1.54) INDEP 0.449 −1.183 0.366 0.508 (0.66) (–1.17) (0.33) (0.63) DUALITY 0.107 0.223* 0.037 0.009 (0.94) (1.69) (0.25) (0.07) LNBSIZE 0.268 0.377 0.005 0.452* (1.22) (1.45) (0.02) (1.90) HBSHARE −0.307* 0.264 0.120 −0.174 (–1.66) (1.17) (0.37) (–0.81) MEETING 0.135 0.247* 0.007 0.080 (1.49) (1.92) (0.05) (0.84) LOSS 0.193*** 0.007 0.050 0.026 (2.60) (0.05) (0.45) (0.31) ST 0.373*** −0.459 0.350* 0.201 (2.69) (–1.48) (1.89) (1.30) GOV_SHARE 0.203 0.203 0.132 0.454** (1.07) (0.66) (0.43) (2.10) GOV_INTERV −0.007 −0.007 −0.028 0.014 (–0.58) (–0.46) (–1.50) (0.92) TOP1 −0.613** −1.063** 0.128 −0.442 (–2.12) (–2.17) (0.29) (–1.43) RPT −0.120 0.093 −0.204 −0.057 (–0.89) (0.41) (–1.03) (–0.41) ORECTA 6.516*** 1.862 8.690*** 4.742** (3.74) (0.45) (4.31) (2.32) CONSTANT 2.755** −0.682 3.590* 2.861** (2.12) (–0.55) (1.74) (2.19) YEAR YES YES YES YES INDUSTRY YES YES YES YES N 6,111 6,111 6,111 6,111 Pseudo R 0.117 0.155 0.181 0.106 The dependent variable in each column is FRAUD_DISCLOSE, FRAUD_MARKET, FRAUD_CAPITAL, and FRAUD_OTHER, respectively. Refer to Table 1 for the deﬁnitions of the variables. ***, ** and * denote sig- niﬁcance at the 1, 5 and 10% levels, respectively. Z-statistics computed with robust standard errors clustered at the ﬁrm level are reported in parentheses. China Journal of Accounting Studies 337 party transactions, a phenomenon not found in NSOEs. Similarly, Jiang et al. (2010) ﬁnd that the occupation of listed company capital by controlling shareholders primarily occurs in NSOEs. This phenomenon occurs because the latter’s governmental back- ground deprives them of capital, which frequently affects NSOEs (Allen, Qian, & Qian, 2005). Consequently, NSOEs are degraded as an ‘Automatic Teller Machine’ for their parent companies or provide guarantees for their parent companies to make loans. Therefore, SOEs are less likely to transgress on rules, resulting in a limited effect of decentralisation on funds related to capital. To test the above theoretical expectation, Table 4 lists the regression results of the effects of government decentralisation upon various types of fraud. When dependent variables are FRAUD_DISCLOSE and FRAUD_MARKET, the coefﬁcients on LAYER are both signiﬁcantly negative at the 5% level. When dependent variables are FRAUD_CAPITAL and FRAUD_OTHER, the coefﬁcients on LAYER are both not signiﬁcant. This proves that government decentralisation decreases information disclosure-related and market transactions-related fraud but has no signiﬁcant effect on capital-related fraud. 5.4. Government decentralisation and corporate fraud: Categorising the level of fraud severity The level of fraud severity directly relates to the economic consequences of fraud. Fol- lowing Chen, Jiang, Liang, and Wang (2011), the backward method is utilised to infer the level of fraud severity according to its consequences. Speciﬁcally, penalties for list- ing ﬁrms’ fraud by supervisory agencies are used to judge the level of severity. First, for the type of penalty, if the ﬁrm suffers no penalty or the penalty is limited only to a condemnation, fraud is not severe. If it contains ﬁnes, warnings, or criticisms, fraud is severe. Second, for ﬁnes, if the penalty is without ﬁnes, fraud is not severe; otherwise, it is severe. Finally, Chen, Jiang, Liang, and Wang (2011) note that there is a selective enforcement issue for supervisory agencies, namely that the penalties are not directly related to the level of severity. Hence, fraud types are further classiﬁed to discern the level of severity. Speciﬁcally, when violation type is fabricating proﬁt, false assets, delayed disclosure, illegal stock purchase, or other frauds, it is not severe; otherwise, it is severe (Chen, Jiang, Liang, and Wang 2011). Because government decentralisation may eliminate the umbrella of government for ﬁrms (Chen, Jiang, Liang, and Wang, 2011), so potentially exposing the ﬁrms to penal- ties, it is not clearly predicted by theory – for this part of the analysis – when ﬁrms might consider fraud with both severe and lenient penalties. Certainly, a higher severity of fraud induces more adverse economic consequences; thus, ﬁrms may strive not to commit fraud with severe penalties. Therefore, theoretically, it is ambiguous whether government decentralisation inhibits more severe fraud or less severe fraud. Thus, an empirical test is attempted for the analysis. In the test, the sample is divided into two, a sample with more severe fraud and no fraud and a sample with less severe fraud and no fraud, allowing effective discernment of which type of fraud is affected by govern- ment decentralisation. Relevant empirical results are presented in Table 5, demonstrat- ing that for the categorisation based on type of penalty, decentralisation largely decreases less-severe fraud, with the Column (2) LAYER being signiﬁcantly negative. For the categorisation based on the amount of ﬁnes, decentralisation largely decreases more-severe fraud, with the Column (3) LAYER being signiﬁcantly negative. Finally, for the categorisation based on the type of fraud, decentralisation largely decreases 338 Liu and Li Table 5. Government decentralisation on corporate fraud: Categorising the level of fraud severity. Type of penalty Amount of ﬁne Type of corporate fraud (1) (2) (3) (4) (5) (6) Variables More severe Less severe More severe Less severe More severe Less severe LAYER −0.096 −0.087* −0.159** −0.061 −0.073 −0.133*** (–1.56) (–1.94) (–2.11) (–1.45) (–1.54) (–2.85) LNMCAP −0.154* −0.222*** −0.038 −0.264*** −0.231*** −0.157*** (–1.91) (–4.25) (–0.51) (–4.92) (–3.55) (–2.74) ROA −1.382** −0.819* −1.990*** −0.504 −1.566*** 0.667 (–2.46) (–1.69) (–3.72) (–1.04) (–3.40) (1.02) LEV 0.024 0.441** 0.167 0.389* 0.311 0.403 (0.10) (2.13) (0.65) (1.88) (1.51) (1.50) MB 0.003 0.011 0.025* −0.000 0.002 0.022 (0.18) (1.04) (1.87) (–0.00) (0.17) (1.59) BIG4 −0.042 −0.284 0.059 −0.359 −0.334 −0.084 (–0.20) (–1.09) (0.31) (–1.24) (–0.96) (–0.42) MANHOLD −99.306 −1.867 −257.421 −2.094 −2.904 −3.877 (–1.12) (–0.81) (–1.24) (–0.90) (–0.97) (–1.03) INDEP 0.437 0.142 1.300 −0.138 0.552 −0.038 (0.46) (0.21) (1.24) (–0.21) (0.79) (–0.04) DUALITY −0.069 0.178 0.030 0.128 0.037 0.194 (–0.45) (1.59) (0.20) (1.15) (0.32) (1.39) LNBSIZE 0.111 0.344* 0.428 0.207 0.226 0.393 (0.34) (1.70) (1.21) (1.05) (0.99) (1.55) HBSHARE −0.251 −0.008 −0.112 −0.054 −0.186 0.093 (–1.35) (–0.04) (–0.59) (–0.29) (–0.92) (0.46) MEETING 0.086 0.169* 0.215 0.125 0.193** 0.031 (0.66) (1.87) (1.53) (1.43) (2.01) (0.28) LOSS 0.195* 0.175** 0.117 0.209*** 0.190** 0.120 (1.78) (2.24) (1.01) (2.72) (2.49) (1.08) ST 0.628*** 0.120 0.605*** 0.142 0.308** 0.298 (3.75) (0.74) (3.50) (0.99) (2.10) (1.59) China Journal of Accounting Studies 339 GOV_SHARE −0.245 0.465** −0.018 0.352* 0.234 0.329 (–0.97) (2.33) (–0.06) (1.82) (1.18) (1.26) GOV_INTERV −0.013 0.004 −0.010 0.001 −0.011 0.015 (–0.83) (0.31) (–0.59) (0.08) (–0.80) (1.01) TOP1 −0.551 −0.692** −0.927* −0.576** −0.608** −0.739* (–1.31) (–2.35) (–1.95) (–2.00) (–1.96) (–1.89) RPT −0.382* −0.025 −0.409 −0.012 −0.011 −0.248 (–1.74) (–0.18) (–1.22) (–0.10) (–0.07) (–1.24) ORECTA 5.321** 7.315*** 7.995*** 5.707*** 6.757*** 4.825* (2.27) (3.86) (3.28) (3.14) (3.76) (1.79) CONSTANT 0.709 2.713** −2.707 4.084*** 2.928** 0.906 (0.37) (2.39) (–1.57) (3.41) (1.98) (0.81) YEAR YES YES YES YES YES YES INDUSTRY YES YES YES YES YES YES N 5,693 5,934 5,672 5,955 5,924 5,703 Pseudo R 0.152 0.114 0.178 0.108 0.122 0.102 The dependent variable is FRAUD. Refer to Table 1 for the deﬁnitions of the variables. ***, ** and * denote signiﬁcance at the 1, 5 and 10% levels, respectively. Z-statistics com- puted with robust standard errors clustered at the ﬁrm level are reported in parentheses. 340 Liu and Li Table 6. Government decentralisation on corporate fraud: The effects of government intervention. Regional government intervention index Regional ﬁnancial deﬁcit degree Regional registered urban unemployment rate (1) (2) (3) (4) (5) (6) High government Low government High government Low government High government Low government Variables intervention intervention intervention intervention intervention intervention LAYER −0.121** −0.073 −0.133** −0.069 −0.139** −0.054 (–2.30) (–1.20) (–2.43) (–1.11) (–2.57) (–0.94) LNMCAP −0.129** −0.302*** −0.122* −0.305*** −0.170*** −0.235*** (–2.07) (–3.92) (–1.96) (–3.75) (–2.69) (–2.84) ROA −1.261** −0.953 −1.337*** −0.747 −0.380 −1.770*** (–2.43) (–1.32) (–2.69) (–0.94) (–0.70) (–2.72) LEV 0.288 0.514* 0.146 0.612** 0.442* 0.432 (1.25) (1.81) (0.63) (1.98) (1.83) (1.51) MB −0.001 0.033** 0.009 0.014 0.017 −0.008 (–0.10) (1.97) (0.79) (0.82) (1.40) (–0.48) BIG4 0.058 −0.417* 0.014 −0.476** 0.069 −1.144*** (0.17) (–1.84) (0.04) (–2.01) (0.27) (–4.00) MANHOLD −8.414 −0.106 −5.561 −0.999 −4.622 −2.469 (–1.47) (–0.04) (–0.75) (–0.38) (–0.66) (–0.93) INDEP 0.227 0.336 0.012 0.190 −0.154 0.597 (0.30) (0.32) (0.02) (0.16) (–0.19) (0.60) DUALITY 0.071 0.182 0.166 −0.045 0.198 −0.046 (0.56) (1.23) (1.31) (–0.26) (1.50) (–0.29) LNBSIZE 0.093 0.695** 0.198 0.299 0.152 0.362 (0.42) (2.06) (0.86) (0.89) (0.60) (1.27) HBSHARE −0.343 0.042 −0.044 −0.051 −0.229 0.125 (–1.00) (0.23) (–0.14) (–0.26) (–0.99) (0.56) MEETING 0.100 0.263** 0.223** 0.044 0.200* 0.092 (0.93) (2.07) (2.03) (0.34) (1.86) (0.70) LOSS 0.071 0.388*** 0.086 0.372*** 0.082 0.340*** (0.73) (3.76) (0.90) (3.37) (0.85) (3.22) ST 0.318** 0.497** 0.301** 0.518** 0.252 0.373* (2.19) (1.99) (1.99) (2.06) (1.53) (1.77) GOV_SHARE 0.532** 0.078 0.498** 0.147 0.146 0.572** (2.43) (0.29) (2.19) (0.52) (0.57) (2.37) China Journal of Accounting Studies 341 GOV_INTERV 0.034* −0.004 0.021 0.001 −0.006 0.005 (1.67) (–0.14) (1.16) (0.03) (–0.39) (0.24) TOP1 −1.246*** −0.051 −1.266*** −0.264 −0.968*** −0.640 (–3.60) (–0.13) (–3.57) (–0.60) (–2.77) (–1.57) RPT −0.191 0.042 −0.107 −0.087 −0.013 −0.192 (–1.25) (0.20) (–0.72) (–0.37) (–0.08) (–0.94) ORECTA 9.543*** 0.425 9.691*** 1.629 6.955*** 7.144*** (4.90) (0.10) (4.73) (0.47) (3.07) (2.90) CONSTANT 1.558 3.021* 1.073 4.629** 2.140 2.909 (1.11) (1.82) (0.75) (2.52) (1.58) (1.56) YEAR YES YES YES YES YES YES INDUSTRY YES YES YES YES YES YES N 3,180 2,931 3,150 2,961 3,340 2,771 Pseudo R 0.096 0.160 0.087 0.172 0.101 0.168 The dependent variable is FRAUD. Refer to Table 1 for the deﬁnitions of the variables. ***, ** and * denote signiﬁcance at the 1, 5 and 10% levels, respectively. Z-statistics com- puted with robust standard errors clustered at the ﬁrm level are reported in parentheses. 342 Liu and Li less-severe fraud, with the Column (6) LAYER being signiﬁcantly negative. Therefore, overall, there is no signiﬁcant disparity among fraud by various levels of severity affected by government decentralisation. The conclusion is most likely formed for two reasons: ﬁrst, as stated previously, under the decentralisation background, there is no obvious preference for SOEs between avoiding severe and not-severe commitment of fraud; second, there is a certain bias in the judgement of level of severity in the paper. 5.5. Government decentralisation and corporate fraud: The effects of government intervention The previous empirical results indicate that the relationship between government decen- tralisation and corporate fraud ﬁts well with the theoretical expectation of the ‘govern- ment intervention hypothesis’. To further support the hypothesis, sub-sample regression is conducted based on the degree of intervention of government over SOEs. Theoreti- cally, a stronger government intervention is associated with a greater decrease in the political costs caused by government decentralisation and thus a greater inﬂuence of government decentralisation upon corporate fraud. To conﬁrm this inference, following Fan et al. (2013), three variables are selected to measure the motive for government intervention: the ‘reducing government intervention over ﬁrms’ index from Fan et al. (2011), regional ﬁnancial deﬁcit degree and regional registered urban unemployment rate. When the level of government intervention is greater than the annual sample med- ian, it is included in the high government intervention group; otherwise, it is included in the low government intervention group. The results are reported in Table 6. The results in Table 6 imply that the coefﬁcients on LAYER are signiﬁcant within only the high government intervention group. Speciﬁcally, in Columns (1), (3) and (5), the coefﬁcients on LAYER are all signiﬁcantly negative at the 5% level. Nevertheless, they are all insigniﬁcant in the subsamples of low government intervention. Thus, the results in Table 6 further conﬁrm the ‘government intervention hypothesis’. 6. Conclusion Understanding the economic consequences of the deepening SOE reforms is critically important. From the perspective of the pyramidal structure of listed SOEs, this paper examines the effect of the degree of government decentralisation upon the probability of SOEs to commit fraud. Theoretically, government decentralisation reduces the politi- cal costs of government intervention, but simultaneously induces agency costs, both of which are vital ingredients for corporate fraud according to the literature. Therefore, government decentralisation might both decrease (‘government intervention hypothe- sis’) and increase corporate fraud probability (‘agency costs hypothesis’). Using the data of A-share listed SOEs in China in 2004–2010, this paper observes empirically that government decentralisation over SOEs signiﬁcantly reduces their fraud probability. Therefore, the relationship between government decentralisation and corpo- rate fraud ﬁts well with the ‘government intervention hypothesis’. Further results pro- ven by the categorisation of fraud types are that government decentralisation largely decreases fraud related to information disclosure and market transactions. However, there is no signiﬁcant difference among fraud activities with various levels of severity affected by government decentralisation. Finally, the effect of government decentralisa- tion upon corporate fraud largely occurs in SOEs in which government intervention is more likely, further conﬁrming the ‘government intervention hypothesis’. China Journal of Accounting Studies 343 The conclusions of this paper provide new empirical evidence for the study of the economic consequences of government decentralisation and the study of factors underlying corporate fraud. The conclusions also support SOE reforms by government decentralisation from the perspective of accounting information quality, thereby embodying important policy implications. The limitation of this paper lies in the exclusion of analysing types of government decentralisation other than the pyramidal structure formed since the 1990s; this necessi- tates future consideration. Acknowledgements The authors appreciate the helpful comments from two anonymous reviewers, Donghua Chen (Associate editor), Pauline Weetman (Language Editor), Liansheng Wu (Joint-editor), and Jason Zezhong Xiao (Joint-editor). Disclosure statement No potential conﬂict of interest was reported by the authors. Funding Hang Liu acknowledges ﬁnancial support from the National Natural Science Foundation of China (grant no. 71402017) and the Program for Liaoning Excellent Talents in University (grant no. WJQ2014035). Xiaorong Li acknowledges ﬁnancial support from the National Natural Science Foundation of China (grant no. 71503283), the Humanities and Social Science Research Project of the Ministry of Education in China (grant no. 14YJC630069), the Social Science Research Project of Beijing (grant no. 15JGC173), the Program for Innovation Research in Central Univer- sity of Finance and Economics, and Zhongcai-Pengyuan Local Finance Investment and Funding Research Institute. Notes 1. The current literature on the Chinese government decentralisation largely refers to the decentralisation in the economic ﬁeld since 1980s after the reform and opening up, whose core is ﬁnance decentralisation from central to local government. Government decentralisa- tion on SOEs is also part of it and is exposed to many studies (Groves, Hong, McMillan, & Naughton, 1994, 1995; Qian, 1996). This paper focuses on the economic consequences of government decentralisation on SOEs. 2. Refer to Fan et al. (2013) for a more detailed description of the formation of state-owned pyramidal structures. 3. The fraud database in CSMAR collects the information related to corporate fraud from the announcements released by the listed ﬁrms committing fraud, the coverage from media des- ignated by the China Securities Regulatory Commission (CSRC), and the announcements released by regulators since 1994. The information of this database includes items such as the time period of fraud, the type of fraud, the type of punishment. 4. In China, if a listed ﬁrm has two consecutive annual losses, market regulators will assign special treatment (ST) status to it. ST ﬁrms face various trading and ﬁnancial restrictions. In addition, if they make losses for one more year, trading will be suspended; if they still make losses in the fourth year, they will be delisted. 5. The index is only revealed by the year 2009, and our sample period is 2004–2010. Therefore, the index of 2010 uses data from 2009 as a replacement. In addition, the index is an inverse index. In other words, a smaller GOV_INTERV indicates a higher degree of government intervention in the company location. 344 Liu and Li 6. Note that the existing study largely concludes that the pyramidal structure will enhance corporate information opacity and that the following empirical results of our paper show that the pyramidal structure will reduce the probability of SOEs to engage in fraud. Thus, theoretically, this explanation does not affect the conclusion of our paper. 7. The listed ﬁrms provide the ownership structure ﬁgures in their annual reports. Through the ownership structure ﬁgures, we can manually identify the pyramidal layers between the con- trolling shareholders and listed ﬁrms. The CSMAR database collects all the ownership struc- ture ﬁgures, so we can download these ﬁgures from CSMAR and manually calculate each ﬁrm’s pyramidal layer. 8. As far as we know, three papers investigate the evolution of NSOEs’ pyramidal structures in China (Chen, Jin, & Liu, 2011; Li, Xin, & Yu, 2008; Liu, Zheng, & Zhu, 2010). The conclusions of these papers generally indicate that NSOEs’ pyramidal structures are more prevalent in regions with worse institutions. Although determinants of pyramidal structures of SOEs and NSOEs are different according to current literature (Fan et al., 2013), we still cannot remove the concern that some unobservable factors affect the usage of pyramidal structures both in SOEs and NSOEs. If these unobservable factors impact on corporate fraud, then the conclusion of our paper cannot be interpreted as government decentralisa- tion. 9. We expound primarily on Column (1)’s regression results of the control variables because the results are our main concern. 10. What requires an explanation is that the coefﬁcient on GOV_INTERV is not signiﬁcant, which is inconsistent with the theory. A possible reason is that the existing variable is not a good measure of the degree of government intervention. Thus, we perform the exploration in two ways: ﬁrst, we set up a dummy variable DUMGI according to the size of GOV_IN- TERV; when ﬁrms’ GOV_INTERV is less than or equal to the annual sample median, it equals 1, otherwise 0. Unreported results show that when the regression model does not contain LAYER, the coefﬁcient on DUMGI is signiﬁcantly positive at the marginal level of 10% (z value = 1.62). In the model containing LAYER, the z value of DUMGI’s coefﬁcient is changed to 1.41. Second, we use the political connection (PC) as another measure of government intervention. According to Fan, Wong, and Zhang (2007), when the Chairman of the board or CEO of the SOE was or is a government ofﬁcial, PC equals 1, otherwise 0. Unreported results show that when not containing LAYER in the regression model, the coef- ﬁcient on PC is signiﬁcantly positive at the 10% level (z value = 1.69). When the regression model contains LAYER, the z value of PC’s coefﬁcient decreases to 1.52. In addition, when we change the dependent variable to FRAUD_DISCLOSE (the most important and common category of corporate fraud), the regression results also conﬁrm our above conclusion, reﬂected by the coefﬁcients on DUMGI and PC being signiﬁcantly positive at the 5% and 10% level, respectively. 11. In our sample, the average of WEDGE is only 4%. References Aghion, P., & Tirole, J. (1997). Formal and real authority in organizations. Journal of Political Economy, 105,1–29. Agrawal, A., & Chadha, S. (2005). Corporate governance and accounting scandals. Journal of Law and Economics, 48, 371–406. Aharony, J., Lee, C.-W. J., & Wong, T. J. (2000). Financial packaging of IPO ﬁrms in China. Journal of Accounting Research, 38, 103–126. Aharony, J., Wang, J., & Yuan, H. (2010). Tunneling as an incentive for earnings management during the IPO process in China. Journal of Accounting and Public Policy, 29,1–26. Allen, F., Qian, J., & Qian, M. (2005). Law, ﬁnance, and economic growth in China. Journal of Financial Economics, 77,57–116. Armstrong, C. S., Jagolinzer, A. D., & Larcker, D. F. (2010). Chief executive ofﬁcer equity incentives and accounting irregularities. Journal of Accounting Research, 48, 225–271. Armstrong, C. S., Larcker, D. F., Ormazabal, G., & Taylor, D. J. (2013). The relation between equity incentives and misreporting: The role of risk-taking incentives. Journal of Financial Economics, 109, 327–350. China Journal of Accounting Studies 345 Bai, J., & Lian, L. (2013). Why do state-owned enterprises over-invest? Government intervention or managerial entrenchment. China Journal of Accounting Studies, 1, 236–259. Beasley, M. S. (1996). An empirical analysis of the relation between the board of director com- position and ﬁnancial statement fraud. The Accounting Review, 71, 443–465. Bebchuk, L. A., Fried, J. M., & Walker, D. I. (2002). Managerial power and rent extraction in the design of executive compensation. The University of Chicago Law Review, 69, 751–846. Bergstresser, D., & Philippon, T. (2006). CEO incentives and earnings management. Journal of Financial Economics, 80,511–529. Burns, N., & Kedia, S. (2006). The impact of performance-based compensation on misreporting. Journal of Financial Economics, 79,35–67. Bushman, R. M., & Piotroski, J. D. (2006). Financial reporting incentives for conservative accounting: The inﬂuence of legal and political institutions. Journal of Accounting and Economics, 42, 107–148. Bushman, R. M., Piotroski, J. D., & Smith, A. J. (2004). What determines corporate trans- parency? Journal of Accounting Research, 42, 207–252. Chen, D., Chen, Y., Li, O. Z., & Ni, C. (2013). Foreign residency rights and corporate frauds. Working Paper. Nanjing University, China, National University of Singapore, Singapore, and Renmin University of China, China. Retrieved from http://ssrn.com/abstract=2310426 Chen, H., Chen, J. Z., Lobo, G. J., & Wang, Y. (2010). Association between borrower and lender state ownership and accounting conservatism. Journal of Accounting Research, 48, 973–1014. Chen, G., Firth, M., Gao, D., & Rui, O. M. (2005). Is China’s securities regulatory agency a toothless tiger? Evidence from enforcement actions. Journal of Accounting and Public Policy, 24, 451–488. Chen, G., Firth, M., Gao, D., & Rui, O. M. (2006). Ownership structure, corporate governance, and fraud: Evidence from China. Journal of Corporate Finance, 12, 424–448. Chen, D., Jiang, D., Liang, S., & Wang, F. (2011). Selective enforcement of regulation. China Journal of Accounting Research, 4,9–27. Chen, D., Jin, X., & Liu, X. (2011). Government quality, social capital and endogenous determi- nants of pyramid structure of family business. China Industrial Economics, (7), 129–139. (In Chinese.) Chen, X., Li, M., Rui, O. M., & Xia, L. (2009). Judiciary independence and the enforcement of investor protection laws: Market responses to the “1/ 15” notice of the Supreme People’s Court of China. China Economic Quarterly, (9), 1–28. (In Chinese.) Cheng, Q. & Lo, K. (2006). Insider trading and voluntary disclosures. Journal of Accounting Research, 44, 815–848. Cheng, Q., Luo, T., & Yue, H. (2013). Managerial incentives and management forecast precision. The Accounting Review, 88, 1575–1602. Cheng, Z., Xia, X., & Yu, M. (2008). Government intervention, pyramidal structures, and invest- ments of local state-owned listed ﬁrms. Management World, (9), 37–47. (In Chinese.) Claessens, S., Djankov, S., Fan, J. P. H., & Lang, L. H. P. (2002). Disentangling the incentive and entrenchment effects of large shareholdings. Journal of Finance, 57, 2741–2771. Dai, Y., Pan, Y., & Liu, S. (2011). Media supervision, government intervention, and corporate governance: Evidence from the perspective of Chinese listed ﬁrms’ ﬁnancial restatements. World Economy,(11), 121–144. (In Chinese.) Dechow, P., Ge, W., & Schrand, C. (2010). Understanding earnings quality: A review of the proxies, their determinants and their consequences. Journal of Accounting and Economics, 50, 344–401. Dechow, P. M., Sloan, R. G., & Sweeney, A. P. (1996). Cases and consequences of earnings manipulations: An analysis of ﬁrms subject to enforcement actions by the SEC. Contempo- rary Accounting Research, 13,1–36. Efendi, J., Srivastava, A., & Swanson, E. P. (2007). Why do corporate managers misstate ﬁnan- cial statements? The role of option compensation and other factors. Journal of Financial Eco- nomics, 85, 667–708. Erickson, M., Hanlon, M., & Maydew, E. L. (2006). Is there a link between executive equity incentives and accounting fraud? Journal of Accounting Research, 44,113–143. Fan, G., X. Wang, & H. Zhu (2011). NBEI index of marketization of China’s provinces 2011 report. Economics Science Press. (In Chinese.) 346 Liu and Li Fan, J. P. H. & Wong, T. J. (2002). Corporate ownership structure and the informativeness of accounting earnings in East Asia. Journal of Accounting and Economics, 33, 401–425. Fan, J. P. H., Wong, T. J., & Zhang, T. (2007). Politically connected CEOs, corporate gover- nance, and post-IPO performance of China’s newly partially privatized Firms. Journal of Financial Economics, 84, 330–357. Fan, J. P. H., Wong, T. J., & Zhang, T. (2013). Institutions and organizational structure: The case of state-owned corporate pyramids. Journal of Law, Economics, and Organization, 29, 1253– Firth, M., Rui, O. M., & Wu, W. (2011). The effects of political connections and state ownership on corporate litigation in China. Journal of Law & Economics, 54, 573–607. Gao, L. & Song, S. (2007). Corporate governance and corporate transparency. Journal of Financial Research,(11), 28–44. (In Chinese.) Groves, T., Hong, Y., McMillan, J., & Naughton, B. (1994). Autonomy and incentives in Chinese state enterprises. Quarterly Journal of Economics, 109, 183–209. Groves, T., Hong, Y., McMillan, J., & Naughton, B. (1995). China’s evolving managerial labor market. Journal of Political Economy, 103, 873–892. Harris, J., & Bromiley, P. (2007). Incentives to cheat: The inﬂuence of executive compensation and ﬁrm performance on ﬁnancial misrepresentation. Organization Science, 18, 350–367. Hong, H., Hu, X., & Guo, C. (2012). Research on the identiﬁcation of listed companies’ ﬁnancial reporting fraud: Based on GONG theory. Accounting Research,(8), 84–90. (In Chinese.) Huang, Q., Cheng, M., Li, W., & Wei, M. (2014). Listing approach, political favours and earn- ings quality: Evidence from Chinese family ﬁrms. China Journal of Accounting Studies, 2, 13–36. Hung, M., Wong, T. J., & Zhang, T. (2012). Political considerations in the decision of Chinese SOEs to list in Hong Kong. Journal of Accounting and Economics, 53, 435–449. Hung, M., Wong, T. J., & Zhang, F. (In press). The value of political ties versus market credibil- ity: Evidence from corporate scandals in China. Contemporary Accounting Research. Jensen, M. C., & Meckling, W. H. (1976). Theory of the ﬁrm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics, 3, 305–360. Jian, M. & Wong, T. J. (2010). Propping through related party transactions. Review of Accounting Studies, 15,70–105. Jiang, G., Lee, C. M. C., & Yue, H. (2010). Tunneling through intercorporate loans: The China experience. Journal of Financial Economics, 98,1–20. Lennox, C., & Pittman, J. A. (2010). Big ﬁve audits and accounting fraud. Contemporary Accounting Research, 27, 209–247. Leuz, C., Nanda, D., & Wysocki, P. D. (2003). Earnings management and investor protection: An international comparison. Journal of Financial Economics, 69, 505–527. Li, Z., Xin, X., & Yu, X. (2008). Financial development, debt ﬁnancing constraint and the pyra- midal structure. Management World, (1), 123–135. (In Chinese.) Liu, H., & Li, X. (2012). Pyramidal structure, tax burden, and ﬁrm value: Evidence from local state-owned enterprises. Management World,(8), 91–105. (In Chinese.) Liu, Q., Luo, L., Zhang, Y., & Chen, H. (2013). Concentration of managerial power, internal con- trol, and accounting information quality. Nankai Business Review, (16), 15–23. (In Chinese.) Liu, Q., Zheng, Y., & Zhu, Y. (2010). The evolution and consequence of Chinese pyramids. Working Paper. Peking University, China, Sun Yat-Sen University, China, and Morgan Stanley. Piotroski, J. D., & Wong, T. J. (2012). Institutions and information environment of Chinese listed ﬁrms. In J.P.H. Fan & R. Morck (Eds.), Capitalizing China, (pp. 201–242). University of Chicago Press. Piotroski, J. D., Wong, T. J., & Zhang, T. (2015). Political incentives to suppress negative infor- mation: Evidence from Chinese listed ﬁrms. Journal of Accounting Research, 53, 405–459. Qian, Y. (1996). Enterprise reform in China: Agency problems and political control. Economics of Transition, 4, 427–447. Quan, X., Wu, S., & Wen, F. (2010). Managerial power, private income, and compensation rigging: Evidence from state-owned listed enterprises in China. Economic Research Journal, (11), 73–87. (In Chinese.) Rao, P., & Jiang, G. (2011). Monetary policy ﬂuctuations, bank credit, and accounting conser- vatism. Journal of Financial Research,(3), 51–71. (In Chinese.) China Journal of Accounting Studies 347 Wang, Q., Wong, T. J., & Xia, L. (2008). State ownership, the institutional environment, and auditor choice: Evidence from China. Journal of Accounting and Economics, 46,112–134. Wang, K., & Xiao, X. (2009). Ultimate government control structures and ﬁrm value: Evidence from Chinese listed companies. China Journal of Accounting Research, 2, 101–122. Wu, L. (2009). State ownership, preferential tax, and corporate tax burdens. Economic Research Journal,(10), 109–120. (In Chinese.) Xia, L., & Chen, X. (2007). Marketization, SOE reform strategy, and endogenously determined corporate governance structure. Economic Research Journal,(7), 82–95. (In Chinese.) Zeng, Q., & Chen, X. (2006). State stockholder, excessive employment, and labor cost. Economic Research Journal,(5), 74–86. (In Chinese.) Zhong, H., Ran, M., & Wen, S. (2010). Government intervention, internal control, and corporate investment. Management World,(7), 98–109. (In Chinese.) Zhu, C., & Lee, C.-W. J. (2008). State ownership and accounting conservatism. Accounting Research,(5), 38–45. (In Chinese.) Appendix 1. An example of pyramidal structure of SOE in China – The ownership structure of Chongqing Changan Automobile Corporation (stock code: 000625) at the 2010 ﬁscal year end SASAC 100% China South Industry Group Corporation 100% China Changan Automobile Group 45.71% Chongqing Changan Automobile Corporation Source: The 2010 annual report of Chongqing Changan Automobile Corporation

Journal

China Journal of Accounting Studies – Taylor & Francis

Published: Oct 2, 2015

Keywords: agency costs; fraud; government decentralisation; political costs; pyramidal layers

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Government decentralisation and corporate fraud: Evidence from listed state-owned enterprises in China

Government decentralisation and corporate fraud: Evidence from listed state-owned enterprises in China

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Government decentralisation and corporate fraud: Evidence from listed state-owned enterprises in China

Government decentralisation and corporate fraud: Evidence from listed state-owned enterprises in China

References (65)

Abstract

Journal

Recommended Articles

There are no references for this article.

Our policy towards the use of cookies