Deferred Taxes and Cost of Debt : Evidence from Japan a

Transcription

1 Deferred Taxes and Cost of Debt : Evidence from Japan a Yumi Inamura b Niigata Universy Shin ya Okuda c Osaka Gakuin Universy a Inamura would like to thank the Ministry of Education, Science, Sports, Culture and Technology (Grant-in-Aid for Young Scientists (B), , ). Okuda would like to thank the Ministry of Education, Science, Sports, Culture and Technology (Grant-in-Aid for Young Scientists (B), , ). b inamura@econ.niigata-u.ac.jp, Mailing Address: c/o Faculty of Economics, Niigata Universy, 5080 Ikarashi-nocyou, Nishi-ku, Niigata Cy, Niigata, , Japan. c Corresponding author. s-okuda@ogu.ac.jp, Fax: , Mailing Address: c/o Osaka Gakuin Universy, Kishibe-Minami, Sua-shi, Osaka , Japan. I

2 Deferred Taxes and Cost of Debt : Evidence from Japan Abstract This paper investigates the information content of deferred taxes in the Japanese debt market. We find that net deferred tax assets are negatively associated wh cost of debt. Moreover, we find that deferred tax liabilies are posively associated wh cost of debt but deferred tax assets are not. This evidence may reflect that is more probable that tax obligation will arise than the tax benefs will realize. Addionally, we find no evidence that deferred tax assets caused by temporary differences and deferred liabilies are significantly different wh respect to their impact on cost of debt. These results suggest that debt investors in the Japanese debt market value the various components of deferred tax assets differently. Keywords Deferred taxes, Cost of debt, Deferred tax assets, Deferred tax liabilies, Valuation allowances JEL codes M41, G1, G14 II

3 Deferred Taxes and Debt Cost: Evidence from Japan 1. Introduction This paper investigates the information content of deferred taxes in the Japanese debt market. Since 1998, Japanese firms have been required to report deferred taxes to harmonize Japanese accounting standards wh International Financial Reporting Standards (IFRS) and the Statement of Financial Accounting Standards (SFAS). 1 Another reason that Japanese firms must report information related to their deferred taxes is that the practice serves as a lifeboat for Japanese banks. In these days, Japanese banks had suffered from nonperforming debt and faced an increased possibily of a deteriorating capal-to-asset ratio. The mandatory reporting of deferred taxes was expected to boost banks capal because they could report a large amount of deferred tax assets caused by nonperforming debt. This study contributes to the lerature in several ways. First, in this study, we extend previous work on deferred taxes by exploring deferred tax assets rather than deferred tax liabilies. Most previous studies (e.g., Givoly & Hayn, 199; Amir et al., 1997; Ayers, 1998) have determined whether deferred tax liabilies have information content. This is because, in many countries, firms tend to report more deferred tax liabilies than assets. This does not necessarily mean that is not worthwhile to examine the information content of deferred tax assets. In fact, experts are more concerned about deferred tax assets than about deferred tax liabilies. For example, SFAS 96 was cricized because of s restrictive requirement for the recognion of deferred tax assets and s amendment by SFAS 109. This suggests that the 1 Before 1998, some firms reported deferred taxes in consolidated financial statements, but a separate standard of accounting for taxes was not issued. 1

4 requirement of recognizing deferred tax assets under SFAS 109 is permissive and firms may report more deferred tax assets. Thus, our analysis has important implications for the information content of deferred tax assets. Second, our analyses focus on Japanese firms deferred tax assets. Japanese firms are more likely to report deferred tax assets than liabilies; thus, the question as to whether tax assets have information content becomes even more salient. Skinner (008) suggest that Japanese regulators use deferred tax accounting as part of a regulatory forbearance strategy and that bank managers use these assets to bolster their banks regulatory capal. Yamori & Kobayashi (007) find that market participants downgraded banks wh greater dependency on deferred tax assets when Resona bank applied for an injection of public funds suggesting that market participants do not regard deferred tax assets as normal assets. Thus, our analyses have important implications regarding the value of information content of noncash assets. Third, whereas most previous work in this domain has investigated whether deferred taxes have information content from a stock market perspective, we build upon extant research by investigating information content from a debt market perspective. For example, Suda (00) and Okuda (004), both of which investigated deferred tax assets, find that Japanese deferred tax assets are posively associated wh stock price. This suggests that figures related to deferred taxes can assist in predicting future cash flow. To reveal possible differences between the stock and debt markets, this paper investigates the information content of deferred taxes from a debt market perspective. Because deferred tax assets are not real assets and cannot be cashed, these credors may not consider them assets. Through our analyses, we find that net deferred tax assets, which are defined

5 deferred tax assets mines deferred tax liabilies, are negatively related to cost of debt. This finding suggests that net deferred tax assets are recognized as assets in the Japanese debt market. Moreover, we find that deferred tax liabilies are posively correlated wh cost of debt but deferred tax assets are not correlated. This evidence may reflect that is more probable that tax obligation will arise than the tax benefs will realize. We also explore the different valuations of deferred tax assets and liabilies. First, we find no evidence that deferred tax assets caused by temporary differences and deferred liabilies are differently associated wh cost of debt. Second, we show that deferred tax assets caused by net operating losses are posively correlated wh cost of debt. Third, we find that when firms manage their earnings, their valuation allowances are less relevant in the debt market. The remainder of this paper is organized as follows. In the next section, we discuss the background of this project and the research hypotheses. In Section 3, we explain research design. We describe the data and the basic statistics in Section 4. Section 5 discusses the results of our analyses. Finally, in Section 6, we summarize our results and describe their implications.. Background and Research Hypotheses Japanese firms tend to recognize deferred tax assets more than liabilies. Some studies find that the Japanese stock market recognizes deferred tax assets as assets. However, there is no evidence that the Japanese debt market recognizes deferred tax assets as assets. If the Japanese debt market recognizes deferred tax assets as assets, In our sample which is consisted by bonds issued in Japan during sample period, the mean of deferred tax assets per total assets is and the median is These figures are larger than those of deferred tax liabilies. 3

6 then net deferred tax assets are negatively related wh the cost of debt, because an increase in assets increases a firm s capal adequacy ratio and reduces s insolvency. This constutes the following hypothesis. Hypothesis I: The net deferred tax assets are negatively related wh the cost of debt. Similar to the IFRS 1 or SFAS 109, Japanese accounting standard requires firms to recognize all deferred tax consequences of taxable temporary differences as deferred tax liabilies. However, Japanese accounting standard also requires firms to recognize all deferred tax consequences of deductible temporary differences and carryforwards as deferred tax assets and reduce them by valuation allowances if they are unlikely to be recognized. These suggest that deferred tax assets are less reliable than deferred tax liabilies. To provide empirical evidence on this issue, we test the following hypothesis in an alternative form. Hypothesis II: Deferred tax assets and liabilies are differently related wh cost of debt. Hypothesis IIa: Net deferred tax assets which are reported negative number are posively related wh cost of debt. But net deferred tax assets which are reported posive number are not related wh cost of debt. Hypothesis IIb: Deferred tax liabilies are related wh cost of debt. But deferred tax assets are not related wh cost of debt. Next, we decompose deferred taxes into net deferred tax assets from temporary 4

7 differences (TD), net operating losses (NOL), and valuation allowances (VA). Amir et al. (1997) suggest that the separation of deferred taxes into s constuent parts provides relevant information. They argue that the valuation of each component is contingent upon the likelihood and expected time to reversal. In Japan, tax losses expired in five (before fiscal year 004) or seven years (after fiscal year 004), 3 which are shorter time frames than those in the Uned States. This suggests that deferred taxes from NOL are less likely to be realized in Japan and may be less valuable in the Japanese debt market. Relative to NOL, the information content of VA is obscure. Similar to SFAS 109, if is possible that even a portion of tax assets will not be realized in the future, Japanese accounting standard requires firms to wre down deferred tax assets by VA. On one hand, VA may be valuable if managers are able to accurately forecast future cash flow and appropriately report VA. On the other hand, VA may be less valuable if a manager uses valuation allowances as an earnings management tool. In light of the above discussion, our hypothesis III s are as follows. Hypothesis III: The components of deferred tax are differently related wh cost of debt. Hypothesis IIIa: Net deferred tax assets from temporary differences (TD) are negatively related wh cost of debt. Hypothesis IIIb: Net deferred tax assets from net operating losses (NOL) are negatively related wh cost of debt. Hypothesis IIIc: Net deferred tax assets from valuation allowances (VA) are posively related wh cost of debt. 3 After fiscal year 01, tax losses expired in nine years in Japan but are restricted only 80%. 5

8 Finally, we discuss the relationship between the likelihood of earnings management and the information content of deferred taxes. Although many researchers of the late 1990s and early 000s have explored whether managers use valuation allowances as earnings management tools (see Miller & Skinner, 1998; Visvanathan, 1998; Bauman et al., 001), none have found evidence for this possibily. However, more contemporary researchers (see Schrand & Wong, 003; Frank & Rego, 006) have suggested that managers do, in fact, use valuation allowances as earnings management tools. The findings from the latter group may suggest that VA may be less valuable for the debt market. Based on these findings, we specify two scenarios for when managers are more likely to use VA as an earnings management tool. First, we explore whether deferred taxes from valuation allowances are posively related wh the cost of debt when the deferred tax from tax loss carryforward does not exceed the deferred tax from valuation allowances. Second, we investigate whether VA are less posively related wh the cost of debt if firms attempt to avoid losses or decreases in earnings. Some studies (e.g., Burgstarler & Dichev, 1997) have suggested that firms are likely to manage their earnings to avoid small losses or earnings decreases. If investors suspect that such firms avoid losses or earnings decreases by using VA, VA may prove to be less valuable and less posively related wh cost of debt. Based on above discussion, we derive following hypothesis. Hypothesis IV: The correlation of deferred taxes from valuation allowances wh the cost of debt varies wh the possibily of earnings management. 6

9 Hypothesis IVa: The deferred taxes from valuation allowances are posively related wh the cost of debt when the deferred tax from valuation allowances exceeds the deferred tax from tax loss carryforward. Hypothesis IVb: The deferred taxes from valuation allowances are less posively related wh the cost of debt if firms are suspected to avoid losses or earnings decreases. 3. Research Design This paper aims to examine the information content of deferred taxes in the debt market. Prior to this discussion, we propose the following basic model. Spread 1Capal- to- Asset Ratio k Control Variables,, 0 k where Spread is a proxy of the cost of debt. 4 Firstly, we expect a posive relationship between the cost of debt and leverage. As Myers (1977) describe, stockholders in leveraged firms have an incentive to transfer wealth from debtholders, even by rejecting investment projects wh posive net present value. Addionally, Black & Scholes (1973) argue that stockholders in leveraged firms are holders of European call options that have an exercise price equal to the face value of the debt. The value of a European call option increases as variance in future firm value increases. As a result, stockholders may choose projects wh negative net present values to ince higher variance in the future value of the firm. It is generally 4 Spread is the inial interest rate of a bond at the bond iniation date minus the interest rate of the Japanese Treasury bond matched by the matury and the bond iniation month, since our sample is bonds issued in Japan. 7

10 accepted that the effects described by Myers (1977) and Black & Scholes (1973) become more pronounced as a firm s leverage ratios increase (Gavish & Kalay, 1983). Therefore, debtholders of highly leveraged firms have an incentive to set high interest rates for protection. Rather than use the leverage ratio in the basic model, we use s inverse, the capal-to-asset ratio for our research design. 5 Given this, we predict a negative relationship between cost of debt and the capal-to-asset ratio. We also include seven control variables in our basic model. We choose these variables on the basis of the background lerature that has focused on debt covenants and the cost of debt as a proxy for cred risk. 6 First, in the tradion of El-Gazzar & Pastena (1991), we include materialy of new debt as one of the control variables. As the amount of the loan grows, debtholders incur a larger default risk. Therefore, there exists a posive relationship between the materialy of new debt and a debtholder s incentive to set the high cost of debt. Second, El-Gazzar & Pastena (1991) also argue that agreements wh long-term maturies pose a greater risk for default, but that debtholders can migate this risk by setting a high interest rate. Therefore, we predict that debtholders impose high costs of debt on a borrower if the matury of a debt is long-term. Third, consistent wh the findings of Graham et al. (008), we predict a negative relationship between a firm s profabily and cost of debt. Since firms wh high 5 We define the capal-to-asset ratio as net assets divided by total assets minus deferred tax assets. We adjust total assets not by deferred tax liabilies but by deferred tax assets because deferred tax liabilies do not affect total assets and only affect the classification of total liabilies and net wealth. 6 Jensen & Meckling (1976), Myers (1977), and Smh & Warner (1979) have developed the Agency Theory of Covenants, which provides a rationale for the presence of covenants in debt contracts (Bradley & Roberts, 004). Bradley & Roberts (004) discuss the relationship between the cost of debt and debt covenants as a proxy for cred risk. 8

11 profabily pose a low default risk for debtholders, they do not have an incentive to impose the high cost of debt on borrowers. Fourth, for reasons similar to those listed in our third prediction, we also expect a negative relationship between interest coverage ratio and cost of debt. Fifth, because large firms pose a low risk for default for debtholders (El-Gazzar & Pastena 1991), we predict a negative relationship between firm size and cost of debt. This prediction is consistent wh the findings of Graham et al. (008) and Shuto & Kagawa (011). Sixth, in line wh Billett et al. s (007) argument that firms wh significant growth opportunies are more likely to face stockholder-debtholder conflicts, we predict a posive relationship between a firm s growth opportunies and cost of debt. Finally, we expect a negative relationship between the existence of collateral and the cost of debt. debtholders can reduce potential losses resulting from borrowers defaults by requiring borrowers to provide collateral (El-Gazzar & Pastena 1991). Wh debtholders potential for loss migated, the use of collateral should negatively relate to the cost of debt. Next, to examine Hypothesis I, we decompose capal to assets ratio into two constuent parts: net deferred taxes and the remaining shareholders equy. We estimate the following regression to investigate how these components are valuated differently in the debt market: Spread NET 0 Year Dummies 1 NDT Control Variables k k, (1) 9

12 where NDT refers to net deferred taxes divided by total assets minus deferred tax assets and NET is the difference between the capal-to-asset ratio and NDT. In equation (1), if the null hypothesis ) is rejected, then is possible that net ( 1 deferred taxes and the remaining shareholders equy may be valuated differently in the market. Next, to test the Hypothesis II, we similarly separate net deferred taxes into s constuent parts in two ways: (a) Posive NDT and Negative NDT and (b) deferred tax assets and deferred tax liabilies. From this, the following equations are derived: Spread NET Control Variables k 0 1 Posive 3 k NDT Year Dummies Negative NDT (a) Spread NET 0 Year Dummies 1 DTA DTL 3 Control Variables k k, (b) where Posive NDT refers to NDT if NDT is posive value and otherwise 0, Negative NDT refers to NDT if NDT is negative value and otherwise 0, DTA is deferred tax assets divided by (total assets minus deferred tax assets), and DTL is deferred tax liabilies divided by (total assets minus deferred tax assets). Hypothesis II is accepted if the sign of is negative in (a) and if the sign of 3 is posive in (b). To examine Hypothesis III, we divide net deferred taxes in to s three elements: deferred taxes from temporary differences, from net operating losses, and from valuation allowances. Then, we divided net deferred taxes from temporary differences into posive part and negative part. From these, we derive the following equations: 10

13 Spread NET 0 k 1 TD Control Variables 3NOL 4VA, (3a) Year Dummies k Spread 0 1NET Posive TD+ 3Negative TD, (3b) NOL VA Control Variables Year Dummies 4 5 k k where TD is net deferred tax assets from temporary differences divided by (total assets minus deferred tax assets), NOL is deferred tax assets from net operation losses divided by (total assets minus deferred tax assets), and VA is deferred tax assets from valuation allowances divided by (total assets minus deferred tax assets). If the signs of and 3 in (3a) are negative and the sign of 4 in (3a) is posive, then Hypothesis III is accepted. If the signs of and 4 in and (3b) are negative and the signs of 3 and 5 in (3b) are posive, then Hypothesis III is accepted. Finally, to test Hypothesis IV, we specify those firms that are likely to manage their earnings wh valuation allowances. First, to evaluate Hypothesis IVa, we classified our sample into three groups of firms that have more deferred taxes from net operation losses than from valuation allowances, have more deferred taxes from valuation allowances than from net operation losses, and that have no net operating losses. Following this classification, we estimate following equations. Spread NET 0 k 1 TD Control Variables NOL k 3 VA Year Dummies 4 (4a-1)(4a-) Spread NET 0 k 1 TD Control Variables 3VA Year Dummies (4a-3) k 11

14 We restrict the sample to NOL above VA in equation (4a-1), VA above NOL in equation (4a-), and NOL =0 in equation (4a-3), We hypothesize that firms wh more deferred taxes from net operation losses are likely to manage their earnings; thus, we hypothesize that the 4 in (4a-1) should be less than 4 in (4a-) or 3 in (4a-3). Next, we test Hypothesis IVb. To specify firms that seek to avoid losses or earnings decreases, we designate firms wh returns-on-assets (ROA) between 0 and 0.01 as small-earnings firms. Similarly, we designate firms whose change of ROA (ROA) is between 0 and 0.01 as those wh small earnings increases. Following this, we derive the following equations. 7 Spread NET 0 k 1 TD Control Variables NOL k 3 VA Year Dummies 4 VA 5 DROA (4b-1) Spread NET 0 k 1 TD Control Variables NOL k 3 VA Year Dummies 4 VA 5 DROA (4b-) Where, DROA is 1 if ROA is 0 < ROA < 0.01 and 0 otherwise and DROA is 1 if ROA is 0 < ROA < 0.01 and 0 otherwise. If the signs of 5 in equation (4b-1) and (4b-) are posive, then Hypothesis IVb is rejected. 4. Data and Basic Statistics 4.1. Sample selection Our sample consist of public, interest-bearing, straight (i.e., nonconvertible) debt 7 We also employ other dummy variable definions to capture possibily of earnings management. See section 5.4. Robustness check. 1

15 issued by Japanese industrial corporations from April 1, 001 to March 31, Since public debt contracts tend to have limed flexibily wh respect to contract renegotiation, debtholders have an incentive to protect themselves by using strategic interest rates at the time the debt contract is made. Addional partial protection can be procured for debtholders if their claim is convertible into equy, as they can easily convert their assets to equy if the firm s risk for default increases (Tirole, 006). Therefore, we focus specifically on public, nonconvertible debt contracts to explore the relationship between cost of debt and deferred taxes. The debt contract data are procured from The Manual of Public Bonds (published by the Japan Securies Dealers Association). This manual contains various kinds of information on public debt, including matury, face value, and interest rates. To collect financial accounting data related to the issuers of the debt, we use the NIKKEI-NEEDS DVD edion database. In the sample period, 1,649 public, nonconvertible debts were issued. Due to incomplete data, we exclude 594 bonds (including bonds issued by banks and bonds wh floating rates). In addion, because of their unique characteristics, we also exclude 165 bonds issued by financial firms. As a result of these exclusions, the final sample consists of 890 public, nonconvertible bond contracts. 4.. Descriptive statistics Panel A of Table 1 presents descriptive statistics for our sample of 890 bond contracts over the 5-year sample period from April 1, 001 to March 31, 006. Definions for all the variables in the table are provided in the Appendix. The mean 8 In Japan, the fiscal year typically starts on April 1 and ends on March 31. According to the usual fiscal year in Japan, our sample period is 5 years from April 1, 001, to March 31,

16 (median) for Spread is about (0.3) percent, wh a standard deviation of The mean (median) of net deferred taxes, NDT (defined by {deferred tax assets deferred tax liabilies} / {total assets deferred tax assets}) is (0.03), suggesting that average firms in Japan have deferred tax assets. NET (defined by {net assets net deferred taxes} / {total assets deferred tax assets}) have a mean (median) of 0.46 (0.01), indicating that NDT represents a relatively large proportion of net assets about 10 percent on average. 9 The mean (median) of deferred tax assets, DTA, is (0.06) wh a standard deviation of 0.019, and the mean (median) of deferred tax liabilies, DTL, is (0.001) wh standard deviation of These results suggest that average Japanese firms have more DTA than DTL, corresponding to the results of NDT. Panel A of Table 1 also presents descriptive statistics associated wh the component variables of deferred taxes: temporary differences (TD), net operating loss (NOL), and valuation allowance (VA). The mean and median of TD are respectively 0.01 and 0.04, wh a standard deviation of Therefore, TD is a large percentage of NDT. The mean (median) of NOL is (0.003) wh a standard deviation of 0.018, and the mean (median) of VA is (0.004) wh a standard deviation of Given that our sample includes firms wh no NOL, this result suggests that firms that do have NOL tend to have a large NOL. Panel A also illustrates descriptive statistics related to our control variables, proxy for the materialy of new debt to bondholders (Material), time period (in years) from the bond iniation date to the time when all borrowed capal must be repaid (Matury), proxies for profabily (ICR, Margin), firm 9 The mean (median) of NDT divided by net assets is (0.100) and the maximum of NDT divided by net assets is Therefore, the amount of NDT to net assets is not small. 14

17 size (Size), proxy for the firm s opportunies for growth (Growth), and whether the bond is secured or not (Secured). Panel B of Table 1 reports the Pearson correlation coefficients between the assorted variables under investigation in this study. Although these correlations are only suggestive of the underlying relationships among our research variables, they appear to be indicative of our expectations. For example, although non-significant, the correlation between Spread and NDT is 0.038, suggesting that firms wh net deferred tax assets are likely to have lower borrowing costs. Spread is shown to be posively correlated wh DTA, suggesting that bondholders do not consider DTA as assets. On the other hand, Spread is posively associated wh DTL, indicating that bondholders consider DTL as debts. In addion, Panel B shows the correlations between the cost of debts (Spread) and the component variables of deferred tax assets: TD, NOL, and VA. The correlation between Spread and TD is but is not significant. This suggests that bondholders may not consider TD as assets. This result is consistent wh the result described above that Spread and DTA have a non-significant relationship. Spread is shown to be posively, significantly correlated wh NOL. This result is not consistent wh our expectation and may suggest that bondholders might think of NOL as a signal for a firm s poor performance. Consistent wh our expectation, Spread is posively related to VA, and the correlation is highly significant. In Panel B of Table 1, we also show the correlations between Spread and each of the control variables. All of these relationships are consistent wh the results of previous research, except for Matury. INSERT TABLE 1 ABOUT HERE 15

18 5. Results 5.1 Basis Results: Test of Hypotheses I and II Table reports the results for Equations (1), (a) and (b) using our full sample of 890 bond contracts. Columns 1 to 3 of Table report the estimated coefficients for these regression equations, along wh their corresponding t-values. Reported t-statistics are adjusted using standard errors corrected for firm-level clustering. First of all, the coefficient of NET (i.e., 1 in equation (1), (a) and (b)) is significantly negative. This result is consistent wh our expectation described in Section 3. As shown in Column 1, the coefficient of NDT (i.e., in equation (1)) is highly significant and negative (1.781, t =.97). This finding suggests that the deferred taxes are negatively associated wh the cost of debt. Addionally, the diference between 1 and is not statistically significant. Therefore, Hypothesis I is supported. Column of Table reports the regression results using two variables, Posive NDT and Negative NDT, to test Hypothesis II. Posive NDT is NDT if NDT > 0 and 0 otherwise. Negative NDT is NDT if NDT < 0 and 0 otherwise. Although the coefficient of Posive NDT (i.e., in equation (a)) is non-significant, the coefficient of Negative NDT (i.e., 3 in equation (a)) is significant wh a negative sign (.15, t =.85). These findings are consistent wh those derived from Column 1 and support Hypothesis IIa. In addion, when DTA and DTL are included instead of NDT (Column 3), the coefficient of DTA (i.e., in equation (b)) is not significant. In contrast, the coefficient of DTL (i.e., 3 in equation (b)) is highly significant and posive (.36, t = 3.44). These findings also consistent wh Hypothesis IIb and support the perspective that deferred tax liabilies have informational value to bondholders. 16

19 We also determine the sign and magnude of the effects from various control variables. First, coefficient of Size is significant and negative in all equations in Table. This suggests that bondholders charge lower interest rates for large firms. Second, we similarly find the coefficient of Matury to be significant and negative at the 5 percent level for all equations in Table. Finally, the effect of Secured on the interest rate on debt is negative. INSERT TABLE ABOUT HERE 5. Analysis of deferred tax components: Test of Hypothesis III Table 3 reports the results of equation (3a) and (3b) regressions for our full sample of 890 bond contracts. As shown in Column 1 of Table 3, the coefficient of TD (i.e., in equation (3a)) is highly significant wh a negative sign ( 1.703, t = 3.38). This finding implies that net deferred taxes from temporary differences are negatively related wh the cost of debt, which is consistent wh Hypothesis IIIa. In contrast, the coefficient of NOL (i.e., 3 in equation (3a)) is significant and posive (5.453, t = 1.76). This finding contradicts Hypothesis IIIb. Moreover, the coefficient of VA (i.e., 4 in equation (3a)) is not significant. This result did not provide support for Hypothesis IIIc. In Column of Table 3, we show the regression results, which included Posive TD and Negative TD instead of TD (i.e., equation (3b)). The coefficient of Posive TD (i.e., in equation (3b)) and Negative TD (i.e., 3 in equation (3b)) are significant wh expected negative signs ( =.09, t = 1.91; 3 = 1.33, t =.53). The difference between and 3, however, is not statistically significant. Similar to the estimation 17

20 result for Column 1, the coefficient of NOL (i.e., 4 in equation (3b)) is significant wh a posive sign (5.308, t = 1.73), which is not consistent wh Hypothesis IIIb. Moreover, the coefficient for VA (i.e., 5 in equation (b)) is posive, but not statistically significant. This result does not provide support for Hypothesis IIIc. These results suggest that deferred taxes from temporary differences have informational value to bondholders. More specifically, bondholders likely regard deferred taxes from posive temporary differences as assets and the deferred taxes from negative temporary differences as debts. Deferred taxes from net operating loss likewise have informational value to bondholders, but as a bad signal regarding cred. Deferred tax assets from valuation allowances do not appear to have informational value to bondholders. We also test a number of control variables in all equations in Table 3. First, the coefficient of Size is significant wh an expected negative sign. Second, we find the coefficient of Matury is negative and marginally significant (p <.10) for both equations. Finally, the effect of Secured on the interest rate on debt is negative. INSERT TABLE 3 ABOUT HERE 5.3 Analysis of deferred taxes from valuation allowance: Test of Hypothesis IV One component of deferred tax, valuation allowance, does not seem to have a significant relationship wh the cost of debt. One of the reasons for this is that bondholders may doubt the possibily of earnings management. Therefore, in addion to the analyses we outlined above, we also investigate whether the relationship between 18

21 deferred taxes from valuation allowances and the cost of debt varies wh the possibily of earnings management. Table 4 reports the regression results of deferred tax components using the reduced sample. In Column 1, we show the regression result using observations for which NOL VA (N = 361). In Column, we show the regression result using observations for which NOL < VA (N = 59). In Column 3, we show the regression result using observations for which NOL =0 (N = 74). The regression result in Column 1 (NOL VA) is qualatively similar to the results reported in Table 3 that use the entire sample. The coefficient of VA is negative and non-significant, which is inconsistent wh our expectation. However, in Column, the coefficient of VA (i.e., 4 in equation (4a-)) is (t = 1.97) and in Column 3, the coefficient of VA (i.e., 3 in equation (4a-3)) is (t = 1.67). These findings is consistent wh Hypothesis IVa, suggesting that deferred tax assets from valuation allowance in firms wh NOL < VA (included NOL = 0) have informational value to bondholders, thereby increasing the interest cost of borrowing. One explanation for this finding may relate to the unlikelihood for earnings management to increase deferred tax assets for firms in which NOL < VA. INSERT TABLE 4 ABOUT HERE We now turn to the test of Hypothesis IVb. Table 5 reports the regression results for deferred tax components wh dummy variables serving as proxies for the existence of earnings management. In Column 1 of Table 5, we use the dummy variable, DROA, which is 1 if ROA is 0 < ROA < 0.01 and 0 otherwise. In Column of Table 5, we use the 19

22 dummy variable, DROA, which is 1 if ROA is 0 < ROA < 0.01 and 0 otherwise. Relative to other firms, firms wh 0 < ROA (ROA) < 0.01 is more likely to manage earnings. Therefore, we use these dummy variables as proxies for the existence of earnings management and include the interaction term VA*DROA (VA*DROA) in the regression model. When (4b-1) and (4b-) are estimated after including the interaction term, the coefficient of VA (i.e., 4 in equations (4b-1) and (4b-)) captures the incremental interest rate for the average firm, while the coefficient of VA*DROA (VA*DROA) (i.e., 5 in equations (4b-1) and (4b-)) captures the incremental influence on VA for the firms in which 0 < ROA (ROA) < Thus, the influence of VA on the interest rate for firms in which 0 < ROA (ROA) < 0.01 is captured by When the interaction term is included (Columns 1 and ), the coefficient of VA remains posive, but non-significant. This finding is not consistent wh Hypothesis IIIc. Moreover, the coefficient of VA*DROA (i.e., 5 in equation (4b-1), Column 1) is negative, but non-significant (4.534, t = 1.60). In contrast, the coefficient of VA*DROA is negative and statistically significant (6.77, t =.6). This result provides support for Hypothesis IVb and suggests that valuation allowances are less posively related wh cost of debt when firms seek to avoid losses or decreases in earnings. INSERT TABLE 5 ABOUT HERE 5.4 Robustness check We also consider several alternative procedures to examine the robustness of our results. First, because of a lack of financial data for multiple components of deferred tax 0

23 assets, we use a reduced sample of 890 bond contracts in the main regression analysis (equations (1), (a) and (b)). However, the results may depend on small sample size. Therefore, we also analyze the relationship between the cost of debt and deferred taxes (equations (1), (a) and (b)) using an extended sample of 1,034 bond contracts issued in the sample period. The results of these addional regressions are qualatively similar to the results described in Table (not reported). Second, to check the robustness of our results reported in Table 5, We also use other definions of DROA and DROA. That is, DROA is dummy variable which 1 if 0 < ROA < and 0 otherwise and DROA is dummy variable which 1 if 0 < ROA < and 0 otherwise. The regression results are nearly identical to those presented in Table 5 (not reported). However, when we use a definion which DROA is 1 if 0 < ROA < 0.05 and 0 otherwise and DROA is 1 if 0 < ROA < 0.05 and 0 otherwise, the coefficients for both VA*DROA and VA*DROA are not significant. Thus, the significance of the estimated coefficient of VA*DROA reported in Table 5 appears to be sensive to the definion of DROA. 6. Conclusion This paper examined the information content of deferred taxes in the Japanese debt market. We find that net deferred tax assets are negatively associated wh cost of debt. Moreover, we find that deferred tax liabilies are posively associated wh cost of debt. We also find deferred tax assets are not statistically correlated wh cost of debt. This evidence may reflect that is more probable that tax obligation will arise than the tax benefs will realize. In addion, we find no evidence that deferred tax assets caused by temporary 1

24 differences are differently associated wh the cost of debt. Further, we show that deferred tax assets caused by net operating losses are posively related to the cost of debt. These findings suggest that debt investors in the Japanese bond market value the components of deferred taxes differently. Addionally, we find that valuation allowances are less posively related wh the cost of debt when firms report more deferred taxes from net operation losses or seek to avoid losses or decreases in earnings decreases, suggesting that the assessment of valuation allowances is affected by a firm s intention to manage earnings.

25 References Amir, E., M. Kirchenheer & K. Willard. (1997). The valuation of deferred taxes. Contemporary Accounting Research, 14(4): Ayers, B. C. (1998). Deferred tax accounting under SFAS No. 109: An empirical investigation of s incremental value-relevance relative to APB No. 11. The Accounting Review, 73(): Burgstahler, D. & I. Dichev. (1997). Earnings management to avoid earnings decreases and losses. Journal of Accounting and Economics, 4: Bauman, C. C., M. P. Bauman & R. F. Halsey. (001). Do Firms Use the Deferred Tax Asset Valuation Allowance to Manage Earnings? The Journal of the American Taxation Association, 3(s-1): 7 48 Black, F. & M. Scholes. (1973). The pricing of options and corporate liabilies. Journal of Polical Economy, 81(3): Billett, M. T., T.D. King & D.C. Mauer Growth opportunies and the choice of leverage, debt matury, and covenants. The Journal of Finance, 6(): Bradley, M. & M. R. Roberts. (004). The structure and pricing of corporate debt covenants. Unpublished paper, Duke Universy. Available at 3

26 El-Gazzar, S. & V. Pastena. (1991). Factors affecting the scope and inial tightness of covenant restrictions in private lending agreements. Contemporary Accounting Research, 8(1): Frank, M. M. & S. O. Rego. (006). Do Managers Use the Valuation Allowance Account to Manage Earnings around Certain Earnings Targets? Journal of the American Taxation Association, 8(1): Gavish, B. & A. Kalay. (1983). On the asset substution problem. Journal of Financial and Quantative Analysis, 18(1): 1 30 Givoly D. & C. Hayn. (199). The Valuation of the Deferred Tax Liabilies: Evidence from stock Market. The Accounting Review, 67(): Graham, J. R., S. Li & J. Qiu. (008). Corporate misreporting and bank loan contracting. Journal of Financial Economics, 89(1): Jensen, M. C. & W. H. Meckling. (1976). The theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics, 3(4): Miller, G. S. & D. J. Skinner Determinants of the Valuation Allowance for Deferred Tax Assets under SFAS No The Accounting Review, 73():

27 Myers, S. C. (1977). Determinants of corporate borrowing. Journal of Financial Economics, 5(): Okuda, S. (004). Valuation of the deferred taxes: Evidence from Japan. Review of Distribution, Communication and Business Administration, 30(1): (In Japanese) Shuto, A. & N. Kagawa. (011). The effect of managerial ownership on the cost of debt: Evidence from Japan. Journal of Accounting, Auding & Finance, 6(): Schrand, C. M. & M. H. Franco Wong. (003). Earnings Management Using the Valuation Allowance for Deferred Tax Assets under SFAS 109. Contemporary Accounting Research, 0(3): Skinner, D. J. (008). The rise of deferred tax assets in Japan: The role of deferred tax accounting in the Japanese banking crisis. Journal of Accounting and Economics, 46(-3): Smh, C. & J. Warner. (1979). On financial contracting: An analysis of bond covenants. Journal of Financial Economics, 7(): Suda, K. (00). On the Determinants of Accounting for Taxes on Income and the Value-relevance of Deferred Tax Assets. Annals of Economics and Business, 5: (In Japanese) 5

28 Tirole, J. (006). The theory of corporate finance. Princeton Universy Press. Vishvanathan, G. (1998). Deferred Tax Valuation and Earnings Management. Journal of Financial Statement Analysis, 3(4): 6-16 Yamori. N. & A. Kobayashi. (007). Wealth Effect of Public Funds Injections to Ailing Banks; Do Deferred Tax Assets and Auding Firms Matter? Japanese Economic Review, 58(4):

29 TABLE 1 Descriptive statistics for full sample Panel A: Distributional properties (N = 890) Variable Mean Std. Dev. Q1 Median Q3 Spread NET NDT DTA DTL TD NOL VA Material Matury ICR Margin Size Growth Secured Panel B: Pearson correlation matrix (N = 890) Variable Spread NET NDT DTA DTL TD NOL Spread 1.00 NET (0.00) 1.00 NDT (0.59) (0.000) 1.00 DTA (0.769) 0.17 (0.000) 0.69 (0.000) 1.00 DTL (0.076) (0.000) 0.81 (0.000) (0.000) 1.00 TD (0.61) (0.000) (0.000) (0.000) (0.000) 1.00 NOL 0.49 (0.000) (0.194) (0.69) 0.16 (0.000) (0.01) (0.000)

30 TABLE 1. Panel B (Continued) Variable Spread NET NDT DTA DTL TD NOL VA (0.000) (0.135) (0.934) (0.013) (0.038) 0.35 (0.000) (0.000) Material 0.09 (0.39) (0.011) (0.109) (0.013) (0.86) (0.154) 0.01 (0.540) Matury 0.83 (0.000) (0.81) (0.001) (0.030) 0.09 (0.006) (0.000) 0.01 (0.000) ICR (0.836) (0.000) (0.108) (0.019) 0.01 (0.730) (0.001) (0.001) Margin 0. (0.000) (0.000) (0.08) (0.001) 0.01 (0.731) 0.09 (0.006) (0.000) Size 0.94 (0.000) (0.000) (0.665) 0.00 (0.965) (0.577) (0.900) (0.094) Growth 0.14 (0.000) (0.055) 0.01 (0.538) (0.733) (0.567) 0.04 (0.479) 0.01 (0.000) Secured (0.000) 0.40 (0.000) (0.000) 0.04 (0.467) 0.40 (0.000) (0.000) 0.57 (0.000) Variable VA Material Matury ICR Margin Size Growth VA 1.00 Material 0.01 (0.53) 1.00 Matury 0.19 (0.000) (0.790) 1.00 ICR 0.17 (0.000) (0.166) (0.036) 1.00 Margin (0.000) 0.00 (0.56) (0.000) 0.16 (0.000) 1.00 Size (0.047) (0.001) 0.35 (0.000) 0.17 (0.000) 0.14 (0.000) 1.00 Growth 0.1 (0.000) (0.736) (0.033) (0.615) (0.103) (0.049) 1.00 Secured 0.04 (0.000) 0.05 (0.463) (0.001) 0.10 (0.000) 0.30 (0.000) (0.000) (0.03) Note: All variables are as defined in the Appendix. 8

31 TABLE Basis regression results (1) (a) (b) Coef. t-stat Coef. t-stat Coef. t-stat Variable NET *** *** *** NDT *** Posive NDT Negative NDT.15.85*** DTA DTL *** Material Matury ** ** ** ICR Margin Size *** *** *** Growth Secured *** *** *** intercept *** *** *** Year dummies Yes Yes Yes Adj. R N Notes: All variables are as defined in the Appendix. Columns 1 to 3 report the results of pooled ordinary least squares regressions. Reported t-values are on an adjusted basis using standard errors corrected for firm-level clustering. ***, **, * denote the 1%, 5%, and 10% levels of significance, respectively. 9

32 TABLE 3 Regression results of deferred tax components (full sample) (3a) (3b) Coef. t-stat Coef. t-stat Variable NET *** *** TD *** Posive TD * Negative TD ** NOL * * VA Material * Matury * * ICR Margin Size *** *** Growth Secured *** *** intercept *** *** Year dummies Yes Yes Adj. R N Notes: All variables are as defined in the Appendix. Columns 1 to report the results of pooled ordinary least squares regressions. Reported t-values are on an adjusted basis using standard errors corrected for firm-level clustering. ***, **, * denote the 1%, 5%, and 10% levels of significance, respectively. 30

33 TABLE 4 Regression results using observations wh NOL > VA, NOL < VA or NOL =0 (4a-1) NOL VA (4a-) NOL < VA (4a-3) NOL = 0 Coef. t-stat Coef. t-stat Coef. t-stat Variable NET *** *** ** TD * ** NOL * VA ** * Material Matury ** ICR *** * Margin Size *** *** *** Growth Secured *** *** intercept *** *** *** Year dummies Yes Yes Yes Adj. R N Notes: All variables are as defined in the Appendix. Columns 1 to 3 report the results of pooled ordinary least squares regressions. Reported t-values are on an adjusted basis using standard errors corrected for firm-level clustering. ***, **, * denote the 1%, 5%, and 10% levels of significance, respectively. 31

34 TABLE 5 Regression results wh dummy variables related to ROA (4b-1) 0 < ROA < 0.01 (4b-) 0 < ΔROA < 0.01 Coef. t-stat Coef. t-stat Variable NET *** *** TD *** *** NOL * * VA DROA*VA DΔROA*VA ** Material * Matury ** ** ICR Margin Size *** *** Growth Secured *** *** intercept *** *** Year dummies Yes Yes Adj. R N Notes: All variables are as defined in the Appendix. Columns 1 to report the results of pooled ordinary least squares regressions. Reported t-values are on an adjusted basis using standard errors corrected for firm-level clustering. ***, **, * denote the 1%, 5%, and 10% levels of significance, respectively. 3

35 Appendix Variable definions Variable Definion 1. Dependent variable Spread = The inial interest rate of the bond at the bond iniation date the interest rate of the Treasury bond matched by the matury and the bond iniation month.. Test variable NET = (net assets net deferred taxes) divided by (total assets deferred tax assets). NDT = Net deferred taxes (= deferred tax assets deferred tax liabilies) divided by (total assets deferred tax assets). Posive NDT = NDT if NDT > 0, 0 otherwise. Negative NDT = NDT if NDT < 0, 0 otherwise. DTA = Deferred tax assets divided by (total assets deferred tax assets). DTL = Deferred tax liabilies divided by (total assets deferred tax assets). TD = Deferred taxes from temporary differences divided by (total assets deferred tax assets). Posive TD = TD if deferred taxes from temporary differences > 0, 0 otherwise. Negative TD = TD if deferred taxes from temporary differences < 0, 0 otherwise. NOL = Deferred tax liabilies from net operating loss divided by (total assets deferred tax assets). VA = Deferred tax liabilies form valuation allowances divided by (total assets deferred tax assets). = 1 if ROA(= net income divided by (total assets deferred tax assets)) is DROA 0 < ROA < 0.01 and 0 otherwise. 1 if ROA is 0 < ROA < 0.01 and 0 otherwise. ROA is (ROA{t}-ROA{t-1}), DΔROA where ROA{t} is the issuer s ROA in year t (t is the latest end of the fiscal year before the bond is issued.) and S{t-1} is the issuer s ROA in year t Control variable Material = The face value of the new bond divided by the book value of the issuer s tangible assets. Matury = The period in years from the bond iniation date to the time all of the borrowed principal must be repaid. ICR = Interest coverage ratio (NIKKEI NEEDS em #FP01084) Profabily = Ordinary prof divided by total sales. Size = Log (total assets deferred tax assets) Growth = (S{t}-S{t-1}) / S{t-1}, where S{t} is the issuer s sales in year t (t is the latest end of the fiscal year before the bond is issued.) and S{t-1} is the issuer s sales in year t-1. Secured = 1 if the bond is secured, 0 otherwise. 33