Abstract
China Journal of aCC ounting StudieS , 2016 Vol . 4, no . 1, 79–103 http://dx.doi.org/10.1080/21697213.2016.1176416 Capitalising or expensing research and development expenditures: a tax perspective explanation* Liangliang Wang department of f inance and accounting, School of economics and Management, Southeast university, nanjing, China ABSTRACT KEYWORDS Capitalisation; financial Tax is frequently an important determinant of accounting choices. reporting costs; ownership This study investigates the impact of tax incentives on certain nature; r&d expenditures; accounting choices associated with research and development (R&D) tax benefits; tax enforcement expenditures. The empirical evidence suggests that firms facing a higher tax rate exhibit a significantly lower capitalisation ratio of R&D expenditures than firms facing a lower tax rate because of the higher tax benefits associated with expensing R&D expenditures. Further, the negative relation between tax rate and the capitalisation ratio of R&D expenditures is more pronounced for firms with lower financial reporting costs, for non-state-owned enterprises, and for firms in districts with weak tax enforcement. These findings contribute to the literature on accounting choices motivated by tax considerations and provide new explanation for the accounting choices associated with R&D expenditures. 1. Introduction Tax is an important determinant of myriad accounting choices (Scholes, Wolfson, Erickson, Maydew, & Shevlin, 2002). This study investigates whether corporate income tax affects a firm’s accounting choice regarding whether to capitalise or expense research and develop - ment (R&D) expenditures. A series of prior studies focused on the choice of a single account- ing method, such as inventory accounting, e.g. first-in first-out or last-in first-out (FIFO and LIFO hereafter), and concluded that tax incentives are an important determinant of a firm’s choice of the LIFO method for inventory accounting (Cushing & Leclere, 1992; Dhaliwal, Frankel, & Trezevant, 1994; Dopuch & Pincus, 1988; Frankel & Trezevant, 1994). However, since then, because there are few single accounting methods that are apt to involve tax-mo- tivated accounting choice, most studies have attempted to focus on the choice of combi- nations of accounting methods by examining a firm’s total accruals, abnormal accruals, or abnormal operating accruals. For example, Guenther (1994), Lopez, Regier, and Lee (1998), and Maydew (1997) use current abnormal accruals to examine tax-motivated income shifting prior to the Tax Reform Act of 1986 in the USA ( TRA 86 hereafter), and there are some similar studies on the Tax Reform Act of 2008 in China (Gai & Hu, 2012; Li, Dong, & Lian, 2011; Li & CONTACT liangliang Wang acwangll@126.com *Paper accepted by Kangtao Ye. © 2016 a ccounting Society of China 80 L. WAnG Zheng, 2010; Wang, Wang, & Gong, 2009). Overall, although there are many studies exam- ining tax-motivated accounting choices, the related research topics remain limited, and the research space in this field remains large. In the process of the convergence of international financial reporting standards, China issued new accounting standards in 2006 (CAS 2006 hereafter). The accounting standard number 6 of CAS2006 (CAS 2006 no. 6) covers intangible assets. CAS 2006 no. 6, distinguishes research expenditures from development expenditures (Article 7). It requires firms to report research expenditures as expenses of the period (Article 8) but it permits firms to report development expenditures as intangible assets if specific conditions are met (Article 9). The term ‘development’ refers to the application of research achievements and other knowledge to a certain plan or design, prior to the commercial production or use, so as to produce any new material, device or product, or substantially improved material, device and product. The permissions for capitalisation set out in CAS 2006 no. 6, are more flexible than the specifications in Sections 54 to 59 of IAS 38. Consequently there is some discretion available to the managers of Chinese companies. In their annual reports Chinese companies typically refer to ‘R&D expense’ and ‘capitalised R&D’ as single line items but do not distinguish research from development (for example, stock code: 002008, annual report of 2008). Consequently in this research paper I refer to ‘R&D expenditures’, following the practices used in the com- pany reports. From the foregoing description it is apparent that the accounting standard CAS 2006 allows, but does not require firms to capitalise R&D expenditures if they meet certain con- ditions. Because the tax law requires firms to deduct R&D expenditures based on the account - ing method used for financial reporting purposes, a firm’s choice between capitalising and expensing R&D expenditures also has significant consequences for a firm’s income tax expenses (Guoshuifa [2008] no. 116), which provides a unique opportunity to study tax-mo - tivated accounting choices. In addition, prior studies on firms’ choices of accounting methods for R&D expenditures are mainly from the perspective of earnings management or the sig- nalling effect (Li, Qu, & Xiao, 2013; Zong, Wang, & Yang, 2009), whereas few studies investigate the associated tax incentives. Overall, my paper examines accounting choice issues from the tax perspective, which may enrich the literature on tax-motivated accounting choices, par- ticularly on the choice of a single accounting method, but may also provide new evidence on the determinants of a firm’s choice of accounting method for R&D expenditures. I apply the following considerations to study the tax incentives of accounting choice for R&D expenditures. First, I surmise that when a firm is deciding on which part of its R&D expenditure should be capitalised, the distinction between research and development cannot be completely determined by reference to CAS 2006. Hence, I surmise that a firm has a certain amount of discretion over the portion of R&D expenditure to be capitalised, which provides the possibility and space for a manager’s opportunistic behaviour. Prior literature has shown that, historically the choice to capitalise R&D expenditures may have been discretionary for firms in France and China (Cazavan-Jeny and Jeanjean, 2006; Li et al., 2013; Zong et al., 2009). Second, prior research has shown the strategic effects of R&D expenditures have become increasingly important to firms in China, inducing firms to invest more in R&D (Zong et al., 2009). It is clear that R&D is important to a firm’s economic revenue; for example, this paper will show that in my research sample the median value of the ratio of R&D expenditures to net profit is 41%. Consequently, the tax benefit incentive may likewise become even more important, and firms may pay more attention to the tax benefit CHInA JOURnAL OF ACCOUnTInG STUDIES 81 when making decisions related to R&D expenditures ( Wang & Wang, 2015). Finally, in terms of the tax law with regard to R&D expenditures (Guoshuifa [2008] no. 116) in China, the new tax rule allows firms to deduct expensed R&D expenditures plus 50% in the current year but requires them to amortise capitalised R&D expenditures plus 50% for no less than 10 years. This would lead to significant differences in the tax consequences between expens- ing and capitalising R&D expenditures. Overall, the above aspects make it possible and meaningful to study the tax-motivated accounting choice associated with R&D expenditures in China. On this basis, I use Chinese companies listed on small and medium-sized board from 2008 to 2011 as my sample, and I examine the relationship between the tax rate and a firm’s capitalisation choice for R&D expenditures. The empirical results suggest that firms facing a higher tax rate exhibit a significantly lower capitalisation ratio of R&D expenditures than firms facing a lower tax rate, which indicates a higher tax benefit of expensing R&D expenditures for firms facing a higher tax rate. Further, I investigate the impact of financial reporting costs, the nature of ownership, and the tax enforcement level on the above relationship. The empirical results suggest that the negative relationship between the tax rate and the capitalisation ratio of R&D expenditures is more pronounced for firms facing lower financial reporting costs, for non-state-owned enterprises, and for firms in districts with weak tax enforcement, indicating a trade-off between tax benefits and other non-tax factors. The additional analysis suggests that the effects of capitalised and expensed R&D expenditures on firm value are quite different among firms with different tax rates; specif- ically, in comparison with firms enjoying a lower tax rate, the amount of capitalised R&D expenditures for firms facing a higher tax rate is lower, whereas the expensed R&D expendi - tures for firms facing a higher tax rate is higher. This finding is consistent with the theoretical expectations of the tax benefits perspective and further validates my tax perspective explanation. My study contributes to the literature in two ways. The prior literature on tax-motivated accounting choices includes the choice of a single accounting method, and the choice of combinations of accounting methods. The literature on the choice of a single accounting method mainly focuses on inventory accounting policies (Dopuch & Pincus, 1988); however, there is limited evidence regarding other single accounting methods. From the perspective of accounting methods for R&D expenditures in China, this paper examines tax-motivated accounting choices, which can enrich the literature on the tax-motivated choice of a single accounting method. The prior literature on the determinants of the accounting choice for R&D expenditures, regarding whether to expense or to capitalise them, has mainly examined the topics from the perspective of earnings management or the signalling effect (Li et al., 2013; Zong et al., 2009). Among these studies, some have indirectly examined the possibility of a tax-based explanation for the treatment of such expenditures; for example, Zong et al. (2009) find that firms with a high level of earnings tend to expense R&D expenditures, and the authors explain this choice from the tax perspective that firms with a high level of earnings can earn more by deducting R&D expenditures. Unlike these studies, I directly explore the association between a firm’s tax rate and the accounting choice for R&D t he Shenzhen Stock exchange once had specific requirements regarding the disclosure of r&d information for the firms listed on small and medium-sized board. Specifically, it required firms listed on small and medium-sized board to disclose their capitalised r&d expenditures, which provides reliable data for my study. 82 L. WAnG expenditures, which can provide direct and reliable evidence regarding the tax perspective explanation. The remainder of this paper is organised as follows. Section 2 reviews the tax-motivated accounting choice literature. Section 3 develops the hypothesis based on the theoretical analysis. Section 4 introduces the research design, including the sample selection procedure and the empirical models. Sections 5 and 6 present the main results, and section 7 reports robustness analyses. Section 8 concludes by discussing the study’s contribution. 2. Literature review Watts and Zimmerman (1990) review the literature on accounting choice and observe that there are two types of studies of accounting choice. The first type focuses on the choice of a single accounting method, e.g. the accounting method for depreciation, inventory or impairment of assets, and the other type focuses on the choice of combinations of account- ing methods that examine total accruals or abnormal accruals. Tax incentives are one of the most important determinants of a firm’s accounting choice (Shackelford and Shevlin, 2001), and the literature on tax-motivated accounting choices can also be classified into two main groups resembling those of Watts and Zimmerman (1990). The first group of studies examines the choice of a single accounting method. This group mainly focuses on the accounting choice for inventory policies. It is generally assumed that the cost or price of inventory generally rises as a result of inflation, and under this assumption, a firm’s choice of the LIFO method – as opposed to FIFO – for inventory would yield lower reported income and expected present value of future tax payments. Based on empirical evidence, Dopuch & Pincus (1988) use the ‘as-if ’ method to investigate the characteristics of firms that choose FIFO or LIFO over the long term, and the authors document that long-term LIFO firms would have foregone significant tax savings during the same time period had they computed their taxes on the basis of the FIFO method, whereas long-term FIFO firms would not have foregone significant tax savings, in which case continuing to use the FIFO method might be an optimal tax choice. Their results support the hypothesis regarding the tax-motivated accounting choice for inventory. More directly, Cushing and Leclere (1992) compare long-time FIFO users with long-time LIFO users by using logistic regression analyses to test whether incorporating a tax savings variable helps to evaluate the LIFO/FIFO choice, and they find that the anticipated tax savings is the primary reason that firms use LIFO, thus providing direct evidence regarding the choice of inventory method. Unlike these two stud- ies, Dhaliwal et al. (1994) investigate the timing of a LIFO liquidation and find that both the probability and the magnitude of a LIFO liquidation are greater for low-tax firms, which also suggests that tax is an important determinant of the inventory accounting choice. Overall, the studies discussed above mainly focus on the inventory accounting method; however, t he tax law of the uSa did not allow firms to use the lifo method to account for inventory for tax reports unless they used the same method for financial reporting purposes. t hus, the impact of the inventory policy on accounting income and taxable income is the same (d opuch & Pincus, 1988). When a firm uses the lifo method to account for inventory, a lifo liquidation occurs if current sales are higher than current purchases; as a result, any inventory not sold in previous periods must be liquidated, which can result in inflated profits because older inventory is typically purchased at a lower price than newer inventory due to inflation. Consequently, a lifo liquidation forces a firm to pay more tax and distribute more dividends (Wang, 2000). CHInA JOURnAL OF ACCOUnTInG STUDIES 83 the evidence based on the choice of other single accounting methods is limited (Dechow, Ge, & Schrand, 2010). The second group of studies examines the choice of combinations of accounting methods. These studies are generally derived from tax reforms and other natural experiments, and the research field is narrow. Based on TRA86 in the USA, Guenther (1994) examines firms’ abnormal current accruals in the year prior to the effective date of the act and finds significant negative abnormal current accruals, which suggests that firms have substantial incentives to defer income. Based on that study, Lopez et al. (1998) further investigate the impact of firms’ prior tax-aggressive behaviour on tax-induced earnings management around TRA86 and find a significant, positive relationship between those two behaviours. Analogously, based on the Tax Reform Act of 2008 in China, a series of studies investigates tax-motivated earnings management, such as Wang et al. (2009), Li & Zheng (2010), Li et al. (2011), and Gai and Hu (2012). In addition to the change in the tax rate due to the tax reform, Boynton, Dobbins, & Plesko (1992) also examine the ee ff ct of the Alternative Minimum Tax, which was enacted in TRA86, and find that firms shifted income to 1986 from 1987 to avoid the 20% tax rate on alternative minimum taxable income. Related studies include those by Dhaliwal and Wang (1992) and Manzon (1992), which are re-examined and reviewed by Choi, Gramlich, and Thomas (2001). After reviewing the above literature, it is clear that although the prior literature investi- gates firms’ accounting choices from the tax perspective, the scope of this literature remains narrow, particularly regarding the choice of a single accounting method. Based on the unique setting of Chinese tax and accounting law on R&D expenditures, as explained in the Introduction, this study investigates whether tax incentives can affect the choice of a single accounting method, which extends the prior literature and provides new empirical evidence by examining tax-motivated accounting choices based on a new accounting method. 3. Development of hypothesis On 15 February 2006, China’s Ministry of Finance issued a new accounting standard, CAS 2006, which was implemented on 1 January 2007 in listed companies. In terms of the accounting methods for R&D, CAS 2006 requires that R&D be divided into research and development phases, and the accounting methods for R&D expenditures during these two phases are very different. Specifically, R&D expenditures during the research phase are to be directly expensed, whereas R&D expenditures during the development phase are allowed to be capitalised when certain requirements are met. This change in the accounting method for R&D expenditures embodies the international convergence of accounting standards ( Wang, Wang, & Yang, 2012) but it is not a complete matching with IAS 38 because IAS 38 is more specific in its requirements for the capitalisation of development expenditure, as explained in the Introduction. To assist in adapting to the new accounting standards and to help firms meet the requirements for tax enforcement, the State Administration of Taxation of China issued Administrative Notice on the Deduction of Firm’s Research and Development in the early 1980s, extant studies, news, and other reports provided evidence that large american firms reported significant accounting earnings but paid little or no tax. in response, the a Mt was enacted in 1986. one component of the a Mt was a book income adjustment that included accounting earnings in the determination of a Mt liability (Choi et al., 2001). 84 L. WAnG Expenditures (Guoshuifa [2008] no. 116) in 2008, which adjusted the deduction policy for R&D expenditures. Specifically, the new tax rule allows firms to deduct expensed R&D expenditures plus 50% in the current year but requires them to amortise capitalised R&D expenditures plus 50% for no less than 10 years. Thus, the tax consequences of capitalised and expensed R&D expenditures are significantly different. Assume that the firm amortises the capitalised R&D expenditures over the next 10 years (the shortest amortisation period); then, if the discount rate i is 10%, the present value of the income tax reduced by capitalised R&D expenditures is 0.92Rτ, which accounts for nearly 61% of that reduced by expensed R&D expenditures. Using my research sample as an example, the median value of R&D expenditures is 34.71 million Yuan, and if we assume that the income tax rate equals 25%, then the tax reduction foregone by capitalising R&D expenditures is 5.08 million Yuan, which is a substantial amount to a firm. Moreover, when the discount rate is greater than 10%, the tax rate decreases, or when an operating loss occurs in the future, the tax reduction foregone by capitalising R&D expenditures is even larger. Klassen, Pittman, and Reed (2004) compare R&D expenditures in the United States and Canada, where the R&D tax credit mechanisms differ from one another. The authors docu- ment that the different tax credit mechanisms have different effects on the R&D investments of firms in the two countries. Unlike the real choice of the amount of R&D investment, the accounting choice for R&D expenditures is more flexible and discretionary, thus, it is more likely to be affected by tax incentives. Moreover, because there is uncertainty in accounting for R&D expenditures that require managerial and accountants’ judgement in China (Zong et al., 2009), this uncertainty provides space for the manager’s accounting choice. In addi- tion, because the choice to expense R&D expenditures is consistent with conservative accounting principles, it is less likely to be adjusted in an audit (Zong et al., 2009). Because the choice to expense R&D expenditures can decrease the tax burden and result in decreased tax revenue for the local tax authorities, the tax authorities will therefore pay more attention to expensed R&D expenditures. In other words, the accounting choice of R&D expenditures is influenced by tax enforcement costs (Desai, Dyck, & Zingales, 2007; Hoopes, Mescall, & Pittman, 2012; Ye & Liu, 2011). The above factors all drive the different accounting choices for R&D expenditures, and under dier ff ent tax benefits, the trade-off between these factors generates cross-sectional differences in the capitalisation ratios of firms’ R&D expenditures. Because the tax deduction foregone by capitalising R&D expenditures is positively corre- lated with a firm’s tax rate, the tax deduction foregone by capitalising R&D expenditures for firms facing a higher tax rate is higher than that of firms facing a lower tax rate. In addi- tion, the incentives to avoid tax per se for firms facing a higher tax rate are higher than those for firms facing a lower tax rate. As a result, the motivation to capitalise R&D expendi - tures for firms facing a higher tax rate is lower than that of firms facing a lower tax rate. In other words, firms facing a higher tax rate are more likely to expense R&D expenditures to obtain the tax benefits. Overall, I propose the following hypothesis: H1. Ceteris paribus, a firm’s tax rate is significantly negatively related to the capitalisation ratio of R&D expenditures. 1.5R t he present value is calculated as follows: ≈ 0.92R . 10∗(1+10%) i=1 CHInA JOURnAL OF ACCOUnTInG STUDIES 85 4. Research design 4.1. Sample selection My initial sample includes all A-share companies listed on the small and medium-sized board from 2008 to 2011, covering 1838 firm-year observations. I selected my sample based on the following process (presented in Panel A of Table 1): (1) I exclude firms in the financial industry because these firms are quite different from those in other industries; (2) I require firms to disclose their R&D information; (3) I also require firms to disclose their capitalisation information regarding R&D expenditures; (4) I exclude firms with missing data regarding the nature of ownership; and (5) I require firms to have no missing data for the other variables. After the above process, I obtain a final sample that consists of 218 firm-year observations. R&D expenditures and capitalisation information are collected from firms’ annual reports, whereas the other data are from the CSMAR and CCER databases. Distributions of my sample by year and industry sectors are shown in Panels B and C of Table 1. 4.2. Empirical models I construct the following model (1) to examine the association between a firm’s tax rate and its capitalisation choice for R&D expenditures: CapRatio = + HTAX + RDI + SIZE + OIACRD + LEV + TBQ + SHRCR 1 2 3 4 5 6 7 + BOARD + BI + MGT + MCOMP + AGE 8 9 10 11 12 (1) + Industry + Year + j j k k where the dependent variable is the capitalisation ratio of R&D expenditures (CapRatio), which is equal to the ratio of capitalised R&D expenditures to total R&D expenditures. Because CapRatio is truncated at 0, I use a Tobit model to estimate this regression. The explanatory variable is the dummy variable for the tax rate (HTAX), which is equal to 1 for firms whose effective tax rate is in the highest 30th percentile and 0 otherwise, where the effective tax rate is calculated following the prior literature ( Wang, 2003; Wu, Wu, & Rui, 2009) and which is equal to the ratio of the current income tax expense to the adjusted earnings before tax (earnings before tax minus investment gains and plus cash dividends and interest revenue). The other variables are control variables that are constructed as follows (Zong et al., 2009). RDI is the intensity of R&D and is equal to the ratio of a firm’s R&D expenditures to sales (An, Shi, & Alcorta, 2006; Li & Xia, 2008). SIZE is the firm’s size and is equal to the natural logarithm of total market value. OIACRD is the operating profit adjusted for capitalised R&D expenditures and is equal to the operating profit minus the capitalised R&D expenditures. i choose companies listed on the small and medium-sized board as my research sample because the Shenzhen Stock exchange once had specific requirements regarding the disclosure of r&d information. Specifically, in the ‘Small and medium listed companies’ information disclosure memorandum no. 28: a guide for management discussion and analysis’, it required firms listed on small and medium-sized board to disclose their total r&d expenditures (capitalised r&d in a separate column), the ratio of r&d expenditures to total sales, and the main results from r&d , including new products, technology, patents (e.g. name, type, and validity period), and their impact on the firm. Wang et al. (2012) also use firms listed on small and medium-sized board as a research sample, and they document that the r&d data disclosed under these rules are more consistent and can better solve the mismatching problem of r&d data found in previous studies. in fact, the accounting choice (capitalisation or expensing) for r&d expenditures can affect both the numerator and the denominator of the effective tax rate measure; in terms of the deduction for r&d expenditures, the capitalisation of r&d expenditures generally causes a firm’s effective tax rate to be higher, which is biased against my findings. 86 L. WAnG Table 1. Sample selection and distribution. Panel A: Sample selection initial sample: all small and medium sized a-share listed companies from 2008 to 2011 1838 excluding firms in financial industry (6) excluding firms not disclosing r&d information (445) f irms not disclosing information on capitalisation of r&d (1157) f irms with missing data for ownership nature (7) f irms with missing data for other variables (5) f inal sample: 218 firm-year observations from 2008 to 2011 218 Panel B: Distribution by year Year 2008 2009 2010 2011 t otal obs. 49 55 59 55 218 Percent. 22.48% 25.23% 27.06% 25.23% 100% Panel C: Distribution by industry industry obs. Percent. Mining 2 0.92% f ood and beverage 15 6.88% t imber and furnishings 1 0.46% Paper and printing 12 5.50% Petrochemicals, rubber, and plastic 22 10.09% electronics 26 11.93% Metal and non-metal 23 10.55% Machinery, equipment, and instrument 40 18.35% Medicine and biological products 22 10.09% o ther manufacturing 8 3.67% Construction 4 1.83% information technology 40 18.35% Wholesale and retail trades 3 1.38% t otal 218 100% LEV is the debt level, which is measured as debt scaled by total assets. TBQ represents a firm’s growth and is specified by the proxy of the firm’s Tobin’s q , which is collected from the CSMAR database. SHRCR is the shareholding ratio of the largest shareholder and is equal to the percentage of shares owned by the largest shareholder. BOARD represents the size of the firm’s board and is measured as the number of directors in the board. BI represents the independence of the board and is equal to the ratio of independent directors to all directors. MGT represents the level of management ownership and is equal to the percentage of shares owned by management. MCOMP represents the incentive level of management and is equal to the natural logarithm of the sum of the top three managers’ salaries. AGE represents the firm’s age and is equal to the number of years that the firm has been listed. Industry are industry dummies that indicate industry sector membership. Year are dummies for years to control for time effects. Table 2 shows the main variables and their definitions. As stated in H1, the greater the tax rate of a firm, the lower the capitalisation ratio of R&D expenditures; thus, I expect a negative sign for the coefficient on HTAX (β ). 5. Capitalising or expensing R&D expenditures: main results 5.1. Descriptive statistics and univariate tests Table 3, Panel A, presents the descriptive statistics for the variables. All the continuous var- iables are winsorised at the 1% and 99% levels. The average value of RDI is 6.1%, with a MIn value of less than 0.1% and a MAX value of 39.3%. The average capitalisation ratio of R&D expenditures (CapRatio) is 17.2%, with substantial differences among all the firms (SD of CHInA JOURnAL OF ACCOUnTInG STUDIES 87 Table 2. d efinitions of the main variables. Variable Definition CapRatio t he ratio of capitalised r&d expenditures to total r&d expenditures. HTAX a dummy variable which is equal to 1 for firms whose effective tax rate is in the highest 30th percentile and 0 otherwise, where the effective tax rate is equal to the ratio of the current income tax expense to the adjusted earnings before tax (earnings before tax minus investment gains and plus cash dividends and interest revenue). RDI t he ratio of a firm’s r&d expenditures to sales. SIZE t he natural logarithm of a firm’s total market value. OIACRD t he operating profit minus the capitalised r&d expenditures. LEV t he debt scaled by total assets. TBQ t he sum of book value of liability and market value of equity, divided by book value of total assets. SHRCR t he percentage of shares owned by the largest shareholder. BOARD t he number of directors in the board. BI t he ratio of independent directors to all directors. MGT t he percentage of shares owned by management. MCOMP t he natural logarithm of the sum of the top three managers’ salaries. AGE t he number of years that the firm had been listed. Industry industry dummies that indicate industry sector membership. Year Year dummies that controlling for time effects. Table 3. d escriptive statistics and univariate tests. Panel A: Descriptive statistics N Mean SD Min. Median Max. CapRatio 218 0.172 0.270 0.000 0.000 0.986 RDI 218 0.061 0.066 0.000 0.043 0.393 HTAX 218 0.298 0.459 0.000 0.000 1.000 SIZE 218 21.903 0.791 20.226 21.840 23.959 OIACRD 218 0.067 0.064 -0.043 0.053 0.282 LEV 218 0.334 0.180 0.029 0.323 0.690 TBQ 218 3.455 2.292 0.954 2.706 11.918 SHRCR 218 0.339 0.140 0.081 0.321 0.716 BOARD 218 8.711 1.678 5.000 9.000 13.000 BI 218 0.363 0.042 0.333 0.333 0.500 MGT 218 0.201 0.219 0.000 0.144 0.724 MCOMP 218 14.004 0.621 12.574 13.971 15.315 AGE 218 3.743 1.824 1.000 4.000 8.000 Panel B: Univariate tests of tax rate on the capitalisation ratio of R&D Variable f irms facing lower tax rate f irms facing higher difference (LTAX) tax rate (HTAX) (t/z-value) CapRatio Mean 0.204 0.098 –0.106*** (–2.67) Median 0.067 0.000 –0.067*** (–3.93) notes: a. difference=HTAX -LTAX; b. *, **, *** indicate two-tailed statistical significance at the 10%, 5%, and 1% levels, respectively. CapRatio = 27.0%). In addition, both the MIn and the median value of CapRatio are 0, which indicates that there is a large proportion of the sample firms that do not capitalise R&D expenditures, which is consistent with the greater tax benefit of expensing R&D expenditures. In my research sample, firms are classified into the higher tax rate group (HTAX =1) if their effective tax rates are in the top 30% of the distribution, and the other firms are classified into the lower tax rate group (HTAX=0). SIZE ranges from 20.226 to 23.959. Firms demonstrate significant variation in OIACRD, LEV, and TBQ. The average value of SHRCR is 33.9%, which 88 L. WAnG indicates a high concentration of firm ownership. Boards consist of between 5 and 13 board members, 36.3% of whom are independent directors. The mean value of MGT is 20.1%, and the MIn and MAX values are 0 and 72.4%, respectively, thus demonstrating significant var - iation. In addition, MCOMP ranges from 12.574 to 15.315. The average listed age is less than 4 years, which is due to the relatively late establishment of the small and medium-sized board. To test whether there is a significant difference in the capitalisation ratios of R&D expendi - tures between firms facing a higher tax rate and a lower tax rate, I conduct parametric and non-parametric one-factor tests. As shown in Panel B of Table 3, the mean capitalisation ratio of R&D expenditures for firms facing a lower tax rate is 20.4%, whereas that of firms facing a higher tax rate is 9.8%; the difference between these two types of firms is –10.6% (higher minus lower), which is significant at the 1% level, thus indicating that firms facing a lower tax rate are more likely to capitalise their R&D expenditures, which supports my hypothesis H1. In addition, the median capitalisation ratio of R&D expenditures for firms enjoying a lower tax rate is 6.7%, whereas that of firms facing a higher tax rate is 0. There is a significant difference (at the 1% level) between these two types of firms, thus further supporting my hypothesis H1. To obtain a clearer picture on the relation between tax rate and the capital- isation ratio of R&D, Figure 1 presents the mean value of the capitalisation ratio of R&D for each of the tax rate groups. In unreported Pearson and Spearman correlation coefficients, HTAX is significantly negatively correlated with CapRatio, which is also consistent with the expectation of my hypothesis H1. 5.2. Multivariate regression results Table 4 reports the multivariate regression results regarding the relationship between a firm’s tax rate and its capitalisation ratio of R&D expenditures. Because the dependent var - iable is truncated, I use the Tobit regression. Column (1) tabulates the results only including industry and year dummies as control variables, and the entire model is significant at the 2 2 1% level (χ =76.19), with pseudo R equal to 26.2%. Column (2) tabulates the results, including all the control variables, and the entire model is also significant at the 1% level (χ =105.97), with the fit effect improved (pseudo R =36.5%). Figure 1. t he capitalisation ratio of r&d between high-tax and low-tax firms. CHInA JOURnAL OF ACCOUnTInG STUDIES 89 Table 4. t he relation between a firm’s tax rate and its capitalisation ratio of r&d . Dependent variable = CapRatio (1) (2) HTAX –0.251*** –0.188** (–2.74) (–2.32) RDI –0.743 (–1.45) SIZE 0.119* (1.89) OIACRD –2.593** (–2.35) LEV 0.045 (0.12) TBQ 0.031 (1.13) SHRCR –0.443 (–1.36) BOARD –0.054* (–1.66) BI –1.464 (–1.24) MGT –0.176 (–0.77) MCOMP 0.011 (0.16) AGE –0.070** (–2.06) intercept 0.293*** –1.389 (4.31) (–0.91) industry Yes Yes Year Yes Yes Sigma_cons 0.373*** 0.334*** (9.37) (10.41) N 218 218 Pseudo R 0.262 0.365 Chi-Square (χ ) 76.19*** 105.97*** notes: t -statistics based on robust standard errors adjusted for clustering by firm are reported in parentheses. *, **, *** indicate two-tailed statistical significance at the 10%, 5%, and 1% levels, respectively. The coefficient of HTAX is negative and significant at the 1% level in column (1). After controlling for other effects, as shown in column (2), the coefficient of HTAX remains negative and significant at the 5% level. These results suggest that compared with firms in a lower tax bracket, firms facing a higher tax rate exhibit a lower capitalisation ratio of R&D expenditures, which implies that firms facing a higher tax rate will obtain more tax benefits by expensing R&D expenditures. Thus, my hypothesis H1 is supported. In addition, the magnitude of the coefficient of HTAX is 18.8% (in column (2)), which is significantly larger than the mean value of the capitalisation ratio in my sample, thus indicating economic significance. The coefficient of RDI is negative but not significant, which is contrary to the findings of prior studies (Aboody & Lev, 1998; Zong et al., 2009), which may be because my sample consists of firms that also disclose their capitalisation information on R&D expenditures. The coefficient of SIZE is positive and significant, which suggests that larger firms exhibit a higher capitalisation ratio of R&D expenditures, which may imply that there are economies of scale, i.e. the innovation efficiency of large firms may be higher such that the R&D investments of large firms are more likely to meet the requirements to capitalise R&D expenditures. The 90 L. WAnG coefficient of OIACRD is negative and significant, which suggests that firms with high prof- itability have relatively weak incentives to increase their earnings through capitalising R&D expenditures (Zong et al., 2009). The coefficient of BOARD is significantly negative, which may either imply that increasing the board size can help to reduce management opportun- ism behaviours (capitalising R&D expenditures), or it reflects that increasing the board size can lead to higher agency costs (Jensen, 1993; Yu & Chi, 2004) and decrease the firm’s inno - vation efficiency. The coefficient of AGE is also significantly negative, which suggests that the capitalisation ratio of R&D expenditures decreases as a firm’s listed age increases. The reason behind this result may be the reversal of accruals at an early stage of listing (Chen & Yuan, 2004), which can lead to a strong motivation not only to increase earnings but also to decrease tax aggressiveness. All the other control variables are not significant, and I do not interpret them here. 6. Capitalising or expensing R&D expenditures: cross-sectional analyses 6.1. Effect of financial reporting costs Scholes et al. (2002) develop a conceptual framework for effective tax planning and note that a tax planner must consider not only the tax costs but also the non-tax costs. Among these non-tax costs, prior studies have shown that financial reporting costs are of great importance (Shackelford & Shevlin, 2001). On the accounting choice for R&D expenditures, if firms want to obtain more tax benefits, then they would choose to expense R&D expendi- tures currently, which can result in lower reported earnings. Therefore, if firms face high financial reporting costs, they may forego the tax benefits to achieve their financial reporting goals. By contrast, firms facing lower financial reporting costs are more likely to achieve their financial reporting goals; as a result, they are more likely to choose the expensing method for R&D expenditures. Overall, an increase in financial reporting costs should result in a lower propensity to expense R&D expenditures for the tax incentives. To test whether firms should trade off between financial reporting costs and tax benefits when making the choice of accounting method for R&D expenditures, I develop the following model (2): CapRatio = + HTAX + HFRC + HFRC ∗ HTAX + RDI + SIZE + OIACRD 1 2 3 4 5 6 + LEV + TBQ + SHRCR + BOARD + BI + MGT 7 8 9 10 11 12 (2) + MCOMP + AGE + Industry + Year + 13 14 j j k k Model (2) augments model (1) with a dummy variable for financial reporting costs (HFRC). To measure a firm’s financial reporting costs, I employ the ‘as-if ’ method. Specifically, I first calculate firms’ net income as if they had expensed all their R&D expenditures (NI ). Second, Adj I compare the earnings benchmark (NI ; e.g. avoiding losses, maintaining growth, and meet- ing analyst forecasts) with actual net income (NI) and as-if net income (NI ). Finally, based Adj on the above comparisons, I define HFRC as 1 (firms facing higher financial reporting costs) in an early study by d opuch & Pincus (1988), who investigate the determinants of the inventory accounting choice, the ‘as-if ’ method is used. in addition to studies on inventory accounting choices, Matsunaga (1995) also uses the ‘as-if ’ method to proxy for the financial reporting costs of implementing employee stock options. Some scholars also employ the ‘as-if ’ method to study the accounting choice for r&d expenditures, for example, Wang et al. (2012) uses this method to study the impact of capitalised and expensed r&d expenditures on firm value. i draw on these studies and employ the ‘as-if ’ method to proxy for a firm’s financial reporting costs. CHInA JOURnAL OF ACCOUnTInG STUDIES 91 if both NI is higher than NI and NI is lower than NI ; otherwise, this measure is 0. Except E Adj E for HFRC, I also add the interaction between HFRC and HTAX (HFRC*HTAX) to the model, which is my main concern. Table 5, column (1), reports the Tobit regression results on the impact of financial reporting costs on the relationship between a firm’s tax rate and its capitalisation ratio of R&D expendi - 2 2 tures. The entire model is significant at the 1% level (χ =140.09), with pseudo R equal to 48.2%.The coefficient of HTAX is significantly negative in column (1), which indicates that the negative relation between a firm’s tax rate and its capitalisation ratio of R&D expenditures holds when the firm faces lower financial reporting costs. The coefficients of HFRC are positive and significant at the 1% level, which implies that firms are more likely to choose the capi- talisation method for R&D expenditures to increase earnings when financial reporting costs are higher, which is consistent with the evidence provided in Zong et al. (2009). The coeffi- cient of the interaction HFRC*HTAX is positive and significant at the 1% level, suggesting that as a firm’s financial reporting costs increase, the negative relationship between the firm’s tax rate and its capitalisation ratio of R&D expenditures is attenuated and even changes. These results imply that firms’ incentives to avoid tax are weakened when they trade off tax benefits against financial reporting costs (Scholes et al., 2002). 6.2. Effect of the nature of firm ownership The prior literature shows that state-owned enterprises are less tax aggressive than non- state-owned enterprises (Wang & Wang, 2015; Wang, Wang, & Peng, 2010; Zheng & Han, 2008). In state-owned enterprises, taxes can be viewed as a dividend to the controlling shareholder, the state, but as a cost to other shareholders (Bradshaw, Liao, & Ma, 2013). Therefore, the controlling shareholder of state-owned enterprises benefits more from less tax aggressiveness. Conversely, in non-state-owned enterprises, taxes are costs to the con- trolling shareholder, the private shareholders, such that the controlling shareholder of non- state-owned enterprises can benefit from tax avoidance activities. Thus, from the perspective of the controlling shareholders, non-state-owned enterprises might be more tax aggressive than state-owned enterprises. In addition, managers of state-owned enterprises are bur- dened with administrative duties other than the firm’s performance, and taxes are regarded as an important part of their duties. For example, Du, Tang, & Young (2012) analyse archival records of the government’s evaluation scores, score adjustments, and evaluation ratings given to Chinese state-owned enterprises by the State-Owned Assets Supervision and Administration Commission of China (SASAC) and show that firms’ social responsibilities (such as employment and taxes, etc.) affect the SASAC’s evaluations. Bradshaw et al. (2013) also document that a firm’s higher tax burden can help managers achieve promotion, thus providing evidence regarding the incentives of managers in state-owned enterprises. Overall, from the perspective of managers, the incentives for managers to avoid taxes in state-owned enterprises are relatively low. By contrast, managers of non-state-owned enterprises are typically the largest shareholders, and the interests between the principal and the agent are more consistent, which can also make firms’ motivation for tax avoidance stronger. In other words, non-state-owned enterprises are more likely to choose their accounting method for R&D expenditures motivated by tax benefits. To compare the effect of firm’s tax rate on the capitalisation choice for R&D expenditures between state-owned enterprises and non-state- owned enterprises, I construct the following model (3): 92 L. WAnG Table 5. Cross-sectional analyses on the relation between a firm’s tax rate and its capitalisation of r&d . Dependent variable = CapRatio Conditional_VAR Conditional_VAR Conditional_VAR =HFRC =SOE =HTE (1) (2) (3) HTAX –0.239*** –0.289*** –0.267*** (–3.00) (–2.88) (–3.02) Conditional_VAR 0.313*** 0.209** 0.178 (3.47) (2.09) (1.45) Conditional_VAR*HTAX 0.436*** 0.247* 0.339** (2.87) (1.80) (2.04) RDI –1.542*** –1.156*** –1.286** (–3.62) (–2.65) (–2.47) SIZE 0.085 0.102 0.110* (1.41) (1.57) (1.75) OIACRD –1.742** –2.138* –2.856*** (–1.99) (–1.97) (–3.11) LEV –0.158 0.119 0.087 (–0.53) (0.33) (0.27) TBQ 0.025 0.025 0.022 (1.09) (1.01) (0.76) SHRCR –0.355 –0.617* –0.426 (–1.23) (–1.86) (–1.34) BOARD –0.045* –0.064** –0.046 (–1.67) (–1.99) (–1.50) BI –0.899 –1.158 –1.286 (–0.86) (–1.08) (–1.13) MGT –0.199 –0.046 –0.138 (–1.02) (–0.19) (–0.62) MCOMP –0.023 0.006 0.000 (–0.35) (0.09) (0.00) AGE –0.052 –0.074** –0.074** (–1.64) (–2.27) (–2.44) intercept –0.456 –0.900 –1.032 (–0.32) (–0.58) (–0.68) industry Yes Yes Yes Year Yes Yes Yes Sigma_cons 0.297*** 0.322*** 0.319*** (11.64) (10.37) (10.83) N 218 218 218 Pseudo R 0.482 0.423 0.412 Chi-square (χ ) 140.09*** 122.86*** 119.72*** notes: t he Conditional_VARs in columns (1), (2), and (3) are HFRC, SOE, and HTE, respectively. t -statistics based on robust standard errors adjusted for clustering by firm are reported in parentheses. *, **, *** indicate two-tailed statistical signifi- cance at the 10%, 5%, and 1% levels, respectively. CapRatio = + HTAX + SOE + SOE ∗ HTAX + RDI + SIZE + OIACRD 1 2 3 4 5 6 + LEV + TBQ + SHRCR + BOARD + BI + MGT 7 8 9 10 11 12 (3) + MCOMP + AGE + Industry + Year + 13 14 j j k k Similar to model (2), model (3) augments model (1) with the dummy variable for state-owned enterprises (SOE) and its interaction with the tax dummy variable (SOE*HTAX), where SOE is coded as 1 for state-owned enterprises and 0 for non-state-owned enterprises. Table 5, column (2), reports the Tobit regression results on the impact of the nature of ownership on the relationship between a firm’s tax rate and its capitalisation ratio of R&D CHInA JOURnAL OF ACCOUnTInG STUDIES 93 2 2 expenditures. The entire model is significant at the 1% level (χ =122.86), with pseudo R equal to 42.3%. The coefficient of HTAX is significantly negative in column (2), which indicates that of non-state-owned enterprises, the negative relation between a firm’s tax rate and the capitalisation ratio of R&D expenditures holds. The coefficient of SOE is positive and signifi- cant at the 5% level, which may be the result of the higher financial reporting costs of state- owned enterprises (Wang et al., 2010; Zheng & Han, 2008). The coefficient of the SOE*HTAX interaction is positive and significant at the 10% level. This result suggests that the negative relationship between a firm’s tax rate and its capitalisation ratio of R&D expenditures is attenuated for state-owned enterprises. Because state-owned enterprises face higher finan - cial reporting costs, and non-state-owned enterprises can obtain more benefits from tax avoidance strategies, state-owned enterprises demonstrate less tax aggressiveness than non-state-owned enterprises. 6.3. Effect of the tax enforcement level Prior studies show that increases in the tax enforcement level can drive firms to be less tax aggressive (Chen, Chen, Cheng, & Shevlin, 2010; Hoopes et al., 2012; Jiang, 2013; Ye & Liu, 2011). On the choice of accounting method for R&D expenditures, although expensing R&D expenditures should be less influenced by the accounting standard and audit adjustments, it will reduce a firm’s tax burden and result in more scrutiny from the tax authorities. Of course, the tax enforcement costs related to expensing R&D expenditures are positively correlated with the tax enforcement level. As Jiang (2013) documents, the Chinese Tax Collection Law allows the tax authorities to ‘check the taxpayer’s books, vouchers, statements, and other relevant materials; go to the taxpayer’s place of production, business premises, and warehouse to check the taxpayer’s taxable commodities, goods, or other property…’ Thus, tax authorities can verify the actual status of firms’ R&D expenditures, including check - ing whether capitalised R&D expenditures meet any and all applicable requirements and confirming the overall level of R&D investment. Therefore, it can be shown that the choice of accounting methods for R&D expenditures is influenced by tax enforcement costs, which are determined by the tax enforcement level of the local tax authorities. According to the framework for effective tax planning proposed by Scholes et al. (2002), tax planners must trade-off tax benefits against non-tax costs. Thus, it can be inferred that the impact of tax benefits on the choice of accounting method for R&D expenditures should be attenuated when firms face a stronger tax enforcement level. To compare the effect of a firm’s tax rate on the capitalisation choice for R&D expenditures between firms in districts with a stronger tax enforcement level and those in districts with a weaker tax enforcement level, I construct the following model (4): CapRatio = + HTAX + HTE + HTE ∗ HTAX + RDI + SIZE + OIACRD 1 2 3 4 5 6 + LEV + TBQ + SHRCR + BOARD + BI + MGT 7 8 9 10 11 12 (4) + MCOMP + AGE + Industry + Year + 13 14 j j k k As with models (2) and (3), model (4) augments model (1) with the dummy variable for tax enforcement level (HTE) and its interaction with the tax dummy variable (HTE*HTAX), and HTE is a proxy for the tax enforcement level. I code HTE as 1 for firms in districts with a stronger tax enforcement level and 0 for firms in districts with a weaker tax enforcement level. To 94 L. WAnG measure the tax enforcement level of a district, I first draw on Mertens (2003) and Ye & Liu (2011) to develop the following tax burden predicted model TAX it = + GDPP + Industry_1 + Industry_2 + (5) 1 it 2 it 3 it it GDP it where TAX is the tax revenue for district i in year t. GDP is the gross domestic product for it it district i in year t. GDPP is the per capita GDP for district i in year t (in logarithmic form), and it Industry_1 and Industry_2 are the proportions of the first industry and second industry it it accounts of GDP, respectively. Based on model (5), I estimate the expected tax burden for district i in year t. Then, I can measure the tax enforcement level by using the difference between the actual and expected value of the tax burden (TE ). It must be noted here that it to distinguish the tax enforcement level of a district, I first calculated the mean value of TE it over the period from 2008 to 2011 (MTE ), then I code HTE as 1 for districts in which MTE is it it in the top 30%, and 0 otherwise. Table 5, column (3), reports the Tobit regression results on the impact of the external tax environment on the relationship between a firm’s tax rate and its capitalisation ratio of R&D expenditures. The entire model is significant at the 1% level (χ =119.72), with pseudo R equal to 41.2%. The coefficient of HTAX is significantly negative in column (3), which indicates that when the external tax enforcement level is weaker, the negative relation between a firm’s tax rate and its capitalisation ratio of R&D expenditures holds. The coefficient of HTE is positive and significant at the 15% level, which implies that the external tax enforcement level is positively correlated with the capitalisation ratio of R&D expenditures. The coefficient of the interaction HTE*HTAX is positive and significant at the 5% level, which suggests that as a firm’s external tax enforcement level increases, the negative relationship between the firm’s tax rate and its capitalisation ratio of R&D expendi - tures is attenuated. These results imply that when the external tax enforcement level increases, the tax enforcement costs due to tax avoidance activities increase. Thus, by trading off tax benefits and tax enforcement costs, firms’ incentives for tax avoidance are weakened (Scholes et al., 2002), which is consistent with Hoopes et al. (2012), Jiang (2013) and Ye & Liu (2011). 7. Further analysis: the effect of tax on the economic consequences of capitalising or expensing R&D expenditures Prior studies have shown that the impact of capitalising R&D expenditures on firm value is different from that of expensing R&D expenditures. For example, Wang et al. (2012) use Chinese small and medium-sized listed companies as a sample and find that capitalising R&D expenditures is significantly positively related to firm value, while expensing R&D expenditures is significantly negatively related to firm value. From the perspective of tax expenses, firms can currently deduct expensed R&D expenditures plus 50%, although they must amortise capitalised R&D expenditures plus 50% over at least 10 years into the future (Guoshuifa [2008] no. 116), which makes a substantial difference to the consequences of the total tax benefits. Based on this difference, one can expect that the market pricing for See previous theoretical analysis section. CHInA JOURnAL OF ACCOUnTInG STUDIES 95 capitalised and expensed R&D expenditures is also different; specifically, for firms facing a higher tax rate, the market pricing for their expensed R&D expenditures is higher than that of firms with a lower tax rate, although the market pricing for their capitalised R&D expendi - tures is lower. To examine this theoretical expectation, I follow Smith, Percy, and Richardson (2001) and Wang et al. (2012) and construct the following regression models: Price = + BPS + EPS + CapRDPS + HTAX + HTAX 1 adj 2 3 4 (6) ∗( BPS + EPS + CapRDPS)+ SIZE + LEV + TBQ + TACC (1) 5 adj 6 7 8 2 9 10 11 + SHRCR + BOARD + BI + MGT + AGE + Industry (2) 12 13 14 15 16 j j + Year + (3) k k Price = + BPS + EPS + ExpRDPS + HTAX + HTAX ∗( BPS + EPS 1 2 adj 3 4 5 6 adj + ExpRDPS)+ SIZE + LEV + TBQ + TACC + SHRCR + BOARD (1) 7 8 2 9 10 11 12 13 (7) + BI + MGT + AGE + Industry + Year + (2) 14 15 16 j j k k Price = + BPS + EPS + CapRDPS + ExpRDPS + HTAX + HTAX 1 adj 2 adj 3 4 5 ∗( BPS + EPS + CapRDPS + ExpRDPS)+ SIZE + LEV + TBQ (1) 6 adj 7 adj 8 9 10 2 11 12 (8) + TACC + SHRCR + BOARD + BI + MGT + AGE (2) 13 14 15 16 17 18 + Industry + Year + (3) j j k k where the dependent variable is Price, which is equal to the closing price of the firm’s stock at the end of April of the next year. BPS and EPS are net assets per share and earnings per share, respectively. BPS and EPS are adjusted net assets per share and adjusted earnings adj adj per share, respectively; the former is equal to the net assets per share minus the capitalised R&D expenditures per share, whereas the latter is equal to the earnings per share plus the expensed R&D expenditures per share. CapRDPS and ExpRDPS are capitalised and expensed R&D expenditures per share, respectively. SIZE represents the firm’s size and is equal to the natural logarithm of the firm’s total shares. TACC is the total accrual and is equal to net income minus cash flow from operating activities, deflated by total assets (Hribar & Collins, 2002). The other variables are as defined as earlier, so I do not repeat the descriptions here. The results of models (6) to (8) are presented in columns (1) to (3) of Table 6. As shown in Table 6, the coefficients of HTAX*CapRDPS are negative and significant at the 5% level in both columns (1) and (3) (based on models (6) and (8)), which suggests that the market pricing of capitalised R&D expenditures for firms facing a higher tax rate is lower than that for firms with a lower tax rate. By contrast, the coefficients of HTAX*ExpRDPS are positive and significant at the 5% level in both columns (2) and (3) (based on models (7) and (8)), which indicates that the market pricing of expensed R&D expenditures for firms facing a higher tax rate is higher than that of firms enjoying a lower tax rate. The above results are consistent with previous theoretical expectations, i.e. the impact of firms’ capitalised R&D and expensed R&D expenditures on firms’ value is different due to their different tax benefits. These findings not only provide new evidence on the economic consequences of firms’ expensed and capitalised R&D expenditures under different tax backgrounds but also con- stitute the basis of theoretical paths for my main hypothesis and further validate my conclusions. 96 L. WAnG Table 6. t he effect of tax rate on the economic consequences of capitalising or expensing r&d expenditures. Dependent variable = Price (1) (2) (3) BPS/BPS 2.247*** 2.360*** 2.161*** adj (5.77) (6.76) (5.42) EPS/EPS 9.212*** 9.165*** 9.120*** adj (3.97) (3.61) (3.84) CapRDPS 12.337 14.343 (1.24) (1.47) ExpRDPS –10.356* –7.920 (–1.81) (–1.55) HTAX 4.503*** 5.117** 3.825** (2.74) (2.40) (2.25) HTAX*BPS/HTAX*BPS –0.629 –1.259** –0.701 adj (–1.46) (–2.34) (–1.64) HTAX*EPS/HTAX*EPS –4.242 –4.722 –4.991* adj (–1.66) (–1.63) (–1.85) HTAX*CapRDPS –26.075** –27.296** (–2.21) (–2.24) HTAX*ExpRDPS 13.894** 13.240** (2.10) (2.07) SIZE –2.814*** –2.999*** –3.006*** (–4.44) (–4.19) (–4.37) LEV 7.131** 6.773** 5.975* (2.44) (2.46) (1.98) TBQ 2.535*** 2.619*** 2.524*** (7.28) (7.54) (7.11) TACC 11.190 11.211 11.631* (1.65) (1.54) (1.67) SHRCR 14.102*** 12.994*** 14.982*** (3.79) (3.46) (3.93) BOARD 0.478 0.515 0.490 (1.30) (1.37) (1.32) BI –0.030 3.630 0.303 (–0.00) (0.24) (0.02) MGT 1.573 1.112 1.737 (0.84) (0.60) (0.90) AGE 0.153 0.096 0.153 (0.57) (0.36) (0.57) intercept 31.014** 32.079* 35.027** (2.15) (1.91) (2.21) industry Yes Yes Yes Year Yes Yes Yes N 218 218 218 a dj. R 0.836 0.831 0.837 f-value 36.64*** 35.53*** 34.79*** notes: t -statistics based on robust standard errors adjusted for clustering by firm are reported in parentheses. *, **, *** indicate two-tailed statistical significance at the 10%, 5%, and 1% levels, respectively. 8. Robustness tests 8.1. Regression results after controlling the sample selection bias I choose companies listed on small and medium-sized board as the sample because the Shenzhen Stock Exchange required a higher standard for information disclosure of R&D expenditures; however, many of the firms in my initial sample did not disclose their CHInA JOURnAL OF ACCOUnTInG STUDIES 97 capitalisation information on their R&D expenditures. For these companies, there are two possible explanations for not disclosing information regarding their R&D expenditures: they did not capitalise any R&D expenditures, or they capitalised R&D expenditures but did not disclose this information. no matter the situation, there may be significant differences in profitability, R&D intensity, asset specificity, or industry distribution between firms that dis- closed and that did not disclose their capitalisation information on R&D expenditures, and ignoring these factors may lead to selection bias in my research design (Heckman, 1979). To mitigate this influence, I draw on the Heckman two-stage model and construct a first-stage Probit model as follows P(Dis_CapRD = 1)= + RDI + MS + ASI + MGT + ETR + EFN + SIZE 1 2 3 4 5 6 7 (9) + ROA + TBQ + LEV + Industry + Year + 8 9 10 j j k k where MS represents a firm’s market power, which is equal to the ratio of the firm’s sales to the sum of all firms’ sales in the industry. ASI represents asset specificity, which is measured as follows (Cushing & McCarty, 1996): ASI = EXP(–0.0801 + 0.1523*REC – 0.5749*INV + 0.6872*PPE + 0.198*ONC), and the greater the value, the lower the degree of the firm’s asset specificity. MGT is equal to the percentage of shares owned by Table 7. regression results when controlling the sample selection bias. Dependent variable Dependent variable = Dis_CapRD = CapRatio (1) (2) (3) (4) (5) RDI 12.539*** HTAX –0.194** –0.238*** –0.295*** –0.270*** (7.09) (–2.44) (–3.02) (–3.17) (–3.06) MS 24.841*** HFRC 0.294*** (2.75) (3.21) ASI 1.374*** HFRC*HTAX 0.389*** (2.99) (2.73) MGT –0.397** SOE 0.174* (–1.97) (1.69) ETR 0.195 SOE*HTAX 0.259** (0.70) (2.03) EFN –0.310 HTE 0.191* (–1.53) (1.69) SIZE –0.065 HTE*HTAX 0.329** (–0.74) (2.17) ROA 3.251** IMR –0.463*** –0.290* –0.395** –0.461*** (2.55) (–2.80) (–1.92) (–2.46) (–3.13) TBQ 0.001 Yes (0.03) o ther controls LEV 1.148*** intercept 0.018 0.407 0.298 0.358 (3.50) (0.01) (0.28) (0.18) (0.24) intercept –7.207*** industry Yes Yes Yes Yes (–3.55) Year Yes Yes Yes Yes industry Yes Sigma_cons 0.323*** 0.292*** 0.313*** 0.307*** Year Yes (10.36) (11.36) (10.37) (10.83) N 1335 N 218 218 218 218 2 2 Pseudo R 0.190 Pseudo R 0.400 0.498 0.450 0.451 2 2 Chi-square (χ ) 231.16*** Chi-square (χ ) 116.15*** 144.76*** 130.75*** 131.12*** notes: t -statistics based on robust standard errors adjusted for clustering by firm are reported in parentheses. *, **, *** indicate two-tailed statistical significance at the 10%, 5%, and 1% levels, respectively. 98 L. WAnG management. ETR is the effective tax rate, as above. EFN represents the need for external financing, following Demirgüç-Kunt and Maksimovic (1998). I use the difference between the growth rate and the sustainable growth rate, where the growth rate is measured as the sales growth rate, whereas the sustainable growth rate is measured as the ratio of ROE to (1–ROE). The other variables are as defined earlier, so I do not repeat their defini- tions here. It should be noted that the research sample for the first-stage regression consists of all firms who disclosed R&D expenditures information over the 2008–2011 period. The results of the first-stage regression are presented in column (1) of Table 7 , which shows that the pseudo R of the entire regression is 19.0%. The coefficients of RDI, MS, ASI, ROA, and LEV are all positive and significant at the 1% level (except for ROA, which is significant at the 5% level). In addition, MGT loads negatively at the 5% level, and the other variables are not significant. Based on the first-stage Probit regression, I calculate the Inverse Mill’s Ratio (IMR hereafter) and include it as a control variable in models (1) to (4). As shown in columns (2) to (5) of Table 7, the coefficients of IMR are all significantly negative at the 10% level or better, which indicates that a firm’s propensity to disclose its R&D capitalisation information is negatively correlated with the capitalisation ratio of R&D expenditures; in other words, firms that disclose their R&D capitalisation information generally have lower capitalisation ratios. These results suggest that it is necessary to control for selection bias. After controlling for selection bias, the results in columns (2) to (4) are the same as those reported in the previous tables, and my conclusions do not change. 8.2. Regression results based on the nominal tax rate The extent to which firms choose to capitalise R&D expenditures can also affect firms’ effec - tive tax rate, which can generate endogeneity in the regression; however, as the capitalisation ratio increases, the effective tax rate likewise increases, which would provide bias against my findings. To alleviate concerns about potential endogeneity, I also employ the nominal tax rate (NTR) of the parent firm to replace the effective tax rate measure, and I report the results in Table 8 (abbreviated results only include the main explanatory variables). As shown in Table 8, neither of the coefficients of NTR is significant in columns (1) to (4) of Table 8. However, the interactions of NTR and HFRC, SOE, and HTE are all significant with the expect signs in Columns (2) to (4), which is consistent with my previous findings. The reasons that the nominal tax rates are not significant may be due to the following. On the one hand, the nominal tax rate of the parent firm cannot effectively reflect the overall tax rate of the con- solidated firm. Taking my research sample as an example, I find that only 24 firms in my sample have a unified nominal tax rate, which accounts for 11% of all 218 firms. In addition, when using the nominal tax rate of the parent firm as a substitute, eight firms still do not have a unified tax rate due to the different tax rates among their branches. On the other a ccording to Chinese Corporate income t ax law, all incorporated affiliates must account for and pay income tax on their own. Because the tax rates of these affiliates generally are not the same, the consolidated financial reports do not disclose all affiliates’ divisional reports, so we cannot calculate the average nominal tax rate (Wang, 2014). t o obtain a larger sample, i use the parent company’s nominal tax rate rather than a uniform tax rate, which could require dropping many companies that have no uniform tax rate. CHInA JOURnAL OF ACCOUnTInG STUDIES 99 Table 8. regression results using the nominal tax rate (abbreviated results). Dependent variable = CapRatio (1) (2) (3) (4) NTR 0.011 0.008 0.002 –0.012 (1.02) (0.77) (0.25) (–0.94) HFRC*NTR 0.046** (2.09) SOE*NTR 0.063*** (2.78) HTE*NTR 0.057*** (3.04) o ther controls Yes intercept –0.628 0.449 –0.437 –0.133 (–0.43) (0.30) (–0.29) (–0.09) industry Yes Yes Yes Yes Year Yes Yes Yes Yes Sigma_cons 0.329*** 0.293*** 0.315*** 0.313*** (10.23) (11.30) (10.25) (10.30) N 210 210 210 210 Pseudo R 0.367 0.478 0.438 0.417 Chi-square (χ ) 102.49*** 133.54*** 122.31*** 116.43*** notes: t -statistics based on robust standard errors adjusted for clustering by firm are reported in parentheses. *, **, *** indicate two-tailed statistical significance at the 10%, 5%, and 1% levels, respectively. hand, the nominal tax rate of the sample lacks variation. Taking the 210 firms of my research sample as an example, the number of firms facing the 10 and 15% tax rates is 30 and 157, respectively, which accounts for almost 90% of all the firms; in addition, the number of firms facing the 25% tax rate is 15, which accounts for 7% of all firms. Overall, the variation in the nominal tax rate is small, which reduces the effectiveness of the estimation (Wooldridge, 2010).Thus, the above two aspects may reduce the reliability and effectiveness of the regres - sion based on the nominal tax rate. However, even so, I still confirm the hypotheses from the perspectives of financial reporting costs, the nature of ownership, and external tax enforcement. 8.3. Regression results relying on alternative effective tax rate measures Because the relationship between the nominal tax rate and the capitalisation ratio of R&D expenditures may not be strictly linear, I employ a dummy variable (HTAX) based on the effective tax rate to measure a firm’s tax rate, which does not require a linear assumption. I also use the effective tax rate per se (continuous variable) for the test, and the results are shown in column (1) of Table 9 (for brevity, I only report the basic test results). In addition to the continuous variable, I also employ three cutoffs (25%, 20%, and 15%) to construct the dummy variable (HTAX) and re -estimate the models. As shown in Table 9, my conclusions are robust to using alternative effective tax rate measures. 100 L. WAnG Table 9. regression results using alternative tax rate measures. Dependent variable = CapRatio HTAX: by other cutoffs (1) (2) (3) (4) HTAX: Continuous Cut off= 25% Cut off= 20% Cut off= 15% HTAX –0.368** –0.180** –0.276** –0.346*** (–2.10) (–2.10) (–2.47) (–2.94) RDI –0.974** –0.781 –0.901* –0.978* (–1.98) (–1.56) (–1.82) (–1.97) SIZE 0.117* 0.112* 0.100 0.097 (1.85) (1.77) (1.58) (1.52) OIACRD –2.620** –2.512** –2.734** –2.711** (–2.45) (–2.31) (–2.48) (–2.53) LEV 0.074 0.035 0.007 0.045 (0.21) (0.10) (0.02) (0.13) TBQ 0.037 0.035 0.041 0.042 (1.39) (1.27) (1.50) (1.54) SHRCR –0.502 –0.457 –0.474 –0.462 (–1.50) (–1.41) (–1.48) (–1.41) BOARD –0.054* –0.051 –0.046 –0.042 (–1.69) (–1.59) (–1.44) (–1.32) BI –1.466 –1.425 –1.348 –1.172 (–1.27) (–1.23) (–1.17) (–1.03) MGT –0.179 –0.194 –0.175 –0.197 (–0.80) (–0.86) (–0.78) (–0.88) MCOMP –0.002 0.004 –0.004 –0.007 (–0.02) (0.06) (–0.06) (–0.11) AGE –0.077** –0.071** –0.065* –0.070** (–2.31) (–2.15) (–1.91) (–2.09) intercept –1.117 –1.198 –0.875 –0.881 (–0.74) (–0.78) (–0.58) (–0.58) industry Yes Yes Yes Yes Year Yes Yes Yes Yes Sigma_cons 0.335*** 0.335*** 0.332*** 0.330*** (10.33) (10.46) (10.46) (10.41) N 218 218 218 218 Pseudo R 0.357 0.360 0.374 0.380 Chi-square (χ ) 103.81*** 104.63*** 108.50*** 110.40*** notes: t -statistics based on robust standard errors adjusted for clustering by firm are reported in parentheses. *, **, *** indicate two-tailed statistical significance at the 10%, 5%, and 1% levels, respectively. 9. Conclusion On 15 February 2006, China issued CAS2006, no. 6, which requires firms to expense their research expenditures for the research period but permits them to capitalise their develop- ment expenditures for the development period when they meet certain requirements. This separation of the accounting treatment of research and development expenditures is con- sistent with the international standard IAS 38, although it is less rigorous in permitting rather than requiring,capitalisation of development expenditure. In practice, Chinese companies continue to refer to ‘R&D’ expenditure as a collective item in their annual reports, and hence the terminology ‘R&D’ has been used in this paper. In line with CAS2006, the State Administration of Taxation likewise issued new deduction methods for R&D expenditures, specifically a new tax rule that allows firms to deduct expensed R&D expenditures plus 50% in the current year, but it also requires them to amortise capitalised R&D expenditures plus CHInA JOURnAL OF ACCOUnTInG STUDIES 101 50% for no less than 10 years into the future. Under this tax policy, the tax benefits between capitalising and expensing R&D expenditures are significantly different, which provides a research opportunity to study issues of tax-motivated accounting choices. On this basis, using Chinese companies listed on small and medium-sized board from 2008 to 2011 as a research sample, I first examined the association between a firm’s tax rate and its capitalisation ratio of R&D expenditures; further, I investigated the impact of financial reporting costs, the nature of ownership, and tax enforcement costs on the above relation- ship. The empirical results suggest that (1) compared with firms facing a lower tax rate, firms facing a higher tax rate exhibit a significantly lower capitalisation ratio of R&D expenditures; (2) compared with firms facing lower financial reporting costs, the negative relationship between the tax rate and the capitalisation ratio of R&D expenditures is weaker for firms facing higher financial reporting costs; (3) compared with state-owned enterprises, the neg- ative relationship between the tax rate and the capitalisation ratio of R&D expenditures for non-state-owned enterprises is more pronounced; and (4) compared with firms in districts with a weaker tax enforcement level, the negative relationship between the tax rate and the capitalisation ratio for firms in districts with a stronger tax enforcement level is attenuated. My findings could not only enrich the literature on accounting choices motivated by taxes but also provide new evidence on the determinants of the choice of accounting methods for R&D expenditures. Acknowledgments The author acknowledges the useful comments and suggestions made by two anonymous reviewers and the editors. Disclosure statement no potential conflict of interest was reported by the author. Funding The author acknowledges the financial support from national nature Science Foundation (71272100, 71372030, and 71502033) and the Fundamental Research Funds for the Central Universities (2242015S20001). References Aboody, D., & Lev, B. (1998). The value relevance of intangibles: The case of software capitalization. Journal of Accounting Research, 36, 161–191. An, T., Shi, H., & Alcorta, L. (2006). An observation and empirical study of R&D behavior of Chinese manufacturing firms: based on a survey of the manufacturing firms in Jiangsu province. Economic Research Journal, (2), 21–30. (in Chinese). Boynton, C. E., Dobbins, P. S., & Plesko, G. A. (1992). Earnings management and the corporate alternative minimum tax. Journal of Accounting Research, 30(Supplement), 131–153. Bradshaw, M. T., Liao, G., & Ma, M. S. (2013). Ownership structure and tax avoidance: Evidence from agency costs of state ownership in China. Available at SSRn 2239837. Cazavan-Jeny, A., & Jeanjean, T. (2006). The negative impact of R&D capitalization: A value relevance approach. European Accounting Review, 15(1), 37–61. 102 L. WAnG Chen, K. C. W., & Yuan, H. (2004). Earnings management and capital resource allocation: Evidence from China's accounting-based regulation of rights issues. The Accounting Review, 79, 645–665. Chen, S. P., Chen, X., Cheng, Q., & Shevlin, T. (2010). Are family firms more tax aggressive than non-family firms? Journal of Financial Economics, 95(1), 41–61. Choi, W. W., Gramlich, J. D., & Thomas, J. K. (2001). Potential errors in detecting earnings management: Reexamining studies investigating the AMT of 1986. Contemporary Accounting Research, 18, 571–613. Cushing, B. E., & Leclere, M. J. (1992). Evidence on the determinants of inventory accounting policy choice. The Accounting Review, 67, 355–366. Cushing, W. W., & McCarty, D. E. (1996). Asset specificity and corporate governance: An empirical test. Managerial Finance, 22, 16–28. Dechow, P., Ge, W., & Schrand, C. (2010). Understanding earnings quality: A review of the proxies, their determinants and their consequences. Journal of Accounting and Economics, 50, 344–401. Demirgüç-Kunt, A., & Maksimovic, V. (1998). Law, finance, and firm growth. The Journal of Finance, 53, 2107–2137. Desai, M. A., Dyck, A., & Zingales, L. (2007). Theft and taxes. Journal of Financial Economics, 84, 591–623. Dhaliwal, D. S., Frankel, M., & Trezevant, R. (1994). The taxable and book income motivations for a LIFO layer liquidation. Journal of Accounting Research, 32, 278–289. Dhaliwal, D., & Wang, S. W. (1992). The effect of book income adjustment in the 1986 alternative minimum tax on corporate financial-reporting. Journal of Accounting and Economics, 15(1), 7–26. Dopuch, n., & Pincus, M. (1988). Evidence on the choice of inventory accounting methods – LIFO versus FIFO. Journal of Accounting Research, 26(1), 28–59. Du, F., Tang, G., & Young, S. M. (2012). Influence activities and favoritism in subjective performance evaluation: Evidence from Chinese state-owned enterprises. The Accounting Review, 87, 1555–1588. Frankel, M., & Trezevant, R. (1994). The year-end LIFO inventory purchasing decision: An empirical test. The Accounting Review, 69, 382–398. Gai, D., & Hu, G. (2012). Research on balancing tax avoidance and cost of financial report: Evidence from Chinese reform of income tax in 2008. Accounting Research, (3), 20–25. (in Chinese). Guenther, D. A. (1994). Earnings management in response to corporate-tax rate changes – Evidence from the 1986 tax-reform act. The Accounting Review, 69(1), 230–243. Heckman, J. J. (1979). Sample selection bias as a specification error. Econometrica: Journal of the Econometric Society, 47(1), 153–161. Hoopes, J. L., Mescall, D., & Pittman, J. A. (2012). Do IRS audits deter corporate tax avoidance? The Accounting Review, 87, 1603–1639. Hribar, P., & Collins, D. W. (2002). Errors in estimating accruals: Implications for empirical research. Journal of Accounting Research, 40(1), 105–134. Jensen, M. C. (1993). The modern industrial revolution, exit, and the failure of internal control systems. The Journal of Finance, 48, 831–880. Jiang, X. (2013). Tax enforcement, tax aggressiveness, and stock price crash risk. Nankai Business Review, (5), 152–160. (in Chinese). Klassen, K. J., Pittman, J. A., & Reed, M. P. (2004). A cross–national comparison of R&D expenditure decisions: Tax incentives and financial constraints. Contemporary Accounting Research, 21, 639–680. Li, D., & Xia, L. (2008). Ownership type, the institutional environment and R&D intensity of Chinese listed firms. Journal of Finance and Economics, (34), 93–104. (in Chinese). Li, L., Qu, X., & Xiao, H. (2013). Capitalization of R&D expenditures: Real signal transmission or earnings management? Journal of Audit & Economics, (1), 60–69. (in Chinese). Li, Z., Dong, Z., & Lian, Y. (2011). Accrual management or real activities management? Evidence based on Chinese tax reform in 2007. Management World, (1), 121–134. (in Chinese). Li, Z., & Zheng, Y. (2010). Study on the earnings management for tax avoidance based on manipulations of accruals and real activities. Journal of Finance and Economics, (6), 80–89. (in Chinese). Lopez, T. J., Regier, P. R., & Lee, T. (1998). Identifying tax-induced earnings management around TRA 86 as a function of prior Tax-Aggressive behavior. Journal of the American Taxation Association, 20, 37–56. Manzon, G. B. (1992). Earnings management of firms subject to the alternative minimum tax. Journal of the American Taxation Association, 14, 88–111. CHInA JOURnAL OF ACCOUnTInG STUDIES 103 Matsunaga, S. R. (1995). The effects of financial-reporting costs on the use of employee stock-options. The Accounting Review, 70(1), 1–26. Maydew, E. L. (1997). Tax-induced earnings management by firms with net operating losses. Journal of Accounting Research, 35(1), 83–96. Mertens, J. B. (2003). Measuring tax effort in Central and Eastern Europe. Public Finance and Management, 3, 530–563. Scholes, M., Wolfson, M., Erickson, M., Maydew, E., & Shevlin, T. (2002). Taxes and Business Strategy: A Planning Approach. Englewood Cliffs, nJ: Prentice Hall. Shackelford, D. A., & Shevlin, T. (2001). Empirical tax research in accounting. Journal of Accounting and Economics, 31(1–3), 321–387. Smith, T., Percy, M., & Richardson, G. D. (2001). Discretionary capitalization of R&D: Evidence on the usefulness in an Australian and Canadian context. Advances in International Accounting, 14, 15–46. Wang, L. (2014). Tax reform and inter-temporal income shifting: Evidence based on book-tax difference. Management World, (11), 105–118. (in Chinese). Wang, L., & Wang, Y. (2015). Does control privatization affect corporate tax aggressiveness? China Accounting and Finance Review, (17), 47–98. (in Chinese). Wang, L., & Wang, Y. (2015). R&D investment and the choice of capital structure: An analysis from the perspective of non-debt related tax shields. China Industrial Economics, (11), 125–140. (in Chinese). Wang, L., Wang, Y., & Yang, Z. (2012). International convergence of accounting standards, R&D expenditure and economic consequences. Journal of Finance and Economics, (38), 49–60. (in Chinese). Wang, Y. (2003). Tax burden of listed companies: Evidence based on size, districts, and industries. Management World, (1), 115–122. (in Chinese). Wang, Y., Wang, L., & Gong, C. (2009). Reform of enterprise income tax, earnings management, and its economic consequences. Economic Research Journal, (3), 86–98. (in Chinese). Wang, Y., Wang, L., & Peng, Y. (2010). Ownership nature of ultimate controller, debt-related tax shields, and capital structure. Economic Research Journal, (9), 122–136. (in Chinese). Wang, Z. (2000). Different types of last-in first-out methods and their applications. Accounting Research, (10), 36–37+45 (in Chinese). Watts, R. L., & Zimmerman, J. L. (1990). Positive accounting theory: A ten year perspective. The Accounting Review, 65(1), 131–156. Wooldridge, J. M. (2010). Introductory Econometrics: A Modern Approach. Beijing: China Renmin University Press (in Chinese). Wu, W., Wu, C., & Rui, M. (2009). Manager’s government background in Chinese listed companies and firm’s tax preferences. Management World, (3), 134–142. (in Chinese). Ye, K., & Liu, H. (2011). Tax enforcement, income tax costs, and earnings management. Management World, (5), 140–148. (in Chinese). Yu, D., & Chi, G. (2004). number of directors, board stability and performance: Theory and experience. Economic Research Journal, (4), 70–79. (in Chinese). Zheng, H., & Han, M. (2008). Tax planning analysis based on listed company with different ownership structure: empirical evidence from state-owned listed company and private listed company in China. China Soft Science, (9), 122–131. (in Chinese). Zong, W., Wang, R., & Yang, Y. (2009). A study on motives of capitalizing R&D expenditures: Evidence from Chinese A-share market. China Accounting Review, (7), 439–454. (in Chinese).
Journal
China Journal of Accounting Studies
– Taylor & Francis
Published: Jan 2, 2016
Keywords: Capitalisation; financial reporting costs; ownership nature; R&D expenditures; tax benefits; tax enforcement
