EVIDENCE ON THE VALUE-RELEVANCE
OF DEFERRED TAX ASSETS AND
DEFERRED TAX LIABILITIES
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA
I dedicate this dissertation to my loving mother Rachel
and to the memory of my father Rosaire
I would like to express my deepest gratitude to my chair Dr. Gary McGill for his guidance and constant support throughout the dissertation. His valuable advice and encouragement were really helpful.
I am grateful to Dr. Rashad Abdel-khalik, Dr. Bipin Ajinkya, and Dr. Ronald Ward for their advice and guidance. Special appreciation goes to Dr. Anwer Ahmed for his inestimable comments. He provided prompt feedback though he was on leave from the University of Florida.
I gratefully acknowledge financial assistance from the Universit6 du Quebec A Montr6al and the Society of Management Accountants of Canada. I also want to thank my colleagues and fellow Ph.D. students particularly Sanjeev Bhojraj, Antonello Callimaci, and Kevan Jensen for their support and help. Special thanks go to Colette Desrochers and Suzanne G6rin for their friendship and encouragement.
Finally, I would like to express my profound appreciation to my spouse and precious friend, Nadi, who provided me love and continuous support. He was always willing to help and give constant encouragement.
TABLE OF CONTENTS
ACKN OW LED GEM EN TS ............................................................................................ iii
LIST OF TABLES ......................................................................................................... vi
LIST OF FIGURES ....................................................................................................... vii
ABSTRA CT ................................................................................................................. viii
I INTRODU CTION AND M OTIVATION ..................................................................... 1
2 ACCOU NTING FOR INCOM E TAXES ...................................................................... 6
3 THEORY AND HYPOTHESES DEVELOPMENT .......................... ........................ I I
A ssets D efinition and D eferred Tax A ssets ................................................................. I I
Liabilities D efinition and D eferred Tax Liabilities ...................................................... 13
Realization of D eferred Taxes .................................................................................... 14
Value-Relevance of Deferred Tax Assets and Deferred Tax Liabilities ....................... 15
Analysis by Industry ................................................................................................... 17
Balance Sheet Deferred Tax Assets and Deferred Tax Liabilities Disclosure .............. 20
Am ounts Recorded ................................................................................................. 20
Econom ic Events Underlying D eferred Taxes .................................................... ... 21
Incom e Tax Footnote Disclosure ................................................................................ 22
Sources of D eferred Tax A ssets .............................................................................. 23
Sources of D eferred Tax Liabilities ........................................................................ 24
Expected Realization Period of D eferred Taxes ...................................................... 24
D isaggregation of PRO ............... ........................................................................ 27
4 RESEARCH DESIGN AND SAW LE SELECTION ................................................ 39
Valuation M odel ......................................................................................................... 41
Hypotheses Testing .................................................................................................... 44
Am ounts Recorded ................................................................................................. 44
Econom ic Events Underlying D eferred Taxes ........................................................ 46
Expected Realization Period of DEP and PRO ....................................................... 47
D isaggregation of PRO .......................................................................................... 48
G ro w th ....................................................................................................................... 5 0
P ro x ie s ................................................................................................................... 5 0
Partitioning of the Sample According to Sales Growth ........................................... 51
Sensitivity Analysis ................................................................................. .................. 52
First Difference M odel ........................................................................................... 52
Year-by-Year Regression ....................................................................................... 53
Partitioning of the Sample According to Growth in Common Equity ...................... 54
EBO Valuation M odel ............................................................................................ 54
Sample Selection and D escriptive Data ....................................................................... 57
Sample Selection .................................................................................................... 57
Sample D escription ................................................................................................ 58
5 EM PIRICAL RESULTS ................................................ ............................................ 67
Descriptive Statistics .................................................................................................. 67
Econom etric Issues ..................................................................................................... 69
Base M odel Estim ated as a Tw o-W ay Fixed-Effects ................................................... 72
Partitioning of the Sample According to Sales Growth ............................................... 78
Sensitivity Analysis .................................................................................................... 81
First-Difference M odel Analysis ............................................................................ 81
Year-by-Year Regression Analysis ......................................................................... 82
EBO Valuation M odel Analysis ............................................................................. 83
Partitioning of the Sample According to Growth in Common Equity ...................... 83
6 SUM M ARY AND CON CLU SION .......................................................................... 110
PEFEREN CES ............................................................................................................ 114
APPEND IX : VARIABLE DEFINITION S .................................................................. 118
BIOGRAPHICAL SKETCH ....................................................................................... 120
LIST OF TABLES
I Summary of predictions for the base model ............................................... ... 60
2 Sales grow th distribution ............................................................................... 61
3 Common equity growth distribution .............................................................. 62
4 SIC code distribution .................................................................................... 63
5 Sam ple selection ........................................................................................... 64
6 Sum m ary statistics ........................................................................................ 65
7 Price-to-book ratio distribution .......................................................... .......... 66
8 Descriptive statistics for selected variables .................................................... 85
9 Pearson correlation coefficients .................................................................... 86
10 Overview of the results ............................................................................... 88
11 Regression results of the two-way fixed-effects model ................................ 90
12 Regression results without controlling for fixed effects ............................... 92
13 Regression results partitioning according to GROW SA ............................ 94
14 Regression results for the first difference model ................................. ........ 98
15 Regression results by year using OLS ....................................................... 100
16 Regression results of two-way fixed-effects EBO model ........................... 104
17 Regression results partitioning according to GROWCE .......................... 106
LIST OF FIGURES
I Classification of deferred taxes ................................................................ 36
2 Sources of deferred tax assets ................................................................... 37
3 Sources of deferred tax liabilities ............................................................. 38
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy
EVIDENCE ON THE VALUE-RELEVANCE OF DEFERRED TAX ASSETS AND DEFERRED TAX LIABILITIES
Chairperson: Dr. Gary McGill
Major Department: Accounting
This paper assesses the value-relevance of the balance sheet classification of current and noncurrent deferred taxes and the disclosure requirements of deferred taxes in the footnote as required under SFAS 109. Reported amounts of deferred tax assets (liabilities) can affect financial statement interpretation and analysis. Investors and creditors can ignore deferred taxes, include them in equity, or consider them as real assets or liabilities. Ratios such as debt to equity, return on assets, and current ratio can be greatly affected.
In this study, I analyze the economic event underlying deferred tax assets (liabilities) in order to obtain information about their realization. I also examine the value-relevance of the income tax disclosure for three major industries: drug, automotive, and computer. Finally, I attempt to isolate the effect of growth on deferred taxes.
I use a model that relates price-to-book ratio to earnings and deferred taxes to perform tests on a sample of companies listed on NYSE and AMEX for the years 1992 to viii
1996. Because the sample is panel data, the model is better estimated by the two-way fixed-effects approach. The findings suggest that, depending on the industry, deferred tax disclosures under SFAS 109 are value-relevant to the market.
The current and noncurrent classification of deferred taxes generally provides valuable information. Also the disaggregation of deferred taxes related to provisions and contra assets into three levels of realization--short-tenn, mid-term, and long-term-appears to be appropriate because it better isolates which components of deferred taxes are considered by the market to be realizable.
Contrary to expectations, controlling for growth generally does not alter the findings. Three explanations exist for these results. First, the proxies for growth used may not be appropriate for the firms in my sample. Second, the market may simply not assess the value-relevance of deferred taxes in light of growth. Finally, the partitioning into high, medium and low growth might not be fine enough to find clear evidence of the effect of growth on the realization of deferred taxes. The study's results are robust since the findings are confirmed by different regression approaches.
INTRODUCTION AND MOTIVATION
Accounting for income taxes has been the subject of many debates over the past thirty years. The application of APB 11, which was effective up to December 1992, resulted in meaningless balance sheet amounts. Several companies reported increasing amounts of deferred tax credits, the impact of which was difficult to assess given the limited disclosure requirements. Many investors disregarded the deferred tax elements in both the balance sheet and the income statement because they could not quantify their effect on future cash flows.
In an attempt to provide more meaningful information, the FASB issued SFAS 109 in 1992. This statement aims to recognize taxes payable or refundable for the current year and future tax consequences of items treated differently in the financial statements and in the tax returns.
Four features of SFAS 109 are noteworthy. First, it permits the recognition of deferred tax assets arising from net operating losses and tax credit carryforwards. Second, it requires that a valuation allowance be estimated to establish an asset amount related to future tax consequences likely to be realized. Third, it specifies that enacted tax rates be used to record all deferred tax assets (liabilities). Finally, it requires the disclosure of temporary differences that comprise significant portions of deferred tax assets (liabilities).
Of primary interest in my thesis is whether certain disclosures related to deferred taxes under SFAS 109 provide value-relevant information to the market. For example, should investors, creditors, and financial analysts consider certain types of deferred tax assets, such as those arising from provision for warranty or from restructuring charges, in their analysis of financial statements? How should they compute current and return on assets ratios? Could users regard the net amount of deferred tax assets or a portion of it as a form of collateral for a loan? Similarly, should investors, bankers, and financial analysts compute the debt-equity ratio using the net amount of deferred tax liabilities? Or, should they ignore this amount as they previously did?
These questions have gained importance because SFAS 109 puts the emphasis on the balance sheet and provides more useful information about cash flows than APB 11. Such information could be obtained from the balance sheet and in the tax footnote. One could argue that numbers disclosed in the balance sheet give only a vague indication as to when firms expect to realize the assets and settle the liabilities. Firms are only required to disclose the net amount of current deferred tax assets (liabilities) and the net amount of noncurrent deferred tax assets (liabilities). On the other hand, the income tax footnote provides additional information to equity investors and creditors since the types of temporary differences that represent significant portions of deferred tax assets (liabilities) are indicated. Consequently, investors and creditors can approximate the timing of reversal of the temporary differences and determine the impact on firms' future cash flows. This extensive disclosure allows the construction of more powerful tests of valuerelevance of deferred income taxes in equity valuation than was previously possible (Chaney and Jeter, 1994; Givoly and Hayn, 1992).
The hypotheses tested in this study aim to validate the value-relevance of two particular pieces of information required by SFAS 109: the balance sheet classification of deferred taxes and the income tax footnote. I investigate whether the market bases its evaluation of security price on the total amount of current and noncurrent deferred tax assets (liabilities) provided in the balance sheet and/or on its assessment of the valuerelevance of the main deferred tax components provided in the tax footnote.
I use a model derived from Ohlson's framework to perform this study. The model relates price-to-book ratio to earnings and deferred taxes.
One motivation to explore this question is that major changes took place with the issuance of SFAS 109. The FASB argues that SFAS 109 corrects a situation where, under APB 11, companies recorded increasing amounts of deferred tax credits. Many users believed that these numbers consisted only of bookkeeping entries with no value. This resulted in a situation where some users of financial statements (financial analysts, bankers etc.) added deferred tax credits to equity and the provision for deferred taxes back to earnings (SFAS 109, par. 72; White et al., 1997). This situation also was documented in prior research (Chaney and Jeter, 1989; Wise, 1986) who reported that all deferred taxes that are shown in the liability section of the balance sheet are unlikely to be paid. It is therefore important to examine whether the reporting requirements under SFAS 109 have provided, as claimed by the FASB, information that is useful in understanding the general effect of income taxes on a particular enterprise (SFAS 109, par. 54).
An additional motivation to undertake this analysis is the fact that amounts of deferred tax assets (liabilities) reported in the balance sheet appear to be material and
could affect financial statements' interpretation and analysis. A preliminary analysis of financial statements reveals that, contrary to popular belief, depreciation is not the only major component of deferred taxes on the balance sheet. Deferred tax components such as pension plan benefits, provision for warranty, and inventory reserves result in substantial amounts of deferred taxes for certain industries. Furthermore, the significance of the depreciation component varies between industries and firms. Timing differences due to depreciation do not always result in deferred tax liabilities. Depending on the industry, it is shown that implementation of the asset-liability approach under SFAS 109 has resulted in recognition of deferred tax assets and that many firms report that a high proportion of these assets are realizable within a relatively short period of time.
Furthermore, this research is motivated by the recent moves of regulators. For example, the International Accounting Standards Committee has issued the International Accounting Standard 12 in October 1996, which prohibits the deferred method and requires the asset-liability method for fiscal years beginning on and after January 1, 1998. Likewise, the Canadian Institute of Chartered Accountants (CICA) requires the assetliability approach in Canada (Handbook, section 3465) which will become effective for fiscal periods beginning on or after January 1, 2000. Canada is the only major jurisdiction in the world to still require the deferred method to account for income taxes. The CICA' s motivation for this change is that the existing recommendations (deferred method) produce measurements of deferred income tax balances which are difficult to interpret for both preparers and users.
Finally, the evidence found in this paper can be useful to standard setters in selecting the appropriate level of disclosure. If it is found that the information is not value-relevant
for investors in setting prices, one could question whether the detailed disclosure requirements under SFAS 109 are worthy, given the costs to produce this information.
My paper extends prior research in the area of deferred taxes (Ayers, 1998; Amir et al., 1997a; Chaney and Jeter, 1994; Givoly and Hayn, 1992). In addition to deferred tax position, I analyze the market valuation of the current/noncurrent balance sheet classification.
My study also examines the value-relevance of the deferred tax components disclosed in the tax footnote. I analyze the economic event underlying these components. Though Amir et al. (1997a) attempt to perform this analysis, they simply rank the components one in relation to the other.
Furthermore, my research differs from these papers in that I attempt to isolate the effect of growth on deferred taxes. Finally, the analysis is performed for three major industries for years 1992 to 1996 inclusively. It emphasizes the need for financial statement users to analyze the income tax disclosure in light of the firm and its context to avoid inappropriate generalizations. Though Amir et al. (1997a) control for industry, they do not consider industry economics in performing their analysis.
The remainder of this paper is structured as follows. Chapter 2 summarizes the accounting for income taxes standards. Chapter 3 presents the theory and hypotheses. Chapter 4 explains the research design and the sampling procedures. Chapter 5 discusses the results. Finally, chapter 6 presents the conclusion, limitations, and considerations for future research.
ACCOUNTING FOR INCOME TAXES
Standard setters have prescribed two methods for accounting for income taxes: the deferred method and the asset-liability method. Both methods consider that the tax expense does not simply represent the amount of taxes payable based on the income of the current year. The methods also require differences between accounting income and taxable income be identified and classified into two categories: permanent differences and timing differences. The former includes items that are not taxable or deductible in any period; the latter represents items that are taxable or deductible, but in periods different from those in which they are recognized for accounting purposes. These timing differences originate in the first of these periods and reverse in subsequent periods.
The computation of timing differences at any point in time poses no particular difficulties. The controversial issue facing standard setters is the selection of a method that will best report these differences in the financial statements given cost-benefit constraints. The method selected has to provide relevant information to the market. This information should possess the basic qualitative characteristics identified in SFAC 2; that is, it should be pertinent and reliable. These considerations have led standard setters to select different methods for the accounting for income taxes over the years. First, they prescribed the deferred method, which was introduced by APB I I and was in effect for more than 20 years from 1967 to 1987. Then, they prescribed the asset-liability method,
which has been in effect since 1989 under SFAS 96 and for years beginning after December 15, 1992, under SFAS 109.
This chapter reviews the different methods for accounting for income taxes. I examine the measurement and reporting requirements of the deferred method, under APB 11, and the asset-liability method, under SFAS 109. SFAS 96 is not studied in this paper since its adoption was optional.
The Deferred Method Under APB I I
The focus of APB I I was on the income statement. The objective was to match tax expense with related revenues and expenses for the year in which they are recognized in pretax financial income. Permanent and timing differences are computed and the resulting income tax expense for the period includes a tax payable or receivable as well as a deferral component. The latter reflects the tax effects of items that will reverse in future periods. Hence, deferred credits or charges represent the tax effects of transactions that reduce or increase taxes currently payable. Deferred taxes are determined on the basis of the tax rates in effect at the time the timing differences originate and are not adjusted for subsequent changes in tax rates or to reflect the imposition of new taxes.
Public companies had to disclose the composition of income tax expense as follows: taxes estimated to be payable for the period, tax effects of timing differences, and tax effects of operating losses. It was necessary to allocate these amounts to income before extraordinary items and to extraordinary items. Companies had to explain the significant variations in the customary relationship between income tax expense and pretax
accounting income, if these differences were not otherwise apparent from the financial statements or from the nature of the entity's business. APB 11 recommended that the nature of significant differences between pretax accounting income and taxable income be disclosed. Public companies had to classify the deferred taxes account into a net current amount and a net noncurrent amount. The current and noncurrent portion of deferred charge and/or credit are determined by the classification of the related asset and liability that gives rise to the temporary difference. Finally, deferred tax debits and deferred tax credits on the balance sheet could be netted if both are classified as current or as noncurrent.
The Asset-Liability Method Under SFAS 109
SFAS 109 aims to recognize in the financial statements taxes payable or refundable for the current year and deferred taxes for the future tax consequences of events that have been recognized in the financial statements or tax returns. The focus is on the balance sheet and deferred tax assets (liabilities) are measured with reference to temporary differences. These differences arise not only from timing of recognition of revenues and expenses for accounting and tax purposes but also from differences in the tax bases of assets or liabilities and their reported amounts in financial statements. The measurement of the deferred tax liabilities for taxable temporary differences and deferred tax assets for deductible temporary differences and operating loss carryforwards is done using the enacted tax rates. Subsequent to their original recognition, deferred tax assets (liabilities)
are adjusted to reflect the effects of changes in tax bases or rates. The effect of the change is part of the tax expense of the period.
SFAS 109 requires public companies to disclose the income tax expense allocated between continuing operations, discontinued operations, and extraordinary items. They also have to report the significant components of income tax expense allocated to continuing operations. A reconciliation using percentages or dollar amounts of the reported amount of income tax expense that would result from applying the domestic federal statutory tax rate to pretax income from continuing operations has to be disclosed along with the nature of each significant reconciling item. Moreover, public companies have to report the gross amounts of deferred tax assets (liabilities) and the approximate tax effect of each type of temporary difference that give rise to a significant portion of deferred tax assets (liabilities). Since SFAS 109 requires that a valuation allowance be established, public companies are required to report the total valuation allowance for deferred tax assets, and the net change in the valuation allowance for the year.
Comparison of the Deferred Method and the Asset-Liability Metho
One of the main differences between the deferred method and the asset-liability method is that the former takes into account timing differences while the latter considers temporary differences which include timing and differences in the tax bases of assets and liabilities and their reported amounts in the financial statements. Hence, some permanent differences under APB I I could give rise to a deferred tax asset or a deferred tax liability under SFAS 109. For example there may be differences in business combinations
between the assigned values and the tax bases of assets and liabilities recognized under the purchase method. Under SFAS 109, these differences constitute temporary differences and result in deferred tax assets (liabilities) whereas under APB 11 they constitute permanent differences with no deferred tax consequences.
Further, the deferred method requires no adjustment to previously recorded deferred tax balances to reflect changes in enacted rates, while the asset-liability approach requires such adjustments. As a result, the asset-liability method reflects the tax effects of liquidating all assets and liabilities on the balance sheet at their carrying amounts, whereas the deferred method provides only for timing differences which have passed through the income statement.
According to APB 11, deferred tax debits are included with other assets. Under SFAS 109, deferred tax assets are supposed to represent an economic benefit. Deferred tax assets are recognized for all deductible temporary differences and are reduced by a valuation allowance if necessary.
APB I I recommended the disclosure of the significant differences between pretax accounting income and taxable income whereas SFAS 109 requires the reporting of the significant components of deferred tax assets (liabilities). Therefore, under SFAS 109, the cumulative effect of these differences is disclosed in addition to the gross amount of deferred tax assets (liabilities).
THEORY AND HYPOTHESES DEVELOPMENT
This paper aims to determine whether the income tax disclosure required by SFAS 109 is useful to investors in equity valuation. Specifically, I test whether the market considers the balance sheet classification and the different types of temporary differences reported in the income tax footnote in equity valuation. The objective of this section is to provide a framework for the analysis of deferred tax assets (liabilities) and to develop hypotheses concerning the market use of the income tax disclosure.
Assets Definition and Deferred Tax Assets
According to the FASB, all elements that are shown on the balance sheet should meet certain criteria. As elements of the balance sheet, deferred tax assets (liabilities) have to conform to the definition of an asset (liability) and be relevant and measurable with sufficient reliability. Recognition is also subject to cost-benefit constraints and a materiality threshold. 1
' These requirements have been set out in SFAC 5 par. 63, which identifies four flindamental recognition criteria subject to cost-benefit constraints and a materiality threshold. First the item meets the definition of an element of financial statements, Second, the item has relevant attributes measurable with sufficient reliability. Third, the item is relevant i.e., makes a difference in user decisions. Fourth, the item is reliable, which is a fiction of representational faithfulness, verifiability, and neutrality.
Assets are probable future economic benefits obtained or controlled by a particular entity as a result of past transactions or events. They include items that eventually result
in net cash inflows to the enterprise.
Deferred tax assets, as computed under SFAS 109, seem to conform to that definition. Deferred tax assets arise from past events or transactions. They represent economic benefits that are expected to be realized through a reduction of future taxable income. Reported amounts are ca Iculated using a method compatible with any other monetary assets. Enacted tax rates expected to apply in the future are used and a valuation allowance is calculated in order to record an amount that is more likely than not to be realized.
Deferred tax assets are recognized for all deductible temporary differences and tax loss carryforwards (figure 1). Even though the realization of these two types of events or transactions has a similar impact on future taxable income, their characteristics differ.
Deferred tax assets resulting from tax loss carryforwards are only dependent on the realization of future taxable income. These amounts represent future economic benefits whether a firm continues to grow or not. Also, these deferred tax assets are not related to any specific item on the balance sheet. Amir and Sougiannis (1997) have examined the market valuation of deferred tax assets arising from tax loss carryforwards. Their results suggest that the market considers these assets in its valuation of firms' securities.
On the other hand, the realization of deferred tax assets resulting from temporary differences depends not only on future taxable income but also on the reversal of temporary differences. Such a reversal might not occur in the foreseeable future because
2 SFAC 6 par. 25 and 28.
of growth. Furthermore, these deferred tax assets relate to particular assets or liabilities reported on the balance sheet. This study addresses the value-relevance of these deferred tax elements.
Liabilities Definition and Deferred Tax Liabilities
Liabilities are defined as probable future sacrifices of economic benefits arising from present obligations of a particular entity to transfer assets or provide services to other entities in the future as a result of past transactions or events. 3,4 Liabilities have three essential characteristics. First, they entail probable future transfer or use of assets at specified or determinable dates, on occurrence of a specified event or on demand. Second, they consist of duty or responsibility that obligates the entity, leaving it little or no discretion to avoid the future sacrifice, Third, they reflect a transaction or other event that has occurred.'
Under SFAS 109, deferred tax liabilities are recognized for all deductible temporary differences (figure 1). They are calculated using enacted tax rates expected to apply to taxable income in future years. According to this view, deferred tax liabilities are an obligation to the government expected to be paid. These liabilities result from past
3 Deferred tax payments could be considered a contingency element as defined by SFAC 5, paragraph 1. They consist of an existing condition, situation or set of circumstances involving uncertainty as to possible gain or loss to an enterprise that will ultimately be resolved when one or more future events occur or fail to occur. Hence deferred tax liabilities will have to be paid if fixed assets are sold at their book value.
4 SFAC 6 par. 35,
5 SFAC 6 par. 36.
transactions and events that were recorded in the financial statements, Companies have little or no discretion to transfer assets to the tax authorities.
As in the case of deferred tax assets, deferred tax liabilities will be realized only if temporary differences reverse. Such reversal might not occur in the foreseeable future because of growth. Deferred tax liabilities also relate to specific assets or liabilities on the balance sheet. This study also addresses the value-relevance of these deferred tax elements.
Realization of Deferred Taxes
SFAS 109 requires deferred tax assets (liabilities) to be recorded at their nominal value ignoring the timing of realization (settlement) of these assets (liabilities).
The measurement of deferred tax assets (liabilities) at historical cost reflects neither the timing of cash flows they generate nor the change in market interest rate or firm's cost of capital. It produces a misleading view of the entity's financial position. Two solutions are available to correct the situation.
The first one is to use a discounted cash flow valuation method. Though this method is useful in assessing the true economic value of the deferred tax assets (liabilities), it implies a prediction as to when the reversal of each difference is likely to occur and the selection of a discount rate. This involves a high degree of subjectivity and might prove an impossible task.
The second approach is to present information that could be useful in assessing amounts and timing of future cash flows. The FASB selected this method in SFAS 109.
With regard to timing of the reversal, the FASB has recommended that deferred tax assets (liabilities) be classified as current and noncurrent according to underlying assets and liabilities. This classification is recommended to avoid the problems encountered with discounting. Furthermore, a greater amount of information is provided in the footnote where deferred tax asset (liability) components are disclosed.
Value-Relevance of Deferred Tax Assets and Deferred Tax Liabilities
Investors, financial analysts, and creditors need to assess the economic value of deferred taxes reported on the balance sheet. Depending on their assessment and perception, they will reach different conclusions regarding the financial position of a given firm.
If users consider deferred tax assets (liabilities) presented in the financial statements as representing future economic benefits (sacrifices), they will make no adjustments when computing different financial ratios. This will be consistent with Amir et al. (1997b) who have developed a theoretical framework based on Feltham and Ohlson (1995) to demonstrate that deferred taxes represent real assets (liabilities).
However, if users perceive deferred tax assets (liabilities) not to represent probable future benefits (sacrifices), they can make two different adjustments. First, they can deduct (add) deferred tax asset (liability) amounts reported in the balance sheet from (to) equity. By doing so, users consider deferred tax assets (liabilities) as unrecognized accumulated expenses (revenues). The rationale is that prior tax expenses were understated (overstated) by the recording of deferred tax assets (liabilities). For example,
when they calculate the debt-equity ratio, deferred tax assets will be deducted from equity while total debt will remain unchanged resulting in a lower debt-equity ratio. When they compute this ratio, deferred tax liabilities will increase equity and decrease total liabilities by an equivalent amount. This will further reduce the debt-equity ratio than in the case of deferred tax assets. With that regard, Comiskey and Mulford (1994) suggest that lenders should seriously consider deducting deferred tax assets in whole or in part in the calculation of tangible net worth.
Second, users can simply ignore deferred tax assets (liabilities). In this situation, deferred tax assets (liabilities) are deducted from total assets (liabilities) without adjusting equity. In the case of deferred tax liabilities, the debt-equity ratio is reduced. However, in the case of deferred tax assets, the debt-equity ratio will not be affected. Similarly, the assessment of deferred taxes may affect the computation of other financial ratios such as the current ratio and the ratio of return on assets.
Furthermore, investors, financial analysts, and creditors may use deferred tax information in a different way. They may examine the income tax footnote to gain insight about other economic events that have affected a firm. Defer-red tax position will not be examined as such but rather its origin. Comiskey and Mulford (1992) indicate that income tax disclosures are a rich source of information that can help loan officers improve both their diagnoses of current cash flows from operations and their forecasts of future cash flows. Hirst and Sevcik (1996) argue that income tax disclosures can be a gold mine for analysts, providing intelligence about future cash flows, the quality of earnings, and a variety of accruals not disclosed elsewhere in the financial statements. Similarly, Weber and Wheeler (1992) argue that since income taxes adhere to the
transaction to which they relate it is possible to extract from the tax footnote additional information.
In summary, depending on its assessment of deferred taxes, the market could reach different conclusions regarding a firm's financial position. These conclusions are based on the market assessment of firms' future growth. Indeed, high growth firms are not likely to realize their deferred tax assets (liabilities) in the foreseeable future since these amounts tend to increase indefinitely. Another important element of the analysis is firm and industry characteristics. This element is discussed below.
Analysis by Indust
The objective of analyzing deferred taxes by industry is to obtain a group that is fundamentally homogeneous. It allows the grouping of firms that have comparable activities and face similar rates of technological change and innovation intensity. The underlying assumption is that companies belonging to a given industry are evaluated similarly by the market.
Performing the tests based on industry offers many other advantages. First, the future realization of deferred tax assets (liabilities) arising from temporary differences depends on their underlying assets. These assets (liabilities) originate from different economic events and transactions. For example, deferred tax assets might arise from the depreciation of fixed assets, restructuring charges, or postemployment benefits. Second, the industry classification permits control for the maturity of the industry. Not all industries are growing. Some have a high level of growth while others are more stable.
Even within a specific industry, firms do not grow at the same rate. Therefore, performing the test by industry allows better isolation of the effect of growth. Finally, literature on industrial organization (McGahan and Porter, 1997; Powell, 1996) emphasizes the importance of controlling for industry.
For the purpose of this research I select three industries that have a major impact on the US economy: drug, automotive and computer. The drug industry is the most research oriented of the US economy. US drug manufacturers produce one-third of all drugs sold worldwide and US medical device manufacturers supply about half of the global market (S&P, 1997). The automotive industry is one of the largest industries in the US. Automakers account directly and indirectly for about 15% of all US jobs and manufacturers of civil and defense aerospace products are one of the largest exporters in the US (S&P, 1997). The US computer industry has a notable worldwide influence since both hardware and software originated in this country. For example, US PC shipment accounts for 38.4% of total worldwide shipments (S&P, 1997).
These industries have different characteristics, which suggest that there might be differences in the types of deferred tax assets (liabilities) and their underlying components. Valuation of deferred tax balances in light of how the industry operates may provide useful information.
Firms belonging to the drug industry invest large amounts in both tangible and intangible assets. R&D innovations drive the business (S&P, 1997). They have a collection of blockbuster drugs that enjoy patent protections. Hence, the drug industry is dominated by a relatively small number of large multinationals that have above average earnings growth (S&P, 1997). In the 90s, many large drug companies have downsized
and restructured resulting in a pretax write-off in excess of 5 billion dollars (Weekly Corporate Growth Report, 1996). In addition to R&D, selling and administrative costs constitute a major portion of expenses (Sosnoff, 1998; S&P, 1997). Further, there are an increasing number of small biotechnology firms that are growing. For these firms, sales and earnings represent only a small fraction of growth potential (Hulbert, 1997; S&P, 1997).
The performance of the automotive industry is traditionally associated with the general economy. Most of the industry costs are fixed. Firms are faced with extreme cost pressures and are constantly bargaining with suppliers, labor unions, trade officials and government regulators (Taylor, 1998; S&P, 1997). Even though the automotive industry comprises a variety of firms, they all require substantial investments in fixed assets to manufacture a growing number of product lines (S&P, 1997). Starting in 1992, some sectors of the automotive industry, automakers and auto parts were recovering while others, aerospace and defense, did not turnaround before 1996 (S&P, 1997).
In the computer industry, the technology and product specifications are widely disseminated and are being continuously upgraded. Products have a short life. Need for capital resources is overcome by focusing on special products and components. In the PC market, the competition is intense and profit margins are declining (S&P, 1997; Stewart, 1996). Since 1991, the computer industry achieved sustainable growth, which is partly due to the favorable general economic conditions (S&P, 1997).
Balance Sheet Deferred Tax Assets and Deferred Tax Liabilities Disclosure
To assess future cash flows related to deferred tax assets (liabilities), investors need to determine the reasonableness of the amount recorded. They can also evaluate the underlying economic transactions when examining the balance sheet disclosure.
Whenever there is a doubt about the asset realization, SFAS 109 requires the recording of a valuation allowance. Companies can recognize deferred tax assets in full if they are likely to generate future taxable income. In attempting to justify the full recognition of deferred tax assets, management is required to consider all sources of evidence such as existing temporary differences, the firm's financial history, its future prospects, and tax planning strategies.
Moreland (1996), Miller and Skinner (1998), and Visvanathan (1997) address this measurement issue. Moreland (1996) and Miller and Skinner (1998) investigate whether management estimates of valuation allowance are modeled in accordance with the provisions of SFAS 109. Their results suggest that the amount of valuation allowance is determined with consideration of the requirements of SFAS 109. Visvanathan (1997) examines whether the amount of valuation allowance is informative about firms' failure profitability. His results indicate that the current amount of valuation allowance is associated with next year profitability. These papers differ from my study because their focus is on the information content of the amount of valuation allowance instead of the
value-relevance of the information related to deferred tax assets (liabilities) in pricing securities.
On the other hand, deferred tax liabilities are recognized fully. Once recognized, deferred tax liabilities remain at their nominal value, adjusted for tax rate changes, until the reversal occurs. Firms do not have to assess the likelihood of settlement of those liabilities.
The first hypothesis aims to determine whether the market uses the aggregate net amount of deferred tax assets (liabilities) in firms' valuation (figure 1). Industry classification should affect this valuation. Deferred taxes have long been considered to relate to fixed assets. Therefore, this first hypothesis will apply particularly to capital intensive industries such as drug and automotive (S&P, 1997). The hypothesis, in its alternative form, is
Hi: In pricing securities, the market values firms' net deferred tax position (DTP) as
a real asset (liability).
Economic Events Underlying Deferred Taxes
SFAS 109 requires public firms to present deferred tax assets (liabilities) in the balance sheet as current or noncurrent elements according to the classification of the related asset or liability that gives rise to the temporary difference. Many companies group current deferred tax assets with prepaid taxes and other prepaid expenses whereas they report noncurrent deferred taxes separately in the balance sheet.
Traditionally a current asset should be realized in the short term and this could trigger the realization of the related deferred tax assets (liabilities). Similarly a
noncurrent asset should not be realized before a certain number of years. Consequently, the realization of the related deferred tax assets (liabilities) should not occur in the short term. As a result, the classification between current and noncurrent provides an estimate as to when taxes will be received or paid and allows the market to discount the amounts accordingly. Moreover, the current and noncurrent disclosures should provide relevant information to the market since these amounts are used separately in the computation of financial ratios. For example, while current deferred taxes are included in current ratio calculations, noncurrent deferred taxes are not. Finally, financial ratios are computed uniformly regardless of the industry classification.
The second hypothesis (figure 1), in its alternative form is
H2: In pricing securities, the market values firms' net current deferred taxes (CDT)
and net noncurrent deferred taxes (NDT) as real assets (liabilities).
Income Tax Footnote Disclosure
Although the current and noncurrent classification is simple, it might not provide the information about the underlying economic events. For example, a noncurrent amount of deferred tax assets originating from depreciation will be grouped with another amount related to warranty provision. Hence, in addition to balance sheet classification, the market might also analyze the main components of deferred tax assets (liabilities) provided in the footnote taking into account the industry in which the firm operates and its future growth.
Sources of Deferred Tax Assets
Deferred tax assets relate to provisions and contra assets, unearned revenues, and to fixed assets (figure 2). Provisions and contra assets result in temporary differences when an expense can be deducted, later, at the time of the disbursement while it is accrued in the financial statements. The temporary difference results in the recognition of deferred tax assets. A typical example arises from the difference between disbursements and the accrual of a warranty provision. This results in a timing difference because although the financial statements include a warranty expense in the determination of income, the tax computation allows a deduction for actual disbursements only.
Unearned revenues can result in deferred tax assets when revenues are included in taxable income earlier than they are recognized for book purposes. Extended warranty payments received fall into this category. In this case, deferred tax assets represent an economic benefit to the company because it will operate in the subsequent year without incurring this particular obligation. According to my preliminary observations, this type of deferred tax assets does not occur frequently (uncommon).
Fixed assets can result in deferred tax assets when depreciation expense is larger than the amounts than can be deducted for tax purpose. Here, deferred tax assets are dependent on predetermined rates enacted by tax laws and arbitrary accounting rules.
Sources of Deferred Tax Liabilities
Deferred tax liabilities relate mostly to fixed assets and to a lesser degree to prepayments allowed for tax purposes and to assets related to revenue accruals that are not taxable (figure 3). With respect to fixed assets, deferred tax liabilities are recorded because the tax depreciation rate is greater than the rate used for accounting income determination. Deferred tax liabilities, related to fixed assets, are the result of enacted tax rules and arbitrary accounting rules.
The prepayments allowed for tax purposes give rise to deferred tax liabilities because an expense is deductible before it is included in the calculation of accounting income. This type of deferred tax liability does not occur frequently (uncommon). An example of such liability is the over funding of a pension plan.
The recognition of revenues in the account gives rise to deferred tax liabilities because revenues are recognized in the accounts before they are fully taxed. One such example is revenue recognition for long-term contracts that are recorded under the percentage of completion method but are not taxed until the project is completed or when the payment is received. In practice, there seems to be few deferred tax liabilities of this category (uncommon).
Elected Realization Period of Deferred Taxes
From the previous discussion, one could identify two main situations where deferred tax assets (liabilities) occur in practice: depreciation methods and accrual of expenses. The analysis of deferred tax assets (liabilities) in the light of the underlying situation
could therefore provide useful information. The next two hypotheses aim to investigate this issue,
Generally, fixed assets represent non-monetary assets. Indeed, accounting principles require that fixed assets be carried at acquisition cost less accumulated depreciation. Hence, amounts presented in the financial statements do not represent future cash flows related to the asset's use or sale. While there is no provision for recognizing increases in value, there is a requirement that carrying amounts be reduced when assets are impaired, i.e. when there is no expectation that reported amounts will be recovered from future operations. In the case of impairment, the amount reported cannot exceed future cash inflows.
The allocation of costs over the useful life of the assets results from the application of some arbitrary rules. However, the total cost of a fixed asset is deductible for both accounting and tax purposes. A larger deduction for tax purpose in earlier years will result in a lesser amount being allowed as a deduction in the future years (figure 3). Hence, even though deferred tax assets (liabilities) resulting from fixed assets are different in nature than other deferred tax assets (liabilities) they should be viewed as real assets (liabilities).
Fixed assets include a variety of elements that are not specifically segregated in the balance sheet. No information is provided with respect to different tax categories and their related book values. Useful lives of fixed assets vary widely according to the industry and accounting policies. Traditionally, fixed asset depreciation resulted in deferred tax liabilities. However, in the recent years, some firms in the computer industry have amortized certain assets at a rate that is higher for accounting than for tax
purposes resulting in the recognition of deferred tax assets instead of deferred tax liabilities. Firms in this industry need to replace frequently their equipment to face stiff competition and the demand for their products, resulting in high amounts of depreciation expense. For instance, Hewlett Packard reported deferred tax assets due to depreciation of $142 million ($110 million) for 1996 (1995).
Deferred tax liabilities arise when the depreciation rate for tax purpose provides an economic incentive for investments by raising the internal rate of return on capital investment projects. Generally, highly capital intensive industries such as automotive report deferred tax liabilities due to differences in depreciation of their fixed assets. For instance, GM reported a deferred tax liability due to depreciation of $4.6 ($4.8) billion in 1996 (199S).
Therefore, an analysis by industry seems to be appropriate. Deferred taxes arising from depreciation should be value relevant for capital intensive industries such as automotive and drug since they result in deferred tax liabilities. This is documented by prior research (Givoly and Hayn, 1992; Chaney and Jeter, 1994) that has reported that deferred tax liabilities arising from depreciation represent real liabilities. As for the computer industry, deferred taxes arising from depreciation might be perceived differently since firms in this industry are less capital intensive and often record deferred tax assets related to depreciation.
The third hypothesis aims to verify whether the income tax footnote disclosure related to fixed assets is useful to investors in equity valuation. The hypothesis in its alternative form is
H3: In pricing securities, the market values firms' deferred tax component arising
from depreciation (DEP) as a real asset (liability).
Provisions and contra assets consist of monetary items that affect directly future cash flows. Deferred tax assets related to these items should represent real assets since the future cash flows with respect to provisions and contra assets are tax deductible and will result in future income tax savings (figure 2).
The amount of provisions and contra assets depends on industry classification. For example, the automotive and computer industries record large amount of warranty, inventory, and doubtful account provisions. Also, the drug and the defense sector of the automotive industries reported, in the 90s, large amounts of provisions related to restructuring charges. In many circumstances, deferred tax assets related to these items exceed deferred tax liabilities resulting from depreciation. In these cases, deferred tax assets should be as value-relevant as deferred tax liabilities. Hence, deferred tax assets related to provisions and contra assets should be particularly relevant for the automotive industry since firms in this industry have reported important amounts of provisions for both warranty and restructuring charges.
The following hypothesis aims to verify whether the income tax footnote disclosure related to accounting provisions and contra assets accounts is useful to investors in equity valuation. The hypothesis in its alternative form is
H4: In pricing securities, the market values firms' deferred tax components arising
from accounting provisions/contra assets (PRO) accounts as real assets. Disaggregation of PRO
The significant deferred tax components, disclosed in the income tax footnote, could provide value-relevant information to investors in their assessment of the timing of
realization of deferred tax assets (liabilities). Temporary differences related to accounting provision or contra assets might reverse within a relatively short period of time or in many years. Therefore, any attempt to properly estimate firm value requires an assessment of the timing of realization of the different components of deferred tax assets (liabilities).
In my sample, I have identified more than fifteen different tax components other than depreciation. They include post-retirement benefits obligations, pension obligations, post-employment benefits, deferred compensation, reserve for inventory, provision for doubtful accounts, provision for warranty, capitalization of intangibles, deferred income, intercompany profits, net operating loss carryforwards, tax credit carryforwards, alternative minimum tax carryforwards, and restructuring charges.
I have grouped all deferred tax components into three levels of realization (figure 2): short-term, mid-term, and long-term. This grouping provides a finer categorization than the current/noncurrent classification required in SFAS 109. It is finer because the grouping is based on the economic event that will trigger the realization of the temporary differences whereas the current/noncurrent classification is solely based on the balance sheet classification of the related asset or liability. Feltham (1972) argues that all decision-makers prefer one information system to another if the former is sufficient for the latter in term of fineness. Also, he argues that a signal is relevant if, when received, it changes the decision either through payoff expectations or expectations about future information. Therefore, the tax footnote disclosure is relevant if it affects security price.
Moreover, this classification allows me to investigate the market valuation of many significant deferred tax components having similar characteristics. It differs from Amir
et al. (1997a) who test the market valuation of seven deferred tax components by ranking them in relation to each other. 6 1 discuss below the proposed classification and explain the composition of each category.
The short-term group includes deferred tax components that are expected to be realized within one year. This time frame is selected because the discounted value within that time frame almost equals the nominal value. Furthermore, practice uses one year as a basis for financial reporting, asset and liability classification, and the going concern
assumption. Timing differences arising from operating activities such as provision for doubtful accounts, inventory reserves, and provision for warranty are related to firms' daily activities.
Provision for doubtful accounts. An account written-off for accounting is deemed realized for tax purpose and the written-off amount becomes tax deductible. In my paper, I assume that the realization of deferred tax assets arising from provision for doubtful accounts occurs in the short-term because a firm has to write-off an account that is not recovered within a year of its recognition.' The deferred tax components related to doubtful accounts are not expected to vary between the industries selected since all firms in these industries realized their revenues through credit sales.
6 The seven components include deferred taxes arising from depreciation and amortization, losses, credits and alternative minimum tax carryforwards, restructuring charges, environmental charges, employee benefits including deferred compensation, pensions, post-retirement benefits. In their tests, they also include the amount of valuation allowance and in one category all other components. SAS 59 requires that auditors evaluate the going concern assumption in relation to a period of one year from the date of the financial statements.
" This is assuming a reasonable period of time for receivable turnover.
Inventor y reserves. In general, a firm records an inventory reserve when goods become obsolete. This reserve is tax deductible for all taxpayers other than those who determine cost using the LIFO method.9 When LIFO is used, the obsolete goods are deemed realized at the time of sale or the destruction. Therefore, timing differences arising from inventory reserve only occur for the LIFO adopters. The realization of deferred tax assets should occur in the short-term because obsolete inventory is either liquidated or destroyed soon after the balance sheet date. The most important inventory reserves for the industries selected are recorded by the computer industry (S&P, 1997).
Provision for warra The nature of product sold determines the terms of the warranty contracts. More expensive products have a longer warranty. For instance, products such as automobiles have a three to five year warranty contract whereas computers only have a year or two warranty coverage. For tax purposes, the economic event that triggers the asset realization is the real expenditure of the accrued amount or part of it.
The realization of this deferred tax component occurs progressively during the course of the warranty contract. Deferred tax assets are realized after a year for one-year coverage warranty contracts or could extend over a longer period for extensive period coverage. For example, in the computer industry, the warranty coverage is typically for a year as opposed to about three years in the automotive industry. Hence, deferred tax assets related to warranty could be generally classified as short-term for computer industry and mid-term for the automotive industry.
9 Reg. Secs. 1.471-2(b) and (c).
The category mid-term includes deferred tax components that are expected to be realized in more than a year but not exceeding five years. This time frame is selected because current standards often require the disclosure in the notes of cash inflows or outflows related to contractual obligations, long-term debts, and long-term investments and loans. 10 Furthermore most users of forecasts are aware of the inability to predict the future over periods longer than five years."
The data analyzed suggests that some timing differences such as restructuring charges, post-employment benefits, and environmental charges are expected to reverse on average within five years after they originate. 12 Generally, these components are less recurrent than the ones included in the short-term category.
Restructuring charges. Restructuring charges are costs incurred or accrued at the time of reorganization of a firm's operations. They include plant closures and related costs, operating losses during phase-out periods, and severance costs if not included in post-employment benefits obligations. For financial reporting, cost of closing a division or discontinuing part of a company's operations are assessed and a provision is recorded.
For tax purposes, the economic event that triggers the asset realization is the real expenditure of the accrued amount or part of it. Unlike other timing differences, the reversal does not depend on growth. Indeed, restructuring charges are generally
nonrecurring items that differ from other expenses such as doubtful accounts. For
10 SFAS 13 Accounting for Leases, SFAS 47 Disclosure of Long-Term Obligations, and SFAS 78 Classification of Obligations that are Callable by the Creditor. Statement of Position (SOP) 92.2 limits financial forecasts to five years. IBM reported, in 1996, that the reversal of its restructuring costs are expected within one to five years.
example, while deferred taxes related to provisions for doubtful account may not reverse because of growth in the account receivable balances, deferred taxes resulting from restructuring charges might reverse because of the nonrecurring nature of the underlying items. Both the drug industry and the defense sector of the automotive industry have undertaken major restructuring of their operations in the early 90s (S&P, 1997) resulting in significant amounts of deferred tax assets.
Postemployment benefits. Postemployment benefits refer to benefits provided to employees or their dependents after employment but before retirement.'13 They differ from postretirement and pension benefits. Postemployment benefits are offered to compensate employees or dependents in situations such as disability, layoff, death, etc. They include, but are not limited to, salary continuation, severance benefits, continuation of health care benefits, and life insurance coverage. Firms estimate the future costs of these benefits and then match it with the services rendered by the targeted employees in current and future periods. Benefits may be paid immediately upon cessation of active employment or over a specific period of time.
The economic event that triggers the asset realization is the payment of the benefits to employees that may occur soon in case of disability or later, in case of layoff or death. It is probable that a large proportion of postemployment benefits relate to lay-off costs subsequent to reorganization or a restructuring of company's activities. This is evidenced by the preliminary analysis that suggests that firms choose to report severance benefits with both restructuring charges and postemployment benefits. Other components of postemployment benefits are probably insignificant.
'" SFAS 112.
Since postemployment benefits relate to restructuring costs, the same observation could be made here. Furthermore, since the automotive industry is the most labor intensive (S&P, 1997), firms within this industry should be more affected.
Amir et al. (1997a) view the deferred tax component postemployment benefits as similar to postretirement and pension benefits and rank it as timing difference to reverse far in the future. In this paper, I consider that the deferred taxes arising from postemployment benefits should be realized in the mid-term.
Environmental charges. Environmental charges result from expenses to be incurred generally within the next five years. For accounting purposes, a provision is recognized to match the revenues of the period. For tax purposes, the expense is deductible only at the time of the disbursement. The deferred tax components related to environmental charges are not expected to vary between the industries selected. Long-termn
The long-term category includes items that are expected to be realized in a period longer than five years and, therefore might be viewed as particularly uncertain. From my preliminary analysis, I infer that this category includes timing differences such as pension plan (defined benefit plan) and postretirement obligations.
Pension plan benefits. There are two types of pension plans: deferred contribution plans and defined benefit plans. In the case of defined contribution plans, no timing difference usually arises because payments made to pensions plans are not only recorded as an expense but are also tax deductible. In the case of defined benefit plans, costs are
recognized as services are rendered by employees. 14,1 Decisions to make payments or to fund the pension plan depend on many factors including legal (payments might be required by the union agreement) and tax considerations (payments to the pension plans are tax deductible and earnings of investments held by plan are not taxable). 16 Hence there could be a difference between the amount expensed and the payments made to the pension plans,
The preliminary analysis suggests that some companies in the automotive industry fund the pension plans with the minimum amount required by law while others in the computer industry fund their plans as costs are accrued. In the latter case, no timing difference arises since costs are accrued at the same level as the funding. For companies that fund only the minimum required by law, the realization of deferred taxes arising from pension plan benefits will occur gradually over the period of employment. Therefore, realization should occur sooner than for postretirement benefits (examined next) but is still far in the future.
Postretirement benefits. For financial reporting, costs are recognized as employees render services." For tax purposes, payments are only deductible for actual benefits paid. Moreover, earnings accumulated in the plan are taxable. Hence contrary to pension plans, companies have no tax incentive to create a separate plan for postretirement benefits. Furthermore, there is no legal requirement to create and fund such a plan. Therefore, companies usually adopt a pay-as-you-go policy. This is particularly true for
14 Generally, no timing difference arise from defined contributions plan because the funding is made as costs are accrued at the end of the year.
" SFAS 87.
16 The amount deductible is subject to some limits.
" SFAS 106.
the automotive industry that has the largest unfunded retiree health care cost liability (S&P, 1997).
The realization of deferred taxes related to postretirement benefits is a function of many factors but the most important is probably the average employee's age or the average number of years before retirement. The average age before retirement exceeds five years in most situations. Therefore, on the average, the realization of deferred taxes should occur at a time that is at least five years from today.
The value-relevance of different deferred tax components should vary across industries. Indeed, components of deferred taxes included in the short-term category should be relevant for valuation purpose for the automotive and the computer industries. As discussed previously, firms in these industries record large amounts of warranty and inventory provisions. On the other hand, many firms in the drug industry have had major restructuring in the 90's resulting in large amounts of deferred taxes related to restructuring charges. Finally, components of pension plan and postretirement benefits included in the long-term category should be value-relevant for labor intensive industries such as the automotive industry.
The fifth hypothesis in its alternative form is
H5: In pricing securities, the market values firms' deferred tax assets arising from the
disaggregation of PRO according to their timing of realization, i.e. short-term
(ST), mid-term (MT), or long-term (LT).
Deferred tax position
(H2) N oncurrent Crrent
Figure 1 Classification of deferred taxes
Deferred tax assets
Provisions/UnearnedFie contra assets (H4) revenues ast
Tax deductible Taxable before Amount deducted
after accrued accrued<
for accounting for accounting amount expensed
Short-term (H5) Extended Depreciation
warranty, Iwarranty (H3)
reserves (uncommon) _______Mid-term (H5)1 restructuring,I
Long-term (H5) pension plan, postretirement
Figure 2 Sources of deferred tax assets
Prepayments Accrued revenues_ l ixed assets
I _I I
Tax deductible Taxable before Amount deducted
before accrued accrued >
for accounting for accounting amount expensed
Over funding Long-term Depreciation
of pension plan contracts (H3)
Figure 3 Sources of deferred tax liabilities
RESEARCH DESIGN AND MANTLE SELECTION
The objective of this chapter is to present the research design used to test the hypotheses developed in chapter three. I use a balance sheet as opposed to an income statement approach to develop the model. The income statement approach used in prior research (Beaver and Dukes, 1972; Rayburn, 1986; Daley, 1994) examines the association between security returns and different measures of earnings: net income, net income excluding deferred tax provision, and cash flows. Generally, the results support the contention that, in setting prices, the market uses deferred taxes. This approach does not allow assessing the value-relevance of balance sheet classification of deferred taxes and the income tax footnote.
Researchers also used the balance sheet approach to examine issues related to deferred income taxes. Three related studies deal with information before SFAS 109 was implemented. In the first, Givoly and Hayn (1992) attempt to assess whether the deferred tax liability is viewed as a real liability by the market using the tax rate reduction of the 1986 Tax Act. The argument is the tax rate decrease lowers the deferred tax liability balance and increases equity. They find that investors consider deferred tax liabilities as a real liability and discount it according to the timing and likelihood of settlement.
Chaney and Jeter (1994) examine the association between security returns and the variation of noncurrent deferred taxes. From their conflicting results, they report that the
change in the noncurrent deferred tax account reflects relevant information for security pricing for at least some subsets of firms.
Finally, Espahbodi, Espahbodi and Tehranian (1995) study the market reaction to pronouncements related to SFAS 96 and 109. They investigate whether the market reacts to the announcements that increase the probability of enactment of SFAS 96 or SFAS 109. As Givoly and Hayn (1992), they also test if the decrease in the deferred tax liability account due to the tax reduction of the 1986 Tax Act results in a market reaction. Their findings support both hypotheses.
These studies are particularly relevant in demonstrating that the amounts of deferred taxes reported in the balance sheet provide valuable information to the market. However, these papers do not attempt to assess the value-relevance of the income tax footnote disclosure. Muller (1997), Ayers (1998), and Amir et al. (1997a) address this issue.
Muller (1997) investigates the valuation implications of U.K. GAAP partial allocation relative to comprehensive allocation deferred income tax disclosures.' He examines the value-relevance of off-balance- sheet deferred tax amounts, the partial allocation amounts, and the partial allocation disclosures in comparison with comprehensive allocation disclosures. According to his findings, the market values offbalance-sheet deferred taxes as liabilities and partial allocation amounts as realizable in the foreseeable future. However, partial allocation disclosure does not seem to have more information content than comprehensive allocation disclosures.
Ayers (1998) investigates whether SFAS 109 provides incremental information over APB 11. He reports that deferred tax liabilities, deferred tax assets, valuation allowance
1 Partial allocation implies recognition of deferred taxes expected to be realized within 5 years (UK).
and adjustments of deferred tax amounts under SFAS 109 increase the value-relevance of deferred tax amounts in financial statements.
Amir et al. (1997a) investigate how investors value firms' net deferred tax position and deferred tax components. Their results suggest that the value of net deferred taxes is slightly higher than its theoretical value of one. Amir et al. (1997a) also find that analyzing individual components of deferred taxes instead of the aggregate amount provides value-relevant information.
Amir et al. (1997a) contributes significantly to the research in this area. However, their paper has several limitations. First, the value-relevance of deferred taxes is related to its ability of predicting future cash flows, i.e. the realization of the deferred tax components. Amir et al. (1997a) did not base their prediction on a well-developed theory about future cash flows. They rather rank the deferred tax components in relation to each other. Second they did not control for future growth. Finally, they only control for industry when performing the test on the market valuation of the net deferred tax
I use a model, derived from the theoretical framework of OhIson (1995), that relates stock price to earnings, book value of common equity, and tax disclosures. Consider the
2AMir et aL (1997a) indicate that they did not perform more tests since controlling for industry did not change their results when testing the market's valuation of firms' deferred tax position.
following valuation equation in which price is expressed as a weighted function of book value and earnings:
Pjt = o)oBVSjt + o)IEPSjt +it (1)
Pit = stock price for firm j at time t,
BVSjt = reported book value of common equity per share for firm j at time t, EPSjt= reported earnings per share for firm j at time t, 4jt = unexplained portion of stock price for firm j at time t.
To determine whether a firm's deferred tax position helps explain price-to-book ratio, BVS is decomposed into adjusted book value of common equity per share, defined as book value of common equity before deferred taxes per share (BVBDT), and deferred tax position per share (DT). This results in the following book value identity: BVSjt=BVBDTjt + DTjt.
Substituting the book value identity in the valuation model, dividing both sides of the equation by BVBDT and adding intercept and error terms yield the following regression equation:
Pjt/BVBDTjt = oti + OIDTjt/BVBDTjt + 41EPSjt/ BVBDTjt + Sjt. (2)
One advantage of this model is that it does not require any assumptions about future cash flows, as with a discounted cash flow model, or about the linkage between earnings and cash flows as with a price-earnings multiple valuation model. It only necessitates that the accounting measurements be consistent with the clean surplus relation (CSR). CSR implies that all changes in book value arise from earnings (E) and dividend
payments net of capital contributions (D) as follows: BVt = BVtj + Et Dt. In general, CSR is assumed to hold and to approximate comprehensive income.
Feltham and Ohlson (1995) identified two situations that may affect book value determination. The first is called unbiased accounting where market value equals book value. This means that the book value reflects the "true" value of equity, i.e. there is no unrecorded goodwill. In this situation the ratio of price-to-book is one. The second is called conservative accounting where the market value exceeds the book value. 3 In this case, there is an amount of unrecorded goodwill that is not recognized in the books resulting in a price-to-book ratio greater than one, The second situation is probably the one that occurs the most in practice. Based on that discussion, a third situation could be contemplated where the book value exceeds the market value. This is a situation where there is a negative goodwill resulting in a price-to book ratio less than one. Troubled companies fall in that category.
In theory the coefficient on earnings and on the different deferred tax variables should be positive or negative but equal to one and the intercept equal to zero. However, in reality it is highly likely that the estimate coefficients differ from their theoretical value because of measurement error and omitted variables from the model.
3 Conservative accounting should not be confused with the conservative principle of accounting. For example, even if a company selects an aggressive accounting method, it may still result in conservative accounting, i.e. the book value will still be lower than the market value.
Equation 2 is the base model used to perform all tests. All variables are defined in the appendix. Table 1 presents a summary of all predictions made for the five hypotheses.
In chapter three, it is argued that the market may view the deferred tax position a as real asset (liability). To test hypothesis one (H1), the notation used in equation 2 is modified to result in the following empirical valuation model:
=j3 al + 0 IDTPjt + ,Ejt + 6jt (3)
Pjt3 = stock price for firm j three months after the fiscal year-end deflated by BVBDTjt, DTPjt = deferred tax position per share for firm j at time t deflated by BVBDTjt Ej EPS for firm j at time t deflated by BVBDTjt, and Sit error term.
The dependent variable is the price-to-book ratio computed using share price at the end of the three months following fiscal year-end. This time period is selected because the deferred tax information becomes available at the date of financial statement release, which generally occurs within three months after fiscal year-end. The variable DTP is based on the difference between reported amounts of deferred tax assets and deferred tax liabilities. A deferred tax position that results in a deferred tax asset (liability) is coded as a positive (negative) number.
Intuitively, the coefficient on E should be positive if earnings can explain a proportion of the ratio of price-to-book value. This prediction is valid for all hypotheses tested,
As discussed in chapter 3, with respect to DTP, three situations may arise. The first is that the coefficient on DTP is positive. This infers that the market views DTP as a real asset (liability). The second is that the coefficient on DTP is negative. In this case, investors deduct (add) deferred tax assets (liabilities) from (to) equity. This means that the market considers a deferred tax asset (liability) position to be detrimental (advantageous) to the firm. Deferred tax assets are considered to be interest-free loan to the government that could be viewed by the market to be the result of poor tax planning (poor management decisions). Similarly, a deferred tax liability position could be considered to represent an interest free loan from the government and therefore valued positively. Finally, the coefficient on DTP is zero if investors simply ignore this information.
As discussed in chapter 3, deferred taxes have long been associated with depreciation. Therefore I expect the coefficient on DTP to be positive for the drug and automotive industries since their investments in fixed assets are important. I also expect coefficients on DTP to be positive for the computer industry since firms in this industry are also capital intensive but at a lesser degree.
Economic Events Underlying Deferred Taxes
In chapter 3, it is hypothesized that the market could use the existing current/noncurrent classification to approximate the effect of deferred taxes on future cash flows. To test hypothesis two (H2), book value of common equity is broken down in BVBDT, CDT and NDT. Equation 2 is modified as follows: Pi3= al~ + f32CDTjt + I33NDTjt + 4IEjt + PSjt (4)
CDTjt current deferred taxes per share for firm j at time t deflated by BVBDTjt, NDTjt = noncurrent deferred taxes per share for firm j at time t deflated by BVBDTjt. All other variables are defined as in equation 3. The variable CDT (NDT) results from the difference between current (noncurrent) deferred tax assets and current (noncurrent) deferred tax liabilities. Deferred tax assets (liabilities) are coded as positive (negative) numbers.
The coefficient on CDT (NDT) is positive if the market considers the current (noncurrent) deferred tax amounts as real assets (liabilities). If the coefficients are negative, then the market adjusts equity accordingly. Similar to DTP, a current (noncurrent) deferred tax asset position is considered to be detrimental to the firm while a current (noncurrent) deferred tax liability position advantageous to the firm. Finally, the coefficients on CDT and NDT are zero if the market ignores this information.
As discussed in chapter 3, industry classification should not impact the market use of the current and noncurrent disclosures. Therefore I expect coefficients on CDT and NDT to be positive for all industries.
Expected Realization Period of DEP and PRO
In chapter 3, it is hypothesized that the market analyzes the income tax footnote and estimates the value-relevance of the significant deferred tax components according to their expected realization and their origin, i.e. whether they result from depreciation or from provisions and contra assets. To test hypotheses three and four (H3, H4), DT from equation 2 is broken down into deferred tax components related to provisions and contra assets, and deferred tax components resulting from depreciation.
The model incorporates two other variables. The first one is the amount of valuation allowance. Since Visvanathan (1997) reports that current amount of valuation allowance is associated with next year's profitability, I include this variable in the model. I also incorporate a variable consisting of the difference between firms' deferred tax position and the total of deferred taxes arising from provisions/contra assets and depreciation. Equation 2 is modified as follows:
PA = (XI + 04DEPjt + 05OTHjt + 036VAjt + J37PROjt + iEjt + 8jr (5)
DEPjt = deferred taxes from depreciation per share for firm j at time t deflated by BVBDTt,
PROjt = deferred taxes related to provisions/contra assets per share for firm j at time t deflated by BVBDTjt,
VAj = valuation allowance per share for firm j at time t deflated by BVBDTjt, OTHjt = difference between DTPjt and the sum of DEPjt and PROjt,
All other variables are defined as in equation 3. Deferred tax assets (liabilities) are coded as positive (negative) numbers. The variable VA is also coded as a negative number since it is a contra asset.
Coefficients on DEP and PRO will be positive if the information is value-relevant to the market. They will be negative if the related deferred tax asset (liability) is viewed by the market to be detrimental (advantageous) to the firm. In this case, investors adjust equity accordingly. Finally if the market ignores the amount of net deferred taxes, the coefficients on PRO and DEP will be zero.
Based on the discussion in chapter 3, coefficients on DEP should be positive for the drug and the automotive industries. As for the computer industry, even though I expect the coefficient to be positive, the recording of deferred tax assets may affect the findings. In addition, I also expect the coefficient on PRO to be positive for all industries. The recording of deferred tax assets related to PRO should be valued as a real asset.
The coefficient on VA should be positive for all industries since VA is expected to be informative about firms' future profitability. The variable OTH should be meaningless since it comprises a variety of deferred tax components, some of which could be realized in the next operating cycle and others in many years. Therefore, no prediction is made about the possible sign of its coefficient.
Disagaregation of PRO
To test hypothesis five (1-15), deferred tax components arising from provisions and contra assets are classified into three categories according to their expected realization
period: ST, MT, and LT. The model also includes DEP, VA and OTH. The following model is tested:
Pjt3= (1 + P34DEPjt + 350THjt + 036VAjt + 038STjt + 039MTjt + i310LTjt +41Ejt+ sjt (6) where
STjt = sum of deferred tax components expected to be realized in the short-term, per share for firm j at time t deflated by BVBDTj,
MTjt = sum of deferred tax components expected to be realized in the mid-term, per share for firm j at time t deflated by BVBDTjt
LTjt = sum of deferred tax components expected to be realized in the long-term, per share for firm j at time t deflated by BVBDTjt
OTHjt = difference between DTPjt and the sum of DEPjt, STit, MTjt, and LTjt All other variables are defined as in equation 3. Deferred tax assets (liabilities) are coded as positive (negative) numbers.
The coefficients on ST, MT, and LT are expected to be positive if the market uses the income tax footnote disclosure to gain insight as to the period of realization of deferred tax components arising from provisions and contra assets. If the market ignores the amount of net deferred taxes, the coefficients on ST, MT, and LT should be zero. Finally, the market adjusts equity if it views negatively the amount of net deferred taxes. In this case, the coefficients on ST, MT, and LT are negative.
Based on the discussion in chapter 3, coefficients on ST should be positive, particularly for the automotive and the computer industries. Coefficients on MT should also be positive for all industries that have reported important amounts of restructuring
charges such as the drug and the automotive industries. Finally, coefficients on LT are expected to be positive for labor intensive industries such as the automotive industry.
As discussed in chapter three, the realization of deferred taxes is dependent on firms' growth. Many proxies could be used to approximate firms' growth. One proxy is the percentage increase in fixed assets. This measure may seem appropriate since deferred taxes have long been associated with depreciation and fixed assets. According to this measure, a firm constantly acquiring or renewing its fixed assets will never recover (pay) its deferred tax assets (liabilities). This proxy would only be valid if all depreciable assets were recorded which is not always the case. For example, in the drug industry, assets reported on the balance sheet do not include R&D expenditures that are substantial. Hence using percentage increase in fixed assets as a proxy for firms' growth will be misleading. Likewise, using book value of common equity as proxy for growth will generate similar problems.
Another proxy for growth used in the literature is return on equity (ROE) (Nelson, 1996). ROE is a comprehensive measure of financial performance and growth potential (S&P, 1997). This proxy might not be appropriate for some of the firms within the industries selected. For example, a pharmaceutical company that reduces significantly its R&D expenditures may show large ROE. Indeed ROE is not an indication of growth.
On the contrary, the company has undermined its ability to develop the new products needed to support future growth.
Firms' growth could also be estimated using price-earnings ratios. However, this proxy may be inadequate since troubled companies will show a high price-earnings ratio irrespective of their growth potential.
Sales increase is another proxy for firms' growth. This proxy appears to be adequate for well-establi shed firms. Though sales increase might not always represent future growth it appears to be an adequate proxy. However, care should be displayed when using this proxy for companies at their development-stage.
Partitioning of the Sample According to Sales Growth
I attempt to control for growth by partitioning the sample in accordance with firm's sales growth. Growth sale (GROWSA) is the one-year historical growth in sales. Table 2 presents the sales growth distribution for the full sample and by industry. As can be observed, for the full sample, about 26% (106 out of 412) of the firms has a sales growth of less than or equal to .05 and about 46% (189 out of 412) has a sales growth higher than .15,
Based on this distribution, I partition the sample into three categories: high growth
(HG) firms with a GROWSA higher than .15, medium growth (MG) firms with a GROWSA higher than .05 but equal or less than. 15, and low growth (LG) firms with a GROWSA equal or less than .05. This partitioning allows me to obtain sub samples for each industry that contain enough observations to perform the regression analysis.
As can be observed, firms in the sample tend to have high GROWSA. To the extent that this partitioning is valid, coefficients on the deferred tax variables should not be significant for the HG and MG categories while they should be significant for the LG category.
First Difference Model
I use a first-difference model, as suggested by Kothari and Zimmerman (1995), to examine the value-relevance of the income tax disclosure.4 A priori, this model may not allow me to infer about the market valuation of deferred taxes. Indeed, the change in the deferred tax elements may be marginal compared to change in book value of common equity. The first difference model is derived by first-differencing the undeflated version of equation 2:
APjt = (XI + CiC2ABVBDTjt +fOIADTjt + Ft (7)
where the symbol A indicates the change in the variables from t and t-1. The change in the adjusted book value of common equity is used as a proxy for ABVBDT. Therefore, earnings enter into the model through ABVBDT. Easton and Harris (1991) suggest including both earnings level and earnings change variables when combining book value and earnings valuation models. Therefore, in addition to earnings, the change in earnings
4 One advantage of this model is that it controls for serial correlation that may exist among variables.
is also added to the model (AE). All variables are deflated using lagged share price three months after fiscal-year end, resulting in the following regression equation: Rjt3 = a1 + ot2AFBVjt + c3AEjt + f31AFDTjt + st (8)
Rjt = change in share price for firm j at time t, deflated by lagged share price three month after fiscal year-end (P3t-1),
AFBVt = first difference in the adjusted book value of common equity for firm j at time t deflated by Pj3t-1 (earnings level),
AEjt = variation in EPS from time t and t-1 for firm j deflated by Pj3t-.. AFDTjt =first difference in the deferred tax position of firm j at time t deflated by Pj3t-1.
This specification is used to test Hl. Similar regression equations are developed to test the other hypotheses. To the extent that the model permits inferences about the value-relevance of deferred taxes, the predictions made for the base model are valid.
My sample includes data for firms from 1992, year of implementation of SFAS 109, to 1996 inclusively. In order to gain some insight about the value-relevance of deferred taxes across years, I estimate the regression equation year-by year using the base model.
Partitioning of the Sample According to Growth in Common Equity
In addition to GROWSA partitioning, I attempt to validate the results by partitioning the sample according to growth in common equity (GROWCE). I compute GROWCE as the one-year historical growth in common equity. Table 3 presents the GROWCE distribution for the full sample and by industry. As can be observed, for the full sample, about 21% (87 out of 412) of the firms has a sales growth of less than or equal to .05 and about 44% (180 out of 412) has a sales growth higher than .15. The HG, MG, and LG categories are formed as the partitioning based on GROWSA.
EBO Valuation Model
To verify whether the results are sensitive to the selected valuation model, I use a price-level model similar to the one used by Abarbanell and Bernard (1994) and Frankel and Lee (1996). The authors of both papers develop their valuation model from the Edwards-Bell-Ohison (EBO) theoretical framework.
Abarbanell and Bernard (1994) relate share price to book value of common equity and measures of future abnormal earnings. They report that book value and earnings forecasts can explain up to 78% of share price. Similarly, Frankel and Lee (1996) regress firm's book value, earnings per share, and firm's value against price. They report that the firm value variable can explain more than 70% of the cross-sectional price variation.
5 Firm's book value and one- and two-year-ahead analyst forecasts of earnings per share determine this value (EBO valuation model).
Accordingly, the model used to perform the sensitivity analysis relates price to current book value and future abnormal earnings: pit = By,1 + DFAEjt (9)
pit = share price at fiscal-year end for firm j at time t, BVjt = book value of common equity per share at fiscal-year end for firm j at time t, DFAEjt = discounted value of future abnormal earnings per share for firm j at time t.
DFAEjt equals the discounted value of the summation of the difference between future earnings and a measure of normal earnings as follows: VX,=1 (1+r)- Et [Xjt, (r BVjt.c-)] (10)
=i+, expected future earnings per share for firm j at time t+, r = measure of cost of capital,
B~t,,= expected book value of common equity per share for firm j at time t+tU- 1.
DFAE is a measure of unrecorded goodwill. It represents the expected future priceto-book premium at time T. It is defined as earnings in excess of a "normal" profit on lagged book value of common equity. Therefore, a firm that does not create wealth will be worth only the value of its invested capital.
For this test, I approximate expectation about future earnings (X) using analysts' forecast of earnings per share. The use of analysts' forecasts assumes that financial analysts are capable of producing future earnings forecasts that respect the clean surplus
relation.6 In addition, a firm' value should be captured within a short period of time. The use of analysts' forecasts of earnings per share as a proxy for expected future earnings divided by lagged book value of common equity per share results in an estimate of the future rate of return on equity.
The EBO valuation model calls for an estimation of abnormal earnings over a period of time that is long enough to capture the price-to-book premium and for a selection of an appropriate measure of cost of capital. I use a five-year forecast horizon. 7 In addition, the cost of capital is approximated as the historical average stock returns for large US firms that are about 15% for the years 1992-1990.
In summary, the DFAE variable is computed as the summation over a five-year forecast horizon of the difference between firms' future return on equity and the cost of capital times the lagged book value of common equity. Substituting BV with the book value identity discussed above results in the following valuation model:
Pj3= 'x0 + ctj BVBDTjt +IDTPjt + 4 1DFAEjt + et(11)
All variables are defined as in equation 3 except that they are not deflated by BVBDT.
6Since analysts incorporate in their earnings forecasts all available information, the forecasts provide a good means to estimate future abnormal earnings. It is highly likely that the market assesses a firm's value based on analyst' forecasts of future earnings. These forecasts should reflect firm and industry growth, general economic conditions, and management's public forecast of earnings. Moreover previous research confirms that financial analysts use financial statement information in their forecasts of earnings (Chugh and Meador, 1984; Williams, Moyes, and Park, 1996). Their expectation about a finn's future earnings should reflect how they value the effect of deferred taxes on a finn's future cash flows. It is difficult to estimate future abnormal earnings beyond five years since analyst forecasts for a long period of time ahead are nonexistent.
This amount is obtained from Ibbotson (1997) Yearbook.
Sample Selection and Descriptive Data
The sample consists of firms listed on the NYSE and the ANIEX as reported by Standard and Poor's in Compustat Company Coverage for 1996. Firms are selected from three major industries using the two-digit SIC code: drugs, automotive, and computer (see distribution in table 4). Data are collected for the years 1992-1996. For 1992 and 1993, 1 only include in the sample the firms that have adopted SFAS 109.
Information about share price and number of shares outstanding at year-end is obtained from the 1996 Compustat tape. Forecasts of earnings per share and book value are taken from Value Line Investment Survey. Financial statements of each company and year selected are obtained from Lexis-Nexis (Edgar Library).
As reported in table 5, the preliminary sample consisted of 915 firm-year observations consisting of 265 (295 and 355) observations in the drug (automotive and computer) industry. A total of 346 observations were dropped from the sample because information about share price or deferred tax position was missing. Firms that report a negative book value of common equity or a net loss (103 observations) were also deleted from the sample. Another 12 observations for which deferred taxes per share are greater than one were excluded from the sample. Finally, in order to correct some econometric issues discussed in next chapter, 42 additional observations had to be eliminated from the sample resulting in total number of observations of 412.
The resulting sample is an unbalanced panel data set of 119 observations for the drug industry, 147 for the automotive industry and 146 for the computer industry. The number of observations per year is 105 observations in 1996, 114 in 1995, 100 in 1994, 65 in 1993 and 28 in 1992.
Table 6 reports descriptive statistics for the three industries for the period 1992-1996. Total assets for firms in the automotive industry are the highest when compared to the drug and computer industries. The proportion of deferred tax assets to total assets (6.3%) is also the highest for the automotive industry. However the current portion of these deferred tax assets accounts for only 36% as opposed to 51% (43%) in the computer (drug) industry.
Deferred tax liabilities represent about 8-10% of total liabilities for the three industries. The current portion is small and ranges from 0.6% for the computer industry to about 3% in the automotive industry. The ratio of book value of common equity before deferred taxes per share is the lowest for the drug industry (8.83), which is consistent with the R&D expenditures being omitted from the balance sheet. This ratio is of 20.52 and 17.44 for the computer and the automotive industries respectively. Growth rates computed on sales and fixed assets are about the same for the drug and computer industries. As for the automotive industry, the growth rates for both sales and fixed assets are the lowest which is consistent with expectations. Finally the valuation allowance in relation to gross deferred tax assets is the highest for the computer industry
(25%) compared with 16% and 5.4% for the drug and the automotive industries respectively. It suggests that firms in the computer industry incurred more losses and were less stable.
Table 7 shows the distribution of the price-to-book ratio (PB), the dependent variable, for the full sample and by industry. As can be observed, firms in the sample have relatively high PB. For the full sample, only 6% (26 out of 412) of the firms have a PB equal or less than 1 while 18% (74 out of 412) have a PB of more than 5. The average PB of 4.90 is the highest for the drug industry as compared to 2.94 and 2.55 for the computer and automotive industries respectively. The high PB values might mitigate the support in favor of the alternative hypotheses. One way to counter this problem is to remove, from the sample, observations with a high PB. However, this will reduce the sample size to a point that will make statistical inferences difficult.
Table 1 Summary of predictions for the base model If net operating If ignored If equity
Hypotheses a Coefficients asset (liability) adjusted
HI: DTP 31 + 0
H2: CDT 32 + 0
NDT 133 + 0
H3: DEP 34 + 0
H4: PRO 137 + 0
H5: ST 138 + 0
MT 139 + 0
LT 310 + 0
OTH 135 N/A N/A N/A
VA 136 + N/A N/A
If it explains
See appendix for variable definitions.
Table 2 Sales growth distribution
Sales growth Total Drug Auto Cptr
(GROWSA) N=412 N=119 N=147 N=146
GROWSA -.20 3 3
-.20 < GROWSA< -.10 12 1 8 3
-.05 < GROWSA < 0 39 11 16 12
0 < GROWSA< .05 38 15 15 8
.05 < GROWSA < .10 53 27 17 9
.15 < GROWSA < .20 60 12 27 21
.20 < GROWSA< .30 57 6 21 30
.30 < GROWSA< .40 28 2 7 19
.40 < GROWSA< .50 12 3 1 8
.50 < GROWSA< .60 14 2 4 8
.60 < GROWSA< .80 6 1 2 3
.80 < GROWSA< 1.00 6 2 1 3
GROWSA 1.00 6 2 1 3
412 119 147 146
Table 3 Common equity growth distribution
Common equity growth Total Drug Auto Cptr
(GROWCE) N=412 N=119 N=147 N=146
GROWCE -.20 9 2 6 1
-.20 < GROWCE< -.10 6 2 3 1
-.10 < GROWCE -.05 7 5 2
-.05 < GROWCE < 0 30 10 10 10
0 < GROWCE_ .05 35 15 10 10
.05 < GROWCE < .10 61 22 21 18
.10 < GROWCE < .15 84 17 48 19
.15 < GROWCE < .20 40 12 17 11
.20 < GROWCE < .30 52 20 11 21
.30 < GROWCE < .40 33 5 7 21
.40 < GROWCE < .50 12 1 3 8
.50 < GROWCE < .60 12 2 2 8
.60 < GROWCE < .80 9 2 7
.80 < GROWCE < 1.00 8 2 2 4
GROWCE 1.00 14 4 3 7
412 119 147 146
Table 4 SIC code distribution
SICa Total Drug Auto Cptr
N=412 N=1 19 N=147 N=146
2830 112 112
3850 7 7
3570 58 58
3660 10 10
3670 78 78
3710 78 78
3720 34 34
3730 7 7
3740 5 5
3750 4 4
3760 13 13
3790 6 6
Total 412 119 147 146
a Each SIC consists of the following: 2830: pharmaceutical preparations and in vitro and in vivo diagnostics, 3850: ophthalmic goods, 3570: computer and office equipment, electronic computers, computer storage devices, computer communication equipment and computer equipment peripheral, 3660: communication equipment, 3670: electronic component accessories, printed circuit boards, semiconductors and related devices and electronic connectors and components, 3 710: motor vehicles and car bodies, truck and bus bodies, motor vehicles and parts, truck trailers and motor homes, 3720: aircraft and parts, aircraft engine, parts and auxiliary equipment, 3730: ship, boat building and repairing, 3740: railroad equipment, 3750: motorcycles, bicycle and parts, 3760: guide missiles and space vehicles and 3790: transportation equipment.
Table 5 Sample selection
Total Drug Auto Cptr
industry 915 265 295 355
price or deferred
tax position (346) (129) (77) (140)
569 136 218 215
value of common
equity (8) (3)
loss (92) (12) (34) (46)
per share > 1 (12) (1) (10) (1)
Incomplete series (42) (4) (19) (19)
Total number of
observations 412 119 147 146
1996 105 29 39 37
1995 114 31 40 43
1994 100 28 34 38
1993 65 19 23 23
1992 28 12 11 5
Total number of
observations 412 119 147 146
Table 6 Summary statistics
Drug Auto Cptr
N=1 19 N=147 N=146
Total assetsa 5318.9 11124.4 3613.6
Total liabilities' 2968.6 9127.5 2203.6
Total fixed assets'a 2458.8 3787.1 1904.3
Total sales' 4320.3 8679.8 3513.6
Gross deferred tax assets 426.3 801.4 430.1
Valuation allowance a 27.1 22.8 94.1
Deferred tax assets' -399.2 778.6 336.0
Current deferred tax assets'10. 174.9 60.4
Current deferred tax liabilities' 5.1 22.3 .1I
Noncurrent deferred tax assets / 57.6 223.4 85.3
Noncurrent deferred tax liabilitieSb -101.8 177.7 63.8
Deferred tax assets/total assets 5.8 % 6.3 % 4.5 %
Valuation allowance/gross deferred tax
assets 15.9% 5.4% 24.9%
Deferred tax liabilities/total liabilities 8.2 % 10.0 % 8.7 %
Current deferred tax assets/deferred tax
assetsb 42.7% 35.9% 51.3%
Noncurrent deferred tax assets/deferred
tax assetSb 17.9% 22.2% 10.6%
Sum of other current and noncurrent
deferred tax assets 39.4% 41.9% 38.1 %
Current deferred tax liabilities/deferred
tax liabilities b 1.9% 2.7% 0.6%
Noncurrent deferred tax liabilities!
deferred tax liabilitieSb 32.8% 44.5% 51.1 %
Sum of other current and noncurrent
deferred tax liabilities 65.3 % 52.8 % 48.3 %
Sales growth 15.0 % 12.7 % 23.6 %
Fixed assets growth 16.3 % 10.4 % 21.8 %
Book value before deferred taxes/share 8.83 17.44 20.52
aIn millions of dollars
b The sum of current and noncurrent deferred tax assets (liabilities) does not equal the total amount. Some companies do not disclose separately the current and noncurrent portions of deferred tax assets (liabilities). Other firms only disclose part of the information. These companies did not provide enough information to allow the reader to establish the missing portions of current and noncurrent deferred tax assets (liabilities).
Table 7 Price-to-book ratio distribution
Price-to-book ratio Total Drug Auto Cptr
(PB) N=412 N=119 N=147 N=146
PB < 1 26 2 10 14
Average PB 3.37 4.90 2.55 2.94
Descriptive statistics and the Pearson correlation coefficients for the variables used in the base model are reported in tables 8 and 9. In table 8, the means and standard deviations for each variable are presented by industry and in total. The P3 variation is the highest for the drug industry. This appears to be consistent with the nature of this industry where share price varies greatly according to discoveries. The variation in E is approximately the same for the drug and the automotive industries (13 and .12 respectively) and it is lowest for the computer industry (08).
All industries have a positive deferred tax position. This observation is contrary to popular belief that deferred taxes generally result in a liability. Furthermore, as expected, the means of the variable DEP are negative for all industries, i.e. on the average, firms report deferred tax liabilities with respect to depreciation. The DEP mean is the lowest for the computer industry (-.02 vs. -.07 and -.05 for the automotive and the drug industries respectively) possibly because many firms in this industry have a large proportion of deferred tax assets. Also, in line with the discussions in chapter 3, deferred taxes arising from restructuring charges (MT) are important for the drug and the
automotive industries as reflected by their means (03 for both industries vs. .009 for the computer industry).
Panel A of table 9 displays Pearson correlation coefficients for the full sample. The correlation coefficients are positive and significant between the variable P3 and all the independent variables except DEP and LT.
Panels B, C and D of table 9 reports the correlation coefficients for the drug, the automotive and the computer industries respectively. For the drug industry, the correlation coefficients between the variable P3 and all the independent variables except NDT, DEP, VA and LT are positive and significant. In the case of DEP, the coefficient is negative and significant.
For the automotive industry, the correlation between P3 and the independent variables are positive and generally significant. The only exceptions are those for P3OTH, P3-VA and P3-LT.
For the computer industry, the correlation coefficients between P3 and the regressors are still positive but some of the coefficients are not significant. A priori this lack of significant correlation is justified since the computer industry is a growing industry. Indeed, investors are likely to be more interested in the future growth opportunities of these firms than in the deferred tax accounts.
Overall, in all panels, the correlation coefficients are generally low among the regressors. Therefore it is unlikely to encounter serious multicollinearity problems. Also, the correlation analysis supports the contention that the value-relevance of deferred taxes has to be tested by industry and not only for the full sample.
In my research, the data set is a panel data that combines time series and cross sectional information. In this type of data set, there are a certain number of cross-units and time periods. Therefore, the usual OLS procedure may not be appropriate because it assumes that the regression parameters do not change over time and that they do not differ between different cross-sectional units. These assumptions are not necessarily valid.
In a typical panel, such as the data set I use in this study, there are a large number of cross-sectional units and only a few time periods. Therefore the focus is on analyzing the heterogeneity of the data rather than the autocorrelation problem, which is generally of less importance even though the statistics may show otherwise.
Indeed, using the Durbin-Watson statistic to detect serial autocorrelation may be misleading. The usual statistic is computed as the summation of the squared difference between the residual at time t and t- I divided by the summation of the squared residual at time t. Therefore the difference of the residuals may be computed between two companies instead of two years for the same firm. For example, assuming that GM follows Ford in the data set and that the data for both companies are available for all 5 years, the Durbin-Watson statistic would be computed for GM in 1992 as the difference between its residual in 1992 and the residual of Ford in 1996.
One way to control for both of these problems is to estimate the model as a two-way fixed-effects model also known as the least squares dummy variable approach (Greene, 1993). The two-way fixed-effects approach is a model in which two dummy variables are incorporated; one controlling for the individual cross-sectional unit effect and the
second one controlling for time-specific effect. This model is estimated using the ordinary least squares method.
If the residual correlations are induced by a fixed-effect, the resulting model may lead to regression residuals with substantially lower intercorrelation, which may increase efficiency of estimation and unbiased standard errors. Further to the extent that the intercepts reflect the effects of correlated omitted variables, the slope coefficients on the explanatory variables are less subject to bias from omitted variables.
The fixed-effect model has some disadvantages. First the effects of any variable that is a constant over time for a given firm will be captured in the intercept. Therefore, the fixed effect of an explanatory variable of interest can be removed. Second the fixedeffect model is costly in terms of degrees of freedom lost. Therefore the significance of some variables may be challenged.
With regard to my study, the two-way fixed-effects model seems a reasonable approach since the difference between units can be viewed as parametric shifts of the regression function. It is likely that the market value of common equity is conditional on firm-specific factors such as future growth opportunities and product differentiation.
Alternatively, the individual effects could be treated as random effects. The main advantage of the random effects model is that it is not costly in terms of degrees of freedom. The random effects model assumes that the individual effects are uncorrelated with other regressors. If the individual effects are correlated with other independent variables, the random effects model may suffer from inconsistencies due to omitted variables. Hausman's test for orthogonality of the random effects and the regressors compares the covariance of the estimates computed with the fixed-effect model and the
random effects model (Greene, 1993).1 If the null hypothesis of orthogonality or no correlation between the regressors and the individual effect is rejected, one cannot use the random effects approach to estimate the model.
In my study, the sample consists of a subset of companies drawn from the entire population of large US firms. It is also reasonable to assume that the individual effects are random effects. I performed the Hausman test and rejected the null hypothesis of orthogonality in most cases. In the situations where the null hypothesis was not rejected, both approaches led to similar results. For this reason, the results are presented only for the two-way fixed-effects model. For comparison purposes, I also report the regression results without controlling for fixed effects (pooled regression).
As indicated in chapter 4, the estimate coefficients on earnings and on the deferred tax variables may differ from their theoretical value of one because of measurement error and omitted variables from the model (Beatty et al., 1996). My findings -- similar to those in other studies -- display such a pattern. Therefore, in the following statistical analysis, the predictions are based on the coefficient signs rather than on magnitude. For that purpose, table 10 presents an overview of the results displaying the sign and the significance of coefficients for all tests except the year-by-year regression. Detailed results are presented in the remainder of this chapter.
1 The covariance (Z) of an efficient estimator with its difference from an inefficient estimator is zero. The distribution of I Mows a chi-squared distribution with k-1 degree of freedom.
Base Model Estimated as a Two-Way Fixed-Effects
Table I I reports the results of testing hypotheses I to 5 using the two-way fixed2
effects model. The results are presented for the model with a common intercept. Therefore the coefficients on the dummy variable have to be evaluated as a shift from the common intercept. One advantage of keeping the intercept in the model is that the regression results give the usual R value.
A general observation derived from table I I is that earnings, E, are generally positive and highly significant. The only exception relates to the computer industry where E is only significant (p-value is .07) in panel B. One explanation here is the fact that the market cares more about other factors such as market share than reported earnings.
Panel A of table I I reports the results of testing the market valuation of DTP. The regression is estimated for the full sample and by industry. The results suggest that the market views DTP as a real asset (liability) for the full sample. This finding is consistent with prior research (Amir et al., 1997a; Chaney and Jeter, 1994; Givoly and Hayn, 1992) that has reported that deferred tax liabilities constituted real liabilities. When the test is performed by industry, the results are mixed. For the drug and the automotive industries, the coefficients are also positive and significant. On the other hand, for the computer industry, the coefficient is negative and not significant suggesting that the market includes DTP in equity. One explanation for this finding relates to the sustainable growth achieved by firms in this industry since 1991. Investors were more interested in the
2 The model has also been estimated without the intercept term and the results did not substantially change.
future growth opportunities of these firms than in financial statement numbers. As indicated in table 6, the average sales and fixed assets growth rates are the highest for the computer industry (23.6% and 21.8% respectively). These findings emphasize the need to perform value-relevance analysis by industry,
Panel B of table 11 presents the regression results of market valuation of the current and noncurrent classification of deferred taxes. The results suggest that this classification affects price-to-book ratio. The coefficients on CDT are positive and significant for the full sample and the drug industry. This finding is consistent with expectations. It suggests that the market views CDT as real assets (liabilities). However, for the automotive and computer industries, the coefficients are not significant (.11 and .33 respectively), although positive, suggesting that the market does not view current deferred tax assets (liabilities) as detrimental (advantageous) to the firm.
The coefficient on NDT is also positive and significant for the full sample and the automotive industry. This finding is consistent with expectations, suggesting that the market views NDT as real assets (liabilities). In the case of the drug industry, though the coefficient is positive, it is not statistically significant. This suggests that investors do not consider NDT assets (liabilities) as detrimental (advantageous) to the firm. Contrary to expectations, the coefficient on NDT is negative and significant for the computer industry. This result indicates that the market views NDT assets (liabilities) as detrimental (advantageous) to the firm and adjusts equity accordingly.
The results for the computer industry indicate that investors treat CDT and NDT differently. One explanation may be that investors view CDT as real assets (liabilities), simply because many firms report CDT along with prepaid taxes and other prepaid
expenses. On the other hand, NDT assets (liabilities) are considered detrimental (advantageous) to the firm in the light of future growth. Therefore, NDT are unlikely to be realized in the foreseeable future. Again, the results suggest that performing the test by industry helps explain the price-to-book ratio.
Panel C of table 11 displays the results of testing the value-relevance of deferred taxes resulting from DEP and PRO. For the full sample and the computer industry, though the coefficients on DEP are not significant, their signs are consistent with expectations. For the automotive industry, the coefficient on DEP is positive and significant suggesting that deferred tax assets (liabilities) related to DEP are real assets (liabilities).
For the drug industry the coefficient on DEP is negative and not significant. This result suggests that investors view DEP as unlikely to be realized and adjust equity accordingly. This could be explained by the fact that, in this industry, fixed assets not only continuously grow but also constitute only a relatively small portion of total expenditures. Some of these expenditures such as R&D are omitted from the balance sheet.
As indicated above the results are different for the computer and the drug industries. While both industries invest large amounts in R&D, they differ in several aspects. In the drug industry, the economic life of a discovery is relatively long and enjoys patent protections. On the other hand, even though patent protections are also available in the computer industry, technologies are continuously modified and upgraded. Therefore, in this industry, the unrecorded R&D do not provide long-term economic benefits to firms.
In line with expectations, coefficients on PRO are positive and significant for the full sample and the automotive industry, implying that deferred taxes arising from PRO constitute real assets affecting price-to-book ratio. For the drug and the computer industries, the coefficients are positive but not significant, suggesting that investors do not consider DTA related to PRO as detrimental to firms within these industries. However, investors are giving less weight to DTA than for firms within the automotive industry.
Contrary to expectations, the coefficient on 0TH is positive and significant for the full sample and the drug industry. This finding is surprising since 0TH includes a number of components such as loss carryforwards, investment tax credits, and other undisclosed items. One plausible explanation is that 0TH is capturing the effect of loss carryforwards and investment tax credits, which represent real assets because they reduce taxes payable in future years. The drug industry has the highest 0TH mean as indicated in table 8 (.08 vs.008 and .05). This may explain why 0TH is significant for the drug industry and not for the automotive and the computer industries.
Finally, the coefficients on VA are generally not significant for the full sample and all industries. The signs of the coefficients are negative implying that valuation allowance is not detrimental to firms within these industries. One explanation may be that investors look at the underlying economic events that led to the recording of the valuation allowance. Miller and Skinner (1998) report that managers set the valuation allowance based on their firms' tax loss and tax credit carryforwards. Therefore, it is possible that investors consider the future tax benefits associated with these carryforwards and not the valuation allowance per se.
Panel D of table 11 reports the findings of testing the value-relevance of deferred tax assets related to ST, MT, and LT. Coefficients on ST are positive and significant for the full sample and the automotive industry. In line with expectations, this finding suggests that deferred taxes arising from ST constitute real assets affecting price-to-book ratio. For the drug and the computer industries, the coefficients are still positive, which is in line with expectations. However, the coefficients are not significant, implying that DTA related to ST are not value-relevant for firms within these industries. One explanation for this result is that investors are more interested in future prospects of these firms than they are with balance sheet numbers.
In line with expectations, the coefficient on MT is positive and significant for the automotive industry. This finding implies that the market values deferred taxes arising from MT as real assets. Coefficients on MT are also positive for the full sample and the drug industry but they are not significant. This finding suggests that DTA related to MT are not value-relevant for firms within these industries. Again it implies that investors are more interested in future growth opportunities than they are with balance sheet items, As for the computer industry, this coefficient is negative and not significant. This suggests that DTA are not value-relevant. However the sign of the coefficient implies that DTA are detrimental to firms within this industry. One possible explanation is that investors examined the underlying economic event that necessitated the recording of the provision.
As discussed in chapter 3, MT relates mostly to restructuring charges, which were particularly important for the drug and the defense sector of the automotive industries. Table 8 indicates that the means of NIT are .03 for both the drug and the automotive
industries and only .009 for the computer industry. These statistics support the above observations except for the drug industry for the reason explained above.
Finally, coefficients on LT are positive but not significant for the full sample, the drug and the automotive industries. This finding suggests that DTA related to LT are not value-relevant for these firms. As indicated in chapter 3, LT comprises deferred taxes related to pension and postretirement benefits, which will not be realized before at least five years. Therefore, investors of these firms seem to have discounted the value of DTA related to LT. However, the positive sign of the coefficients for the drug and the automotive industries reflects the fact that both industries are labor intensive. Table 8 indicates that the mean LT is the highest for the automotive firms (.05), followed by the drug firms (01) and the lowest for the computer firms (0004).
For the computer industry, the coefficient is negative and not significant. Firms in this industry usually fund their pension plans as costs are accrued resulting in no temporary differences. Therefore the main component of LT for these firms relates to postretirement benefits. As indicated above, this amount is small when compared to the drug and the automotive industries. This may explain the sign of this coefficient.
Finally, panel D presents the results for the variables DEP, OTH and VA. As expected, the results are similar to the findings discussed for panel C.
Table 12 reports the results for the model without controlling for the fixed effects. The t-tests reflect White corrections for heteroscedasticity. These results are only reported as an indication of what would be the results if one does not properly control for the individual fixed effects in pooling all data together. The results are generally in line with the base model. One noteworthy result in panels C and D of table 12 is that DEP is
highly significant when the tests are performed by industry while they were generally not significant when the fixed-effect model is used (see panels C and D of table 11).
To conclude, the findings suggest that depending on the industry, deferred tax disclosures under SFAS 109 provide value-relevant information to the market. The balance sheet disclosure is generally useful for the automotive industry, which is a more traditional sector of the US economy. Furthermore for this industry, the classification of deferred tax components into ST, MT, and LT appear to be appropriate since it better isolates which components of deferred taxes are considered by the market to be realizable. As for the drug industry, only the balance sheet disclosures provide valuerelevant information to the market. Finally, for the computer industry, disclosures under SFAS 109 do not generally provide any valuable information. These results emphasize the need to control for industry when assessing the value-relevance of the income tax disclosures.
Partitionina of the Sample According to Sales Growth
In this section, I examine the results arising from the partitioning of the sample into HG, MG, and LG. As indicated in chapter 4, this partitioning allows me to obtain sub samples that contain enough observations to perform the regression analysis.
Panel A of table 13 reports the results of testing the market valuation of DTP using the partition sample. The regression is run for the full sample and by industry. The model is estimated using the usual OLS procedure without controlling for the individual cross-sectional unit and for time-specific effects. The partitioning seems to have
eliminated most of these fixed-effects. Indeed, I tested the null hypothesis that the fixedeffects are the same across all units and most of the time I could not reject the null hypothesis. Where I rejected the null hypothesis, estimating the model using the twoway fixed-effects model did not significantly change the results.
A general observation derived from table 13 is that earnings, E, are positive and highly significant for the HG firms. The only exception relates to the drug industry. For the MG and LG groups, the coefficients on E are also generally significant except for the automotive and the computer industries for the MG and LG categories respectively.
Panel A of table 13 reports the results of testing the market valuation of DTP. Contrary to predictions made with respect to growth, coefficients on DTP are positive and significant for the full sample and the computer industry. This result suggests that DTP constitutes real assets (liabilities) regardless of growth.
For the MG and LG firms, the results are generally consistent with the base model. All coefficients are positive and strongly significant except for LG firms included in the drug and the computer industries,
Panel B of table 13 presents the regression results of market valuation of the current and noncurrent classification of deferred taxes. The partitioning modifies the results obtained with the base model. For the HG firms, the findings are not completely in line with expectations with respect to growth. Though many coefficients are not significant, some remain so. For example, the coefficient on CDT is positive and significant for the drug industry and the coefficient on NDT is positive and significant for the computer industry. A priori, there is no plausible explanation as to why these coefficients are significant, particularly because these firms have a GROWS A higher than. 15. Therefore
it is highly unlikely that deferred tax disclosure could affect price-to-book ratio. For the MG firms, coefficients on CDT are generally not significant although they are generally significant for NDT whereas for the LG firms, the results are mixed.
Panel C of table 13 displays the results of testing the value-relevance of deferred taxes resulting from DEP and PRO. Contrary to expectations with respect to growth, the results for the full sample suggest that deferred tax components related to DEP and PRO affect price-to-book ratio for the HG but not for the LG firms. However, the results are mitigated when the tests are performed by industry.
For the HG firms, coefficients on DEP for the automotive and the computer industries are positive and significant. Also coefficients on PRO, OTH and VA are significant for the computer industry.
For the MG and LG firms in the automotive industry, the coefficients on PRO are positive and highly significant. These coefficients are also positive and significant only for the MG firms in the drug and the computer industries. Furthermore, contrary to expectations with respect to growth, the coefficient on OTH is positive and generally significant for the MG firms. Finally, coefficients on VA for the MG and LG groups vary according to the industry.
Panel D of table 13 reports the findings of testing the value-relevance of deferred taxes arising from ST, MT, and LT. Overall the partitioning and the industry
classification provide mixed results that are sometimes contrary to expectations with respect to growth. For example, for the HG firms, the coefficients on ST are positive but not significant for the full sample and for the drug and the automotive industries while it is significant for the computer industry. The coefficients are still positive for the MG
firms but contrary to the HG firms, they are significant for the full sample and for the drug and the automotive industries. For the LG firms, this coefficient is only positive and significant for the automotive industry.
In summary, contrary to expectations, growth does not generally affect the findings obtained in the base model. Many reasons could explain these results. First, GROWSA might not be an appropriate proxy for growth for firms in my sample. Indeed, the dependent variable, price-to-book ratio, may in itself be a proxy for growth. As indicated in the literature (Palepu et al., 1997), price-to-book ratios should vary across firms according to differences in their future returns on equity, growth in book value, and risk. Second, GROWSA ignores long-term prospects such as order trends and unfilled orders and cancellations (S&P, 1997). Third, the market might simply not assess the valuerelevance of deferred taxes in light of growth. Finally, the partitioning HG, MG, and LG might not be fine enough to find clear evidence of the effect of growth on the realization of deferred tax components.
First-Difference Model Analysis
As a test on the specification of the base model, I run the regression using a firstdifference model. The findings are reported in table 14. The results generally support the conclusions made in the base model. One noteworthy exception to the results is that the coefficients on DEP are not significant for the full sample and for all industries. One
explanation is that when using a first-difference model, independent variables are measured with error resulting in coefficients biased toward zero (Kothari and Zimmerman, 1995). Another explanation is that the variables in the first-difference model are small compared to their levels. Consequently it may be that their impact on returns could not be detected statistically.
Year-by-Year Regression Analysi
As an additional check on the specification of the base model, the analysis is performed separately for each year. The findings are reported in table 15. The coefficients on DTP are significant for 1992 and 1995 whereas the coefficients on CDT are significant for 1992 and 1995 and the coefficient on NDT is only significant in 1995. The value-relevance of PRO and the disaggregation of PRO into ST, MT, and LT is confirmed for 1992, 1995 and marginally for 1996.
Overall, the data for 1992, 1995 and 1996 seem to drive the results. One explanation is that investors examined 1992 financial statements more carefully because it was the first year of the implementation of SFAS 109. The findings for 1995 and 1996 suggest that investors, becoming more familiar with the standard, started to investigate more thoroughly the footnote disclosure and its implication for cash flows. Finally, it appears that analyzing the results by year ignoring industry classification provides limited valuerelevant information.
EBO Valuation Model Analysis
The results of testing the value-relevance of deferred taxes with the EBO-valuation model are presented in table 16. The model is estimated using the two-way fixed-effects approach.
Panels A and B of table 16 report the results of the market valuation of balance sheet disclosures. The results are generally highly significant particularly for the full sample and the automotive industry. These results are consistent with the ones obtained in the base test.
Panels C and D of table 16 show the findings for testing the value-relevance of DEP and the disaggregation of PRO into ST, MT and LT. Again, the results are generally consistent with the base model.
Though there are inconsistencies in the results, one common feature that remains valid is that conducting an analysis by industry provides valuable insights. Furthermore, the findings in the base test seem to be robust since they are generally supported by the results obtained using another valuation model.
Partitioning of the Sample According to Growth in Common Equity
The results of testing the value-relevance of deferred taxes by partitioning the sample according to GROWCE are presented in table 17. Overall the results obtained are similar to the ones using GROWSA. Therefore it seems that GROWCE is not a better proxy for growth than GROWSA. Indeed, GROWCE ignores intangible assets such as goodwill,
employment knowledge, quality of management, and trade secrets and patents that are important unrecorded assets particularly for the drug and the computer industries.
Table 8 Descriptive statistics for selected variables
Variables a.b Total Drug Auto Cptr
N=412 N=1 19 N=147 N=146
P3 3.37 4.90 2.55 2.94
(2.39) (2.70) (1.65) (2.16)
E .15 .18 .15 .13
(.11) (.13) (.12) (.08)
DT'P .05 .06 .06 .03
(.13) (.10) (.18) (.07)
CDT o05 .06 .05 .04
(.07) (.05) ('10) (.04)
NDT .0008 .002 .01 -.01
(.09) (.09) (.12) (.06)
PRO .09 .08 .14 .05
(.14) (.07) (.21) (.06)
DEP -.05 -.05 -.07 -.02
(.08) (.05) (. 11) (.03)
0TH .04 .08 .008 .05
(. 11) (.12) (.09) (.12)
VA .04 .04 .02 .05
(.08) (.08) (.04) (.10)
ST .05 .03 .06 .04
(.09) (.04) (.14) (.04)
MT .02 .03 .03 .009
(.06) (.05) (.07) (.03)
LT .02 .01 .05 .0004
(.06) (.02) (.10) (.03)
a Means (standard deviations in parentheses) for variables used in the base model. b See appendix for variable definitions.
Table 9 Pearson correlation coefficients
Panel A: Full sample (N=412) a,b
E DTP CDT NDT PRO DEP OTH VA ST MT LT
P3 .56* .36* .32* .25* .13* .07 .31 .18* .14* .15* -.05
E 1 .41* .36* .30* .22* .20* .03 -.03 .22* .13* .05
DTP 1 .65* .82* .67* .16* .26* .11* .60* .41* .25*
CDT 1 .16* .56* .02 .04 .02 .64* 39* -.03
NDT 1 .50 .15* .29* .14* .34* .27* .37*
PRO 1 -.37* -.17* .09** .81* .60* .50*
DEP 1 -.04 .01 -.22* -.26* -.28*
OTH 1 .70* -.09** -.05 -.19*
VA 1 .13 .08** -.06
ST 1 .32* .07
MT 1 -.01
a*Significant at the level of .05 or less.
*Significant at the level of. 10.
b Subscripts for firm and time have been omitted for clarity.
Panel B: Drug industry (N=119) a,b
E DTP CDT NDT PRO DEP OTH VA ST MT LT
P3 .62* .32* .49* .07 .35* -.40* .30* .15 .40* .18** -.02
E 1 .20* .55* -.09 .36* -.24* -.02 -.09 .53* .05 .14
DTP 1 .37* .86* .40* .01 .56* -.01 .21* .29* .23*
CDT 1 -.14 .72* -.29* -.15* -.21' .48* .49* .31*
NDT 1 .03 .17** .68* .11 -.05 .05 .08
PRO 1 -.32 -.15 .004 .38* .84* .55*
DEP 1 -.15 .04 -.06 -.34* -.11
OTH 1 .63* -.02 -.13 -.13
VA 1 .03 .005 -.05
ST 1 -.11 -.23*
MT 1 .49*
a*Significant at the level of .05 or less.
*Significant at the level of .10.
b Subscripts for firm and time have been omitted for clarity.
Table 9 continued
Panel C: Auto industry (N=147) a,b
E DTP CDT NDT PRO DEP OTH VA ST MT LT
P3 .67* .56* .37* .54* .28* .37* .04 .05 .31* .17* .008
E 1 .56* .32* .57* .25* .49* -.13 -.10 .26* .13 .07
DTP 1 .74* .81* .76* .24* .06 .33* .71* .48* .25*
CDT 1 .26* .56* .10 .17' .30* .70* .40* -.11
NDT 1 .66* .23* -.11 .24 .48* .37* .46*
PRO 1 -.32* -.25* .35* .86* .60* .45*
DEP 1 -.08 -.12 -.23* -.19* -.22*
OTH 1 .43* -.14** -.08 -.28*
VA 1 .37* .29* -.004
ST 1 .46* .03
MT 1 -.11
a *Significant at the level of .05 or less.
**Significant at the level of. 10.
b Subscripts for firm and time have been omitted for clarity.
Panel D: Cptr industry (N=146) a,b
E DTP CDT NDT PRO DEP OTH VA ST MT LT
P3 .41* .37* .22* .30* .09 .19* .34* .25* .06 .03 .06
E 1 .30* .22* .30* .11 .07 .17* .08 .12 .13 -.07
DTP 1 .46* .84* .44* .18* .50* .19* .28* .25* .27*
CDT 1 .09 .50* .07 .01 .004 .58* .10 .20*
NDT 1 .25* .01 .61* .28* .01 .26* .20*
PRO 1 -.30* .06 .27* .83* .53* .42*
DEP 1 -.09 -.15** -.08 -.32* -.15*
OTH 1 .85* -.02* .10 .03
VA 1 .24* .17* .06
ST 1 .19* .23*
MT 1 -.33*
a*Significant at the level of .05 or less.
**Significant at the level of. 10.
b Subscripts for firm and time have been omitted for clarity.
Table 10 Overview of the results
HI H2 H3 114 H5
_______________DTP CDT NDT DEP PRO ST MT LT
If net operating asset (liability) + + + + + + + + If ignored 0 0 0 0 0 0 0 0
If equity adjusted
Two-way fixed-effects- ---Full + + + + + + + +
Drug + + + + + + +
Auto + + + + + +
Computer + + + +
No control for fixed-effect
Full + + + + + + +
Drug + + ++ + +
Auto + +~ +~ + + + +
Computer +~ +~ + +~ + +
Full +~ + + +- + +
Drug + + + + +
Auto +- + + +
Full + +++ + + + +
Drug + + + + + +
Auto + +_ + +
Computer + ___ + + +
a'See appendix for variable definitions. Coefficient signs shown in the table. b Significant at the level of .05 or less.
*Significant at the level of 10.
Table 10 continued
HI H2 H3 H4 H5
DTP CDT NDT DEP PRO ST MT LT
If net operating asset (liability) + + + + + + + + If ignored 0 0 0 0 0 0 0 0
If equity adjusted .. .
HG: Full + + + + + + + +
Drug + + + + +
Auto + + + + + + +
Computer +* + + +* +* + +
MG: Full + + + + + +
Drug + + + +* +*Auto + + +* +* +" + +
Computer + + +** + + + +
LG: Full +* +* + + +
Drug + + + + +
Auto + + + + + +
Computer + + + + +
HG: Full + +* + + + +
Drug + + + + + +
Auto + + + +* + + + +
Computer + + + +* + + + +
MG: Full + + + + + +
Drug + + + + +
Auto +* + + + + + +
Computer + + + +* + +
LG: Full + + + +
Drug+ + + + + +
Auto + + + + + +
Computer + + + +- +
a See appendix for variable definitions. Coefficient signs shown in the table. b *Significant at the level of .05 or less.
Significant at the level of. 10.
Predictions are only valid for the LG group. For the HG group, the coefficients should not be significant while they should be mixed for the MG group.
Table 11 Regression results of the two-way fixed-effects model
Panel A: Deferred tax position a,b
Pit,= Zocj + j x +Oj3DTPjt + IE jt+6j
Full 412 3.69 6.28
R2=.88 (.0001) (.0001)
Drug 119 3.77 11.21
R2=.89 (.03) (.0001)
Auto 147 3.58 4.74
R2=.87 (.001) (.0009)
Cptr 146 -.22 1.71
R2=.89 (.94) (.28)
a See appendix for variable definitions. Numbers in parentheses are p-values. b ao: full: -2.99 to 8.14, drug: -6.32 to 4.88, auto: -1.61 to 3.42, cptr: -2.75 to 9.21.
ai: full: -. 15 to .03, drug: -1.00 to -.03, auto: -.40 to. 10, cptr: .04 to .49.
Panel B: Classification of deferred taxes a,b
P- = o + o +I3aCDTjt + 33NDTjt + 41Et +6jt
Sample N 32 P3
Full 412 5.49 2.72 6.46
R2=.88 (.008) (.02) (.0001)
Drug 119 8.37 2.77 11.25
R- -.89 (.07) (.14) (.0001)
Auto 147 3.33 4.39 4.46
R2=.87 (.11) (.001) (.002)
Cptr 146 4.51 -10.06 2.97
R2=.90 (.33) (.03) (.07)
a See appendix for variable definitions. Numbers in parentheses are p-values. b a0 : full: -3.12 to 8.22, drug: -5.65 to 5.19, auto: -1.60 to 3.35, cptr: -3.53 to 10.11.
a,: full: -. 13 to .04, drug: -.98 to -.04, auto: -.40 to .09, cptr: .06 to .53.
Table 11 continued
Panel C: Deferred taxes arising from provisions and depreciation a,b
Pj= o + cX34DEPjt + 03SOTHjt 36 VA + 7PROjt + Et +sj
Sample N 134 35 36 07 1
Full 412 1.63 5.70 -1.34 3.60 7.06
R2=.89 (.49) (.0001) (.40) (.0007) (.0001)
Drug 119 -6.29 5.22 -1.09 2.35 11.01
R2=.90 (.28) (.02) (.76) (.37) (.0001)
Auto 147 4.46 2.10 -4.01 3,97 4.90
R2=.87 (.06) (.44) (.26) (.0005) (.0008)
Cptr 146 15.93 -5.19 -6.22 4.35 2.59
2-=.90 (.11) (.26) (.10) (.29) (.12)
a See appendix for variable definitions. Numbers in parentheses are p-values. b 0o: full: -5.12 to 5.51, drug: -7.44 to 2.68, auto: -1.92 to 3.30, cptr: -3.59 to 8.89.
a1: full: -.26 to -.04, drug: -1.10 to -.19, auto: -.49 to .01, cptr: .01 to .36.
Panel D: Deferred tax components a,b
P + (Xi + 134DEPt 13SOTHjt + 36 VAt + P38STjt + 39MTjt + 131oLTjt + 41Ejt +sj
Sample N 04 05 06 38 09 0I0 11
Full 412 2.21 5.48 -.94 8.93 .91 .51 6.79
R=.90 (.35) (.0002) (.56) (.0001) (.69) (.77) (.0001)
Drug 119 -6.56 5.06 -1.07 4.25 1.53 3.68 10.76
R2=. 90 (.28) (.03) (.78) (.65) (.73) (.70) (.0001)
Auto 147 4.68 1.58 -3.11 7.05 4.04 1.36 5.14
R=.88 (.05) (.56) (.38) (.002) (.05) (.42) (.0005)
C],tr 146 14.89 -2.75 -4.29 8.22 -4.54 -4.37 2.25
R =.90 (.14) (.60) (.32) (.16) (.77) (.67) (.19)
a See appendix for variable definitions, Numbers in parentheses are p-values. ba0: full: -5.12 to 5.92, drug: -7.60 to 2.63, auto: -2.20 to 3.17, cptr: -3.47 to 8.74.
al: full: -.21 to -.006, drug: -1.10 to -.20, auto: -.43 to .05, cptr: .03 to .36.