The Rise of Deferred Tax Assets in Japan: The Role of Deferred Tax Accounting in the Japanese Banking Crisis*

Douglas J. Skinner

Graduate School of Business University of Chicago

dskinner@chicagogsb.edu

April 2008; original version February 2005

Abstract

This paper provides evidence on the role that accounting for deferred taxes played in the recent financial crisis among the major Japanese banks, dramatized by the collapse of Resona Bank in 2003. Upon adoption of deferred tax accounting in fiscal 1998, the major Japanese banks immediately recognized net deferred tax assets (DTAs) of 6.6 trillion (approximately $55 billion), or about 29% of bank equity. Without these assets, these banks would have been insolvent. The evidence supports the conclusion that the Japanese Government and bank regulators used deferred tax accounting as part of a regulatory forbearance strategy, and that bank managers responded by using deferred tax assets to supplement their banks regulatory capital. The results have implications for international accounting standard-setters because they illustrate how economic incentives can distort the way that accounting rules are adopted and used. Key words: Deferred tax assets; Japanese banks; regulatory forbearance; regulatory capital arbitrage; bank regulation. JEL Codes: G18, G21, M41.

*I am grateful to Kenichi Akiba, Ned Akov, Anne Beatty, Christa Bouwman, Mitsuhiro Fukao, Michelle Hanlon, Shigeo Ishigaki, Kazuo Kato, Gregory Miller, Naoko Nemoto, Tsuyoshi Oyama, Junichi Suzuki, Ross Watts, anonymous referees, and workshop participants at the AFBC in Sydney, JAR/LBS conference, the NBER Japan Project Meeting, Florida, Houston, Northwestern, and Stanford for helpful comments and discussions, and to Shinnosuke Iizuka, Masaki Mizutani, Manabu Sakagami, Yoshiko Imai-Suga, Tomomi Sugiura, and especially Tomohito Suga for helpful assistance, translation help, and discussions regarding Japanese banks and Japanese accounting rules. All errors and inaccuracies are my own. Financial support from the University of Chicago, Graduate School of Business is gratefully acknowledged.

1

1. Introduction

This paper provides evidence on accounting for deferred taxes by Japanese banks from

fiscal 1998, when deferred tax accounting was first adopted by the banks, through fiscal 2003.

The main goal of the paper is to document the role of deferred taxes in the decade-long crisis in

the Japanese banking sector. I investigate whether: (1) deferred taxes were used by the Japanese

Government and bank regulators at the Ministry of Finance (MOF) and Financial Services

Agency (FSA) as a tool of regulatory forbearance; that is, to give the major Japanese banks

the appearance of financial health when in fact many were insolvent, (2) managers of the banks

themselves used the discretion inherent in deferred tax accounting to practice regulatory capital

arbitrage; that is, to manage reported levels of regulatory capital in such a way as to avoid

falling below minimum capital thresholds.

I present two types of evidence in the paper. First, I provide aggregate evidence on

changes in the financial strength of the major Japanese banks from 1982 to 2003 and show how

deferred tax assets (hereafter, DTAs) became a significant part of the banks regulatory capital.

The bursting of the bubble in Japanese stock and real estate prices in the early 1990s and the

subsequent economic malaise caused a steady decline in the financial strength of these banks.

During the first part of the 1990s, the banks were able to maintain necessary levels of regulatory

capital by realizing gains on their large holdings of equity securities. This source of financial

slack ran out in about 1998.

The timing and circumstances of the introduction of deferred tax accounting in Japan are

consistent with its use as a tool of regulatory forbearance. The rule was introduced in 1998,

when the banks other sources of regulatory capital had been depleted. The rule was introduced

at a time when Japanese accounting rules were set by a committee of the MOF, was issued in

2

conjunction with Japanese law that allowed deferred taxes to be included on corporate balance

sheets for the first time, and was introduced quickly, with only four months between the initial

exposure draft and the final statement. Perhaps most important, the rule was accompanied by a

decision by regulators at the MOF to allow banks to include DTAs in their regulatory capital.

This decision was unique bank regulators in other countries either do not recognize DTAs at all

or impose a stringent cap. The rule was immediately adopted (earlier than required) by almost

all Japanese banks. Because of the unusually rapid rule-making process, little accounting or

auditing guidance accompanied the new rule, making it difficult for auditors to verify the

reasonableness of the banks DTAs.1 When first recognized in fiscal 1998, net DTAs at the

major Japanese banks totaled 6.6 trillion, or about 29% of shareholders equity; all of the major

Japanese banks would have fallen below minimum capital levels without this accounting

change.2 The recognition of these DTAs allowed the Government to characterize these banks as

healthy, helping to rationalize its decision to inject 7.5 trillion in new capital.

Deferred tax assets continued to be an important source of regulatory capital for Japanese

banks in the years after this, most notably in fiscal 2002 when the major banks DTAs totaled

7.5 trillion, or 60% of equity. The collapse of Resona Bank likely signaled the beginning of the

end for DTAs in Japanese banks. Resona failed in May 2003 when its auditors refused to sign

off on its DTAs, causing it to fall below minimum capital levels. The Government immediately

stepped in to save the bank, injecting 1.96 trillion in new capital. In the period after Resona

there has been a decline in the major banks use of DTAs, although DTAs still represent a

1 Accounting for deferred taxes provides managers with significant discretion because the realizability of DTAs depends on managers assessments of their firms ability to generate sufficient levels of future taxable income (Miller and Skinner, 1998; Schrand and Wong, 2003; Dhaliwal, Gleason, and Mills, 2004). Deferred taxes provide an especially difficult challenge for auditors because there is no objective evidence they can use to verify managers estimates of the realizability of DTAs. 2 6.6 trillion is equivalent to about $55 billion using the exchange rate in effect in late 1998 (about 120/$). This exchange rate can be used as a rough approximation throughout the period of this study.

3

significant fraction of equity for some banks.3 Overall, the evidence supports claims by

economists that accounting helped to prolong the Japanese financial crisis by allowing regulators

and politicians to mask the true extent of the banks economic problems,4 consistent with the

economic consequences literature (Watts and Zimmerman, 1986; Holthausen and Leftwich,

1983).

The second type of evidence that I present is a bank-level cross-sectional analysis of the

regulatory capital arbitrage hypothesis as applied to deferred tax accounting in Japanese banks.5

I find that in F1998 Japanese banks initially recognized large gross DTAs that were mainly

attributable to their large current and past losses and loan loss provisions. Given the relatively

stringent Japanese tax laws, these DTAs largely reflect the economic situation of these banks

rather than accounting discretion. However, I also find that managers of these banks recognized

relatively small valuation allowances against these DTAs, which meant that the net DTAs

recognized by these banks were also large, increasing their regulatory capital. The banks

recognized these relatively large net DTAs in spite of large and persistent losses, weak financial

positions, and poor future prospects, factors that normally limit the ability of entities to recognize

DTAs. In addition, the future earnings necessary to justify the net DTAs recognized by the

banks in F1998 were substantially larger than the earnings they subsequently realized; further,

these differences are negatively related to the banks regulatory capital positions, indicating that

managers of weaker banks made more aggressive DTA choices (correlation of -.46).

3 The ratio of net DTAs to Tier I capital was 49% for SMFG, one of the Japanese megabanks, at March 31, 2005. Beginning in F2005, the FSA has limited the extent to which the major Japanese banks can include DTAs as part of Tier I regulatory capital (the limit is 40% in F2005, 30% in F2006, and 20% in F2007). 4 For example, see Fukao (1998, 2000), Hoshi (2001), Hoshi and Kashyap (2001). For evidence on regulatory forbearance during this period, and in particular that bank regulators allowed banks to evergreen loans to impaired borrowers (zombies) to avoid the realization of losses on their balance sheets, see Peek and Rosengren (2005). 5 Suda (2002) reports evidence broadly similar to mine, although that studys sample and data is less comprehensive. Gee and Mano (2006) also discuss the role of deferred tax accounting in Japanese banks, focusing on the Japanese mega banks from 2002 to 2004.

4

Cross-sectional regressions estimated for F1999 through F2003 provide further support

for the regulatory capital arbitrage hypothesis. While gross DTAs are, as expected under GAAP,

driven largely by past and current profitability, the banks net DTAs are consistently negatively

related to their adjusted regulatory capital positions, indicating that banks with larger capital

deficiencies recognized larger net DTAs. This result holds after controlling for other factors that

affect the realizability of banks DTAs, such as bank size and profitability.

In the end, it is clear that DTAs provided an important source of regulatory capital for

Japanese banks, and that less well-capitalized banks utilized DTAs to a larger extent than

healthier banks. The results raise the question of why managers of these weak banks, which

were at times reporting large and persistent losses, were not required to report larger valuation

allowances against their banks DTAs, and why banks in Japan, unlike those elsewhere in the

world, were permitted to include these assets in their regulatory capital. Based on the evidence

here, it is hard to escape the conclusion that these decisions occurred in Japan as part of the

Governments regulatory forbearance towards the banks.

The evidence that I report illustrates why changing accounting rules is not sufficient to

change financial reporting practice. While the Japanese adopted a deferred tax accounting rule

that is based on and similar to the IFRS rule (IAS-12) and the U.S. rule (SFAS-109), the

evidence here shows that the effect of the Japanese rule was quite different, with the banks

recognizing large net DTAs in spite of relatively clear evidence of financial difficulties, largely

because regulatory and political factors provided incentives for bank regulators and managers to

implement the rule in such a way as to achieve their own regulatory ends. This result should

give pause to those who believe that the adoption of IFRS around the world is likely to lead to

standardization of accounting practice.

5

The evidence also raises questions about whether Japanese GAAP rules are equivalent

to IFRS, as Japanese regulators have recently suggested in discussions with the European

Commission (EC). This is an important issue, because Japanese companies rely almost

exclusively on European capital markets for external debt financing.6 At the time the EC started

requiring IFRS (2005), it gave Japan until July 2008 to convince it that Japanese GAAP was

equivalent to IFRS, which has stimulated a great deal of standard-setting activity in Japan.7

The evidence that I provide clearly illustrates that, at least with respect to deferred taxes, the

Japanese adoption of rules that on the surface appear similar to those in other regimes does not

necessarily result in the equivalence of accounting practice.

The evidence also shows that accounting rules, when used for regulatory purposes, are

likely to be affected by political and regulatory incentives. Japan is not alone in this regard.

There is evidence that regulators of U.S. Savings and Loans (S&Ls) practiced regulatory

forbearance during the 1980s when many S&Ls were in serious financial difficulties, and used

regulatory accounting principles (RAP) to help achieve this purpose.8 The difference is that in

Japan regulators decisions affected Japanese GAAP, and so overall accounting practice, rather

than regulatory accounting rules in particular.

Section 2 provides further discussion of the regulatory forbearance argument and the role

of deferred tax accounting, as well as aggregate evidence on the role of deferred taxes in

6 See, for example, Gao (2007). 7 For example, the Accounting Standards Board of Japan (ASBJ) and IASB announced an agreement known as the Tokyo Agreement in August 2007 to accelerate the convergence of Japanese GAAP. For a summary of the issues, see CESR (2007). 8 See, for example, Blacconiere (1991) and Blacconiere et al. (1991). The use of DTAs to supplement the regulatory capital of Japanese banks (and the associated regulatory forbearance) is similar to the use of supervisory goodwill by US S&Ls during the early 1980s. See FDIC (1997). Another example is the way that US banks adopted SFAS-106 in the early 1990s, as documented by Ramesh and Revsine (2001). In 1993, the last year SFAS-106 could be adopted, a change in banking regulations forced a closer link between regulatory reporting and GAAP rules. Ramesh and Revsine show that banks regulatory capital levels systematically affected both how they adopted this rule (by either writing off OPEB costs all at once or amortizing the cost), and the timing of their adoption of SFAS-109 which potentially cushioned the blow of SFAS-106.

6

Japanese banks. Section 3 discusses empirical predictions for the bank-level tests. Section 4

reports large sample empirical evidence. Section 5 concludes.

2. Deferred Tax Accounting and Regulatory Forbearance

Regulatory forbearance implies that the Japanese Government, in conjunction with bank

regulators at the MOF and (later) the FSA, chose not to enforce bank capital requirements and

instead implicitly supported insolvent banks until their economic fortunes improved. Whether

this policy decision was optimal from an economic standpoint is beyond the scope of this paper,

and I take the Governments decision to support the banks as given.9 Instead, my interest is in

the role that deferred tax accounting played in the Governments forbearance strategy.

This section describes the background, circumstances, and effects of the adoption of

deferred tax accounting in Japan, in particular by the major Japanese banks. Section 2.1 provides

evidence on the evolution of the regulatory capital of the major Japanese banks from the early

1980s to the late 1990s, and shows that conventional sources of bank regulatory capital had

largely been depleted by 1998. Section 2.2 discusses how the deepening financial crisis forced

the Government and regulators to take a number of measures to boost the regulatory capital of

the banks, and how accounting, particularly deferred tax accounting, played a role in this

process. Section 2.3 discusses how the deferred tax rule was effectively set by the Governments

bank regulators, rather than by independent standard-setters, and how both the timing and

manner of the rules adoption are consistent with its use as a tool of regulatory forbearance.

2.1 Analysis of the capital strength of major Japanese banks, 1982-2003. Figure 1 plots the major Japanese banks aggregate stockholders equity, stock of available

unrealized gains on securities, and the cumulative after-tax amount of realized gains on these

9 See Hoshi and Kashyap (2004) for discussion and more references.

7

securities from F1982 through F2002.10 For regulatory purposes the banks are required to

maintain a minimum level of Tier I (core) capital, which corresponds fairly closely to

stockholders equity.11 The cumulative after-tax amount of realized gains shows the extent to

which the banks rely on the gain realizations as a source of capital. Figure 2 plots the major

components of equity from F1993 through F2003. More detailed numbers are provided in

Appendix 1.

Figure 1 shows that the major banks stockholders equity grew steadily during the

1980s. At the end of F1989 cumulative realized gains on securities represented less than 10% of

total equity. At the same time, the banks accumulated large unrealized gains on their holdings of

investment securities. These gains grew from about 9.6 trillion in F1982 (about 1.7 times

equity) to 55 trillion in F1988 (about 3.6 times equity), largely due to the increase in Japanese

equity prices.

This picture changes dramatically in the early 1990s. From F1990 through F1994 total

stockholders equity for the banks stays virtually constant at 21 trillion, with little change in

either paid-in capital or retained earnings. The banks earnings, however, were increasingly due

to realizations of securities gains. These realizations began in the late 1980s, and offset the

banks declining operating earnings. Prior to F1989, realized gains (losses) on investment

securities were small and largely insignificant. In F1989, however, the banks realized gains of 10 Cumulative realized gains on securities are calculated by tax-effecting the gains (losses) assuming a tax rate of 40% and then cumulating, beginning with fiscal 1982. 11 Under BIS rules there are two types of bank capital, referred to as Tier I and Tier II capital. Tier I capital consists of common stock, preferred stock, capital surplus, minority interest, and retained earnings, minus goodwill, certain other intangibles and (in countries other than Japan) all or most DTAs. Tier II capital includes items such as subordinated debt and convertible bonds as well as loan loss provisions. A risk-based capital ratio is computed by dividing these amounts by risk-weighted assets. Under BIS rules (the Basel Accord) Tier I capital ratio must be at least 4% of risk-weighted assets while the sum of Tier I and Tier II capital must be at least 8% of this same amount, with Tier II capital limited to 100% of Tier I capital. Montgomery (2005, pp. 28-29) argues that the Tier I capital requirement is more stringent for Japanese banks because the components of Tier II capital are more easily managed by bank managers. As discussed later, capital requirements are slightly different (more onerous) for banks with international operations than for banks that do not have international operations. Note also that there are differences in the definition of regulatory capital across the two sets of banks.

8

2.8 trillion, allowing them to report net earnings of 2.2 trillion. This began a period during

which the major banks relied on the realization of gains to offset poor operating performance, as

shown in the following table (amounts in trillions of yen):

F1989 F1990 F1991 F1992 F1993 F1994 F1995 F1996 F1997

Realized gains

2.8

1.8

2.1

0.5

2.2

4.0

3.7

3.4

4.3

Pre-tax earnings, including the gains

3.0

2.3

2.0

1.3

0.8

0.1

(3.1)

(0.1)

(5.2)

Net Income

2.2

1.3

0.9

0.5

0.4

(0.1)

(3.6)

(0.1)

(5.3)

A number of authors interpret the banks gain realizations during this period as income

smoothing (Fukao, 1998; Hoshi and Kashyap, 1999; Shrieves and Dahl, 2003). The practice of

strategically realizing gains and immediately repurchasing the same securities to preserve cross-

shareholdings is known in Japan as fukumi keiei (hidden asset management).12 In the last three

years of this period (F1995-F1997) even these large gain realizations were not sufficient to offset

the banks operating losses.

Cumulative realized gains on securities totaled 8.3 trillion on an after-tax basis by

F1994, and represented about 39% of equity and 72% of retained earnings. Over the next three

years the banks retained earnings fell from 11.6 trillion in F1994 to only 1.9 trillion in F1997

(see Figure 2). By F1997 cumulative realized gains of 15.2 trillion exceeded total shareholders

12 It is unlikely that the banks large scale sales of securities during this period were motivated by liquidity needs. As discussed by a number of authors (e.g., Fukao, 1998; Hoshi and Kashyap, 1999; Shrieves and Dahl, 2003), the securities sold by the banks during this period were largely shares that represented the banks long-standing holdings of the equity of their corporate lending customers, especially group (keiretsu) companies. To maintain these relationships and avoid possible retaliatory sales of the banks own shares, the banks immediately bought back the shares they sold at current market prices. Given the tax consequences of these transactions (the gains were subject to capital gains tax at rates of about 50%), Shrieves and Dahl (2003) argue that these transactions actually resulted in net cash outflows to the banks. In November 2000 the JICPA passed a rule preventing the banks from recording gains on sales of stock that were repurchased shortly thereafter.

9

equity of 13.6 trillion, meaning that the banks would have had negative equity without the gain

realizations. Total shareholders equity in F1997 was below that of F1996.

By F1997 the major Japanese banks had reached a turning point. Consistently poor

operating performance combined with a substantial decline in the amount of available unrealized

capital gains had largely sapped them of their financial strength. Overall financial sector

difficulties were highlighted by the collapse of a major bank (Hokkaido Takushoku Bank, a city

bank) and two large securities firms (Sanyo Securities and Yamaichi Securities) in late 1997.

2.2 The Crisis of 1998 and the introduction of deferred tax accounting

The economic and financial situation continued to worsen during 1998 and in October of

that year the Government passed legislation that made public monies totaling 60 trillion (about

12% of GDP) available for capital injections into weak but solvent banks, to nationalize failing

banks, and to protect bank depositors. Soon after this two more major banks Long Term Credit

Bank of Japan and Nippon Credit Bank failed and were nationalized. This subsection

discusses why the Government decided to inject capital into the banks and how accounting was

used to reduce the political costs of the Governments decision.

Allowing more banks to fail at this time was not an acceptable solution for the Japanese

Government. First, in 1996, because of increasing concerns about the stability of the financial

system, the Government removed the cap on bank deposit insurance, essentially guaranteeing

bank deposits. This made the Government responsible for the financial consequences of bank

failures. Second, because of the central role of the banks in Japans economic system, the failure

of any significant number of banks (especially the main city banks, which sat at the heart of

the Japanese corporate groups) would have had serious consequences for the rest of the

economy, which was already troubled. The failure of these banks would likely have led to the

10

failure of many of their corporate borrowers, with a number of associated adverse economic

consequences.13 Bank failures, and corporate bankruptcies more generally, were almost unheard

of in Japan at this time.14

The decision to inject capital into troubled banks in 1998 was politically difficult for the

Government. In 1995 and 1996 the Government used taxpayer money to bail out many jusen

(financial institutions similar to credit unions), a decision that was unpopular with the public.15

The Governments handling of the financial crisis led to large losses for the ruling Liberal

Democratic Party (LDP) in upper house elections in the summer of 1998, which forced the

incumbent government to resign. This made the Government reluctant to directly inject capital

into failed banks.

Nevertheless, to avoid further bank failures, in late 1998 the Government injected

approximately 7.5 trillion into the major banks. Given this decision, the issue before the

Government was how to portray the capital injection in a way that would be palatable to the

public. This meant characterizing the banks as undercapitalized but solvent, and so required the

banks to meet regulatory capital requirements.16 One way of achieving this would have been for

the Government to lower the required regulatory capital threshold (8% of risk-weighted assets, at

least half of which must be Tier I capital). However, this would have been costly politically

13 Banks had incentives to continue to lend to weak and effectively insolvent corporate borrowers to prevent them from having to recognize the associated loan losses. Peek and Rosengren (2005) provide evidence that: (1) firms in Japan were more likely to receive additional bank credit if they were in poor financial condition, and (2) this practice (ever-greening) was more prevalent among banks with relatively low regulatory capital and more extensive corporate affiliations. Caballero et al. (2006) discuss how the continued extension of credit to otherwise insolvent corporate borrowers (zombies) is part of regulatory forbearance. 14 Prior to mid-1995, no Japanese commercial bank had failed in the postwar period (Peek and Rosengren, 2001). See Hoshi and Kashyap (2001, pp. 111-112) for a discussion of the convoy system under which bank regulators required strong banks to merge with any weak banks in danger of failing. According to Hoshi and Kashyap (2001, Table 6.1), there were three bankruptcies of firms listed on the Tokyo Stock Exchange between 1987 and 1996, compared to 117 on the NYSE over the same period. 15 According to Hoshi and Kashyap (2001, Ch. 8) the use of public funds to solve the jusen crisis in 1995 outraged the public. Also see Fukao (2003). 16 The Rapid Recapitalization Act (under which the capital injections were made) specifically required the banks to be healthy.

11

because the 8% threshold is part of global BIS (Basel) regulatory standards and so would have

been a clear admission both to the Japanese public and to the international financial community

that Japanese banks were too weak to meet international capital standards.17

Given this backdrop, the Government needed a way to convince the public that the major

banks were healthy, in spite of their financial difficulties. Changing accounting rules to achieve

this purpose was a relatively low cost solution because at this time the Government, acting

through the MOF, set both regulatory accounting practice and GAAP rules. Given the relatively

large information and contracting costs in the political and regulatory arenas (Watts and

Zimmerman, 1986; Holthausen and Leftwich, 1983) it is unlikely that the Japanese public had

the incentive to undo the effect of these accounting changes.

The Government, acting through the MOF, made three accounting changes around this

time, all of which had the effect of increasing the banks regulatory capital. First, in early 1998

the MOF changed banking regulations to allow the banks to use either the cost method or the

lower of cost or market (LCM) method to account for investment securities. Previously, the

LCM method had been required. Because, as described above, the banks had largely realized all

of the unrealized gains on their investment portfolios and then immediately repurchased the

same securities (to preserve group cross-holdings), their investment portfolios were largely

marked-to-market. Given the continuing decline in Japanese stock prices during this period, the

LCM rule would have forced the banks to record losses on their portfolios, further reducing Tier

I capital.

17 Such a move would have likely increased the Japan premium a premium Japanese banks with international operations had to pay relative to their competitors in the U.S. and Europe on interbank borrowings. This premium emerged a couple of years earlier when the Japanese banks problems first became apparent to the international banking community, and varied as a function of how effectively the Government was seen to be in dealing with the problems (Peek and Rosengren, 2001).

12

Second, in March 1998, the Government passed an Act modifying the Commercial Code

to allow financial institutions and certain large companies to revalue their holdings of land, a

practice not otherwise permitted. This allowed these entities to record revaluation gains, because

in many cases the land had been held for so long that even their post-crash values were above

book value (Hoshi and Kashyap, 2001, p. 276). The effect of this law was to increase the banks

shareholders equity and increase regulatory capital.18 In addition, by allowing certain

companies to revalue their holdings of land and so report artificially strong balance sheets, the

banks were able to banks renew loans to these firms and avoid recognition of loan losses (Peek

and Rosengren, 2005).19

For the Japanese public, the most visible problem with the banks was their failure to deal

with their bad loans, which had persisted since the early 1990s when the bursting of the bubble in

real estate prices first created problems for borrowers. Consequently, another important political

goal for the Government was to show that the capital injections were being used to help resolve

the banks bad loan problems.

From the regulators standpoint, deferred tax accounting was advantageous in this regard

because it allowed them to make an implicit deal with the banks: in exchange for recognizing

higher loan loss provisions, the regulators agreed to allow the banks to recognize partially

offsetting DTAs.20 As discussed in Section 3.1, given the way Japanese tax law treated loan

losses, this naturally led weaker banks (with larger non-performing loan problems) to recognize

larger DTAs, and so provided a natural accounting hedge for the banks capital deficiencies.

18 Specifically, the banks could include 45% of the revaluation gains in Tier II capital. The land revaluation law did not directly increase Tier I capital of the banks. 19 The companies to which the law applied were typically large corporate borrowers in particular industries, such as real estate development and construction, to which the banks had leant heavily in the 1980s bubble years and which suffered the most when real estate collapsed in the early 1990s. 20 A number of current and former Bank of Japan officials characterize this implicit deal as providing the banks with a soft landing in dealing with their bad loan problems.

13

Importantly, because the deferred tax effects are run through income and then into retained

earnings, DTAs represent Tier I capital, a regulatory treatment unique to Japan.21

The effects of the land revaluations and the adoption of deferred tax accounting are

shown in Figures 2 and 3 and Appendix 1. The land revaluation reserve appears in equity for the

first time in F1998 at an amount of 1.4 trillion, or a little over 10% of total stockholders equity

at the previous year-end. The introduction of deferred tax accounting had an even more

substantial effect on the banks stockholders equity. In total, net DTAs of 8.9 trillion were

recognized by Japanese banks at the F1998 year-end, of which the major banks recognized 6.6

trillion, close to half of total equity at the previous year-end (Figure 3 and Appendix 1).22

Together, these two components accounted for about 35% of bank equity at the F1998

year end, with DTAs playing the major role. In addition, at the same time the Government

injected 7.3 trillion into the major banks, of which 6.1 trillion increased equity (see Figure 2).

Without these two accounting changes and the injection of capital, bank capital at the end of

F1998 would have been only 8.7 trillion, considerably lower than the already low 13.6 trillion

at the previous year-end.

To clearly demonstrate the effect of deferred tax accounting on the banks regulatory

capital positions, I computed Tier I capital before inclusion of net DTAs and the capital

21 BIS guidelines specifically indicate that assets should be measured conservatively and that regulators are expected to adjust GAAP-based financial statements for both intangible assets (including goodwill) and deferred tax assets (BIS, 2000). In the U.S., the Feds risk-based capital guidelines limit the amount of DTAs included in Tier I capital to the lesser of: (i) the amount of DTAs expected to be realized within one year, and (ii) 10 percent of Tier I capital. These guidelines also exclude goodwill and certain other intangible assets from Tier I capital. 22 When the amount of the land revaluation was first included in equity in March 1999, the law required it to be shown net of tax, with the corresponding DTL included in the liabilities section of the balance sheet. The law also specifically indicated that because of the statutory and temporary nature of the land revaluation, this DTL amount is to be recorded separate from the entitys other deferred tax assets and liabilities. Consistent with this, deferred tax footnotes (including the table that details the components of deferred tax assets and liabilities) exclude mention of the land revaluation DTL, with all disclosure related to the land revaluation included in a separate footnote. In addition, Japanese entities do not net the land revaluation DTL against the other DTAs and DTLs. Because of this special nature of this DTL and its separation from all other deferred tax amounts, it is ignored in computing the appropriate level of the valuation allowance for DTAs.

14

injections for the 15 banks that received capital injections from the Government in early 1999.23

For 12 of the 15 banks, Tier I capital would have fallen below the 4% minimum without the

inclusion of net DTAs, and all of the 15 banks would have fallen below the 8% total capital

minimum without the inclusion of net DTAs.24 Thus, without the DTAs all of Japans major

banks would have fallen below minimum regulatory capital requirements, which would have

precluded the capital injections, causing them to fail. This would have led to a crisis of

unprecedented magnitude given the size and significance of these banks.

2.3 Was deferred tax accounting part of the overall reform of accounting standards in Japan?

The way that deferred tax accounting was adopted in Japan supports the argument that its

adoption was part of the Governments regulatory forbearance strategy. First, at this time

Japans accounting rules were effectively set by the Government, a result of Japans traditional

code law system.25 Second, the rule was pushed through quickly (in comparison to the speed

of the normal accounting standard-setting process) and adopted early by virtually all Japanese

banks. Third, as discussed above, bank regulators at the MOF allowed the banks to include

DTAs in Tier I capital, a decision that is contrary to usual BIS practice and that was unique to

Japan.26

23 14 of the 15 banks that received capital injections were major banks. The other bank was Yokohama Bank, the largest regional bank at the time. Bank-level data on regulatory capital and capital injections are from Fukao (2000) and Nakaso (1999). 24 All of these banks were designated as banks with international operations (international banks); as such, the minimum total capital requirement was 8%, consistent with BIS rules internationally. In contrast, the Government imposed a less stringent 4% requirement on Japanese banks with no international operations (domestic banks). 25 Japans system is known as the Triangular Legal System because financial reporting rules are determined by three related sets of laws, the Commercial Code, the Securities and Exchange Law, and the Corporate Income Tax Law (JICPA, 2006, Accounting and Disclosure System in Japan, available at http://www.jicpa.or.jp/n_eng/e-account.html). Japan has been moving away from this traditional code law type of system and in 2001 set up an independently funded private sector standard-setter, the Accounting Standards Board of Japan, modeled after the U.S. FASB. The link between the Japanese tax code and financial reporting rules has become progressively weaker, beginning in the mid-1990s. 26 It is possible that the Japanese banks DTAs had economic value, and so were appropriately included in Tier I capital. However, because they view the value of these assets as uncertain bank regulators around the world almost universally exclude all or most DTAs from Tier I capital, along with goodwill and most other intangibles (see, e.g.,

15

In 1998, Japanese accounting standards were set by the Business Accounting Deliberation

Council (BADC). This body was a committee of the MOF; as such it was likely to be influenced

by the Governments regulatory goals. The timing and manner in which the deferred tax rule

was adopted is consistent with this regulatory influence. In June 1998, the MOF and Ministry of

Justice issued a joint report reinterpreting the Commercial Code to allow deferred tax assets and

liabilities to be recognized on corporate balance sheets. Also in June 1998, the BADC issued its

initial exposure draft on deferred tax accounting which was followed, in October 1998, by a final

statement.27 The statement went in effect in F1999 (the fiscal year ended March 2000) but

allowed earlier adoption. Virtually all Japanese banks adopted the rule early, in F1998, even

though the rule was issued only five months before the fiscal period.28 The fact that deferred tax

accounting was adopted so quickly made it difficult for auditors, with no previous experience of

deferred tax accounting, to assess the reasonableness of the banks deferred tax accounting

choices in that year.

In 1996, the Japanese Government announced a package of financial reforms known

collectively as the Big Bang. This included a mandate to modernize Japanese accounting

standards, and to adopt standards that were accepted world-wide. Thus, an alternative

interpretation is that deferred tax accounting was adopted in F1998 as part of the accounting

Big Bang. While possible, the fact that other accounting reforms were adopted after deferred

tax accounting makes this explanation less likely. Fair value accounting for investment

securities was not adopted until F2001, even though this rule would seem to be especially

BIS 2000). In addition, as I discuss below, it is hard to argue that the Japanese banks DTAs were unusually valuable in fact, my evidence shows the opposite and supports the conclusion that, especially at this time, the Japanese banks DTAs had little economic value. 27 In contrast, the BADC issued an exposure draft on accounting for financial instruments in June 1998, but did not require this new accounting until F2001. 28 According to the Bank of Japan (1999), 134 banks adopted the rule early at the March 1999 year end and only four regional banks did not adopt the rule at this time.

16

important in an economy where banks, other financial institutions, and many non-financial

companies hold large portfolios of investment securities.29

3. Empirical Predictions

Section 2 provides arguments and evidence to support the view that deferred tax

accounting was introduced in Japan as part of the Governments regulatory forbearance strategy.

This section motivates more specific empirical tests of the regulatory forbearance argument and

the related prediction that managers of Japanese banks practiced regulatory capital arbitrage.

The regulatory forbearance argument implies that the inclusion of DTAs in banks

regulatory capital was inappropriate because these assets were of questionable economic value.

It is possible that Japanese banks DTAs did, in fact, have economic value and were

appropriately included as part of regulatory capital. Consequently, the first part of the empirical

analysis assesses the realizability of the banks DTAs. Section 3.1 discusses the Japanese GAAP

rule on deferred taxes and relevant Japanese tax law.

Closely related to regulatory forbearance is the argument that Japanese bank managers

managed their banks capital levels to meet regulatory requirements, a practice known as

regulatory capital arbitrage. Section 3.2 discusses previous evidence on regulatory capital

arbitrage in Japanese banks and develops this prediction in the context of deferred tax

accounting.

3.1 Japanese GAAP rules for deferred tax assets and the relevant tax law

The Japanese accounting rule for deferred taxes (described further in Appendix 2) is

similar to SFAS-109 in the U.S. in requiring recognition of both deferred tax assets and liabilities

along with a valuation allowance that reflects the extent to which any resulting DTAs are not

29 Conversely, given the fact that Japan is traditionally a code law country, with close ties between the tax code and accounting rules, it is not clear why deferred tax accounting would be a pressing concern.

17

realizable. The most subjective part of this rule is the determination of the valuation allowance,

which involves assessing the realizability of the entitys DTAs. SFAS-109 states that a valuation

allowance is to be recognized if, based on the weight of available evidence, it is more likely

than notthat some portion or all of the deferred tax assets will not be realized (FASB, 1992,

para. 17 (e), emph. in original).

The Japanese GAAP rule does not have an analogous clear statement of how the level of

the valuation allowance should be determined. Instead, as discussed in more detail in Appendix

2, in November 1999 the Japanese Institute of CPAs (JICPA) issued guidance regarding the

evidence necessary to justify the realizability of DTAs. This guidance became the de facto

standard. It classifies entities into five categories based on their taxable income and past

earnings stability. Entities with sufficient taxable income and/or stable past profitability (the

first two categories) are permitted to recognize DTAs in full. Because of their persistent losses

during this period few Japanese banks satisfied these conditions and so fell instead into a third

category. For entities in this category DTAs are limited to an amount that can be justified based

on a schedule of the entitys temporary differences, subject to an overall limitation that DTAs not

exceed the tax benefits of taxable income expected over the next five years. This made it crucial

for the banks to project sufficient taxable income over this period.

Japanese tax law, at least as it pertains to banks, is generally more stringent than that in

the U.S. Loan losses are not deductible until the loans are actually disposed of, meaning that the

borrower goes bankrupt (a rare occurrence in Japan) or the loan is sold. This meant that there is

a relatively long lag between when loan losses are recognized for book and tax purposes, which

leads to relatively large DTAs.

18

During the period of this study, Japanese law allowed tax losses to be carried forward for

five years with no carryback. These rules are also more restrictive than those in the U.S. As

with U.S. GAAP, tax loss carryforwards automatically generate DTAs. It then becomes a matter

of managerial judgment as to whether the associated DTAs are realizable and to what extent a

valuation allowance is necessary. Previous research shows that U.S. managers typically set the

valuation allowance to track the extent of loss carryforwards, indicating that the realizability of

DTAs attributable to carryforwards is questionable.30 This tendency should be stronger in Japan

given the more restrictive carryforward rules.

Another significant source of deferred taxes for Japanese banks during this period was

the unrealized gains and losses on investment securities that resulted from the banks adoption of

fair value accounting (required beginning in F2001). Under this accounting rule, the tax effects

of the unrealized gains and losses included in equity result in deferred tax liabilities and assets.

Given the magnitude of Japanese banks stockholdings, these tax effects are relatively large.

This discussion suggests that Japanese banks, especially before the introduction of fair

value accounting, will have relatively large gross DTAs given the likely magnitude of the

temporary differences associated with their loan loss provisions and their tax loss carryforwards.

The recognition of these gross DTAs is largely non-discretionary. The more difficult accounting

question, involving significant managerial discretion, is whether and to what extent these DTAs

are realizable, and so the empirical tests (Section 4) separately analyze the determinants of the

banks gross DTAs, valuation allowances, and net DTAs.

3.2 Bank managers regulatory capital arbitrage incentives

Regulatory capital arbitrage occurs when bank managers exploit the discretion available

under banking regulations to report regulatory capital levels that satisfy the necessary 30 See, e.g., Miller and Skinner (1998) or Schrand and Wong (2003).

19

thresholds.31 Ito and Sasaki (2002) show that in the early 1990s, when BIS rules were adopted in

Japan, bank managers responded by issuing subordinated debt and curtailing new loan activity,

to help their banks meet Tier II capital requirements. This was necessary to offset sharp declines

in the value of the banks portfolios of investment securities (see Section 2.1) that would

otherwise have reduced the banks Tier II capital. Shrieves and Dahl (2003) show that over the

period from 1989 to 1996 managers of Japanese banks timed the realization of securities gains

and managed reported loan loss provisions to smooth reported income and bolster their banks

regulatory capital levels, consistent with the discussion in Section 2.1.

I investigate whether managers of Japanese banks exploit their discretion under deferred

tax accounting rules to practice regulatory capital arbitrage. The specific argument that I test is

that Japanese banks deferred tax accounting choices are driven by bank managers incentives to

increase reported levels of regulatory capital in the face of shortages of other components of that

capital. This leads to the cross-sectional prediction that the relative level of banks net DTAs is

inversely related to regulatory capital levels, other factors held constant.

4. Empirical Evidence

4.1 Sample and preliminary evidence

To generate a sample of banks with available data for F1998 I identify all Japanese banks

on the Compustat Global Financial Services database, sort them by total assets, and choose the

largest 100 banks. This procedure captures all of the major Japanese banks and most of the

31 A number of papers investigate how managers of U.S. banks trade off the incentives provided by regulatory capital regulations, tax incentives, and financial reporting incentives. For example, see Moyer (1990), Scholes, Wolfson, and Wilson (1990), Beatty, Chamberlain, and Magliolo (1995), and Collins, Shackelford, and Wahlen (1995).

20

regional banks and so includes the majority of Japanese banks, and the very large majority on a

value-weighted basis.32

Ninety-two banks have sufficient financial statement detail (hand collected from their

financial statements) to be included in the sample. Three regional banks did not adopt deferred

tax accounting until F1999 and so are removed from this part of the analysis. I omit Bank of

Tokyo-Mitsubishi (BTM) and Mitsubishi Trust & Banking because these banks (part of the same

group) were listed in New York and used U.S. GAAP. I also remove Long Term Credit Bank

because it failed in late 1998, distorting its numbers. These exclusions leave 86 banks: eight city

banks, one long term credit bank, six trust banks, and 71 regional banks.

Table 1, Panel A reports descriptive statistics on the size, profitability, and deferred tax

positions of these banks in F1998. The primary source of financial statement data is the NEEDS

database from Nikkei (Nihon Keizai Shimbun, America, Inc.). These are large banks, with mean

(median) total assets of 7,608 billion (2,761 billion).33 Japanese banks initially recognized

DTAs that are large relative to their equity and assets. Mean (median) net DTAs are 96 billion

(25 billion) compared to mean (median) stockholders equity of 322 billion (114 billion). At

the mean (median), these DTAs represent 34% (21%) of the banks stockholders equity and

1.1% (0.9%) of total assets.

Table 1, Panel A reports three measures of bank profitability: current return-on-assets

(ROA), average ROA over the three prior years, and the number of losses reported by the bank

32 This initial set includes the nine main city banks in existence at that time, three long-term credit banks (IBJ, Long Term Credit/Shinsei, and Nippon Credit/Aozora), the major trust banks, as well as many of the regional banks (including both the tier one and tier two regional banks). According to the Bank of Japan (1999), there were a total of 144 Japanese banks at this time, including 64 regional banks and 61 Tier 2 regional banks. 33 Using an exchange rate of 120/$, these amounts translate into $63 billion ($23 billion). The largest Japanese banks are also large by world standards. The largest five Japanese banks included here (Fuji, Sumitomo, Sakura, Dai-Ichi Kangyo, and Sanwa) had total assets ranging from $408 billion to $483 billion in F1998. The Bank of Tokyo Mitsubishi (which I exclude) had total assets of $618 billion in F1998. In comparison, Citibank had total assets of $669 billion in F1998.

21

over the three prior years.34 These numbers make it clear that Japanese banks initially

recognized relatively large net DTAs in spite of poor current and prior profitability. Mean

(median) ROA in F1998 is -0.78% (-0.38%); 55% of the banks report losses in this year. Mean

(median) average past ROA is -0.02% (0.11%); 46% of the banks report losses on average over

the past three years. The poor current and past profitability of the banks raises questions about

the realizability of their DTAs unless these banks were expecting a significant reversal of

fortune, an unlikely prospect given the state of the Japanese economy at this time. I present

evidence on these banks profitability in the five years after F1998 in Table 3.

Panel B of Table 1 presents details of the banks deferred tax numbers. The banks gross

DTAs are similar in size to the net DTAs reported in Panel A, with a mean (median) of 103

billion (25 billion), because these banks generally do not report significant DTLs or valuation

allowances. Only 30 banks report DTLs, which have a mean (median) value of 0.8 billion (0),

and only 16 banks report valuation allowances, which have a mean (median) value of 5.6

billion (0).

Schrand and Wong (2003) provide evidence on the deferred tax positions of 235 US

banks upon their adoption of SFAS-109 in 1992. DTAs outweigh DTLs for US banks as well.

As in Japan, this is mainly due to differences between the book and tax treatment of loan losses,

although this temporary difference is larger for Japanese banks because of the more stringent

deductibility requirements of Japanese tax law. Schrand and Wong report that while all of the

banks in their sample report DTAs, 98% also report DTLs. Schrand and Wong also find that US

banks recognize DTLs that are typically 60-70% of their DTAs. The differences in the extent of

these banks DTLs presumably result from differences in the banks book-tax situations. The US

34 Throughout the paper, ROA is calculated to exclude the effects of deferred tax accounting by using pretax earnings in the numerator and lagged total assets adjusted to exclude DTAs in the denominator.

22

banks larger levels of DTLs relative to DTAs increase the realizability of DTAs, reducing the

need for a valuation allowance. This makes it harder to understand why the Japanese banks have

little or no valuation allowances.35

As expected, the principal source of the Japanese banks DTAs are temporary differences

due to the banks loan loss provisions, which represent a mean (median) 65% (68%) of DTAs.

Tax loss carryforwards account for a mean (median) 9% (0%) of DTAs; 18 banks report this

component. These numbers are similar to those for US banks. Schrand and Wong (2003) report

that, on average, the temporary difference related to loan loss provisions accounts for about 60%

of these banks DTAs while tax loss carryforwards represent around 5% of DTAs.

Schrand and Wong (2003) report that 39% of the banks in their sample report a valuation

allowance, which averages about 11% of the underlying gross DTAs. Only eight Japanese banks

(20%) report a valuation allowance, which average only 3% of gross DTAs. Unlike differences

between the underlying DTAs and DTLs, these differences are harder to explain as being due to

differences in the tax laws, and support the prediction that managers of Japanese banks were

relatively aggressive in their initial recognition of net DTAs.

Because the major banks differ along several dimensions from regional banks, Table 2

compares gross and net DTA numbers, valuation allowances, bank capital ratios, and

35 The numbers for the Bank of Tokyo-Mitsubishi (BTM), which uses US GAAP, are quite different to those of the other major Japanese banks. In F1998, it reports net DTAs that are generally smaller than those of the other Japanese banks (about .5% of total assets and 14% of equity), has a valuation allowance that represents about 19% of its DTAs and recognizes DTLs that represent 47% of DTAs. The main source of BTMs DTLs is unrealized gains on investment securities, which were not recognized at this time under Japanese GAAP. BTM uses Japanese GAAP numbers for purposes of calculating its regulatory capital ratios. Using these numbers (for which footnote data is not available), the net DTA ratios increase to .8% of assets and 24% of equity, supporting the idea that Japanese GAAP and/or its interpretation/enforcement allows for more liberal recognition of DTAs. An important caveat, however, is that its Japanese deferred tax numbers exclude DTLs related to unrealized gains on investment securities.

23

profitability data for the 15 major banks in the sample to those for the 71 regional banks.36 Data

on bank capital ratios is also from Nikkei. I collect and report data on both Tier I capital ratios

and overall capital ratios (the numerator of the overall capital ratio is the sum of Tier I and Tier

II capital net of certain adjustments).

Unsurprisingly, the regional banks are much smaller than the major banks. Using either

total assets or stockholders equity, the median major bank is about ten times larger than the

median regional bank. There is some evidence that gross DTAs are proportionately larger at the

major banks mean (median) gross DTAs deflated by stockholders equity for the major banks

is 48.9% (31.7%) versus 34.4% (16.7%) for the regional banks, although only the difference in

medians is statistically significant. The larger gross DTAs of the major banks are likely due to

their poorer current and past profitability as shown in the bottom of the table, current and past

ROA numbers, as well as the count of past losses, indicate substantially worse profitability at the

major banks (differences statistically significant at the 1% level or better).

The major banks have proportionately larger loss carryforwards and smaller loan losses

than the regional banks. Because loss carryforwards are usually viewed as less realizable than

other components of banks DTAs, this implies that these banks would record larger valuation

allowances than the regional banks. There is some evidence of this in Table 2; the mean

(median) valuation allowance for the major banks is .098 (.009) versus .008 (0) for the regional

banks, although only the difference in medians is statistically significant. These differences are

small in economic terms, however, which means the major banks report larger net DTAs than the

36 Historically in Japan the various types of banks the city banks, long-term credit banks, trust banks, regional banks, and sogo (second-tier regional banks) perform different economic functions and serve different types of customers, the result of banking regulation (Hoshi and Kashyap, 2001). The city banks lent mainly to large corporate customers and operated large branch networks, usually in urban areas. These banks sat at the center of the main corporate groups (keiretsu) in Japan, often with their affiliated trust banks. In contrast, the regional banks typically operated as full service banks in limited geographical areas, usually one or two prefectures.

24

regional banks. While mean net DTAs are the same when deflated by equity, the difference in

medians is substantial (28.7% versus 16.7%) and statistically significant at the 5% level (two-

tailed). The differences are more evident when I deflate by total assets, with differences in

means and medians both significant at the 1% level. This suggests that managers of the major

banks were more aggressive in their recognition of net DTAs than managers of the regional

banks.

The evidence also shows that the major banks core capital is substantially weaker than

that of the regional banks.37 At the end of F1996 [F1997], the major banks had mean (median)

Tier I capital ratios of 5.3% (4.8%) [5.4% (5.0%)], compared to corresponding ratios of 6.7%

(6.5%) [7.2% (6.8%)] for the regional banks, differences that are statistically significant at better

than the 1% level under two-tailed tests. In contrast, the total capital ratios of the two sets of

banks are approximately the same, with no statistically significant differences. For example, at

the end of F1997, the major banks had mean (median) total capital ratios of 10.4% (9.8%),

compared to 10.0% (9.9%) for the regional banks.38

4.2 Is the optimism inherent in managers deferred tax choices related to their banks regulatory

capital positions?

To this point, the evidence shows that when Japanese banks adopted deferred tax

accounting: (1) they recognized gross DTAs that were large relative to their total assets and

stockholders equity, (2) they recorded little/no valuation allowances, resulting in relatively large

net DTAs, (3) they were largely unprofitable and had been so for some time, (4) the major banks

recognized larger net DTAs than the regional banks, in spite of being less profitable with less

37 Because the data on the regulatory capital levels of most banks is missing from the Nikkei database in F1998, I compare the banks capital levels using the numbers reported at year-end F1997 and F1996. 38 Tier II capital is defined slightly differently for banks with international operations (most of the major banks are in this category) than for other banks (most of the regional banks are in this category). For example, Tier II capital for banks in the latter category excludes unrealized gains on investment securities.

25

Tier I capital. Further, as documented in Section 2.2, 15 major banks received capital injections

from the Government in early 1999. Prior to receiving these capital injections, all of these banks

would have fallen below minimum regulatory capital requirements without the recognition of

DTAs. All of this evidence is consistent with regulatory forbearance. This subsection provides a

further test of this argument by testing: (i) whether managers deferred tax accounting choices

implied expectations about future earnings that were systematically too optimistic, and (ii)

whether any observed over-optimism is negatively related to the banks regulatory capital

positions, as predicted under the regulatory capital arbitrage hypothesis.

Unfortunately, analysts earnings forecast data, which could have served as an unbiased

estimate of managers expectations, are not available for these banks. As an alternative measure

of managers expectations of future earnings, I use the banks realized earnings over the five

years after F1998 and compare this amount to the earnings managers implicitly expected over

this period, as indicated by their banks net DTAs. This allows me to assess the reasonableness

of managers claims as to the realizability of their banks DTAs.

Table 3 reports three variables. First, I compute the minimum taxable income necessary

over the next five years to support the net DTAs recognized by the banks in F1998.39 To obtain

this number, I divide each banks net DTAs by an assumed 50% tax rate, which approximates the

corporate tax rate in Japan at this time. I compare this amount to the (pretax) earnings the banks

actually realized over the next five years, and compute the difference between these amounts.

These three amounts are available for 70 banks (almost all of the major banks merged with one

39 I use a five year period because: (1) this was the tax loss carryforward period in Japan at this time (Section 3.1), (2) JICPA guidelines on assessing the realizability of DTAs also require income to be estimated over this period (Section 3.1), and (3) it seems unlikely that managers could reliably estimate earnings over a period longer than five years.

26

another over the five year period, which reduces the available sample). All numbers are deflated

by total assets in fiscal 1998, adjusted to exclude DTAs.

The results in Table 3 show that the net DTAs recognized by bank managers in F1998

implied that they expected their banks to earn a mean (median) total ROA of 1.89% (1.42%)

over the subsequent five years. While modest in absolute terms, these amounts are large relative

to the banks current and past earnings, summarized in Table 1 (for example, median F1998

ROA was -0.38%). These numbers are also substantially larger than what the banks

subsequently earned. These banks earned a mean (median) total ROA of -0.15% (0.03%) over

the five years after F1998, with half of the banks reporting overall losses. Thus, differences

between the two sets of numbers are large the mean (median) difference is -2.04% (-1.72%) of

total assets, with managers of 60 banks (86%) being overly optimistic. This is strong evidence

that managers estimates of bank earnings were systematically too optimistic. Moreover, the

correlation between these earnings forecast errors and the banks regulatory capital positions

(measured using the Tier I capital ratios in March 1997) is -.46 (significantly different from zero

at the 1% level), implying that the managerial optimism embedded in the DTA numbers is

inversely related to regulatory capital levels.40 Managers of banks with the weakest capital

positions were the most optimistic.

The table also reports the results of this analysis separately for the major and regional

banks, although only six major banks have available data. Managers of the major banks are

especially optimistic: the larger net DTAs recognized by these banks represent average expected

40 The correlation is even stronger if I measure the banks regulatory slack as the difference between their total capital and the relevant regulatory minimum (8% for banks with international operations and 4% for other banks). It is possible that managers expectations at this time were not biased because of regulatory capital incentives but rather were systematically optimistic because of the underlying macroeconomic environment. However, by this time the Japanese economy was clearly mired in a slump, making a rebound unlikely. In addition, macro-driven systematic optimism does not explain the negative cross-sectional correlation between managers over-optimism and regulatory capital slack.

27

future earnings of 3.81% of total assets (median, 3.22%) compared to subsequent realized

performance of -2.04% (median, -1.54%), for an average difference of 5.85% (median 5.33%).

These differences are clearly substantial in economic terms and are statistically significant at the

5% level. Differences for the regional banks are also positive (mean (median) of 1.68%

(1.59%)) and statistically significant at the 1% level, but clearly smaller than those for the major

banks (differences between the two sets of are also significant better than the 1% level).

4.3 Are managers deferred tax choices related to their banks regulatory capital positions?

This subsection provides further evidence on the regulatory capital arbitrage hypothesis

by examining the prediction that the net DTAs recognized by bank managers are negatively

related to their banks regulatory capital slack. I first report results for F1998, the year that

deferred tax accounting is adopted by the banks (see Table 4), followed by results for F1999

through F2003 (Tables 5 and 6).

For F1998, Table 4 reports cross-sectional regressions of three dependent variables the

banks gross DTAs, valuation allowances, and net DTAs on control variables and a proxy for

the managers regulatory capital arbitrage incentives. The DTA variables are deflated by total

assets (adjusted to remove the effects of DTAs) while the valuation allowance is deflated by

gross DTAs. These regressions control for the nature of the banks DTAs (the fraction due to

loan losses and tax loss carryforwards), the banks size (log of total assets), and their current and

past profitability (current and average past pretax ROA). To proxy for managers regulatory

capital incentives, I include the banks March 1997 Tier I capital ratio (labeled BIS397). I use this

proxy because March 1999 (F1998) Tier I capital levels are not available from Nikkei for most

banks (the same is true in March 1998).41 Because of the differences between major and

41 As an alternative measure of regulatory capital slack, I also computed the difference between the banks total regulatory capital ratios and the appropriate regulatory minimum (8% for banks with international operations and

28

regional banks, I also include a regional bank dummy. Data are available for 69 banks. In the

regressions that examine managers deferred tax accounting choices (the valuation allowance

and net DTAs), I also include the banks average realized future ROA, computed over as many

of the next five years as are available, to proxy for managers contemporaneous earnings

expectations which should drive their recognition decisions under GAAP. Because of

heteroskedasticity due to the large differences in bank size, the t-statistics are computed using

White (1980) heteroskedasticity-consistent standard errors.

The first regression in Table 4 shows that the most important explanatory variable for

gross DTAs is past bank profitability this variable is strongly negatively related to gross DTAs

and is largely responsible for the adjusted R-squared of 85%.42 This is expected because banks

with poorer past profitability and larger loan loss provisions naturally record larger gross DTAs.

I also find that larger banks and (after controlling for size) regional banks report lower gross

DTAs, on average. The coefficient on the regulatory capital variable is positive and significant (t

= 3.36) indicating that banks with stronger capital positions tend to report larger gross DTAs.

The next regression assesses the determinants of the banks valuation allowances. Unlike

gross DTAs, managers exercise discretion over the valuation allowance. This regression again

has good fit, with an R-squared of 63%. The results indicate that banks with relatively more

DTAs due to loan losses, higher past (and current) profitability, and higher expected future

profitability have lower valuation allowances. These results are all consistent with more

realizable DTAs leading to lower valuation allowances, as the normal application of GAAP

would imply. I also find, however, that after controlling for these factors there is a positive

4% otherwise). The results are largely consistent with those reported in Table 4 (although in the net DTA regression, the regulatory capital variable is significantly negative, consistent with regulatory capital arbitrage). 42 The correlation between average past ROA and current ROA is .78, making it hard to determine the separate contribution of these variables. Because these variables are so highly collinear, I include only past ROA in the reported regressions. Inferences are very similar when either or both variables are included.

29

relation between regulatory capital slack and the valuation allowance (t = 2.46), implying that

managers of weaker banks set the valuation allowance lower (and so record larger net DTAs),

consistent with the regulatory capital arbitrage argument.

The banks net DTAs reflect the banks (largely non-discretionary) gross DTA positions

and their managers valuation allowance decisions. This is the most critical variable since it

directly measures the contribution of DTAs to core capital. The only significant variables in this

regression are past profitability which, although negative and highly significant (t = -4.37), is not

as important as in the gross DTA regression, and future ROA, which is positive and significant (t

= 2.04), implying that managers report higher net DTAs when they expect their banks to be more

profitable in the future, as expected under GAAP. The coefficient on regulatory capital,

although negative, is not statistically significant, inconsistent with regulatory capital arbitrage.

I next examine the banks deferred tax accounting choices from F1999 through F2003,

the last year for which I have data. Table 5 reports descriptive statistics for each of these years,

while Table 6 reports annual cross-sectional regressions analogous to those reported in Table 4.

In Table 5 I again present the numbers separately for the major and regional banks because of the

significant differences between these banks, differences that become more pronounced beginning

in F2001 when the major banks merged to form five large banking groups (megabanks). For

these years, I report the banks Tier I capital ratios along with an adjusted core capital ratio

which excludes the effects of DTAs. This variable shows what the banks core capital position

would have been without DTAs; I also report (in parentheses) the number of banks that would

have fallen below the regulatory core capital minimum without deferred taxes. In the

30

regressions, this variable serves as a measure of managers regulatory capital incentives because

it directly measures the banks regulatory slack.43

Table 5 shows a number of trends in the banks financial performance and deferred tax

positions over this period along with differences between the major and regional banks. It is

clear that DTAs continue to comprise a significant part of the banks capital in F1999: the major

(regional) banks average Tier I capital ratio is 6.8% (7.4%) with net DTAs and 5.1% (4.7%)

without net DTAs; 15 banks would have fallen below the 4% threshold without DTAs. The

numbers in F2000 are similar to those in F1999. In F2001, the situation worsens appreciably: the

major banks report losses that average about 1.5% of assets in this year and the banks core

capital position (before considering DTAs) worsens noticeably due to the losses and (for the

major banks) the introduction of fair value accounting. All four of major banks and 17 regional

banks would have had capital deficiencies in F2001 without DTAs. Consistent with regulatory

capital arbitrage, the banks net DTA positions increase significantly in this year, especially for

the major banks. For the major banks net DTAs represent a mean (median) 59% (60%) of equity

and represent around half of Tier I capital in this year.

The major banks continue to report losses in F2002, eroding Tier I capital even further; at

the same time, these banks net DTAs increase to a mean (median) 78% (79%) of shareholders

equity and contribute to well over half of their Tier I capital. While some of this year-to-year 43 I have also calculated regulatory slack relative to the banks required total capital ratios. To do this, I take the difference between the banks total capital levels and the appropriate regulatory threshold (8% for international banks and 4% for domestic banks). The results in Table 6 are generally weaker when I use this alternative measure of slack. This is expected because net DTAs comprise Tier I capital and because the Tier I capital requirement is typically more difficult to achieve (Montgomery, 2005). Consistent with this, when I compare the domestic and international banks, I find no statistically significant differences in their Tier I capital ratios. However, I find that banks with international operations have significantly higher (at better than the 1% level) total capital levels than domestic banks (expected since Tier II capital is defined more broadly for these banks), but that differences in the two sets of banks total regulatory slack (defined as their total capital levels net of the regulatory minima) are significant in the opposite direction. That is, banks with international operations have significantly lower levels of total regulatory slack than the domestic banks domestic banks have relatively high levels of total regulatory slack (median 5.5%) compared to banks with international operations (median 2.9%) and their own levels Tier I capital slack (median 2.8%).

31

variation is due to (largely non-discretionary) variation in the banks gross DTAs, the net DTAs

reflect bank managers judgments about realizability in spite of their worsening financial

condition, the major banks report larger net DTAs, consistent with regulatory capital

arbitrage/forbearance. Notice that the extent to which the major banks DTAs are due to loss

carryforwards increases and the proportion due to loan losses falls, which also suggests that their

realizability declines. Finally, in F2003 the banks profitability and capital positions improve,

and both the major and regional banks reduce their net DTAs. Thus, these numbers show that

changes in the banks net DTAs tend to be inversely related to their financial performance and

the realizability of the DTAs, a correlation that is inconsistent with what we would expect under

GAAP but consistent with regulatory capital arbitrage.44

Table 6 reports annual cross-sectional regressions for F1999 through F2003. I use the

same dependent variables as in Table 4. To measure managers regulatory capital incentives, I

use the banks Tier I capital ratios adjusted to remove the effects of DTAs, which directly

measures regulatory slack. To proxy for expectations about future earnings, I again include a

future realized earnings measure although as the years progress from F1999 to F2003 there are

fewer available years of future ROA.45 For the years F2001-F2003, after fair value accounting

for investment securities was required in Japan, the regressions also include a variable that

captures the extent to which the banks DTAs (DTLs) are attributable to unrealized losses (gains)

on investment securities. Other variables are defined identically to those in Table 4. The gross

DTA regressions are reported in Panel A, the valuation allowance regressions in Panel B, and the

net DTA regressions in Panel C. I discuss the results year by year.

44 Another possibility consistent with GAAP is that improvements in profitability led to utilization of tax loss carryforwards and so to declines in DTAs. 45 For the F1999 regressions, future ROA is averaged over as many years as each bank has available from F2000-F2003 (maximum of four), for the F2000 regressions, future ROA is averaged over as many years as are available from F2001-F2003 (maximum of three), etc., with no future ROA data available for the F2003 regressions.

32

Similar to Table 4, the F1999 and F2000 gross DTA regressions indicate that past

profitability (again highly correlated with current profitability, which I exclude from the

regressions) is the most important determinant of the banks gross DTA positions, as expected

under GAAP. The adjusted capital variable is not significant in these regressions. This variable

is, however, negative and statistically significant at the 1% level in the net DTA regressions in

both years (t-statistics of -2.6 and -3.8) suggesting that weaker banks report higher net DTAs,

other factors held constant, consistent with the regulatory capital arbitrage argument. The past

profitability variable is also significant in the net DTA regressions, although with smaller

coefficients than in the gross DTA regressions.

The evidence in favor of regulatory capital arbitrage becomes stronger in F2001 and

F2002. Once again, in the gross DTA regressions the coefficient on past profitability is negative

and highly significant while that on adjusted capital is also negative but only significant in

F2001. The adjusted capital variable is not significant in the valuation allowance regressions. In

fact, the only variable that is significant in these regressions is past profitability, which is

negative, suggesting that less profitable banks have higher valuation allowances, consistent with

GAAP. The adjusted regulatory capital variable is negative and significant in the net DTA

regressions, again consistent with the regulatory capital arbitrage argument (in these two years

the t-statistics on the regulatory capital arbitrage variables are -4.6 and -3.2). The unrealized

gain/loss variable is highly significant in the net DTA regressions in both F2001 and F2002,

suggesting that the tax effects of these items have an impact on the banks net DTA positions. In

F2002 (but not F2001) net DTAs are significantly larger for major banks than for the regional

banks (t = -3.4), other things held constant.

33

In F2003, average past profitability has a less important effect on deferred taxes, with t-

statistics of only -1.8 in the gross DTA regression, -2.0 in the valuation allowance regression,

and -0.2 in the net DTA regression. More importantly, however, the coefficient on the adjusted

capital variable is again significantly negative in the net DTA regression (t = -2.8), consistent

with regulatory capital arbitrage. Similar to F2002, the regional bank dummy is significantly

negative in the net DTA regression, implying that the major banks report larger net DTAs, other

factors held constant.

To summarize, these results show that average past profitability is consistently

significantly negatively related to the banks gross DTAs, as expected under GAAP. This

variable is also negatively related to the banks valuation allowances, suggesting that less

profitable banks report higher valuation allowances, also consistent with the normal application

of GAAP. However, I also find that the adjusted capital variable is consistently and significantly

negatively related to the banks net DTAs, consistent with regulatory capital arbitrage. There is

also evidence in F2002 and F2003, after the emergence of the megabanks, that these banks

report higher net DTAs than other banks. Neither of these results holds as strongly in the gross

DTA regressions, consistent with them reflecting managerial decisions under regulatory capital

arbitrage.

5. Conclusion

I argue that deferred tax accounting was introduced in Japan in 1998 at least partly as a

tool of regulatory forbearance, to help the banks meet regulatory capital requirements. This

interpretation is supported by the following: (1) the Japanese Government (including bank

regulators) could not allow any significant number of banks, especially the major banks, to fail,

largely because of the central role these banks played in the economy, (2) during the period from

34

F1989 to F1997 the banks realized large gains on their holdings of investment securities, but by

1998 this source of regulatory slack had largely been used up, (3) deferred tax accounting was

adopted by an accounting rule-making body which was part of the Ministry of Finance, the bank

regulator at the time, (4) the Government and its regulators introduced a number of other

accounting and regulatory changes at this time that enabled the banks to increase their regulatory

capital, including a decision to allow the banks to include DTAs in regulatory capital, a policy

unique to Japan and inconsistent with how BIS rules are implemented elsewhere, (5) because of

the relatively stringent Japanese tax laws, deferred tax accounting enabled the weakest banks

(with the largest loan losses and poorest profitability) to recognize the largest DTAs, and so

provided an effective way of increasing regulatory capital for the weakest banks, and (6)

deferred tax accounting was introduced contemporaneously with a large capital injection from

the Government into the major banks; without the increase in regulatory capital due to DTAs,

these banks would have been insolvent, complicating the politics of the Governments decision

to inject capital.

Consistent with this view, I show that the weakest banks did in fact recognize the largest

net DTAs, that many of these banks recognized these assets in spite of large current and past

losses, that the expected future earnings necessary to justify the banks DTAs largely failed to

materialize, and that this over-optimism was inversely related to the strength of the banks

regulatory capital positions. These results are confirmed in cross-sectional regressions, which

analyze the determinants of the banks gross DTAs, valuation allowances, and net DTAs. While

the banks gross DTAs and valuation allowances are both well-explained by current and past

profitability, after controlling for other determinants their net DTAs are inversely related to

35

adjusted regulatory capital, consistent with regulatory capital arbitrage on the part of bank

managers.

Overall, the evidence supports the idea that deferred tax accounting was introduced by

Japanese regulators to help encourage weaker banks to face up to their bad loan problems by

offering a soft landing. The evidence raises the question of whether, given the prolonged

financial crisis in Japan, politicians and regulators were justified in using DTAs as a way of

managing t