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

Learn More →

Fiscal centralization, government control and corporate tax burden: Evidence from China

Fiscal centralization, government control and corporate tax burden: Evidence from China China Journal of Accounting Studies, 2013 Vol. 1, Nos. 3–4, 168–189, http://dx.doi.org/10.1080/21697221.2013.870367 Fiscal centralization, government control and corporate tax burden: Evidence from China a b Jun Liu * and Feng Liu a b School of Business, Sun Yat-Sen University, China; School of Management, Xiamen University, China and Center for Accounting, Finance and Institutions, Sun Yat-Sen University, China This paper examines the relationship between government control and the tax burden of firms in China. We develop a new corporate tax burden measurement taking turn- over taxes into account, because in China turnover taxes actually constitute the main component of tax burden. We find that the tax burden of state-owned enterprises (SOEs) is lower than non-SOEs, indicating that non-SOEs are facing tax discrimina- tion. Among SOEs, the tax burden of local SOEs is higher than that of central SOEs, and the lower the local governments’ level, the higher the tax burden of SOEs under their control. We interpret these findings as the result of local governments’ tax competition and tax grabbing behaviors under China’s current highly centralized fiscal system. In addition, we find that our results are mainly caused by firms’ differences in tax refunds and the Value-Added Tax (VAT) burden. Keywords: fiscal centralization; government control; turnover tax burden; corporate tax burden 1. Introduction Kornai (1980) points out that in a socialist economy, the government is usually paternalistic to the SOEs. Although China’s market-oriented economy reform has continued for about 30 years, ownership discrimination against non-SOEs is still an indisputable reality in the economic environment. Calomiris, Fisman, and Wang (2010) find a negative effect of government ownership on returns at the announcement of the sale of government-owned shares in China, indicating that the benefits of government ownership outweigh the efficiency costs. Other researchers have proven that SOEs have the advantages over non-SOEs in entering an industry (Chen, Yu, Wang, & Lai, 2008), borrowing from banks (Fang, 2007; Jiang & Li, 2006; Lu, Zhu, &Fan, 2009), and obtaining subsidies from the government (Pan, Dai, & Li, 2009). Tax is also an important channel by which the government affects firms’ value. However, so far only a few studies have examined the relationship between government control and firms’ tax burden in China, and the results are mixed. Zheng and Han (2008) find that SOEs tend to be more conservative in tax avoidance and have higher effective tax rate (ETR) than non-SOEs. Wu (2009) examines the effects of state ownership on corporate tax burdens and finds that corporate income tax burdens increase with state ownership. However, the research of Cao, Liu, and Zhang (2009) finds that for the firms located in Chinese preferential tax zones, central SOEs bear a *Corresponding author. Email: cnjamesjunliu@gmail.com Paper accepted by Liansheng Wu. © 2013 Accounting Society of China China Journal of Accounting Studies 169 lower tax burden than local SOEs and non-SOEs, but there is no significant difference in tax burdens between local SOEs and non-SOEs. Thus, what effect government control has on corporate tax burdens remains an open question. The 1994 tax sharing reform in China is an attempt by the central government to recentralize the fiscal authority. The new tax system encourages local governments to alleviate their financial difficulties by tax competition (Shen & Fu, 2006) and tax grabbing (Zhou, 2005). SOEs generally have high lobbying power because of their close relationship with the government derived from ownership, and besides, as SOEs often undertake many social functions of the government, they are usually offered more preferential policies as compensation (Wu, Wang, Luo, & Gillis, 2012). Therefore, SOEs are more likely to obtain tax preferences and looser tax enforcement in govern- ments’ tax competition, and to avoid governments’ tax grabbing. Thus, the tax burden of SOEs should be lower than non-SOEs. Furthermore, under the current highly centralized fiscal system, lower level local governments are often in worse financial conditions. Because local SOEs are usually under the strong control of local govern- ments, they are more likely to be grabbed than central SOEs. So the tax burden of local SOEs should be higher than that of central SOEs, and the lower the level of government, the higher the tax burden of SOEs. With a sample of listed companies in the China A-share market during 2003–2006, we examine the tax burden of firms with different types of ultimate owners in China. According to the Interim Regulations on Supervision and Management of Enterprises’ State-owned Assets promulgated by the State Council of the PRC, SOEs are under the supervision and administration of the authorities, which are established to implement the responsibilities, and rights as owners, of central, provincial, and city governments. We classify SOEs into three categories: central SOEs, provincial SOEs, and city SOEs. Provincial and city SOEs can be referred to as local SOEs collectively. In contrast to prior tax burden studies, which generally focus on Enterprise Income Tax (EIT), we develop a new corporate tax burden measurement based on tax cash flows. This measure includes the turnover tax burden, which is actually the main component of corporate tax burden in the special Chinese taxation and business environment. We find that the tax burden of SOEs is significantly lower than non-SOEs, indicating that non-SOEs are facing tax discrimination. Among SOEs, the tax burden of central SOEs is lower than local SOEs. The tax burden of provincial SOEs is also lower and not significantly different from central SOEs, but the tax burden of city SOEs is higher than other SOEs and is not significantly different from non-SOEs. In addition, we divide the tax burden into two parts, tax payments and tax refunds, and find that tax refunds of SOEs are generally higher than non-SOEs, but there is no significant difference in tax payments, indicating that our findings are mainly driven by the differences in tax refunds from the government. We also separate the corporate tax burden into three main components: VAT, Business Tax (BT) and EIT, and find that the VAT burden of SOEs is lower than non-SOEs, but there are no significant differences in BT and EIT burdens. This indicates that the findings of this paper are mainly derived from differences in VAT burdens and, more importantly, research on corporate tax burdens will be biased if it only focuses on EIT. Our paper makes several contributions. First, we examine the effect of government control on corporate tax burdens and find evidence of tax discrimination against non-SOEs. This provides proof of government paternalism (Kornai, 1980) from a tax burden perspective and enriches the literature on the effect of political connections on corporate tax burdens (Adhikari, Derashid, & Zhang, 2006; Faccio, 2010). Second, we 170 Liu and Liu introduce a new corporate tax burden measurement based on tax cash flows, taking the turnover tax burden into account. This new measurement makes the corporate tax burden research in China more complete and instructive. Third, most prior tax burden studies ignore differences in tax burdens between SOEs controlled by different levels of government. In this paper we divide SOEs into three categories by the hierarchical level of their ultimate owners and provide a more detailed picture of the effect of government control on corporate tax burdens. Fourth, the findings of this paper provide an explanation for the negative relationship between control by lower level governments and firm value (Xia & Fang, 2005), and also provide new evidence to support the shift from the ‘helping hand’ to the ‘grabbing hand’ of local governments under the highly centralized fiscal system (Chen, Hillman, & Gu, 2002). The rest of this paper is organized as follows: Section 2 provides institutional background on the Chinese tax system and analyzes the mechanisms of government control on corporate tax burdens and presents our main hypotheses. Section 3 describes differences between the Chinese and American tax systems, explains why we need to take turnover taxes into consideration in corporate tax burden research, and then describes our new tax burden measurement and research model in greater detail. Section 4 presents our main empirical results and also a range of robustness checks. Section 5 concludes. 2. Institutional background and hypothesis development 2.1. Institutional background China’s current tax system was established by the 1994 Tax Sharing Reform. The core content of this reform was to reset the tax revenue sharing system between central and local governments. The new tax revenue sharing system re-categorized taxes into central taxes, local taxes and shared taxes. This tax sharing system is still in operation today and has exerted a profound influence on the relationship between the central and local governments. Among the three most important sources of tax revenues (VAT, BT, EIT), VAT is a shared tax, 75% for the central government and 25% for local govern- ments; BT is a local tax, which only belongs to local governments; and EIT is a shared tax, 60% for the central government and 40% for local governments. To ensure the smooth and effective implementation of the tax sharing system, significant changes also took place in China’s tax collection system in 1994. First, two sets of tax authorities, national and local tax bureaus, were set up to collect different types of taxes. National tax bureaus were established for collecting central and shared taxes, and are administered directly by the upper level national tax bureaus. Local tax bureaus are in charge of collecting local taxes and are under the dual leadership of both the local governments and the higher level tax bureaus. Before 1994, the central government had no tax collection authorities and had to rely totally on local authorities to collect and remit tax revenues. The new tax collection system changed this practice and ensured the revenues of central government be collected fully and in time. Second, in order to get sufficient support from local governments, a tax revenue refund system was designed to protect the interests of local governments. The revenues of the Two Taxes (VAT and Consumption Tax), which used to be the main sources of local revenue, would be refunded to the localities based on the amounts in the year 1993. In addition, 30% of the incremental central revenue from the Two Taxes would also be refunded each year. Actually, a large portion of the central-to-local transfer payment is just the refund of these tax revenues after the tax sharing reform. According China Journal of Accounting Studies 171 to Ma and Yu (2003), an average of 62.1% of the central-to-local transfer payment is the tax revenue refund during 1998–2001. 2.2. Hypothesis development The 1994 Tax Sharing Reform is a fiscal reform with an obvious centralizing tendency (Zhou, 2006). Through this reform, central government took back most of the discretionary power on revenue collection, which was gradually devolved to the localities in the 1980s. However, this reform did not change the fiscal expenditure structure between central and local governments, local governments were still undertak- ing most local public expenditures. The reform thus resulted in wide gaps between local fiscal revenues and expenditures. While fiscal expenditure is impossible to cut down, increasing revenue becomes the only way left to bridge the gap. Tax, as the biggest and highest quality portion of revenue, has inevitably received great attention from local governments. There are usually two ways for local governments to increase tax revenue. The first method is tax competition. The new tax system has determined the tax revenue distribution and refund system between central and local governments, which will stimulate local governments to provide various preferential policies to attract mobile capital to their administrative region so that they can develop the economy and get more tax surplus. Shen and Fu (2006) point out that tax competition between local governments was usually performed through providing preferential tax policies and loose tax enforcement. Under China’s current tax system, tax preferences are often shown as tax rate reductions, tax amount reductions and tax base reductions, among which the first two are especially common (Wang, 2003). Tax rate reductions are usually conducted based on the location or industry of enterprises. In China, many tax-related issues need to be authorized by relevant government authorities, such as whether an enterprise is located in a preferential tax zone; whether it is categorized in a preferential tax industry and which preferential tax rate is to be chosen. We believe that local governments have wide discretionary powers on whether an enterprise receives a tax rate reduction. Tax amount reductions are often shown as tax refunds from local governments. Although tax refunds unauthorized by the central government have been forbidden since 2002, in reality, local governments still refund a lot of tax revenue to enterprises as tax preferences. The categories of taxes involved include not only EIT, but also some turnover taxes, such as VAT and BT. As well as preferential tax policies, loose tax enforcement is also a common tax competition method. Under China’s current tax collection system, local tax bureaus are under the dual leadership of local governments and the higher level tax bureaus. However, the appointment of chief tax officers is mainly decided by local governments and even the salaries of employees and administrative funds are provided by local governments. So, it is safe to say that local governments have strong power to influence local tax bureaus, if necessary. National tax bureaus are nominally only under the leadership of the upper level national tax bureaus, but actually are also strongly influenced by local governments, because there are no special judicial agencies for tax affairs in China’s tax system. Only when tax bureaus get full cooperation from local judicial agencies will they fulfill their tax collection and administration duties (Xu et al., 2001; Ye & Lin, 2007). Guo and Li (2009) also provide evidence for the influence of local governments on tax 172 Liu and Liu enforcement. They find that local governments are in competition for VAT revenue, which is collected by national tax bureaus. The second method is tax grabbing. With a clear tax sharing and refund system, the 1994 tax reform hardened the budget constraints of local governments. When facing fiscal stress, local governments will grab economic resources from lower level govern- ments and enterprises within their precincts. Zhou (2005) develops a concept of ‘Inverted Soft Budget Constraints’ to describe and explain this behavior of local governments. One important characteristic of China’s current tax system is ‘high tax rate, loose tax enforcement’ (Mao, 2003), so there is much room for local governments to strengthen tax enforcement and, if necessary, local governments do not hesitate to do so. Lv and Fan (2006) find that the tax enforcement efficiency of most provinces in China has improved after 1994 and fiscal pressure is an important cause. Furthermore, some local governments, especially some governments at grassroot levels, even grab revenues by assigning tax fees on the enterprises under their jurisdiction (Zhou, 2005). The primary goal of modern firms is to maximize the wealth or value of sharehold- ers (Fama & Jensen, 1983; Jensen, 1986; Jensen & Meckling, 1976; Ross, 1973). In China, most enterprises have established management evaluation systems based on operating performance and firm value. So managers generally have incentives to improve operating performance by reducing tax fees. However, for enterprises with different government control backgrounds, their abilities to get tax benefits from governments’ tax competition are not the same, and the probabilities of governments grabbing them are also different. SOEs have natural political ties because of the government ownership. Kornai (1992) points out that, in a socialist economy, the ruling party, government, and SOEs always get along inextricably with each other, and the executives of SOEs are often cross appointed as the officers of others. In China, top executives of SOEs are usually ranked in the administrative hierarchy in the same way as government officials, and are appointed by governments or higher-level party organizations. The cross-appointment between executives of SOEs and government officers is also very common, but this phenomenon does not exist between governments and non-SOEs. Therefore, the close relationship between SOEs and the government will endow SOEs with higher lobbying power. In addition, SOEs in China often shoulder many policy burdens, such as employment, public utility, etc. As a compensation, SOEs are usually offered more preferential policies by the government. Lin and Tan (1999) argue that the policy bur- dens are the root cause of government’s soft budget constraint on SOEs. Wu et al. (2012) find that SOEs’ size is negatively associated with their effective tax rate (ETR) and non-SOEs’ size is positively associated with ETR, when these firms are not subject to any preferential tax status. They explain the finding as the result of SOEs’ high lob- bying power and the compensation from the government. Therefore, we can reason- ably infer that SOEs have greater advantages than non-SOEs in seeking tax preferences and loose tax enforcement, and have lower extent of tax grabbing. This discussion leads to our first hypothesis: H1a: The tax burden of SOEs is lower than that of non-SOEs. Furthermore, because of the vague property rights and the dual roles of the executives, government control over SOEs usually proves to be ‘weak control’ in ownership and ‘strong control’ in administration (He, 1998). The operating decisions made by executives of SOEs often deviate from the goal of maximizing shareholder China Journal of Accounting Studies 173 wealth and tend to satisfy the evaluation criteria for government officials, such as, GDP growth and tax revenue contribution. By contrast, the executives of non-SOEs are prob- ably large shareholders or the founder and their family. They are more likely to make decisions for shareholders’ benefits and have more incentives to avoid taxes. Zheng and Han (2008) find that the tax avoidance of SOEs is more conservative than other firms. Wu, Wang, Lin, Li, and Chen (2007) also find that non-SOEs are more likely to change registration locations to avoid a greater tax burden. So, there is also the possibility that the tax burden of SOEs is not lower than that of non-SOEs. Following this discussion, we state a competing hypothesis of H1a: H1b: The tax burden of SOEs is not lower than that of non-SOEs. Under the current highly centralized fiscal system, most local governments are in a bad financial condition and have a strong motivation to improve tax revenues. For local SOEs, local governments are usually large shareholders, so local governments can levy more tax revenues by interfering in the operations of local SOEs or even assigning tax fees directly to them. But central SOEs are different. Although they are spread out all over China, they are directly under the administration of the central government, enjoy a high hierarchical status and so are less likely to be affected by local governments. In addition, central SOEs have the autonomy to invest in other regions or even migrate to other places, which also helps them to get more tax preferences and looser tax enforce- ment. Hence, the tax burden of local SOEs is probably higher than that of central SOEs. Furthermore, there is also a possibility that the tax burden differs between local SOEs controlled by different levels of local governments. The 1994 tax reform did not touch on the fiscal relationship between governments below provincial level. Up to now, most local governments below the provincial level still employ a mixed tax revenue distribution system based on a fixed quota and sharing rate (Cai, 2007). Because the fiscal relationship between local governments is always determined completely by higher level governments, when facing financial stress provincial governments will shift the pressure to city level governments and city level govern- ments will follow suit to shift it further downward (Zhou, 2008). As a result, lower- level governments usually face even tougher financial conditions and have a strong motivation to grab tax revenues from enterprises. So the tax burden of local SOEs con- trolled by lower level governments may be higher than that of local SOEs controlled by higher ones. We summarize this argument as our second hypothesis: H2: The tax burden of local SOEs is higher than that of central SOEs, and the lower the local governments’ level, the higher the tax burden of SOEs under their control. 3. Research design and sample selection 3.1. The content of corporate tax burden Up till now, most of the corporate tax burden research in China has focused on EIT, ignoring turnover taxes. However, we argue that corporate tax burden should include all taxes paid to tax authorities. Panel A of Table 1 shows the major sources of China’s tax revenue after the 1994 tax sharing reform. Taxes paid by firms, especially the turnover taxes of VAT and BT account for a large proportion of total revenue. In recent years, the proportion of EIT is increasing, but is still lower than turnover taxes. From Panel B, we can see that individual income tax, social insurance tax and EIT have been 174 Liu and Liu Table 1. Tax revenue structure of US and China, by major sources, as a percentage of the total. Panel A: Major sources of China’s tax revenues 1994–2010 Individual income Year VAT Consumption tax BT EIT tax Others Total 1994 52.48% 10.18% 13.41% 13.56% 1.43% 8.93% 100.00% 1995 50.87% 9.47% 14.55% 13.85% 2.20% 9.05% 100.00% 2000 48.55% 6.93% 14.89% 13.98% 5.21% 10.44% 100.00% 2005 48.17% 5.46% 13.71% 17.85% 6.78% 8.02% 100.00% 2010 42.26% 9.24% 15.24% 17.54% 6.61% 9.12% 100.00% Panel B: Major sources of US tax revenues 1960-2010 Year Individual EIT Social Excise Others Total income insurance tax tax tax 1960 44.00% 23.20% 15.90% 12.60% 4.30% 100.00% 1970 46.90% 17.00% 23.00% 8.10% 5.00% 100.00% 1980 47.20% 12.50% 30.50% 4.70% 5.10% 100.00% 1990 45.20% 9.10% 36.80% 3.40% 5.50% 100.00% 2000 49.60% 10.20% 32.20% 3.40% 4.60% 100.00% 2010 41.50% 8.90% 40.00% 3.10% 6.50% 100.00% Note: Data Sources of Panel A: China Statistical Yearbook 2011, Tax Yearbook of China 1994-2010;Data Sources of Panel B: U.S. Congressional Budget Office; Office of Management and Budget. the main sources of US federal revenue since 1960. Individual income tax and social insurance tax, which are paid mainly by individuals, contribute more than 80% of the total nowadays, and while the proportion of EIT revenue paid by corporations is gradu- ally decreasing, it still accounts for nearly 10%. From Table 1, we can see that the main difference in the tax systems between the US and China is that under the current US tax system, government revenue mainly comes from the taxes paid by individuals. The taxes paid by firms are less important and EIT is the major tax paid by firms. For China, however, government revenue mainly comes from the taxes paid by firms, especially turnover taxes. World Bank (2006) surveys the investment climate of 120 cities in China and finds that the turnover tax burden of firms in China is significantly higher than their EIT burden. The smallest gap exists in Southeast China, where firms’ VAT burden is 3.5 times as high as their EIT burden. The largest gap exists in Northwest China, where VAT burden is 8.2 times as high as the EIT burden. Therefore, we argue that it is not appropriate to merely focus on EIT in the corpo- rate tax burden research in China, and it is necessary to incorporate turnover taxes into the calculation of corporate tax burden. 3.2. Tax burden measurement The measurements used in prior corporate tax burden research are often ETRs, which are widely used in western studies to measure the tax burden of EIT. Only Lou (2007), Wang and Liu (2012) and Yang, Ding, and Wu (2000), construct new measurements and study the burden of taxes other than EIT, but their measurements still cannot indicate the total tax burden of the firm. To overcome the defects of these existing mea- surements, we develop a new corporate tax burden measurement based on cash flows: China Journal of Accounting Studies 175 TaxNCF Taxburden1 ¼ Sales where TaxNCF is net cash outflows for tax payments and is calculated as ‘Payments of all types of taxes – Refunds of taxes’. Under China’s GAAP, the item of Payments of all types of taxes in the cash flow statement includes almost all taxes paid by firms, such as BT, VAT, Consumption Tax, EIT, educational surcharge, stamp duties, and so on. And the item of Refunds of taxes includes almost all tax refunds the firm receives. Therefore, the difference between these two items, TaxNCF,reflects the net expenditure of the firm on tax. The value of Sales is taken from the item of sales in the income statement. By definition, tax burden equals the amount of tax divided by the amount of taxable economic source. For firms in normal operating conditions, the economic source to pay tax fees is sales revenue. So we choose Sales as the denominator of the tax burden measurement. It should be noted that because the cash flow statement is prepared on a cash basis, the value of TaxNCF may not be exactly the amount of tax expense that should be recognized in the income statement on the accrual basis. But according to China’sTax Law, after the tax assessable period ends, the tax should be paid to tax authorities within 15 days or an even shorter time, so we believe the value of TaxNCF is very close to that of tax expense. To be more prudent, we also form another corporate tax burden measurement to mitigate the difference between tax cash flows and tax expenses: TaxNCF avg Taxburden2 ¼ Sales where TaxNCF_avg is the three-year moving average of the TaxNCFs of year t–1, t and t+1. Other variables are as defined above. 3.3. Empirical model We set up the following model to test our hypotheses: Taxburden ¼ b þ b UltCtrls þ b Controls þ e (1) i;t 0 1 i;t 2 i;t where Taxburden is Taxburden1 (or Taxburden2, each tested in separate regressions). it UltCtrls are dummy variables for each type of ultimate owner, namely SOE, Central- gov, Localgov, Provgov, and Citygov. Controls are the control variables. According to prior corporate tax burden studies (Adhikari et al., 2006; Derashid & Zhang, 2003; Gupta & Newberry, 1997; Holland, 1998; Stickney & McGee, 1982; Zimmerman, 1983), we include Capint, Invint, Size, Leverage and ROA as control variables in the model to control for the capital intensity, inventory intensity, size, leverage and profitability of firms. Since VAT is levied on the basis of value added during the course of goods production or services provision, firms with high gross margins probably pay more VAT. Therefore, we add Grossmargin into the model to control for the influence of gross margin on tax burden. Furthermore, membership of a business group, cross-industry operations and cross-region operations are all helpful for the firms to avoid or evade taxes through transfer pricing, so we introduce three dummy variables – Group, CRSind and CRSrgn – to control for these characteristics of firms. In addition, for a long time the tax preference policies in China have been based on regions, so classifying firms by the regions where they are located has become a 176 Liu and Liu convention in corporate tax burden research. We divide China’s territory into five regions by different tax preference policies, specifically, (a) the Special Economic Zones and Shanghai Pudong New District (SEZs); (b) East Region; (c) Central Region; (d) West Region; and (e) Northeast Region. Then we employ a separate indicator for each region, Location, to control for the tax policy differences between regions. In addition, because in China’s tax system the main tax categories of different industries are probably not the same, and tax preference policies can differ between industries and years, we also employ separate indicators for Industry and Year as control variables. Details of variable definitions are shown in Table 2. 3.4. Sample selection Since the 1994 Tax Reform, China’s tax system has changed several times, but it was relatively stable during the 2002–2007 period. Since the calculation of Taxburden2 needs data from prior and later accounting periods, we choose the listed companies on the main board of China’s A share stock market during 2003–2006 as the initial Table 2. Variable definitions. Variable Definition Taxburden Taxburden1 = TaxNCF/sales, where TaxNCF=Payments of all types of tax-Refunds of taxes Taxburden2 = (Three-year moving average of the TaxNCFs of year t–1, t and t+1)/ sales Ultctrls SOE = A dummy variable that equals 1 for firms controlled by the government, and 0 otherwise. Centralgov = A dummy variable that equals 1 for firms controlled by the central government, and 0 otherwise. Localgov = A dummy variable that equals 1 for firms controlled by local governments, and 0 otherwise. Provgov = A dummy variable that equals 1 for firms controlled by provincial governments, and 0 otherwise. Citygov = A dummy variable that equals 1 for firms controlled by city governments, and 0 otherwise. controls Capint = Year-end net value of fixed assets/Year-end total assets Invint = Year-end net value of inventory/Year-end total assets Size = Natural logarithm of year-end total assets Leverage = Year-end total liabilities/Year-end total assets ROA = Net income/Year-end total assets Grossmargin = (Sales-cost of sales)/sales Group = A dummy variable that equals 1 if the firm is a member of a business group, and 0 otherwise. CRSind = A dummy variable that equals 1 if a firm’s sales come from more than one industry, and 0 otherwise. CRSrgn = A dummy variable that equals 1 if a firm’s sales come from more than one region, and 0 otherwise. Location = A separate indicator for the regions of SEZs, East Region; Central Region; West Region; Northeast Region. Industry = A separate indicator for industries classified according to the Guidelines for the Industry Classification of Listed Companies issued by China Securities Regulatory Commission. Firms in the manufacturing industry are classified by their two-digital code. Year = A separate indicator for the years from 2003 to 2006. China Journal of Accounting Studies 177 Table 3. Sample selection procedure and composition. Panel A: Sample selection procedure Year 2003 2004 2005 2006 Total Initial sample 1,268 1,356 1,352 1,435 5,411 Less: financial companies 15 15 15 20 65 Less: companies whose 53 3 3 14 ultimate owners cannot be identified Less: companies whose 228 234 235 235 932 ultimate owners changed during the sample period Less: companies which 300 297 281 278 1,156 have been ST or PT since listed Less: observations with 322 368 302 312 1,304 missing values Final Sample 398 439 516 587 1,940 Panel B: Sample distribution by type of ultimate owner and year Year 2003 2004 2005 2006 Total Percentage Non-SOEs 88 106 145 170 509 26.24% Central SOEs 90 102 111 119 422 21.75% Provincial SOEs 120 125 144 158 547 28.20% City SOEs 100 106 116 140 462 23.81% Total 398 439 516 587 1,940 100.00% sample. We then exclude: (a) financial companies whose industrial code is ‘I’ in the Guidelines for the Industry Classification of Listed Companies issued by China Securities Regulatory Commission; (b) companies whose ultimate owners cannot be identified; (c) companies whose ultimate owners have changed from one type to another during the sample period, because the tax preferences enjoyed by the companies will not change immediately even after their ultimate owners have changed; (d) companies that have been specially treated (ST) or particularly transferred (PT) since they are listed, because ST and PT companies are often not in normal operating conditions, and their tax burdens are also likely to be abnormal; and (e) observations with missing values. Panel A of Table 3 presents the sample selection procedure, the final sample includes 1940 firm-year observations. Panel B of Table 3 reports the year distribution of the observations for the four types of firms: non-SOEs, central SOEs, provincial SOEs, and city SOEs. The numbers of them are similar, and the number of SOE observations accounts for 73.76% (21.75%+28.20%+23.81%) of the total, indicat- ing that the economy of China is still dominated by SOEs. All data used in this paper are from the China Stock Market Research (CSMAR) database. 4. Empirical results 4.1. Descriptive statistics Table 4 reports the descriptive statistics of the sample. It shows that the mean and median Taxburden1 of non-SOEs are both larger than those of the full sample, while the mean and median Taxburden1 of central SOEs are smaller than those of the full sample. Compared with non-SOEs, the mean and median Taxburden1 are higher for local SOEs, 178 Liu and Liu Table 4. Descriptive statistics. Variables Full sample Non-SOE SOE Central SOE Local SOE Provincial SOE City SOE Taxburden1 0.062 0.063 0.062 0.051 0.067 0.065 0.070 [0.051] [0.052] [0.051] [0.041] [0.055] [0.055] [0.056] (0.059) (0.055) (0.060) (0.053) (0.063) (0.066) (0.058) Taxburden2 0.063 0.064 0.063 0.051 0.067 0.065 0.070 [0.053] [0.057] [0.051] [0.041] [0.055] [0.052] [0.057] (0.057) (0.054) (0.059) (0.051) (0.061) (0.066) (0.055) Capint 0.305 0.277 0.314 0.288 0.326 0.321 0.331 [0.283] [0.271] [0.287] [0.250] [0.304] [0.298] [0.309] (0.177) (0.148) (0.186) (0.186) (0.185) (0.190) (0.179) Invint 0.170 0.180 0.167 0.164 0.168 0.158 0.181 [0.138] [0.134] [0.139] [0.152] [0.132] [0.119] [0.142] (0.141) (0.140) (0.142) (0.116) (0.151) (0.150) (0.152) Size 21.470 21.170 21.570 21.550 21.580 21.740 21.400 [21.390] [21.110] [21.480] [21.340] [21.520] [21.650] [21.390] (0.908) (0.776) (0.928) (1.127) (0.832) (0.884) (0.724) Leverage 0.480 0.483 0.479 0.471 0.483 0.464 0.505 [0.492] [0.502] [0.490] [0.479] [0.492] [0.470] [0.528] (0.168) (0.162) (0.170) (0.176) (0.168) (0.169) (0.163) ROA 0.044 0.045 0.044 0.045 0.044 0.048 0.038 [0.037] [0.041] [0.036] [0.034] [0.036] [0.043] [0.031] (0.047) (0.042) (0.048) (0.050) (0.047) (0.049) (0.045) Grossmargin 0.241 0.249 0.239 0.223 0.246 0.259 0.230 [0.206] [0.219] [0.203] [0.193] [0.207] [0.218] [0.198] (0.143) (0.134) (0.146) (0.135) (0.150) (0.163) (0.130) Group 0.791 0.768 0.799 0.893 0.760 0.887 0.610 [1.000] [1.000] [1.000] [1.000] [1.000] [1.000] [1.000] (0.407) (0.422) (0.401) (0.309) (0.427) (0.317) (0.488) CRSind 0.914 0.941 0.904 0.900 0.906 0.901 0.911 [1.000] [1.000] [1.000] [1.000] [1.000] [1.000] [1.000] (0.281) (0.236) (0.294) (0.300) (0.292) (0.299) (0.285) CRSrgn 0.926 0.986 0.904 0.960 0.881 0.887 0.874 [1.000] [1.000] [1.000] [1.000] [1.000] [1.000] [1.000] (0.262) (0.117) (0.294) (0.197) (0.324) (0.317) (0.332) Observations 1940 509 1431 422 1009 547 462 Note: The medians are reported in square brackets and standard errors are reported in parentheses. All variables are winsorized at the 1st and 99th percentiles. in which city SOEs bear the highest tax-burden among firms controlled by different types of owners. Results are similar if we use Taxburden2 as the measurement. The descriptive statistics for the control variables show that the mean and median capital intensity of non-SOEs are lower than those of SOEs, consistent with the fact that most non-SOEs in China cluster in labor-intensive or technology-intensive indus- tries. Central SOEs have lower capital intensity than local SOEs. Inventory intensity is higher in non-SOEs and lower in SOEs. In addition, the size of non-SOEs is smaller than that of SOEs. Provincial SOEs have the largest size and central SOEs next, while city SOEs have the smallest size among SOEs. Firms controlled by different types of owners have a similar debt ratio. Profitability is highest in provincial SOEs and lowest in city SOEs. We also learn that 80% of the sample is controlled by a group. This percentage is smaller in non-SOEs and larger in central/provincial SOEs. More than 90% of the sample firms operate in multiple industries and regions. China Journal of Accounting Studies 179 Table 5 reports the correlations for the variables. The two measurements for tax burden, Taxburden1 and Taxburden2, are highly correlated, consistent with our conjec- ture that the tax amount from the cash flow statement is similar to the tax expense. The SOE dummy is negatively correlated with tax burden but is not significant. It seems that the tax burden of the entire SOE sample is not significantly different from that of non-SOEs. But Centralgov has a negative correlation with tax burden while Localgov has a positive correlation, indicating that the tax burden is smaller for central SOEs and larger for local SOEs. Provgov is positively correlated with tax burden but is not signif- icant, while the correlation between Citygov and tax burden is positive and significant, suggesting that city SOEs have a larger tax burden. In addition, Capint is positively correlated with tax burden, possibly due to the regulation that VAT associated with new facilities is not deductible. Invint is negatively correlated with Taxburden1. The correla- tion of Size and tax burden is significantly positive, consistent with the ‘Political Cost’ argument. Leverage is negatively correlated with tax burden, implying that financial leverage acts as a tax shield. The profitability indicators (ROA and Grossmargin) are both positively correlated with tax burden. In addition, the correlations between CRS- ind/CRSrgn and tax burden are significantly negative, suggesting that cross-industry and cross-region firms bear less tax. The correlation between Group and tax burden is not significant. 4.2. Multiple regressions To test the hypotheses in a multivariate setting, we first estimate model (1) using ordinary least squares (OLS). Here and throughout, all the regressions are estimated using robust standard errors that are adjusted for clustering at the firm level. Results presented in Table 6 show that the two tax burden indicators, Taxburden1 and Taxbur- den2, yield similar estimates from the model with high goodness-of-fits observed from Adj-R , indicating that Taxburden1 and Taxburden2 are good measurements of corporate tax burden. Columns (1) and (4) show that the coefficients on SOE are signif- icantly negative, suggesting that the tax burden of SOEs is lower than that of non-SOEs, consistent with H1a. To test the differences in tax burdens among firms controlled by different levels of government, we use dummy variables representing the identity of the ultimate owners in the model. In Table 6, Columns (2) and (5), we document a significant negative coefficient of the central SOE indicator, Centralgov, indicating that central SOEs bear a lower tax burden than non-SOEs. The coefficient of the local SOE indicator, Localgov, is negative but insignificant. But when we classify local SOEs into provincial SOEs and city SOEs in Columns (3) and (6), the coefficient of Centralgov is significantly negative with a p-value<0.01, while the provincial SOE indicator, Provgov, is also negatively related to tax burden, significant at the 5% level. The coefficient of the city SOE indicator, Citygov, is insignificant. These results suggest that the lower tax burden of SOEs is mainly driven by central SOEs and provincial SOEs, while the tax burden of city SOEs is similar to that of non-SOEs. To test H2 and further analyze the relationship between government control and corporate tax burden, we remove the non-SOE sample from the full sample and re-estimate model (1) using only the SOE sample. Central SOEs are used as the base group when comparing the tax burden between firms controlled by different levels of government. Table 7 displays the results. Columns (1) and (4) show that Localgov is positively related to tax burden, significant at the 5% level. It suggests that local SOEs 180 Liu and Liu Table 5. Pearson and Spearman correlation matrix. Taxburden1 Taxburden2 SOE Centralgov Localgov Provgov Citygov Capint Invint Size Leverage ROA Grossmargin Group CRSind CRSrgn *** *** *** *** *** *** *** *** *** *** *** *** Taxburden1 0.9557 –0.0087 –0.1248 0.0954 0.0313 0.0788 0.2023 –0.2043 0.0769 –0.2000 0.3077 0.6740 –0.0235 –0.0786 –0.1700 *** *** *** *** *** *** *** *** *** *** *** *** Taxburden2 0.9472 –0.0238 –0.1365 0.0917 0.0225 0.0837 0.1955 –0.2015 0.0801 –0.2026 0.3120 0.6812 –0.0258 –0.0656 –0.1766 *** *** *** *** *** ** *** ** ** ** *** SOE –0.0011 –0.0126 0.3145 0.6209 0.3737 0.3334 0.0688 –0.0515 0.1884 –0.0113 –0.0455 –0.0543 0.0338 –0.0577 –0.1376 *** *** *** *** *** *** *** *** *** *** Centralgov –0.1045 –0.1092 0.3145 –0.5489 –0.3304 –0.2948 –0.0725 0.0201 0.0078 –0.0258 –0.0236 –0.0737 0.1325 –0.0253 0.0683 *** *** *** *** *** *** *** *** *** *** *** Localgov 0.0854 0.0792 0.6209 –0.5489 0.6019 0.5370 0.1204 –0.0619 0.1595 0.0113 –0.0206 0.0130 –0.0796 –0.0300 –0.1775 *** *** *** *** ** *** *** *** ** ** *** *** Provgov 0.0276 0.0209 0.3737 –0.3304 0.6019 –0.3503 0.0515 –0.0952 0.2005 –0.0654 0.054 0.0452 0.1471 –0.0282 –0.0935 *** *** *** *** *** *** *** *** *** *** *** Citygov 0.0710 0.0708 0.3334 –0.2948 0.5370 –0.3503 0.0869 0.0279 –0.0248 0.0823 –0.0811 –0.0325 –0.2488 –0.0053 –0.1094 *** *** *** ** *** ** *** *** *** *** *** *** *** *** Capint 0.2248 0.2329 0.0918 –0.0510 0.1230 0.0569 0.0841 –0.5001 0.1006 –0.1190 0.0849 0.1134 –0.0348 –0.0783 –0.1152 * * ** * *** *** *** *** *** Invint –0.0426 –0.0355 –0.0409 –0.0254 –0.0151 –0.0552 0.0406 –0.5278 –0.0298 0.2869 –0.0841 –0.1554 –0.0228 0.0355 0.1536 *** *** *** ** *** *** * *** *** *** ** Size 0.0984 0.1037 0.1962 0.0507 0.1309 0.1864 –0.0435 0.1623 0.0075 0.3291 0.1109 –0.0334 0.0358 –0.0085 –0.0506 *** *** *** *** *** *** *** *** *** *** *** ** Leverage –0.1967 –0.1981 –0.0090 –0.0293 0.0163 –0.0597 0.0821 –0.1064 0.2890 0.3038 –0.3083 –0.2948 –0.1133 0.0784 0.0577 *** *** ** *** *** *** *** *** *** *** ** ROA 0.3071 0.3142 –0.0129 0.0050 –0.0155 0.0491 –0.0700 0.1115 –0.0596 0.1528 –0.3182 0.3724 0.0217 –0.0861 –0.0459 *** *** *** *** ** *** *** *** *** *** Grossmargin 0.6697 0.6736 –0.0298 –0.0677 0.0297 0.0754 –0.0449 0.1316 –0.0931 –0.0029 –0.3096 0.3603 –0.0209 –0.0405 –0.0979 *** *** *** *** * *** Group –0.0036 0.0032 0.0338 0.1325 –0.0796 0.1471 –0.2488 –0.0243 –0.0093 0.0410 –0.1110 0.0204 –0.0118 0.0006 0.0142 *** *** ** *** *** *** ** *** CRSind –0.1035 –0.0943 –0.0577 –0.0253 –0.0300 –0.0282 –0.0053 –0.0898 0.0365 –0.0243 0.0868 –0.1173 –0.0554 0.0006 0.0743 *** *** *** *** *** *** *** *** ** ** ** * *** *** CRSrgn –0.1830 –0.1850 –0.1376 0.0683 –0.1775 –0.0935 –0.1094 –0.1370 0.0579 –0.0473 0.0564 –0.0400 –0.1021 0.0142 0.0743 *** ** * Note: Pearson correlations are reported on the bottom left and Spearman correlations on the upper right. , and denote statistical significance at the 1%, 5%, 10% levels, respectively. China Journal of Accounting Studies 181 Table 6. Multiple regression: types of ultimate owners and corporate tax burden (full sample). Dependent variable: Taxburden1 Dependent variable: Taxburden2 Variables (1) (2) (3) (4) (5) (6) * ** SOE –0.0059 –0.0074 (–1.919) (–2.298) *** *** *** *** Centralgov –0.0116 –0.0123 –0.0130 –0.0137 (–2.900) (–3.063) (–3.180) (–3.344) Localgov –0.0036 –0.0052 (–1.106) (–1.501) ** ** Provgov –0.0084 –0.0101 (–2.174) (–2.468) Citygov 0.0014 –0.0001 (0.356) (–0.030) *** *** *** *** *** *** Capint 0.0371 0.0353 0.0332 0.0425 0.0407 0.0386 (3.696) (3.515) (3.276) (3.999) (3.826) (3.609) *** *** *** *** *** *** Invint 0.0421 0.0431 0.0416 0.0444 0.0454 0.0438 (2.950) (3.014) (2.903) (3.186) (3.247) (3.148) ** ** ** ** ** ** Size 0.0033 0.0033 0.0040 0.0035 0.0035 0.0042 (1.984) (2.020) (2.372) (2.047) (2.084) (2.429) * * ** * * ** Leverage –0.0194 –0.0194 –0.0215 –0.0202 –0.0202 –0.0223 (–1.844) (–1.848) (–2.023) (–1.883) (–1.888) (–2.066) ROA 0.0356 0.0380 0.0353 0.0474 0.0498 0.0471 (1.102) (1.192) (1.112) (1.501) (1.598) (1.520) *** *** *** *** *** *** Grossmargin 0.2421 0.2407 0.2432 0.2302 0.2289 0.2314 (17.955) (17.652) (17.581) (17.299) (17.025) (16.947) Group 0.0021 0.0030 0.0052 0.0027 0.0037 0.0058 (0.642) (0.918) (1.531) (0.822) (1.080) (1.699) ** ** ** * CRSind –0.0097 –0.0099 –0.0101 –0.0074 –0.0076 –0.0078 (–2.009) (–2.054) (–2.094) (–1.597) (–1.644) (–1.687) CRSrgn –0.0064 –0.0048 –0.0046 –0.0056 –0.0041 –0.0038 (–1.080) (–0.821) (–0.786) (–0.925) (–0.680) (–0.644) *** *** *** *** *** *** Constant –0.1087 –0.1106 –0.1243 –0.1109 –0.1127 –0.1266 (–3.114) (–3.179) (–3.551) (–3.132) (–3.200) (–3.570) Location YES YES YES YES YES YES Industry YES YES YES YES YES YES Year YES YES YES YES YES YES Observations 1,940 1,940 1,940 1,940 1,940 1,940 Adj-R 0.596 0.598 0.601 0.615 0.617 0.620 Note: All t-statistics, reported in parenthesis, are based on standard errors adjusted for heteroskedasticity and *** ** * firm-level clustering. , and denote statistical significance at the 1%, 5%, 10% levels, respectively. bear a significantly higher tax burden than central SOEs. Among local SOEs, the coefficient on Provgov is insignificant while the coefficient of Citygov is significantly positive with a p-value <0.01, as shown in Columns (2) and (5), indicating that the tax burden of provincial SOEs is not much different from that of central SOEs, while city SOEs have a much higher tax burden than central SOEs. We further remove the central SOE sample and retain the local SOE sample to compare the tax burden of provincial SOEs and city SOEs. Provincial SOEs are used as the base group and the results are displayed in Columns (3) and (6) of Table 7. The coefficient of Citygov is positive and significant at the 10% level, suggesting that the tax burden of city SOEs is higher than provincial SOEs. Thus, H2 is supported by the results. 182 Liu and Liu Table 7. Multiple regression: Types of ultimate owners and corporate tax burden (SOE subsample). Dependent variable: Taxburden1 Dependent variable: Taxburden2 Variables (1) (2) (3) (4) (5) (6) ** ** Localgov 0.0089 0.0088 (2.241) (2.165) Provgov 0.0053 0.0050 (1.275) (1.179) *** * *** * Citygov 0.0141 0.0075 0.0141 0.0080 (2.883) (1.712) (2.868) (1.801) *** *** *** *** *** *** Capint 0.0352 0.0334 0.0415 0.0388 0.0369 0.0470 (3.205) (3.019) (2.846) (3.497) (3.329) (3.195) ** ** ** ** Invint 0.0353 0.0334 0.0285 0.0371 0.0353 0.0287 (2.248) (2.133) (1.428) (2.334) (2.228) (1.436) ** *** ** *** Size 0.0043 0.0050 0.0034 0.0048 0.0055 0.0041 (2.304) (2.638) (1.325) (2.549) (2.894) (1.624) Leverage –0.0167 –0.0184 –0.0170 –0.0175 –0.0193 –0.0163 (–1.383) (–1.510) (–1.126) (–1.430) (–1.561) (–1.054) ** * ROA 0.0548 0.0515 0.0959 0.0478 0.0444 0.0938 (1.489) (1.399) (2.022) (1.295) (1.207) (1.943) *** *** *** *** *** *** Grossmargin 0.2372 0.2403 0.2399 0.2252 0.2283 0.2267 (14.935) (14.800) (11.992) (14.504) (14.374) (11.235) Group 0.0018 0.0044 0.0013 0.0024 0.0051 0.0023 (0.412) (0.997) (0.282) (0.538) (1.132) (0.513) CRSind –0.0064 –0.0066 –0.0074 –0.0046 –0.0048 –0.0062 (–1.328) (–1.356) (–1.221) (–0.999) (–1.030) (–1.045) CRSrgn –0.0053 –0.0051 –0.0038 –0.0053 –0.0051 –0.0032 (–0.908) (–0.884) (–0.550) (–0.896) (–0.872) (–0.463) *** *** ** *** *** ** Constant –0.1394 –0.1557 –0.1184 –0.1473 –0.1641 –0.1328 (–3.492) (–3.874) (–2.189) (–3.725) (–4.135) (–2.522) Location YES YES YES YES YES YES Industry YES YES YES YES YES YES Year YES YES YES YES YES YES Observations 1,431 1,431 1,009 1,431 1,431 1,009 Adj-R 0.635 0.638 0.644 0.652 0.655 0.660 Note: Central SOEs are the base group in Columns (1), (2), (4) and (5). Provincial SOEs are the base group in Columns (3) and (6). All t-statistics, reported in parenthesis, are based on standard errors adjusted for heter- *** ** * oskedasticity and firm-level clustering. , and denote statistical significance at the 1%, 5%, 10% levels, respectively. 4.3. Additional tests 4.3.1. An examination of tax burden components: tax payments and tax refunds Our tax burden measures consist of two components: tax payments and tax refunds. To examine whether the differences in tax burden come from tax payments or tax refunds, we use two variables, Taxpay and Taxrfd, as the proxies for tax payments and tax refunds, respectively, and compare the tax burden components in firms controlled by different owners. Taxpay is computed as tax payments divided by sales, while Taxrfd is measured as tax refunds divided by sales. Taxpay and Taxrfd are used as the explained variables to re-estimate model (1). The results are presented in Table 8. Table 8 shows that when using Taxpay as the dependent variable, the coefficient of SOE is not significant, suggesting that there is no difference in tax payments between SOEs and non-SOEs. Centralgov is negatively associated with Taxpay, significant at China Journal of Accounting Studies 183 Table 8. Multiple regression: types of ultimate owners and corporate tax payments / refunds. Dependent variable: Taxpay Dependent variable: Taxrfd Variables (1) (2) (3) (4) (5) (6) *** SOE –0.0019 0.0039 SOE (–0.745) (2.669) ** ** ** ** Centralgov –0.0069 –0.0072 0.0049 0.0052 (–2.130) (–2.232) (2.359) (2.539) ** Localgov 0.0000 0.0035 (0.008) (2.230) *** Provgov –0.0024 0.0060 (–0.748) (3.014) Citygov 0.0025 0.0008 (0.764) (0.480) *** *** *** ** ** ** Capint 0.0265 0.0250 0.0239 –0.0109 –0.0106 –0.0095 (3.142) (2.996) (2.824) (–2.432) (–2.310) (–2.124) *** *** *** Invint 0.0331 0.0340 0.0332 –0.0070 –0.0072 –0.0064 (2.775) (2.855) (2.771) (–1.040) (–1.058) (–0.951) * * ** Size 0.0027 0.0027 0.0031 –0.0002 –0.0002 –0.0006 (1.906) (1.944) (2.119) (–0.299) (–0.309) (–0.794) ** ** ** Leverage –0.0118 –0.0118 –0.0128 0.0078 0.0078 0.0089 (–1.251) (–1.251) (–1.345) (2.001) (2.006) (2.245) ROA 0.0179 0.0199 0.0186 –0.0189 –0.0193 –0.0179 (0.655) (0.742) (0.691) (–1.283) (–1.306) (–1.226) *** *** *** Grossmargin 0.2423 0.2411 0.2424 0.0067 0.0069 0.0055 (19.605) (19.406) (19.354) (1.160) (1.182) (0.941) Group 0.0010 0.0018 0.0029 –0.0006 –0.0008 –0.0019 (0.392) (0.715) (1.116) (–0.364) (–0.456) (–1.133) ** ** ** CRSind –0.0092 –0.0094 –0.0095 –0.0016 –0.0015 –0.0014 (–2.236) (–2.272) (–2.288) (–0.759) (–0.750) (–0.690) CRSrgn –0.0045 –0.0032 –0.0030 0.0028 0.0025 0.0024 (–0.872) (–0.618) (–0.596) (1.410) (1.266) (1.209) *** *** *** * Constant –0.0780 –0.0796 –0.0865 0.0216 0.0219 0.0292 (–2.630) (–2.696) (–2.893) (1.340) (1.360) (1.797) Location YES YES YES YES YES YES Industry YES YES YES YES YES YES Year YES YES YES YES YES YES Observations 1,940 1,940 1,940 1,940 1,940 1,940 Adj-R 0.622 0.624 0.625 0.158 0.159 0.167 Note: All t-statistics, reported in parenthesis, are based on standard errors adjusted for heteroskedasticity and *** ** * firm-level clustering , and denote statistical significance at the 1%, 5%, 10% levels, respectively. the 5% level, indicating that the tax payments of central SOEs are lower than non-SOEs. The coefficients of Localgov, Provgov and Citygov are insignificant, implying that there are no significant differences in tax payments between local SOEs and non-SOEs, or between provincial/city SOEs and non-SOEs. When Taxrfd is used as the dependent variable, the coefficient on SOE is positive and significant, suggesting that the tax refunds of SOEs are higher than non-SOEs. Both Centralgov and Localgov are positively related to Taxrfd, indicating that among SOEs the tax refunds of central and provincial SOEs are significantly higher than non-SOEs. More specifically, the coefficient of Provgov is positive and significant, while the coefficient of Citygov is insignificant, suggesting that the main difference in tax refunds between local SOEs and non-SOEs is driven by provincial SOEs. Overall, 184 Liu and Liu the results in Table 8 indicate that the impact of government control on corporate tax burdens mainly comes from their influence on tax refunds, implying that the prohibition against unauthorized tax refunds has not been fully implemented. 4.3.2. An examination by major tax categories: VAT, business tax and EIT The corporate tax burden of Chinese firms consists of three major types of tax, VAT, BT and EIT. In order to identify from which types of tax the difference in tax burden arises, we further separate the tax burden into VAT burden (VATburden), BT burden (BTburden) and EIT burden (EITburden). The variables are defined as follows and all financial data needed are from the CSMAR database: VATburden ¼ðTaxNCF  EITNCF  B&SNCFÞ=Sales BTburden ¼ BTNCF=Sales EITburden ¼ EITNCF=Sales where EITNCF, B&SNCF and BTNCF are the net cash outflows of EIT, Business Taxes and Surcharges, and BT. EITNCF =(Income tax axDeferred income tax – ΔIn- come tax payable); B&SNCF =(Business taxes and surcharges – ΔBusiness taxes and surcharges payable), Business taxes and surcharges payable = Tax payable – Income tax payable – VAT payable; BTNCF = BT – ΔBT payable. Other variables are the same as defined above. The change (Δ) is computed between year t and t–1. These measurements are used to re-estimate model (1) and the results presented in Table 9 are similar to those in Table 6 when the VAT indicator (VATburden) is used as the dependent variable. The VAT burden of SOEs is lower than non-SOEs. Among SOEs, the VAT burden of central and provincial SOEs is significantly lower than non- SOEs and there is no significant difference between local SOEs and non-SOEs. But the BT burden (BTburden) shows no significant difference between non-SOEs and SOEs controlled by different levels of government. In terms of the EIT burden, the difference between SOEs and non-SOEs is not significant. This indicates that we may find differ- ent results if we ignore turnover taxes when analyzing the corporate tax burden. Central SOEs bear a lighter EIT burden than non-SOEs, consistent with the findings of Cao et al. (2009). Differences in EIT burdens between provincial SOEs, city SOEs and non- SOEs are not significant. These results suggest that the findings presented in Section 4.2 are mainly driven by the value added tax burden. It also suggests that results would be biased if we only focus on EIT when analyzing the tax burden of a firm. 4.4. Robustness checks 4.4.1. Deleting loss firms Firms do not need to pay EIT when they incur a loss and their tax burden may be biased due to their abnormal operating conditions. Hence, we remove the 157 firm-year observations with a loss and re-estimate model (1). The results are similar to the previous findings (see Table 10). China Journal of Accounting Studies 185 Table 9. Multiple regression: types of ultimate owners and corporate tax categories. Dependent variable: VATburden Dependent variable: BTburden Dependent variable: EITburden Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) ** SOE –0.0059 –0.0001 –0.0008 (–2.347) (–0.108) (–0.537) * * ** ** Centralgov –0.0055 –0.0061 –0.0016 –0.0014 –0.0040 –0.0040 (–1.691) (–1.771) (–1.490) (–1.370) (–2.473) (–2.468) ** Localgov –0.0061 0.0006 0.0005 (–2.266) (0.554) (0.316) *** Provgov –0.0104 0.0015 0.0003 (–3.100) (1.259) (0.171) Citygov –0.0015 –0.0002 0.0007 (–0.487) (–0.207) (0.382) Controls YES YES YES YES YES YES YES YES YES ** ** * *** *** *** Constant 0.0074 0.0075 –0.0052 –0.0197 –0.0202 –0.0177 –0.0644 –0.0654 –0.0660 (0.250) (0.254) (–0.176) (–2.024) (–2.087) (–1.837) (–3.415) (–3.533) (–3.629) Observations 1,753 1,753 1,753 1,013 1,013 1,013 1,766 1,766 1,766 Adj-R 0.444 0.444 0.449 0.639 0.641 0.642 0.525 0.528 0.528 *** ** * Note: All t-statistics, reported in parenthesis, are based on standard errors adjusted for heteroskedasticity and firm-level clustering. , and denote statistical significance at the 1%, 5%, 10% levels, respectively. 186 Liu and Liu Table 10. Multiple regression: Types of ultimate owners and corporate tax burden (removing loss firms). Dependent variable: Dependent variable: Taxburden1 Taxburden2 Variables (1) (2) (3) (4) (5) (6) * ** SOE –0.0052 –0.0067 (–1.668) (–2.076) *** *** *** *** Centralgov –0.0118 –0.0124 –0.0131 –0.0138 (–2.813) (–2.950) (–3.100) (–3.231) Localgov –0.0024 –0.0040 (–0.725) (–1.154) * * Provgov –0.0067 –0.0082 (–1.676) (–1.951) Citygov 0.0019 0.0002 (0.486) (0.049) Controls YES YES YES YES YES YES *** *** *** *** *** *** Constant –0.1089 –0.1115 –0.1232 –0.1130 –0.1156 –0.1270 (–3.048) (–3.132) (–3.432) (–3.152) (–3.242) (–3.541) Observations 1,783 1,783 1,783 1,783 1,783 1,783 Adj-R 0.611 0.614 0.616 0.630 0.633 0.635 Note: All t-statistics, reported in parenthesis, are based on standard errors adjusted for heteroskedasticity and *** ** * firm-level clustering. , and denote statistical significance at the 1%, 5%, 10% levels, respectively. 4.4.2. Median regressions Although we winsorize all variables at their 1st and 99th percentile values, we also estimate the model using median regressions with respect to departures from the median. We obtain similar results as above (not reported). 4.4.3. Alternative indicators Our second tax burden indicator, Taxburden2, is computed using the three-year moving average of net tax cash flows as the numerator. To be consistent with the way we calculate Taxburden2, we use the moving-average sales of year t–1, t and t+1 as the denominator to re-calculate Taxburden2 as the explanatory variable. The untabulated results are similar to those obtained in the previous section. We also use the ratio of active debt to assets as an alternative debt ratio and return on equity as an alternative to return on assets as control variables in the regression model. The untabulated results remain the same. 5. Conclusion In this paper, we examine the effect of government control on the corporate tax burden of A-share listed companies in China during 2003–2006. We introduce a new corporate tax burden measurement based on tax cash flows: the ratio between net cash outflows of tax payments and the sales of the corporation. This allows us to include the turnover tax burden of firms, which is important because turnover taxes are the main component of taxes for most firms in China. We also classify SOEs into three categories by the level of government as their ultimate owners. The results show that the tax burden of SOEs is significantly lower than non-SOEs, indicating that in China non-SOEs face tax China Journal of Accounting Studies 187 discrimination. Among the SOEs, the tax burden of central SOEs is lower than that of local SOEs. The tax burden of provincial SOEs is also lower and not significantly different from that of central SOEs, but the tax burden of city SOEs is higher than that of other SOEs and is not significantly different from that of non-SOEs. Furthermore, we find that the results of this paper are mainly derived from differences in tax refunds and the VAT burden of firms, indicating that the prohibition against unauthorized tax refunds has not been fully implemented, and more importantly, the results are biased if corporate tax burden research only focuses on EIT. In summary, the findings of this paper show that government control is helpful for firms to obtain a lower tax burden and the effect of control by different levels of government on corporate tax burdens is not the same. We interpret these findings as the result of tax competition and tax grabbing by local governments under the highly centralized fiscal system in China. This also indicates that the development of China’s market economy is still at a low level and China still has a long way to go to transform its government functions and provide a fair economic environment for enterprises. Acknowledgements This study is supported by the National Science Foundation of China (71272079) and the Ministry of Education (MOE) Project of Key Research Institute of Humanities and Social Sciences in Universities (11JJD790032). We would like to offer most sincere thanks to Kangtao Ye for his constructive comments on this paper. We are also very grateful for the valuable comments and suggestions made by Liansheng Wu, Jason Xiao, two anonymous reviewers, and the editors. We would like to thank the organizers and other participants of the First Annual Conference of the China Journal of Accounting Studies in Chengdu, China,2012. We also appreciate the excellent help from the language advisor, John Nowland from City University of Hong Kong. Notes 1. As reported in the Audit Report (No. 12 of 2010) published by China’s National Audit Office (CNAO), in order to attract investment, several local governments introduced their own policies of tax reduction, and refunded a large amount of tax revenue to enterprises in 2008 and 2009. 2. It is generally thought that private capital also has high mobility, but overall the investment climate in China for private capitals is poor. Entrepreneurs of non-SOEs are very cautious about investing in other regions unless they have friends in the local governments there (Lu & Pan, 2009). Some local governments even employ the ‘coax and coerce’ strategy for private capital (Zhou, 2008). Therefore, the mobility of private capital is probably not as high as generally expected and it is not as helpful to lower corporate tax burden as imagined. 3. Based on the analysis above, the tax expenses here and after include VAT paid by enterprises. 4. In China, SEZs include Shenzhen, Zhuhai, Shantou, Xiamen, and Hainan province. We also include Shanghai Pudong New District in the SEZs because of its special economic and political importance; The East Region includes nine provinces (municipalities) of Beijing, Tianjin, Hebei, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, and Guangdong; The Central Region includes six provinces of Shanxi, Anhui, Jiangxi, Henan, Hubei, and Hunan; The West Region includes 12 provinces (autonomous regions and municipalities) of Chongqing, Sichuan, Guizhou, Yunnan, Xizang, Shaanxi, Gansu, Ningxia, Qinghai, Xinjiang, Neimenggu, and Guangxi; The Northeast Region includes three provinces of Heilongjiang, Jilin, and Liaoning. 188 Liu and Liu References Adhikari, A., Derashid, C., & Zhang, H. (2006). Public policy, political connections, and effective tax rates: Longitudinal evidence from Malaysia. Journal of Accounting and Public Policy, 25, 574–595. Cai, H. Y. (2007). Study on fiscal relationship among the local governments in China (in Chinese). Beijing: China Financial & Economic Publishing House. Calomiris, C. W., Fisman, R., & Wang, Y. (2010). Profiting from government stakes in a command economy: Evidence from Chinese asset sales. Journal of Financial Economics, 96, 399–412. Cao, S. J., Liu, X., & Zhang, W. J. (2009). Fiscal decentralization, local government competition, and corporate effective tax. The Journal of World Economy (in Chinese), 4,69–83. Chen, B., Yu, J., Wang, X. J., & Lai, J. Q. (2008). An empirical study on the development of private listing corporation in China. Securities Market Herald (in Chinese), 4,42–47. Chen, K., Hillman, A. L., & Gu, Q. Y. (2002). Fiscal re-centralization and behavioral change of local governments: From the helping hand to grabbing hand. China Economic Quarterly (in Chinese), 4,111–130. Derashid, C., & Zhang, H. (2003). Effective tax rates and the ‘industrial policy’ hypothesis: Evidence from Malaysia. Journal of International Accounting, Auditing and Taxation, 12, 45–62. Faccio, M. (2010). Differences between politically connected and nonconnected firms: A cross-country analysis. Financial Management, 39, 905–928. Fama, E., & Jensen, M. (1983). Separation of ownership and control. Journal of Law and Economics, 26, 301–325. Fang, J. X. (2007). Ownership, institutional environment and capital allocation. Economic Research Journal (in Chinese), 12,82–92. Guo, J., & Li, T. (2009). A study on the competition in taxation between China’s local governments. Management World (in Chinese), 11,54–64. Gupta, S., & Newberry, K. (1997). Determinants of the variability of corporate effective tax rates: Evidence from longitudinal data. Journal of Accounting and Public Policy, 16,1–34. He, J. (1998). An empirical analysis of listed companies’ governance structure. Economic Research Journal (in Chinese), 5,51–58. Holland, K. (1998). Accounting policy choice: The relationship between corporate tax burden and company size. Journal of Business Finance and Accounting, 25, 265–288. Jensen, M. (1986). Agency cost of free cash flow, corporate finance, and takeovers. American Economic Review, 76, 323–329. Jensen, M., & Meckling, W. (1976). Theory of the firm: Managerial behavior, agency costs, and ownership structure. Journal of Financial Economics, 3, 305–360. Jiang, W., & Li, B. (2006). Institutional environment, state ownership and bank lending discrimination. Journal of Financial Research (in Chinese), 11,116–126. Kornai, J. (1980). Economics of shortage, Vol. B. New York: North-Holland Publishing Co. Kornai, J. (1992). The socialist system: The political economy of communism. Princeton: Princeton University Press. Lin, J. Y., & Tan, G. (1999). Policy burdens, accountability, and the soft budget constraint. The American Economic Review, 89, 426–431. Lou, Q. (2007). The Analysis of influential factors of tax burden of listed companies: the empirical data from Shanghai and Shenzhen Stock Markets. Communication of Finance and Accounting (in Chinese), 3,6–8. Lu, M., & Pan, H. (2009). Political connection: The development of private entrepreneurs and enterprises (in Chinese). Beijing: Peking University Press. Lu, Z. F., Zhu, J. G., & Fan, Z. (2009). Monetary contractions, bank discrimination, and investors benefit loss of private listed companies. Journal of Financial Research (in Chinese), 8, 124–136. Lv, B. Y., & Fan, Y. (2006). The progress and characters among provinces of tax collection and administration efficiency after the tax revenue sharing system. The Journal of World Economy (in Chinese), 10,69–77. Ma, S. Y., & Yu, H. X. (2003). Inter-governmental transfers and regional economic convergence in China. Economic Research Journal (in Chinese), 3,26–33. China Journal of Accounting Studies 189 Mao, Y. S. (2003). Low tax rate, strict tax enforcement: The direction of tax reform. Taxation Research (in Chinese), 9,2. Pan, Y., Dai, Y. Y., & Li, C. X. (2009). Political connections and government subsidies of companies in financial distress: Empirical evidence from Chinese ST listed companies. Nan- kai Business Review (in Chinese), 5,6–17. Ross, S. (1973). The economic theory of agency: The principal’s problem. American Economic Review, 63, 134–39. Shen, K. R., & Fu, W. L. (2006). Tax competition, region game and their efficiency of growth. Economic Research Journal (in Chinese), 6,16–26. Stickney, C. P., & McGee, V. E. (1982). Effective corporate tax rates: The effect of size, capital intensity, leverage and other factors. Journal of Accounting and Public Policy, 1, 125–152. Wang, S. R., & Liu, N. (2012). Statistical research on the influences of tax preferential policies upon industries’ net profit based on annual report of listed companies. Journal of Beijing Technology and Business University (Social Science) (in Chinese), 2, 97–102. Wang, Y. M. (2003). The relation of income tax burden of listed companies and size, area and industry. Management World (in Chinese), 1,115–122. World Bank (2006). China – governance, investment climate, and harmonious society: Competitiveness enhancement for 120 cities in China. Washington, DC: World Bank. Wu, L., Wang, Y., Lin, B. X., Li, C., & Chen, S. (2007). Local tax rebates, corporate tax burdens, and firm migration: Evidence from China. Journal of Accounting and Public Policy, 26, 555–583. Wu, L. (2009). State ownership, preferential tax, and corporate tax burdens. Economic Research Journal (in Chinese), 10, 109–120. Wu, L., Wang, Y., Luo, W., & Gillis, P. (2012). State ownership, tax status and size effect of effective tax rate in China. Accounting and Business Research, 42,97–114. Xia, L. J., & Fang, Y. Q. (2005). Government control, institutional environment and firm value: Evidence from the Chinese securities market. Economic Research Journal (in Chinese), 5, 40–51. Xu, S. D., Yang, W. L., Xu, J. G., You, K. J., Du, M. K., Ma, Q. L., Yao, J. Q., Xu, S. Q., Tang, G. L., Pang, F. X., & Tie, W. (2001). Research on the division of tax power. Taxation Research (in Chinese), 3,2–7. Yang, Z. G., Ding, L., & Wu, B. Z. (2000). An empirical study of value-added tax & income tax burden on enterprise. Economic Research Journal (in Chinese), 12,26–35. Ye, Z. R., & Lin, H. (2007). Tax competition alienation and tax right institution innovation. Taxation Research (in Chinese), 2,18–22. Zheng, H. X., & Han, M. F. (2008). Tax planning analysis based on listed company with different ownership structure: The empirical evidence from state-owned listed companies and private listed companies in China. China Soft Science (in Chinese), 9, 122–131. Zhou, X. G. (2005). Inverted soft budget constraint: Extrabudgetary resource-seeking in local governments. Social Science in China (in Chinese), 2, 132–143. Zhou, F. Z. (2006). A decade of tax sharing: The system and its evolution. Social Science in China (in Chinese), 6, 100–115. Zhou, L. A. (2008). Local governments in transition: Official incentives and Governance (in Chinese). Shanghai: Shanghai People’s Publishing House. Zimmerman, J. L. (1983). Taxes and firm size. Journal of Accounting and Economics, 5, 119–149. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png China Journal of Accounting Studies Taylor & Francis

Fiscal centralization, government control and corporate tax burden: Evidence from China

China Journal of Accounting Studies , Volume 1 (3-4): 22 – Dec 1, 2013

Loading next page...
 
/lp/taylor-francis/fiscal-centralization-government-control-and-corporate-tax-burden-scx4YqcP0Y

References (47)

Publisher
Taylor & Francis
Copyright
© 2013 Accounting Society of China
ISSN
2169-7221
eISSN
2169-7213
DOI
10.1080/21697221.2013.870367
Publisher site
See Article on Publisher Site

Abstract

China Journal of Accounting Studies, 2013 Vol. 1, Nos. 3–4, 168–189, http://dx.doi.org/10.1080/21697221.2013.870367 Fiscal centralization, government control and corporate tax burden: Evidence from China a b Jun Liu * and Feng Liu a b School of Business, Sun Yat-Sen University, China; School of Management, Xiamen University, China and Center for Accounting, Finance and Institutions, Sun Yat-Sen University, China This paper examines the relationship between government control and the tax burden of firms in China. We develop a new corporate tax burden measurement taking turn- over taxes into account, because in China turnover taxes actually constitute the main component of tax burden. We find that the tax burden of state-owned enterprises (SOEs) is lower than non-SOEs, indicating that non-SOEs are facing tax discrimina- tion. Among SOEs, the tax burden of local SOEs is higher than that of central SOEs, and the lower the local governments’ level, the higher the tax burden of SOEs under their control. We interpret these findings as the result of local governments’ tax competition and tax grabbing behaviors under China’s current highly centralized fiscal system. In addition, we find that our results are mainly caused by firms’ differences in tax refunds and the Value-Added Tax (VAT) burden. Keywords: fiscal centralization; government control; turnover tax burden; corporate tax burden 1. Introduction Kornai (1980) points out that in a socialist economy, the government is usually paternalistic to the SOEs. Although China’s market-oriented economy reform has continued for about 30 years, ownership discrimination against non-SOEs is still an indisputable reality in the economic environment. Calomiris, Fisman, and Wang (2010) find a negative effect of government ownership on returns at the announcement of the sale of government-owned shares in China, indicating that the benefits of government ownership outweigh the efficiency costs. Other researchers have proven that SOEs have the advantages over non-SOEs in entering an industry (Chen, Yu, Wang, & Lai, 2008), borrowing from banks (Fang, 2007; Jiang & Li, 2006; Lu, Zhu, &Fan, 2009), and obtaining subsidies from the government (Pan, Dai, & Li, 2009). Tax is also an important channel by which the government affects firms’ value. However, so far only a few studies have examined the relationship between government control and firms’ tax burden in China, and the results are mixed. Zheng and Han (2008) find that SOEs tend to be more conservative in tax avoidance and have higher effective tax rate (ETR) than non-SOEs. Wu (2009) examines the effects of state ownership on corporate tax burdens and finds that corporate income tax burdens increase with state ownership. However, the research of Cao, Liu, and Zhang (2009) finds that for the firms located in Chinese preferential tax zones, central SOEs bear a *Corresponding author. Email: cnjamesjunliu@gmail.com Paper accepted by Liansheng Wu. © 2013 Accounting Society of China China Journal of Accounting Studies 169 lower tax burden than local SOEs and non-SOEs, but there is no significant difference in tax burdens between local SOEs and non-SOEs. Thus, what effect government control has on corporate tax burdens remains an open question. The 1994 tax sharing reform in China is an attempt by the central government to recentralize the fiscal authority. The new tax system encourages local governments to alleviate their financial difficulties by tax competition (Shen & Fu, 2006) and tax grabbing (Zhou, 2005). SOEs generally have high lobbying power because of their close relationship with the government derived from ownership, and besides, as SOEs often undertake many social functions of the government, they are usually offered more preferential policies as compensation (Wu, Wang, Luo, & Gillis, 2012). Therefore, SOEs are more likely to obtain tax preferences and looser tax enforcement in govern- ments’ tax competition, and to avoid governments’ tax grabbing. Thus, the tax burden of SOEs should be lower than non-SOEs. Furthermore, under the current highly centralized fiscal system, lower level local governments are often in worse financial conditions. Because local SOEs are usually under the strong control of local govern- ments, they are more likely to be grabbed than central SOEs. So the tax burden of local SOEs should be higher than that of central SOEs, and the lower the level of government, the higher the tax burden of SOEs. With a sample of listed companies in the China A-share market during 2003–2006, we examine the tax burden of firms with different types of ultimate owners in China. According to the Interim Regulations on Supervision and Management of Enterprises’ State-owned Assets promulgated by the State Council of the PRC, SOEs are under the supervision and administration of the authorities, which are established to implement the responsibilities, and rights as owners, of central, provincial, and city governments. We classify SOEs into three categories: central SOEs, provincial SOEs, and city SOEs. Provincial and city SOEs can be referred to as local SOEs collectively. In contrast to prior tax burden studies, which generally focus on Enterprise Income Tax (EIT), we develop a new corporate tax burden measurement based on tax cash flows. This measure includes the turnover tax burden, which is actually the main component of corporate tax burden in the special Chinese taxation and business environment. We find that the tax burden of SOEs is significantly lower than non-SOEs, indicating that non-SOEs are facing tax discrimination. Among SOEs, the tax burden of central SOEs is lower than local SOEs. The tax burden of provincial SOEs is also lower and not significantly different from central SOEs, but the tax burden of city SOEs is higher than other SOEs and is not significantly different from non-SOEs. In addition, we divide the tax burden into two parts, tax payments and tax refunds, and find that tax refunds of SOEs are generally higher than non-SOEs, but there is no significant difference in tax payments, indicating that our findings are mainly driven by the differences in tax refunds from the government. We also separate the corporate tax burden into three main components: VAT, Business Tax (BT) and EIT, and find that the VAT burden of SOEs is lower than non-SOEs, but there are no significant differences in BT and EIT burdens. This indicates that the findings of this paper are mainly derived from differences in VAT burdens and, more importantly, research on corporate tax burdens will be biased if it only focuses on EIT. Our paper makes several contributions. First, we examine the effect of government control on corporate tax burdens and find evidence of tax discrimination against non-SOEs. This provides proof of government paternalism (Kornai, 1980) from a tax burden perspective and enriches the literature on the effect of political connections on corporate tax burdens (Adhikari, Derashid, & Zhang, 2006; Faccio, 2010). Second, we 170 Liu and Liu introduce a new corporate tax burden measurement based on tax cash flows, taking the turnover tax burden into account. This new measurement makes the corporate tax burden research in China more complete and instructive. Third, most prior tax burden studies ignore differences in tax burdens between SOEs controlled by different levels of government. In this paper we divide SOEs into three categories by the hierarchical level of their ultimate owners and provide a more detailed picture of the effect of government control on corporate tax burdens. Fourth, the findings of this paper provide an explanation for the negative relationship between control by lower level governments and firm value (Xia & Fang, 2005), and also provide new evidence to support the shift from the ‘helping hand’ to the ‘grabbing hand’ of local governments under the highly centralized fiscal system (Chen, Hillman, & Gu, 2002). The rest of this paper is organized as follows: Section 2 provides institutional background on the Chinese tax system and analyzes the mechanisms of government control on corporate tax burdens and presents our main hypotheses. Section 3 describes differences between the Chinese and American tax systems, explains why we need to take turnover taxes into consideration in corporate tax burden research, and then describes our new tax burden measurement and research model in greater detail. Section 4 presents our main empirical results and also a range of robustness checks. Section 5 concludes. 2. Institutional background and hypothesis development 2.1. Institutional background China’s current tax system was established by the 1994 Tax Sharing Reform. The core content of this reform was to reset the tax revenue sharing system between central and local governments. The new tax revenue sharing system re-categorized taxes into central taxes, local taxes and shared taxes. This tax sharing system is still in operation today and has exerted a profound influence on the relationship between the central and local governments. Among the three most important sources of tax revenues (VAT, BT, EIT), VAT is a shared tax, 75% for the central government and 25% for local govern- ments; BT is a local tax, which only belongs to local governments; and EIT is a shared tax, 60% for the central government and 40% for local governments. To ensure the smooth and effective implementation of the tax sharing system, significant changes also took place in China’s tax collection system in 1994. First, two sets of tax authorities, national and local tax bureaus, were set up to collect different types of taxes. National tax bureaus were established for collecting central and shared taxes, and are administered directly by the upper level national tax bureaus. Local tax bureaus are in charge of collecting local taxes and are under the dual leadership of both the local governments and the higher level tax bureaus. Before 1994, the central government had no tax collection authorities and had to rely totally on local authorities to collect and remit tax revenues. The new tax collection system changed this practice and ensured the revenues of central government be collected fully and in time. Second, in order to get sufficient support from local governments, a tax revenue refund system was designed to protect the interests of local governments. The revenues of the Two Taxes (VAT and Consumption Tax), which used to be the main sources of local revenue, would be refunded to the localities based on the amounts in the year 1993. In addition, 30% of the incremental central revenue from the Two Taxes would also be refunded each year. Actually, a large portion of the central-to-local transfer payment is just the refund of these tax revenues after the tax sharing reform. According China Journal of Accounting Studies 171 to Ma and Yu (2003), an average of 62.1% of the central-to-local transfer payment is the tax revenue refund during 1998–2001. 2.2. Hypothesis development The 1994 Tax Sharing Reform is a fiscal reform with an obvious centralizing tendency (Zhou, 2006). Through this reform, central government took back most of the discretionary power on revenue collection, which was gradually devolved to the localities in the 1980s. However, this reform did not change the fiscal expenditure structure between central and local governments, local governments were still undertak- ing most local public expenditures. The reform thus resulted in wide gaps between local fiscal revenues and expenditures. While fiscal expenditure is impossible to cut down, increasing revenue becomes the only way left to bridge the gap. Tax, as the biggest and highest quality portion of revenue, has inevitably received great attention from local governments. There are usually two ways for local governments to increase tax revenue. The first method is tax competition. The new tax system has determined the tax revenue distribution and refund system between central and local governments, which will stimulate local governments to provide various preferential policies to attract mobile capital to their administrative region so that they can develop the economy and get more tax surplus. Shen and Fu (2006) point out that tax competition between local governments was usually performed through providing preferential tax policies and loose tax enforcement. Under China’s current tax system, tax preferences are often shown as tax rate reductions, tax amount reductions and tax base reductions, among which the first two are especially common (Wang, 2003). Tax rate reductions are usually conducted based on the location or industry of enterprises. In China, many tax-related issues need to be authorized by relevant government authorities, such as whether an enterprise is located in a preferential tax zone; whether it is categorized in a preferential tax industry and which preferential tax rate is to be chosen. We believe that local governments have wide discretionary powers on whether an enterprise receives a tax rate reduction. Tax amount reductions are often shown as tax refunds from local governments. Although tax refunds unauthorized by the central government have been forbidden since 2002, in reality, local governments still refund a lot of tax revenue to enterprises as tax preferences. The categories of taxes involved include not only EIT, but also some turnover taxes, such as VAT and BT. As well as preferential tax policies, loose tax enforcement is also a common tax competition method. Under China’s current tax collection system, local tax bureaus are under the dual leadership of local governments and the higher level tax bureaus. However, the appointment of chief tax officers is mainly decided by local governments and even the salaries of employees and administrative funds are provided by local governments. So, it is safe to say that local governments have strong power to influence local tax bureaus, if necessary. National tax bureaus are nominally only under the leadership of the upper level national tax bureaus, but actually are also strongly influenced by local governments, because there are no special judicial agencies for tax affairs in China’s tax system. Only when tax bureaus get full cooperation from local judicial agencies will they fulfill their tax collection and administration duties (Xu et al., 2001; Ye & Lin, 2007). Guo and Li (2009) also provide evidence for the influence of local governments on tax 172 Liu and Liu enforcement. They find that local governments are in competition for VAT revenue, which is collected by national tax bureaus. The second method is tax grabbing. With a clear tax sharing and refund system, the 1994 tax reform hardened the budget constraints of local governments. When facing fiscal stress, local governments will grab economic resources from lower level govern- ments and enterprises within their precincts. Zhou (2005) develops a concept of ‘Inverted Soft Budget Constraints’ to describe and explain this behavior of local governments. One important characteristic of China’s current tax system is ‘high tax rate, loose tax enforcement’ (Mao, 2003), so there is much room for local governments to strengthen tax enforcement and, if necessary, local governments do not hesitate to do so. Lv and Fan (2006) find that the tax enforcement efficiency of most provinces in China has improved after 1994 and fiscal pressure is an important cause. Furthermore, some local governments, especially some governments at grassroot levels, even grab revenues by assigning tax fees on the enterprises under their jurisdiction (Zhou, 2005). The primary goal of modern firms is to maximize the wealth or value of sharehold- ers (Fama & Jensen, 1983; Jensen, 1986; Jensen & Meckling, 1976; Ross, 1973). In China, most enterprises have established management evaluation systems based on operating performance and firm value. So managers generally have incentives to improve operating performance by reducing tax fees. However, for enterprises with different government control backgrounds, their abilities to get tax benefits from governments’ tax competition are not the same, and the probabilities of governments grabbing them are also different. SOEs have natural political ties because of the government ownership. Kornai (1992) points out that, in a socialist economy, the ruling party, government, and SOEs always get along inextricably with each other, and the executives of SOEs are often cross appointed as the officers of others. In China, top executives of SOEs are usually ranked in the administrative hierarchy in the same way as government officials, and are appointed by governments or higher-level party organizations. The cross-appointment between executives of SOEs and government officers is also very common, but this phenomenon does not exist between governments and non-SOEs. Therefore, the close relationship between SOEs and the government will endow SOEs with higher lobbying power. In addition, SOEs in China often shoulder many policy burdens, such as employment, public utility, etc. As a compensation, SOEs are usually offered more preferential policies by the government. Lin and Tan (1999) argue that the policy bur- dens are the root cause of government’s soft budget constraint on SOEs. Wu et al. (2012) find that SOEs’ size is negatively associated with their effective tax rate (ETR) and non-SOEs’ size is positively associated with ETR, when these firms are not subject to any preferential tax status. They explain the finding as the result of SOEs’ high lob- bying power and the compensation from the government. Therefore, we can reason- ably infer that SOEs have greater advantages than non-SOEs in seeking tax preferences and loose tax enforcement, and have lower extent of tax grabbing. This discussion leads to our first hypothesis: H1a: The tax burden of SOEs is lower than that of non-SOEs. Furthermore, because of the vague property rights and the dual roles of the executives, government control over SOEs usually proves to be ‘weak control’ in ownership and ‘strong control’ in administration (He, 1998). The operating decisions made by executives of SOEs often deviate from the goal of maximizing shareholder China Journal of Accounting Studies 173 wealth and tend to satisfy the evaluation criteria for government officials, such as, GDP growth and tax revenue contribution. By contrast, the executives of non-SOEs are prob- ably large shareholders or the founder and their family. They are more likely to make decisions for shareholders’ benefits and have more incentives to avoid taxes. Zheng and Han (2008) find that the tax avoidance of SOEs is more conservative than other firms. Wu, Wang, Lin, Li, and Chen (2007) also find that non-SOEs are more likely to change registration locations to avoid a greater tax burden. So, there is also the possibility that the tax burden of SOEs is not lower than that of non-SOEs. Following this discussion, we state a competing hypothesis of H1a: H1b: The tax burden of SOEs is not lower than that of non-SOEs. Under the current highly centralized fiscal system, most local governments are in a bad financial condition and have a strong motivation to improve tax revenues. For local SOEs, local governments are usually large shareholders, so local governments can levy more tax revenues by interfering in the operations of local SOEs or even assigning tax fees directly to them. But central SOEs are different. Although they are spread out all over China, they are directly under the administration of the central government, enjoy a high hierarchical status and so are less likely to be affected by local governments. In addition, central SOEs have the autonomy to invest in other regions or even migrate to other places, which also helps them to get more tax preferences and looser tax enforce- ment. Hence, the tax burden of local SOEs is probably higher than that of central SOEs. Furthermore, there is also a possibility that the tax burden differs between local SOEs controlled by different levels of local governments. The 1994 tax reform did not touch on the fiscal relationship between governments below provincial level. Up to now, most local governments below the provincial level still employ a mixed tax revenue distribution system based on a fixed quota and sharing rate (Cai, 2007). Because the fiscal relationship between local governments is always determined completely by higher level governments, when facing financial stress provincial governments will shift the pressure to city level governments and city level govern- ments will follow suit to shift it further downward (Zhou, 2008). As a result, lower- level governments usually face even tougher financial conditions and have a strong motivation to grab tax revenues from enterprises. So the tax burden of local SOEs con- trolled by lower level governments may be higher than that of local SOEs controlled by higher ones. We summarize this argument as our second hypothesis: H2: The tax burden of local SOEs is higher than that of central SOEs, and the lower the local governments’ level, the higher the tax burden of SOEs under their control. 3. Research design and sample selection 3.1. The content of corporate tax burden Up till now, most of the corporate tax burden research in China has focused on EIT, ignoring turnover taxes. However, we argue that corporate tax burden should include all taxes paid to tax authorities. Panel A of Table 1 shows the major sources of China’s tax revenue after the 1994 tax sharing reform. Taxes paid by firms, especially the turnover taxes of VAT and BT account for a large proportion of total revenue. In recent years, the proportion of EIT is increasing, but is still lower than turnover taxes. From Panel B, we can see that individual income tax, social insurance tax and EIT have been 174 Liu and Liu Table 1. Tax revenue structure of US and China, by major sources, as a percentage of the total. Panel A: Major sources of China’s tax revenues 1994–2010 Individual income Year VAT Consumption tax BT EIT tax Others Total 1994 52.48% 10.18% 13.41% 13.56% 1.43% 8.93% 100.00% 1995 50.87% 9.47% 14.55% 13.85% 2.20% 9.05% 100.00% 2000 48.55% 6.93% 14.89% 13.98% 5.21% 10.44% 100.00% 2005 48.17% 5.46% 13.71% 17.85% 6.78% 8.02% 100.00% 2010 42.26% 9.24% 15.24% 17.54% 6.61% 9.12% 100.00% Panel B: Major sources of US tax revenues 1960-2010 Year Individual EIT Social Excise Others Total income insurance tax tax tax 1960 44.00% 23.20% 15.90% 12.60% 4.30% 100.00% 1970 46.90% 17.00% 23.00% 8.10% 5.00% 100.00% 1980 47.20% 12.50% 30.50% 4.70% 5.10% 100.00% 1990 45.20% 9.10% 36.80% 3.40% 5.50% 100.00% 2000 49.60% 10.20% 32.20% 3.40% 4.60% 100.00% 2010 41.50% 8.90% 40.00% 3.10% 6.50% 100.00% Note: Data Sources of Panel A: China Statistical Yearbook 2011, Tax Yearbook of China 1994-2010;Data Sources of Panel B: U.S. Congressional Budget Office; Office of Management and Budget. the main sources of US federal revenue since 1960. Individual income tax and social insurance tax, which are paid mainly by individuals, contribute more than 80% of the total nowadays, and while the proportion of EIT revenue paid by corporations is gradu- ally decreasing, it still accounts for nearly 10%. From Table 1, we can see that the main difference in the tax systems between the US and China is that under the current US tax system, government revenue mainly comes from the taxes paid by individuals. The taxes paid by firms are less important and EIT is the major tax paid by firms. For China, however, government revenue mainly comes from the taxes paid by firms, especially turnover taxes. World Bank (2006) surveys the investment climate of 120 cities in China and finds that the turnover tax burden of firms in China is significantly higher than their EIT burden. The smallest gap exists in Southeast China, where firms’ VAT burden is 3.5 times as high as their EIT burden. The largest gap exists in Northwest China, where VAT burden is 8.2 times as high as the EIT burden. Therefore, we argue that it is not appropriate to merely focus on EIT in the corpo- rate tax burden research in China, and it is necessary to incorporate turnover taxes into the calculation of corporate tax burden. 3.2. Tax burden measurement The measurements used in prior corporate tax burden research are often ETRs, which are widely used in western studies to measure the tax burden of EIT. Only Lou (2007), Wang and Liu (2012) and Yang, Ding, and Wu (2000), construct new measurements and study the burden of taxes other than EIT, but their measurements still cannot indicate the total tax burden of the firm. To overcome the defects of these existing mea- surements, we develop a new corporate tax burden measurement based on cash flows: China Journal of Accounting Studies 175 TaxNCF Taxburden1 ¼ Sales where TaxNCF is net cash outflows for tax payments and is calculated as ‘Payments of all types of taxes – Refunds of taxes’. Under China’s GAAP, the item of Payments of all types of taxes in the cash flow statement includes almost all taxes paid by firms, such as BT, VAT, Consumption Tax, EIT, educational surcharge, stamp duties, and so on. And the item of Refunds of taxes includes almost all tax refunds the firm receives. Therefore, the difference between these two items, TaxNCF,reflects the net expenditure of the firm on tax. The value of Sales is taken from the item of sales in the income statement. By definition, tax burden equals the amount of tax divided by the amount of taxable economic source. For firms in normal operating conditions, the economic source to pay tax fees is sales revenue. So we choose Sales as the denominator of the tax burden measurement. It should be noted that because the cash flow statement is prepared on a cash basis, the value of TaxNCF may not be exactly the amount of tax expense that should be recognized in the income statement on the accrual basis. But according to China’sTax Law, after the tax assessable period ends, the tax should be paid to tax authorities within 15 days or an even shorter time, so we believe the value of TaxNCF is very close to that of tax expense. To be more prudent, we also form another corporate tax burden measurement to mitigate the difference between tax cash flows and tax expenses: TaxNCF avg Taxburden2 ¼ Sales where TaxNCF_avg is the three-year moving average of the TaxNCFs of year t–1, t and t+1. Other variables are as defined above. 3.3. Empirical model We set up the following model to test our hypotheses: Taxburden ¼ b þ b UltCtrls þ b Controls þ e (1) i;t 0 1 i;t 2 i;t where Taxburden is Taxburden1 (or Taxburden2, each tested in separate regressions). it UltCtrls are dummy variables for each type of ultimate owner, namely SOE, Central- gov, Localgov, Provgov, and Citygov. Controls are the control variables. According to prior corporate tax burden studies (Adhikari et al., 2006; Derashid & Zhang, 2003; Gupta & Newberry, 1997; Holland, 1998; Stickney & McGee, 1982; Zimmerman, 1983), we include Capint, Invint, Size, Leverage and ROA as control variables in the model to control for the capital intensity, inventory intensity, size, leverage and profitability of firms. Since VAT is levied on the basis of value added during the course of goods production or services provision, firms with high gross margins probably pay more VAT. Therefore, we add Grossmargin into the model to control for the influence of gross margin on tax burden. Furthermore, membership of a business group, cross-industry operations and cross-region operations are all helpful for the firms to avoid or evade taxes through transfer pricing, so we introduce three dummy variables – Group, CRSind and CRSrgn – to control for these characteristics of firms. In addition, for a long time the tax preference policies in China have been based on regions, so classifying firms by the regions where they are located has become a 176 Liu and Liu convention in corporate tax burden research. We divide China’s territory into five regions by different tax preference policies, specifically, (a) the Special Economic Zones and Shanghai Pudong New District (SEZs); (b) East Region; (c) Central Region; (d) West Region; and (e) Northeast Region. Then we employ a separate indicator for each region, Location, to control for the tax policy differences between regions. In addition, because in China’s tax system the main tax categories of different industries are probably not the same, and tax preference policies can differ between industries and years, we also employ separate indicators for Industry and Year as control variables. Details of variable definitions are shown in Table 2. 3.4. Sample selection Since the 1994 Tax Reform, China’s tax system has changed several times, but it was relatively stable during the 2002–2007 period. Since the calculation of Taxburden2 needs data from prior and later accounting periods, we choose the listed companies on the main board of China’s A share stock market during 2003–2006 as the initial Table 2. Variable definitions. Variable Definition Taxburden Taxburden1 = TaxNCF/sales, where TaxNCF=Payments of all types of tax-Refunds of taxes Taxburden2 = (Three-year moving average of the TaxNCFs of year t–1, t and t+1)/ sales Ultctrls SOE = A dummy variable that equals 1 for firms controlled by the government, and 0 otherwise. Centralgov = A dummy variable that equals 1 for firms controlled by the central government, and 0 otherwise. Localgov = A dummy variable that equals 1 for firms controlled by local governments, and 0 otherwise. Provgov = A dummy variable that equals 1 for firms controlled by provincial governments, and 0 otherwise. Citygov = A dummy variable that equals 1 for firms controlled by city governments, and 0 otherwise. controls Capint = Year-end net value of fixed assets/Year-end total assets Invint = Year-end net value of inventory/Year-end total assets Size = Natural logarithm of year-end total assets Leverage = Year-end total liabilities/Year-end total assets ROA = Net income/Year-end total assets Grossmargin = (Sales-cost of sales)/sales Group = A dummy variable that equals 1 if the firm is a member of a business group, and 0 otherwise. CRSind = A dummy variable that equals 1 if a firm’s sales come from more than one industry, and 0 otherwise. CRSrgn = A dummy variable that equals 1 if a firm’s sales come from more than one region, and 0 otherwise. Location = A separate indicator for the regions of SEZs, East Region; Central Region; West Region; Northeast Region. Industry = A separate indicator for industries classified according to the Guidelines for the Industry Classification of Listed Companies issued by China Securities Regulatory Commission. Firms in the manufacturing industry are classified by their two-digital code. Year = A separate indicator for the years from 2003 to 2006. China Journal of Accounting Studies 177 Table 3. Sample selection procedure and composition. Panel A: Sample selection procedure Year 2003 2004 2005 2006 Total Initial sample 1,268 1,356 1,352 1,435 5,411 Less: financial companies 15 15 15 20 65 Less: companies whose 53 3 3 14 ultimate owners cannot be identified Less: companies whose 228 234 235 235 932 ultimate owners changed during the sample period Less: companies which 300 297 281 278 1,156 have been ST or PT since listed Less: observations with 322 368 302 312 1,304 missing values Final Sample 398 439 516 587 1,940 Panel B: Sample distribution by type of ultimate owner and year Year 2003 2004 2005 2006 Total Percentage Non-SOEs 88 106 145 170 509 26.24% Central SOEs 90 102 111 119 422 21.75% Provincial SOEs 120 125 144 158 547 28.20% City SOEs 100 106 116 140 462 23.81% Total 398 439 516 587 1,940 100.00% sample. We then exclude: (a) financial companies whose industrial code is ‘I’ in the Guidelines for the Industry Classification of Listed Companies issued by China Securities Regulatory Commission; (b) companies whose ultimate owners cannot be identified; (c) companies whose ultimate owners have changed from one type to another during the sample period, because the tax preferences enjoyed by the companies will not change immediately even after their ultimate owners have changed; (d) companies that have been specially treated (ST) or particularly transferred (PT) since they are listed, because ST and PT companies are often not in normal operating conditions, and their tax burdens are also likely to be abnormal; and (e) observations with missing values. Panel A of Table 3 presents the sample selection procedure, the final sample includes 1940 firm-year observations. Panel B of Table 3 reports the year distribution of the observations for the four types of firms: non-SOEs, central SOEs, provincial SOEs, and city SOEs. The numbers of them are similar, and the number of SOE observations accounts for 73.76% (21.75%+28.20%+23.81%) of the total, indicat- ing that the economy of China is still dominated by SOEs. All data used in this paper are from the China Stock Market Research (CSMAR) database. 4. Empirical results 4.1. Descriptive statistics Table 4 reports the descriptive statistics of the sample. It shows that the mean and median Taxburden1 of non-SOEs are both larger than those of the full sample, while the mean and median Taxburden1 of central SOEs are smaller than those of the full sample. Compared with non-SOEs, the mean and median Taxburden1 are higher for local SOEs, 178 Liu and Liu Table 4. Descriptive statistics. Variables Full sample Non-SOE SOE Central SOE Local SOE Provincial SOE City SOE Taxburden1 0.062 0.063 0.062 0.051 0.067 0.065 0.070 [0.051] [0.052] [0.051] [0.041] [0.055] [0.055] [0.056] (0.059) (0.055) (0.060) (0.053) (0.063) (0.066) (0.058) Taxburden2 0.063 0.064 0.063 0.051 0.067 0.065 0.070 [0.053] [0.057] [0.051] [0.041] [0.055] [0.052] [0.057] (0.057) (0.054) (0.059) (0.051) (0.061) (0.066) (0.055) Capint 0.305 0.277 0.314 0.288 0.326 0.321 0.331 [0.283] [0.271] [0.287] [0.250] [0.304] [0.298] [0.309] (0.177) (0.148) (0.186) (0.186) (0.185) (0.190) (0.179) Invint 0.170 0.180 0.167 0.164 0.168 0.158 0.181 [0.138] [0.134] [0.139] [0.152] [0.132] [0.119] [0.142] (0.141) (0.140) (0.142) (0.116) (0.151) (0.150) (0.152) Size 21.470 21.170 21.570 21.550 21.580 21.740 21.400 [21.390] [21.110] [21.480] [21.340] [21.520] [21.650] [21.390] (0.908) (0.776) (0.928) (1.127) (0.832) (0.884) (0.724) Leverage 0.480 0.483 0.479 0.471 0.483 0.464 0.505 [0.492] [0.502] [0.490] [0.479] [0.492] [0.470] [0.528] (0.168) (0.162) (0.170) (0.176) (0.168) (0.169) (0.163) ROA 0.044 0.045 0.044 0.045 0.044 0.048 0.038 [0.037] [0.041] [0.036] [0.034] [0.036] [0.043] [0.031] (0.047) (0.042) (0.048) (0.050) (0.047) (0.049) (0.045) Grossmargin 0.241 0.249 0.239 0.223 0.246 0.259 0.230 [0.206] [0.219] [0.203] [0.193] [0.207] [0.218] [0.198] (0.143) (0.134) (0.146) (0.135) (0.150) (0.163) (0.130) Group 0.791 0.768 0.799 0.893 0.760 0.887 0.610 [1.000] [1.000] [1.000] [1.000] [1.000] [1.000] [1.000] (0.407) (0.422) (0.401) (0.309) (0.427) (0.317) (0.488) CRSind 0.914 0.941 0.904 0.900 0.906 0.901 0.911 [1.000] [1.000] [1.000] [1.000] [1.000] [1.000] [1.000] (0.281) (0.236) (0.294) (0.300) (0.292) (0.299) (0.285) CRSrgn 0.926 0.986 0.904 0.960 0.881 0.887 0.874 [1.000] [1.000] [1.000] [1.000] [1.000] [1.000] [1.000] (0.262) (0.117) (0.294) (0.197) (0.324) (0.317) (0.332) Observations 1940 509 1431 422 1009 547 462 Note: The medians are reported in square brackets and standard errors are reported in parentheses. All variables are winsorized at the 1st and 99th percentiles. in which city SOEs bear the highest tax-burden among firms controlled by different types of owners. Results are similar if we use Taxburden2 as the measurement. The descriptive statistics for the control variables show that the mean and median capital intensity of non-SOEs are lower than those of SOEs, consistent with the fact that most non-SOEs in China cluster in labor-intensive or technology-intensive indus- tries. Central SOEs have lower capital intensity than local SOEs. Inventory intensity is higher in non-SOEs and lower in SOEs. In addition, the size of non-SOEs is smaller than that of SOEs. Provincial SOEs have the largest size and central SOEs next, while city SOEs have the smallest size among SOEs. Firms controlled by different types of owners have a similar debt ratio. Profitability is highest in provincial SOEs and lowest in city SOEs. We also learn that 80% of the sample is controlled by a group. This percentage is smaller in non-SOEs and larger in central/provincial SOEs. More than 90% of the sample firms operate in multiple industries and regions. China Journal of Accounting Studies 179 Table 5 reports the correlations for the variables. The two measurements for tax burden, Taxburden1 and Taxburden2, are highly correlated, consistent with our conjec- ture that the tax amount from the cash flow statement is similar to the tax expense. The SOE dummy is negatively correlated with tax burden but is not significant. It seems that the tax burden of the entire SOE sample is not significantly different from that of non-SOEs. But Centralgov has a negative correlation with tax burden while Localgov has a positive correlation, indicating that the tax burden is smaller for central SOEs and larger for local SOEs. Provgov is positively correlated with tax burden but is not signif- icant, while the correlation between Citygov and tax burden is positive and significant, suggesting that city SOEs have a larger tax burden. In addition, Capint is positively correlated with tax burden, possibly due to the regulation that VAT associated with new facilities is not deductible. Invint is negatively correlated with Taxburden1. The correla- tion of Size and tax burden is significantly positive, consistent with the ‘Political Cost’ argument. Leverage is negatively correlated with tax burden, implying that financial leverage acts as a tax shield. The profitability indicators (ROA and Grossmargin) are both positively correlated with tax burden. In addition, the correlations between CRS- ind/CRSrgn and tax burden are significantly negative, suggesting that cross-industry and cross-region firms bear less tax. The correlation between Group and tax burden is not significant. 4.2. Multiple regressions To test the hypotheses in a multivariate setting, we first estimate model (1) using ordinary least squares (OLS). Here and throughout, all the regressions are estimated using robust standard errors that are adjusted for clustering at the firm level. Results presented in Table 6 show that the two tax burden indicators, Taxburden1 and Taxbur- den2, yield similar estimates from the model with high goodness-of-fits observed from Adj-R , indicating that Taxburden1 and Taxburden2 are good measurements of corporate tax burden. Columns (1) and (4) show that the coefficients on SOE are signif- icantly negative, suggesting that the tax burden of SOEs is lower than that of non-SOEs, consistent with H1a. To test the differences in tax burdens among firms controlled by different levels of government, we use dummy variables representing the identity of the ultimate owners in the model. In Table 6, Columns (2) and (5), we document a significant negative coefficient of the central SOE indicator, Centralgov, indicating that central SOEs bear a lower tax burden than non-SOEs. The coefficient of the local SOE indicator, Localgov, is negative but insignificant. But when we classify local SOEs into provincial SOEs and city SOEs in Columns (3) and (6), the coefficient of Centralgov is significantly negative with a p-value<0.01, while the provincial SOE indicator, Provgov, is also negatively related to tax burden, significant at the 5% level. The coefficient of the city SOE indicator, Citygov, is insignificant. These results suggest that the lower tax burden of SOEs is mainly driven by central SOEs and provincial SOEs, while the tax burden of city SOEs is similar to that of non-SOEs. To test H2 and further analyze the relationship between government control and corporate tax burden, we remove the non-SOE sample from the full sample and re-estimate model (1) using only the SOE sample. Central SOEs are used as the base group when comparing the tax burden between firms controlled by different levels of government. Table 7 displays the results. Columns (1) and (4) show that Localgov is positively related to tax burden, significant at the 5% level. It suggests that local SOEs 180 Liu and Liu Table 5. Pearson and Spearman correlation matrix. Taxburden1 Taxburden2 SOE Centralgov Localgov Provgov Citygov Capint Invint Size Leverage ROA Grossmargin Group CRSind CRSrgn *** *** *** *** *** *** *** *** *** *** *** *** Taxburden1 0.9557 –0.0087 –0.1248 0.0954 0.0313 0.0788 0.2023 –0.2043 0.0769 –0.2000 0.3077 0.6740 –0.0235 –0.0786 –0.1700 *** *** *** *** *** *** *** *** *** *** *** *** Taxburden2 0.9472 –0.0238 –0.1365 0.0917 0.0225 0.0837 0.1955 –0.2015 0.0801 –0.2026 0.3120 0.6812 –0.0258 –0.0656 –0.1766 *** *** *** *** *** ** *** ** ** ** *** SOE –0.0011 –0.0126 0.3145 0.6209 0.3737 0.3334 0.0688 –0.0515 0.1884 –0.0113 –0.0455 –0.0543 0.0338 –0.0577 –0.1376 *** *** *** *** *** *** *** *** *** *** Centralgov –0.1045 –0.1092 0.3145 –0.5489 –0.3304 –0.2948 –0.0725 0.0201 0.0078 –0.0258 –0.0236 –0.0737 0.1325 –0.0253 0.0683 *** *** *** *** *** *** *** *** *** *** *** Localgov 0.0854 0.0792 0.6209 –0.5489 0.6019 0.5370 0.1204 –0.0619 0.1595 0.0113 –0.0206 0.0130 –0.0796 –0.0300 –0.1775 *** *** *** *** ** *** *** *** ** ** *** *** Provgov 0.0276 0.0209 0.3737 –0.3304 0.6019 –0.3503 0.0515 –0.0952 0.2005 –0.0654 0.054 0.0452 0.1471 –0.0282 –0.0935 *** *** *** *** *** *** *** *** *** *** *** Citygov 0.0710 0.0708 0.3334 –0.2948 0.5370 –0.3503 0.0869 0.0279 –0.0248 0.0823 –0.0811 –0.0325 –0.2488 –0.0053 –0.1094 *** *** *** ** *** ** *** *** *** *** *** *** *** *** Capint 0.2248 0.2329 0.0918 –0.0510 0.1230 0.0569 0.0841 –0.5001 0.1006 –0.1190 0.0849 0.1134 –0.0348 –0.0783 –0.1152 * * ** * *** *** *** *** *** Invint –0.0426 –0.0355 –0.0409 –0.0254 –0.0151 –0.0552 0.0406 –0.5278 –0.0298 0.2869 –0.0841 –0.1554 –0.0228 0.0355 0.1536 *** *** *** ** *** *** * *** *** *** ** Size 0.0984 0.1037 0.1962 0.0507 0.1309 0.1864 –0.0435 0.1623 0.0075 0.3291 0.1109 –0.0334 0.0358 –0.0085 –0.0506 *** *** *** *** *** *** *** *** *** *** *** ** Leverage –0.1967 –0.1981 –0.0090 –0.0293 0.0163 –0.0597 0.0821 –0.1064 0.2890 0.3038 –0.3083 –0.2948 –0.1133 0.0784 0.0577 *** *** ** *** *** *** *** *** *** *** ** ROA 0.3071 0.3142 –0.0129 0.0050 –0.0155 0.0491 –0.0700 0.1115 –0.0596 0.1528 –0.3182 0.3724 0.0217 –0.0861 –0.0459 *** *** *** *** ** *** *** *** *** *** Grossmargin 0.6697 0.6736 –0.0298 –0.0677 0.0297 0.0754 –0.0449 0.1316 –0.0931 –0.0029 –0.3096 0.3603 –0.0209 –0.0405 –0.0979 *** *** *** *** * *** Group –0.0036 0.0032 0.0338 0.1325 –0.0796 0.1471 –0.2488 –0.0243 –0.0093 0.0410 –0.1110 0.0204 –0.0118 0.0006 0.0142 *** *** ** *** *** *** ** *** CRSind –0.1035 –0.0943 –0.0577 –0.0253 –0.0300 –0.0282 –0.0053 –0.0898 0.0365 –0.0243 0.0868 –0.1173 –0.0554 0.0006 0.0743 *** *** *** *** *** *** *** *** ** ** ** * *** *** CRSrgn –0.1830 –0.1850 –0.1376 0.0683 –0.1775 –0.0935 –0.1094 –0.1370 0.0579 –0.0473 0.0564 –0.0400 –0.1021 0.0142 0.0743 *** ** * Note: Pearson correlations are reported on the bottom left and Spearman correlations on the upper right. , and denote statistical significance at the 1%, 5%, 10% levels, respectively. China Journal of Accounting Studies 181 Table 6. Multiple regression: types of ultimate owners and corporate tax burden (full sample). Dependent variable: Taxburden1 Dependent variable: Taxburden2 Variables (1) (2) (3) (4) (5) (6) * ** SOE –0.0059 –0.0074 (–1.919) (–2.298) *** *** *** *** Centralgov –0.0116 –0.0123 –0.0130 –0.0137 (–2.900) (–3.063) (–3.180) (–3.344) Localgov –0.0036 –0.0052 (–1.106) (–1.501) ** ** Provgov –0.0084 –0.0101 (–2.174) (–2.468) Citygov 0.0014 –0.0001 (0.356) (–0.030) *** *** *** *** *** *** Capint 0.0371 0.0353 0.0332 0.0425 0.0407 0.0386 (3.696) (3.515) (3.276) (3.999) (3.826) (3.609) *** *** *** *** *** *** Invint 0.0421 0.0431 0.0416 0.0444 0.0454 0.0438 (2.950) (3.014) (2.903) (3.186) (3.247) (3.148) ** ** ** ** ** ** Size 0.0033 0.0033 0.0040 0.0035 0.0035 0.0042 (1.984) (2.020) (2.372) (2.047) (2.084) (2.429) * * ** * * ** Leverage –0.0194 –0.0194 –0.0215 –0.0202 –0.0202 –0.0223 (–1.844) (–1.848) (–2.023) (–1.883) (–1.888) (–2.066) ROA 0.0356 0.0380 0.0353 0.0474 0.0498 0.0471 (1.102) (1.192) (1.112) (1.501) (1.598) (1.520) *** *** *** *** *** *** Grossmargin 0.2421 0.2407 0.2432 0.2302 0.2289 0.2314 (17.955) (17.652) (17.581) (17.299) (17.025) (16.947) Group 0.0021 0.0030 0.0052 0.0027 0.0037 0.0058 (0.642) (0.918) (1.531) (0.822) (1.080) (1.699) ** ** ** * CRSind –0.0097 –0.0099 –0.0101 –0.0074 –0.0076 –0.0078 (–2.009) (–2.054) (–2.094) (–1.597) (–1.644) (–1.687) CRSrgn –0.0064 –0.0048 –0.0046 –0.0056 –0.0041 –0.0038 (–1.080) (–0.821) (–0.786) (–0.925) (–0.680) (–0.644) *** *** *** *** *** *** Constant –0.1087 –0.1106 –0.1243 –0.1109 –0.1127 –0.1266 (–3.114) (–3.179) (–3.551) (–3.132) (–3.200) (–3.570) Location YES YES YES YES YES YES Industry YES YES YES YES YES YES Year YES YES YES YES YES YES Observations 1,940 1,940 1,940 1,940 1,940 1,940 Adj-R 0.596 0.598 0.601 0.615 0.617 0.620 Note: All t-statistics, reported in parenthesis, are based on standard errors adjusted for heteroskedasticity and *** ** * firm-level clustering. , and denote statistical significance at the 1%, 5%, 10% levels, respectively. bear a significantly higher tax burden than central SOEs. Among local SOEs, the coefficient on Provgov is insignificant while the coefficient of Citygov is significantly positive with a p-value <0.01, as shown in Columns (2) and (5), indicating that the tax burden of provincial SOEs is not much different from that of central SOEs, while city SOEs have a much higher tax burden than central SOEs. We further remove the central SOE sample and retain the local SOE sample to compare the tax burden of provincial SOEs and city SOEs. Provincial SOEs are used as the base group and the results are displayed in Columns (3) and (6) of Table 7. The coefficient of Citygov is positive and significant at the 10% level, suggesting that the tax burden of city SOEs is higher than provincial SOEs. Thus, H2 is supported by the results. 182 Liu and Liu Table 7. Multiple regression: Types of ultimate owners and corporate tax burden (SOE subsample). Dependent variable: Taxburden1 Dependent variable: Taxburden2 Variables (1) (2) (3) (4) (5) (6) ** ** Localgov 0.0089 0.0088 (2.241) (2.165) Provgov 0.0053 0.0050 (1.275) (1.179) *** * *** * Citygov 0.0141 0.0075 0.0141 0.0080 (2.883) (1.712) (2.868) (1.801) *** *** *** *** *** *** Capint 0.0352 0.0334 0.0415 0.0388 0.0369 0.0470 (3.205) (3.019) (2.846) (3.497) (3.329) (3.195) ** ** ** ** Invint 0.0353 0.0334 0.0285 0.0371 0.0353 0.0287 (2.248) (2.133) (1.428) (2.334) (2.228) (1.436) ** *** ** *** Size 0.0043 0.0050 0.0034 0.0048 0.0055 0.0041 (2.304) (2.638) (1.325) (2.549) (2.894) (1.624) Leverage –0.0167 –0.0184 –0.0170 –0.0175 –0.0193 –0.0163 (–1.383) (–1.510) (–1.126) (–1.430) (–1.561) (–1.054) ** * ROA 0.0548 0.0515 0.0959 0.0478 0.0444 0.0938 (1.489) (1.399) (2.022) (1.295) (1.207) (1.943) *** *** *** *** *** *** Grossmargin 0.2372 0.2403 0.2399 0.2252 0.2283 0.2267 (14.935) (14.800) (11.992) (14.504) (14.374) (11.235) Group 0.0018 0.0044 0.0013 0.0024 0.0051 0.0023 (0.412) (0.997) (0.282) (0.538) (1.132) (0.513) CRSind –0.0064 –0.0066 –0.0074 –0.0046 –0.0048 –0.0062 (–1.328) (–1.356) (–1.221) (–0.999) (–1.030) (–1.045) CRSrgn –0.0053 –0.0051 –0.0038 –0.0053 –0.0051 –0.0032 (–0.908) (–0.884) (–0.550) (–0.896) (–0.872) (–0.463) *** *** ** *** *** ** Constant –0.1394 –0.1557 –0.1184 –0.1473 –0.1641 –0.1328 (–3.492) (–3.874) (–2.189) (–3.725) (–4.135) (–2.522) Location YES YES YES YES YES YES Industry YES YES YES YES YES YES Year YES YES YES YES YES YES Observations 1,431 1,431 1,009 1,431 1,431 1,009 Adj-R 0.635 0.638 0.644 0.652 0.655 0.660 Note: Central SOEs are the base group in Columns (1), (2), (4) and (5). Provincial SOEs are the base group in Columns (3) and (6). All t-statistics, reported in parenthesis, are based on standard errors adjusted for heter- *** ** * oskedasticity and firm-level clustering. , and denote statistical significance at the 1%, 5%, 10% levels, respectively. 4.3. Additional tests 4.3.1. An examination of tax burden components: tax payments and tax refunds Our tax burden measures consist of two components: tax payments and tax refunds. To examine whether the differences in tax burden come from tax payments or tax refunds, we use two variables, Taxpay and Taxrfd, as the proxies for tax payments and tax refunds, respectively, and compare the tax burden components in firms controlled by different owners. Taxpay is computed as tax payments divided by sales, while Taxrfd is measured as tax refunds divided by sales. Taxpay and Taxrfd are used as the explained variables to re-estimate model (1). The results are presented in Table 8. Table 8 shows that when using Taxpay as the dependent variable, the coefficient of SOE is not significant, suggesting that there is no difference in tax payments between SOEs and non-SOEs. Centralgov is negatively associated with Taxpay, significant at China Journal of Accounting Studies 183 Table 8. Multiple regression: types of ultimate owners and corporate tax payments / refunds. Dependent variable: Taxpay Dependent variable: Taxrfd Variables (1) (2) (3) (4) (5) (6) *** SOE –0.0019 0.0039 SOE (–0.745) (2.669) ** ** ** ** Centralgov –0.0069 –0.0072 0.0049 0.0052 (–2.130) (–2.232) (2.359) (2.539) ** Localgov 0.0000 0.0035 (0.008) (2.230) *** Provgov –0.0024 0.0060 (–0.748) (3.014) Citygov 0.0025 0.0008 (0.764) (0.480) *** *** *** ** ** ** Capint 0.0265 0.0250 0.0239 –0.0109 –0.0106 –0.0095 (3.142) (2.996) (2.824) (–2.432) (–2.310) (–2.124) *** *** *** Invint 0.0331 0.0340 0.0332 –0.0070 –0.0072 –0.0064 (2.775) (2.855) (2.771) (–1.040) (–1.058) (–0.951) * * ** Size 0.0027 0.0027 0.0031 –0.0002 –0.0002 –0.0006 (1.906) (1.944) (2.119) (–0.299) (–0.309) (–0.794) ** ** ** Leverage –0.0118 –0.0118 –0.0128 0.0078 0.0078 0.0089 (–1.251) (–1.251) (–1.345) (2.001) (2.006) (2.245) ROA 0.0179 0.0199 0.0186 –0.0189 –0.0193 –0.0179 (0.655) (0.742) (0.691) (–1.283) (–1.306) (–1.226) *** *** *** Grossmargin 0.2423 0.2411 0.2424 0.0067 0.0069 0.0055 (19.605) (19.406) (19.354) (1.160) (1.182) (0.941) Group 0.0010 0.0018 0.0029 –0.0006 –0.0008 –0.0019 (0.392) (0.715) (1.116) (–0.364) (–0.456) (–1.133) ** ** ** CRSind –0.0092 –0.0094 –0.0095 –0.0016 –0.0015 –0.0014 (–2.236) (–2.272) (–2.288) (–0.759) (–0.750) (–0.690) CRSrgn –0.0045 –0.0032 –0.0030 0.0028 0.0025 0.0024 (–0.872) (–0.618) (–0.596) (1.410) (1.266) (1.209) *** *** *** * Constant –0.0780 –0.0796 –0.0865 0.0216 0.0219 0.0292 (–2.630) (–2.696) (–2.893) (1.340) (1.360) (1.797) Location YES YES YES YES YES YES Industry YES YES YES YES YES YES Year YES YES YES YES YES YES Observations 1,940 1,940 1,940 1,940 1,940 1,940 Adj-R 0.622 0.624 0.625 0.158 0.159 0.167 Note: All t-statistics, reported in parenthesis, are based on standard errors adjusted for heteroskedasticity and *** ** * firm-level clustering , and denote statistical significance at the 1%, 5%, 10% levels, respectively. the 5% level, indicating that the tax payments of central SOEs are lower than non-SOEs. The coefficients of Localgov, Provgov and Citygov are insignificant, implying that there are no significant differences in tax payments between local SOEs and non-SOEs, or between provincial/city SOEs and non-SOEs. When Taxrfd is used as the dependent variable, the coefficient on SOE is positive and significant, suggesting that the tax refunds of SOEs are higher than non-SOEs. Both Centralgov and Localgov are positively related to Taxrfd, indicating that among SOEs the tax refunds of central and provincial SOEs are significantly higher than non-SOEs. More specifically, the coefficient of Provgov is positive and significant, while the coefficient of Citygov is insignificant, suggesting that the main difference in tax refunds between local SOEs and non-SOEs is driven by provincial SOEs. Overall, 184 Liu and Liu the results in Table 8 indicate that the impact of government control on corporate tax burdens mainly comes from their influence on tax refunds, implying that the prohibition against unauthorized tax refunds has not been fully implemented. 4.3.2. An examination by major tax categories: VAT, business tax and EIT The corporate tax burden of Chinese firms consists of three major types of tax, VAT, BT and EIT. In order to identify from which types of tax the difference in tax burden arises, we further separate the tax burden into VAT burden (VATburden), BT burden (BTburden) and EIT burden (EITburden). The variables are defined as follows and all financial data needed are from the CSMAR database: VATburden ¼ðTaxNCF  EITNCF  B&SNCFÞ=Sales BTburden ¼ BTNCF=Sales EITburden ¼ EITNCF=Sales where EITNCF, B&SNCF and BTNCF are the net cash outflows of EIT, Business Taxes and Surcharges, and BT. EITNCF =(Income tax axDeferred income tax – ΔIn- come tax payable); B&SNCF =(Business taxes and surcharges – ΔBusiness taxes and surcharges payable), Business taxes and surcharges payable = Tax payable – Income tax payable – VAT payable; BTNCF = BT – ΔBT payable. Other variables are the same as defined above. The change (Δ) is computed between year t and t–1. These measurements are used to re-estimate model (1) and the results presented in Table 9 are similar to those in Table 6 when the VAT indicator (VATburden) is used as the dependent variable. The VAT burden of SOEs is lower than non-SOEs. Among SOEs, the VAT burden of central and provincial SOEs is significantly lower than non- SOEs and there is no significant difference between local SOEs and non-SOEs. But the BT burden (BTburden) shows no significant difference between non-SOEs and SOEs controlled by different levels of government. In terms of the EIT burden, the difference between SOEs and non-SOEs is not significant. This indicates that we may find differ- ent results if we ignore turnover taxes when analyzing the corporate tax burden. Central SOEs bear a lighter EIT burden than non-SOEs, consistent with the findings of Cao et al. (2009). Differences in EIT burdens between provincial SOEs, city SOEs and non- SOEs are not significant. These results suggest that the findings presented in Section 4.2 are mainly driven by the value added tax burden. It also suggests that results would be biased if we only focus on EIT when analyzing the tax burden of a firm. 4.4. Robustness checks 4.4.1. Deleting loss firms Firms do not need to pay EIT when they incur a loss and their tax burden may be biased due to their abnormal operating conditions. Hence, we remove the 157 firm-year observations with a loss and re-estimate model (1). The results are similar to the previous findings (see Table 10). China Journal of Accounting Studies 185 Table 9. Multiple regression: types of ultimate owners and corporate tax categories. Dependent variable: VATburden Dependent variable: BTburden Dependent variable: EITburden Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) ** SOE –0.0059 –0.0001 –0.0008 (–2.347) (–0.108) (–0.537) * * ** ** Centralgov –0.0055 –0.0061 –0.0016 –0.0014 –0.0040 –0.0040 (–1.691) (–1.771) (–1.490) (–1.370) (–2.473) (–2.468) ** Localgov –0.0061 0.0006 0.0005 (–2.266) (0.554) (0.316) *** Provgov –0.0104 0.0015 0.0003 (–3.100) (1.259) (0.171) Citygov –0.0015 –0.0002 0.0007 (–0.487) (–0.207) (0.382) Controls YES YES YES YES YES YES YES YES YES ** ** * *** *** *** Constant 0.0074 0.0075 –0.0052 –0.0197 –0.0202 –0.0177 –0.0644 –0.0654 –0.0660 (0.250) (0.254) (–0.176) (–2.024) (–2.087) (–1.837) (–3.415) (–3.533) (–3.629) Observations 1,753 1,753 1,753 1,013 1,013 1,013 1,766 1,766 1,766 Adj-R 0.444 0.444 0.449 0.639 0.641 0.642 0.525 0.528 0.528 *** ** * Note: All t-statistics, reported in parenthesis, are based on standard errors adjusted for heteroskedasticity and firm-level clustering. , and denote statistical significance at the 1%, 5%, 10% levels, respectively. 186 Liu and Liu Table 10. Multiple regression: Types of ultimate owners and corporate tax burden (removing loss firms). Dependent variable: Dependent variable: Taxburden1 Taxburden2 Variables (1) (2) (3) (4) (5) (6) * ** SOE –0.0052 –0.0067 (–1.668) (–2.076) *** *** *** *** Centralgov –0.0118 –0.0124 –0.0131 –0.0138 (–2.813) (–2.950) (–3.100) (–3.231) Localgov –0.0024 –0.0040 (–0.725) (–1.154) * * Provgov –0.0067 –0.0082 (–1.676) (–1.951) Citygov 0.0019 0.0002 (0.486) (0.049) Controls YES YES YES YES YES YES *** *** *** *** *** *** Constant –0.1089 –0.1115 –0.1232 –0.1130 –0.1156 –0.1270 (–3.048) (–3.132) (–3.432) (–3.152) (–3.242) (–3.541) Observations 1,783 1,783 1,783 1,783 1,783 1,783 Adj-R 0.611 0.614 0.616 0.630 0.633 0.635 Note: All t-statistics, reported in parenthesis, are based on standard errors adjusted for heteroskedasticity and *** ** * firm-level clustering. , and denote statistical significance at the 1%, 5%, 10% levels, respectively. 4.4.2. Median regressions Although we winsorize all variables at their 1st and 99th percentile values, we also estimate the model using median regressions with respect to departures from the median. We obtain similar results as above (not reported). 4.4.3. Alternative indicators Our second tax burden indicator, Taxburden2, is computed using the three-year moving average of net tax cash flows as the numerator. To be consistent with the way we calculate Taxburden2, we use the moving-average sales of year t–1, t and t+1 as the denominator to re-calculate Taxburden2 as the explanatory variable. The untabulated results are similar to those obtained in the previous section. We also use the ratio of active debt to assets as an alternative debt ratio and return on equity as an alternative to return on assets as control variables in the regression model. The untabulated results remain the same. 5. Conclusion In this paper, we examine the effect of government control on the corporate tax burden of A-share listed companies in China during 2003–2006. We introduce a new corporate tax burden measurement based on tax cash flows: the ratio between net cash outflows of tax payments and the sales of the corporation. This allows us to include the turnover tax burden of firms, which is important because turnover taxes are the main component of taxes for most firms in China. We also classify SOEs into three categories by the level of government as their ultimate owners. The results show that the tax burden of SOEs is significantly lower than non-SOEs, indicating that in China non-SOEs face tax China Journal of Accounting Studies 187 discrimination. Among the SOEs, the tax burden of central SOEs is lower than that of local SOEs. The tax burden of provincial SOEs is also lower and not significantly different from that of central SOEs, but the tax burden of city SOEs is higher than that of other SOEs and is not significantly different from that of non-SOEs. Furthermore, we find that the results of this paper are mainly derived from differences in tax refunds and the VAT burden of firms, indicating that the prohibition against unauthorized tax refunds has not been fully implemented, and more importantly, the results are biased if corporate tax burden research only focuses on EIT. In summary, the findings of this paper show that government control is helpful for firms to obtain a lower tax burden and the effect of control by different levels of government on corporate tax burdens is not the same. We interpret these findings as the result of tax competition and tax grabbing by local governments under the highly centralized fiscal system in China. This also indicates that the development of China’s market economy is still at a low level and China still has a long way to go to transform its government functions and provide a fair economic environment for enterprises. Acknowledgements This study is supported by the National Science Foundation of China (71272079) and the Ministry of Education (MOE) Project of Key Research Institute of Humanities and Social Sciences in Universities (11JJD790032). We would like to offer most sincere thanks to Kangtao Ye for his constructive comments on this paper. We are also very grateful for the valuable comments and suggestions made by Liansheng Wu, Jason Xiao, two anonymous reviewers, and the editors. We would like to thank the organizers and other participants of the First Annual Conference of the China Journal of Accounting Studies in Chengdu, China,2012. We also appreciate the excellent help from the language advisor, John Nowland from City University of Hong Kong. Notes 1. As reported in the Audit Report (No. 12 of 2010) published by China’s National Audit Office (CNAO), in order to attract investment, several local governments introduced their own policies of tax reduction, and refunded a large amount of tax revenue to enterprises in 2008 and 2009. 2. It is generally thought that private capital also has high mobility, but overall the investment climate in China for private capitals is poor. Entrepreneurs of non-SOEs are very cautious about investing in other regions unless they have friends in the local governments there (Lu & Pan, 2009). Some local governments even employ the ‘coax and coerce’ strategy for private capital (Zhou, 2008). Therefore, the mobility of private capital is probably not as high as generally expected and it is not as helpful to lower corporate tax burden as imagined. 3. Based on the analysis above, the tax expenses here and after include VAT paid by enterprises. 4. In China, SEZs include Shenzhen, Zhuhai, Shantou, Xiamen, and Hainan province. We also include Shanghai Pudong New District in the SEZs because of its special economic and political importance; The East Region includes nine provinces (municipalities) of Beijing, Tianjin, Hebei, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, and Guangdong; The Central Region includes six provinces of Shanxi, Anhui, Jiangxi, Henan, Hubei, and Hunan; The West Region includes 12 provinces (autonomous regions and municipalities) of Chongqing, Sichuan, Guizhou, Yunnan, Xizang, Shaanxi, Gansu, Ningxia, Qinghai, Xinjiang, Neimenggu, and Guangxi; The Northeast Region includes three provinces of Heilongjiang, Jilin, and Liaoning. 188 Liu and Liu References Adhikari, A., Derashid, C., & Zhang, H. (2006). Public policy, political connections, and effective tax rates: Longitudinal evidence from Malaysia. Journal of Accounting and Public Policy, 25, 574–595. Cai, H. Y. (2007). Study on fiscal relationship among the local governments in China (in Chinese). Beijing: China Financial & Economic Publishing House. Calomiris, C. W., Fisman, R., & Wang, Y. (2010). Profiting from government stakes in a command economy: Evidence from Chinese asset sales. Journal of Financial Economics, 96, 399–412. Cao, S. J., Liu, X., & Zhang, W. J. (2009). Fiscal decentralization, local government competition, and corporate effective tax. The Journal of World Economy (in Chinese), 4,69–83. Chen, B., Yu, J., Wang, X. J., & Lai, J. Q. (2008). An empirical study on the development of private listing corporation in China. Securities Market Herald (in Chinese), 4,42–47. Chen, K., Hillman, A. L., & Gu, Q. Y. (2002). Fiscal re-centralization and behavioral change of local governments: From the helping hand to grabbing hand. China Economic Quarterly (in Chinese), 4,111–130. Derashid, C., & Zhang, H. (2003). Effective tax rates and the ‘industrial policy’ hypothesis: Evidence from Malaysia. Journal of International Accounting, Auditing and Taxation, 12, 45–62. Faccio, M. (2010). Differences between politically connected and nonconnected firms: A cross-country analysis. Financial Management, 39, 905–928. Fama, E., & Jensen, M. (1983). Separation of ownership and control. Journal of Law and Economics, 26, 301–325. Fang, J. X. (2007). Ownership, institutional environment and capital allocation. Economic Research Journal (in Chinese), 12,82–92. Guo, J., & Li, T. (2009). A study on the competition in taxation between China’s local governments. Management World (in Chinese), 11,54–64. Gupta, S., & Newberry, K. (1997). Determinants of the variability of corporate effective tax rates: Evidence from longitudinal data. Journal of Accounting and Public Policy, 16,1–34. He, J. (1998). An empirical analysis of listed companies’ governance structure. Economic Research Journal (in Chinese), 5,51–58. Holland, K. (1998). Accounting policy choice: The relationship between corporate tax burden and company size. Journal of Business Finance and Accounting, 25, 265–288. Jensen, M. (1986). Agency cost of free cash flow, corporate finance, and takeovers. American Economic Review, 76, 323–329. Jensen, M., & Meckling, W. (1976). Theory of the firm: Managerial behavior, agency costs, and ownership structure. Journal of Financial Economics, 3, 305–360. Jiang, W., & Li, B. (2006). Institutional environment, state ownership and bank lending discrimination. Journal of Financial Research (in Chinese), 11,116–126. Kornai, J. (1980). Economics of shortage, Vol. B. New York: North-Holland Publishing Co. Kornai, J. (1992). The socialist system: The political economy of communism. Princeton: Princeton University Press. Lin, J. Y., & Tan, G. (1999). Policy burdens, accountability, and the soft budget constraint. The American Economic Review, 89, 426–431. Lou, Q. (2007). The Analysis of influential factors of tax burden of listed companies: the empirical data from Shanghai and Shenzhen Stock Markets. Communication of Finance and Accounting (in Chinese), 3,6–8. Lu, M., & Pan, H. (2009). Political connection: The development of private entrepreneurs and enterprises (in Chinese). Beijing: Peking University Press. Lu, Z. F., Zhu, J. G., & Fan, Z. (2009). Monetary contractions, bank discrimination, and investors benefit loss of private listed companies. Journal of Financial Research (in Chinese), 8, 124–136. Lv, B. Y., & Fan, Y. (2006). The progress and characters among provinces of tax collection and administration efficiency after the tax revenue sharing system. The Journal of World Economy (in Chinese), 10,69–77. Ma, S. Y., & Yu, H. X. (2003). Inter-governmental transfers and regional economic convergence in China. Economic Research Journal (in Chinese), 3,26–33. China Journal of Accounting Studies 189 Mao, Y. S. (2003). Low tax rate, strict tax enforcement: The direction of tax reform. Taxation Research (in Chinese), 9,2. Pan, Y., Dai, Y. Y., & Li, C. X. (2009). Political connections and government subsidies of companies in financial distress: Empirical evidence from Chinese ST listed companies. Nan- kai Business Review (in Chinese), 5,6–17. Ross, S. (1973). The economic theory of agency: The principal’s problem. American Economic Review, 63, 134–39. Shen, K. R., & Fu, W. L. (2006). Tax competition, region game and their efficiency of growth. Economic Research Journal (in Chinese), 6,16–26. Stickney, C. P., & McGee, V. E. (1982). Effective corporate tax rates: The effect of size, capital intensity, leverage and other factors. Journal of Accounting and Public Policy, 1, 125–152. Wang, S. R., & Liu, N. (2012). Statistical research on the influences of tax preferential policies upon industries’ net profit based on annual report of listed companies. Journal of Beijing Technology and Business University (Social Science) (in Chinese), 2, 97–102. Wang, Y. M. (2003). The relation of income tax burden of listed companies and size, area and industry. Management World (in Chinese), 1,115–122. World Bank (2006). China – governance, investment climate, and harmonious society: Competitiveness enhancement for 120 cities in China. Washington, DC: World Bank. Wu, L., Wang, Y., Lin, B. X., Li, C., & Chen, S. (2007). Local tax rebates, corporate tax burdens, and firm migration: Evidence from China. Journal of Accounting and Public Policy, 26, 555–583. Wu, L. (2009). State ownership, preferential tax, and corporate tax burdens. Economic Research Journal (in Chinese), 10, 109–120. Wu, L., Wang, Y., Luo, W., & Gillis, P. (2012). State ownership, tax status and size effect of effective tax rate in China. Accounting and Business Research, 42,97–114. Xia, L. J., & Fang, Y. Q. (2005). Government control, institutional environment and firm value: Evidence from the Chinese securities market. Economic Research Journal (in Chinese), 5, 40–51. Xu, S. D., Yang, W. L., Xu, J. G., You, K. J., Du, M. K., Ma, Q. L., Yao, J. Q., Xu, S. Q., Tang, G. L., Pang, F. X., & Tie, W. (2001). Research on the division of tax power. Taxation Research (in Chinese), 3,2–7. Yang, Z. G., Ding, L., & Wu, B. Z. (2000). An empirical study of value-added tax & income tax burden on enterprise. Economic Research Journal (in Chinese), 12,26–35. Ye, Z. R., & Lin, H. (2007). Tax competition alienation and tax right institution innovation. Taxation Research (in Chinese), 2,18–22. Zheng, H. X., & Han, M. F. (2008). Tax planning analysis based on listed company with different ownership structure: The empirical evidence from state-owned listed companies and private listed companies in China. China Soft Science (in Chinese), 9, 122–131. Zhou, X. G. (2005). Inverted soft budget constraint: Extrabudgetary resource-seeking in local governments. Social Science in China (in Chinese), 2, 132–143. Zhou, F. Z. (2006). A decade of tax sharing: The system and its evolution. Social Science in China (in Chinese), 6, 100–115. Zhou, L. A. (2008). Local governments in transition: Official incentives and Governance (in Chinese). Shanghai: Shanghai People’s Publishing House. Zimmerman, J. L. (1983). Taxes and firm size. Journal of Accounting and Economics, 5, 119–149.

Journal

China Journal of Accounting StudiesTaylor & Francis

Published: Dec 1, 2013

Keywords: fiscal centralization; government control; turnover tax burden; corporate tax burden

There are no references for this article.