Vous êtes sur la page 1sur 27

The Valuation of Deferred Taxes*

ELI AMIR, Columbia University

MICHAEL KIRSCHENHEITER, Columbia University

KRISTEN WILLARD, Columbia University

We examine the value relevance of deferred tax components disclosed under SFAS
No. 109. We classify deferred tax components into seven categories: depreciation and
amortization; losses and credits carried forward; restructuring charges; environmental
charges; employee benefits; valuation allowance required hy SFAS No. 109; and all
other components. We find that separating deferred taxes into components provides
value relevant information. In particular, the valuation coefficient on deferred tax lia-
bilities from depreciation and amortization is close to zero, reflecting investors' expec-
tations that firms will continue to invest in depreciable assets reducing the probability
of future reversal. Also, deferred taxes from restructuring charges have valuation coef-
ficients larger than other deferred tax components, reflecting the higher likelihood of
reversal in the short run. Finally, we find that the net realizable value of deferred taxes
from losses and credits carried forward is negatively associated with stock prices. This
result suggests that investors do not expect part of these carryforwards to he utilized,
although we cannot rule out the possibility that model misspecification is driving this
result.

Condense
Les impots report6s resultent de l'ecart entre l'impot exigible et la charge constatee au
titre des impots de l'exercice. II n'y a done d'impots reportes que lorsque les normes
d'information financiere different des exigences en matifere d'information fiscale. II
arrive souvent que les utilisateurs des etats financiers ne s'entendent pas sur la methode
la plus appropriee pour evaluer une societe dont le hilan contient des impots reportes
actifs et passifs. Certains sont d'avis que les impots reportes nets representent des oblig-
ations financieres et, par cons6quent, que leur valeur doit etre actualis6e de la meme
maniere que les autres obligations financieres a long terme. D'autres invoquent le fait
qu'il est frequent que les impots reportes passifs (resultant, par exemple, d'ecarts tem-
porels lies a l'amortissement) ne soient jamais regies, de sorte que la valeur comptable
de l'avoir des actionnaires de l'entreprise devrait etre augmentee des impots reportes
passifs nets (et diminuee des impots report6s actifs nets). La majorite des utilisateurs des
etats financiers paraissent cependant r6agir a la complexite du probleme en ignorant car-
rement les impots report6s.

* We want to thank Lane Daley (editor), Jerry Feltham, Trevor Harris, Jim Ohlson, and
seminar participants at the University of British Columbia, Columbia University, and the
University of California (Los Angeles) for their helpful comments. Jing Liu and Xiaojun
Zhang deserve special thanks for their research assistance. Eli Amir is grateful to the
KPMG Peat Marwick Foundation for its financial support.

Contemporary Accounting Research Vol. 14 No. 4 (Winter 1997) pp. 597-622 ©CAAA
598 Contemporary Accounting Researcb

Les auteurs se donnent pour objectif de proposer un cadre de rdfdrence simple per-
mettant de comprendre le role des impots report6s nets, et des 616ments qui les com-
posent, dans I'fitablissement de la valeur de I'entreprise. Us d6tnontrent aussi que Tin-
formation relative aux impots report6s foumie aux termes du SFAS n° 109 est pertinente
£i r^tablissement de cette valeur. Us examinent plus pr^cisement si revaluation des 616-
ments d'impot report6 d6pend de la probabilit6 de r6sorption. A partir de I'information
foumie sur les 616ments d'impot report6, conform6ment aux exigences du SFAS n° 109,
ils 6valuent la probabilit6 de r6sorption de chaque 616ment et elaborent des pr6visions
quant au sens et ^ I'importance de la valeur de ces 616ments. Ils d^terminent ensuite la
mesure dans laquelle revaluation des 616ments d'impot report6 depend de la probabil-
ity de r6sorption. Ils font l'hypothdse que les impots report6s qui sont plus susceptibles
de se r6sorber it la p6riode subs6quente contdbuent davantage & la valeur de I'entreprise
(en valeur absolue) que les 616ments d'impot report6 qui sont moins susceptibles de se
r6sorber k court terme. Enfin, ils d6terminent la pertinence de la provision pour 6valua-
tion d'actif pr6sent6e conform6ment au SFAS n° 109 dans la d6termination de la valeur.
Pour examiner revaluation des impots report6s telle qu'elle est 6tablie par le
march6, les auteurs ont recours h un modMe d'6valuation dans lequel la valeur
marchande de l'avoir des actionnaires est une fonction Iin6aire des actifs d'exploitation
nets, des actifs financiers nets et des b6n6fices d'exploitation anormaux actuels et dif-
f6r6s (d6finis comme 6tant les ben6fices d'exploitation r6els diminu6s des b6n6fices
d'exploitation pr6vus Equivalents au cout du capital de I'entreprise multipli6 par les act-
ifs nets d'exploitation diff6r6s). Les auteurs font des impots report6s passifs nets une
cat6gorie distincte d'actifs (n6gatifs) dans leur modfele d'6valuation et estiment l'inci-
dence de chaque cat6gorie diff6rente d'actifs sur la valeur marchande de l'avoir des
actionnaires.
L'6chantillon choisi par les auteurs englobe toutes les entreprises faisant partie des
500 soci6t6s ouvertes de Fortune (a l'exclusion des institutions financieres et des
soci6t6s de services publics) et figurant dans la base de donn6es Compact Disclosure et
les fichiers industriels de Compustat. Les auteurs ont recueilli des donn6es sur les dif-
f6rents 616ments d'impot report6 de chaque entreprise pour les ann6es 1992, 1993 et
1994, sous r6serve de l'adoption par I'entreprise du SFAS n° 109. Leur 6chantillon
d6finitif comporte 243 observations pour 1992, 459 observations pour 1993 et 412
observations pour 1994, soit un total de 1 114 entreprises-ann6es.
Les auteurs constatent qu'un dollar d'actif d'exploitation net est, en moyenne, 6val-
u6 a un dollar, qu'un dollar d'actif financier net apporte a la valeur de I'entreprise moins
d'un dollar, et qu'un dollar d'impots report6s nets est 6valu6 h un peu plus d'un dollar.
Selon eux, revaluation h moins d'un dollar relative a l'actif financier net serait
attribuable k la correlation entre l'endettement net et la valeur actualis6e nette des b6n6-
fices d'exploitation anormaux futurs prevus. Une diminution dans les actifs financiers
nets (c'est-a-dire une augmentation dans l'endettement net) peut indiquer des ben6fices
anormaux futurs plus faibles en raison de la non-rentabilit6 de Sexploitation. En con-
s6quence, une mesure erron6e de la valeur actualis6e nette des b6n6fices d'exploitation
anormaux futurs peut donner lieu h un coefficient d'6valuation des actifs financiers nets
inf6rieur h un.
Les auteurs 6tudient de plus prfes la valeur et la pertinence des 616ments d'impot
report6, qu'ils classent en sept categories : 1) l'amortissement; 2) les pertes, les cr6dits
et les impots minimums de remplacement faisant l'objet d'un report prospectif ; 3) les
charges de restructuration ; 4) les charges Ii6es h l'environnement ; 5) les avantages
sociaux consentis aux employ6s, y compris la r6mun6ration diff6r6e, les regimes de
retraite et les actifs fiscaux resultant de l'adoption des SFAS n°^ 106 et 112 (comptabil-
isation par les employeurs des avantages post6rieurs au depart h la retraite et avantages
postedeurs a l'emploi) ; 6) la provision pour evaluation d'actif exig6e par le SFAS
The Valuation of Deferred Taxes 599

n° 109 et 7) tous les autres elements. Les auteurs estiment que le partage des impots
report6s en ses 616nients foumit de 1'information pertinente h revaluation de l'entre-
prise. Le coefficient d'evaluation appliqu6 aux impots report6s passifs decoulant de
l'amortissement, notamment, se rapproche de z6ro. Ce r6suUat reflate le fait que les
investisseurs s'attendent a ce que l'entreprise continue d'investir dans des actifs amor-
tissables, rdduisant ainsi la probabilit6 de resorption future. Les investisseurs croient
done que ces passifs sont sur6valu6s en raison du fait que la valeur des r6sorptions
futures n'est pas actuaiisee, de sorte que la valeur comptable de I'avoir des actionnaires
est sous-6valuee.
Les auteurs constatent 6galement que les impots reportds relatifs aux charges de
restructuration ont des coefficients d'evaluation superieurs a ceux des autres elements
d'impot reporte. Cette constatation reflate, au moins en partie, une probability plus
grande de rdsorption a court terme. Enfin, les auteurs constatent que la valeur de reali-
sation nette des impots reportes d6coulant des pertes et des credits faisant l'objet de
reports prospectifs (impots reportSs actifs provenant des reports prospectifs nets de la
provision pour Evaluation d'actif du SFAS n° 109) varie inversement au cours des
actions, ce qui donne a penser que les investisseurs, en moyenne, ne s'attendent pas &
ce que ces reports prospectifs soient utilises, bien qu'ils ne puissent Eliminer la possi-
bility qu'une erreur dans la definition du modele mfene a ce resultat.
Globalement, ces resultats, qui ne sont pas sensibles aux modifications apportees
aux caracteristiques du module, coincident avec revaluation par les investisseurs des
impots reportes, compte tenu du moment de la r^sorption de ces impots report^s. A cet
egard, ces r6sultats sont done conformes a l'affirmation du FASB selon laquelle les
impots report6s nets devraient etre comptabilis6s comme tout autre Element d'actif et de
passif du bilan.

Deferred taxes arise when current tax payable differs from recorded income tax
expense and, as such, only exist where financial reporting standards do not
coincide with tax-reporting requirements. We seek, in this study, to provide a
simple framework for understanding the role of net deferred taxes and their
components in equity valuation. We also provide evidence regarding the incre-
mental value relevance of deferred tax information. Specifically, we examine
whether the valuation of deferred tax components depends on the probability of
reversal. Using disclosures on deferred tax components, required by Statement
of Financial Accounting Standards (SEAS) No. 109, we assess the likelihood of
reversals for each component and form expectations about the direction and
magnitude of these components' valuation. We then determine the extent to
which the valuation of deferred tax components depends on the probability of
reversal. Based on the model described later, we expect that deferred taxes that
are more likely to reverse in the next period will contribute more to firm value
(in absolute value) than those deferred taxes that are less likely to reverse soon.
Finally, we assess the value relevance of the valuation allowance disclosed
under SEAS No. 109.
To examine the market valuation of deferred taxes, we employ a variant of
the Feltham and Ohlson (1995) framework. Feltham and Ohlson (F&O) derive
the market value of equity as a linear function of the current values of net oper-
ating assets, net financial assets, and abnormal operating earnings (defined as
600 Contemporary Accounting Research

actual operating earnings minus expected operating earnings equal to the firm's
cost of capital times lagged net operating assets). We introduce net deferred tax
liabilities as a distinct category of (negative) assets in the valuation equation
and estimate the effect of each different asset class on the market value of the
firm's equity.
Results of our analysis suggest that, controlling for abnormal operating
earnings, net operating assets, and net financial assets, net deferred taxes help
explain cross-sectional variation in firms' market value of equity. An examina-
tion of the value relevance of deferred tax components shows that deferred tax
assets that are related to restructuring charges, which are more likely to reverse
in the following period, have larger valuation coefficients than deferred tax
assets from environmental charges and from employee benefits, which reverse
over a longer period of time. In addition, we find that deferred tax liabilities
from depreciation and amortization have relatively small valuation coeffi-
cients, indicating that these deferred tax liabilities are, on average, overstated
(and book value of equity understated) due to the lack of discounting and the
long-term nature of their reversal. We also find that deferred tax assets related
to losses and credits carried forward and the related SFAS No. 109 valuation
allowance are not value relevant, reflecting investors' assessment of the low
likelihood of utilizing these tax assets in the future.
We proceed with a review of the relevant literature on the role of account-
ing for income taxes in valuation. Then we describe the valuation model
employed in the analysis, after which we discuss the sample selection and data
collection procedures. Next we analyze our results and finally we offer some
concluding remarks.

Background and literature review


Financial statement users often disagree as to the most appropriate method for
valuing a firm that has deferred tax assets and liabilities on its balance sheet, l
Some claim that net deferred taxes represent obligations to pay taxes in the
future, and hence, should be regarded as financial liabilities. As such , these lia-
bilities should be offset against the firm's other long-term net financial assets.
Proponents of this method often argue that if the temporary differences, which
gave rise to the deferred tax liabilities, are not expected to reverse (settle) in the
near future, these liabilities should be discounted similar to other long-term
financial obligations, taking into account the expected time to achieve reversal
and the cost of borrowing.2
Others argue that many deferred tax liabilities (e.g., deferred taxes result-
ing from depreciation and temporary amortization differences) are never set-
tled; hence, net deferred tax liabilities should be added to (and net deferred tax
assets should be subtracted from) the firm's book value of shareholders' equi-
ty. Consistent with this approach. Statement of Standard Accounting Practice
(SSAP) No. 15 issued by the Accounting Standards Committee in the United
Kingdom (ASC 1985) requires companies to adopt a partial interperiod tax
The Valuation of Deferred Taxes 601

allocation method, that is, to recognize only those deferred taxes that are
expected to materialize in the foreseeable future (3-5 years). This partial recog-
nition effectively regards long-term temporary differences as part of equity.
Most financial statement users, however, appear to respond to the complexity
of the issue by ignoring deferred taxes altogether.^
In 1992, the Financial Accounting Standards Board (FASB) issued SFAS
No. 109, which modified the accounting for income taxes (FASB 1992). Three
major differences between SFAS No. 109 and the preceding accounting rule
under Accounting Principles Board (APB) Opinion No. 11 (APB 1967) are rel-
evant to this study. First, measurement of deferred taxes under SFAS No. 109 is
based on the applicable tax rate, which is the tax rate that is expected to apply
at the time the asset or liability is expected to be realized (liability method).
This requirement means that the effect of any tax rate changes is recognized at
the time the new tax law is enacted. Second, SFAS No. 109 is more liberal than
the preceding rules (APB Opinion No. 11 and SFAS No. 96) in the recognition
of deferred tax assets. All deductible differences, tax losses, and credits must
give rise to a deferred tax asset, which is reduced by a valuation allowance
when necessary. Finally, instead of disclosing the main components of deferred
income tax expenses (income statement approach), financial disclosures must
include the main components of the net deferred tax balance (a balance sheet
approach).
Interperiod tax allocation has been addressed in previous empirical studies.
Beaver and Dukes (1972) find that unexpected stock returns are more highly
correlated with unexpected earnings measures that include tax deferrals than
with unexpected earnings measures that do not. They conclude that the infor-
mation used to set stock prices includes eamings that are based on interperiod
tax allocation accounting. These findings have recently been confirmed by
Chaney and letter (1992), who find a negative association between deferred tax
expenses and stock returns using 1982-83 data. In addition, they decompose
deferred tax expenses into recurring and nonrecurring items and find that recur-
ring items are valued more negatively than nonrecurring items. They conclude
that this decomposition provides incremental information to investors.
In an attempt to assess whether investors view the deferred tax liability as
a "real" liability, Givoly and Hayn (1992) analyze the relation between firm
characteristics and unexpected stock returns around events related to the 1986
Tax Reform Act (TRA). They find a positive association between stock retums
around the TRA and the reduction in the deferred tax liability implied by the
change in the tax rate. They conclude that investors view the deferred tax lia-
bility as a real liability and discount it according to the likelihood of the liabil-
ity's settlement.
We believe, however, that limitations in their data and methodology leave
the conclusions of Givoly and Hayn 1992 open to further investigation. For
instance, Givoly and Hayn use pve-SFAS No. 109 COMPUSTAT disclosures
that are incomplete. In many cases, COMPUSTAT fails to record short-term
602 Contemporary Accounting Research

deferred tax assets, introducing bias into the deferred tax figures. In addition,
the prevailing accounting rules under APB Opinion No. 11 were stricter than
those under SEAS No. 109, and prevented firtns frotn recognizing deferred tax
assets from credits carried forward and from certain deductibles. These restric-
tions may introduce another type of bias into the measures of deferred tax lia-
bilities. Moreover, Givoly and Hayn use indirect measures of the likelihood of
deferred tax liability settlement (the probability of future losses). Finally, they
measure unexpected returns over 252 trading days from September 1984 to
September 1986, about half of the trading days during this two-year period. A
return model over such a long window requires a control for pre-tax unexpect-
ed earnings over that period. If more profitable firms have, on average, larger
deferred tax liabilities, then Givoly and Hayn's results may be explained by the
lack of a proper control for unexpected earnings.
Ohlson and Penman (1992) measure the association between returns and
accounting information over increasingly long windows of one to ten years.
They also disaggregate income and book value of equity into various compo-
nents, including tax expense and deferred tax liabilities, respectively. They find
that the regression coefficients on the tax components are consistently smaller
than the coefficients on other income and balance sheet components. They
interpret (on page 570) this result as "consistent with the notion that the
accounting measurement of deferred tax liabilities is inherently more complex
than the measurement of other assets and liabilities."
Our study contributes to the literature in several ways. In our analysis, we
use post-SEAS No. 109 disclosures for the period 1992-94. These disclosures
include the main components of the deferred tax liability and are more com-
prehensive than the preceding disclosures under APB Opinion No. 11. In addi-
tion, we categorize deferred tax components according to the likelihood of set-
tlement and examine directly the valuation differences between reversing and
nonreversing deferred taxes. We also examine the value relevance of the
deferred tax valuation allowance required under SEAS No. 109.
Our study focuses on the relation between the net deferred tax liability (and
its components) and the market value of equity (a price-level model). Unlike
Beaver and Dukes (1972), Chaney and Jeter (1992), and Ohlson and Penman
(1992), who use return earnings association analysis, a price-level model com-
bines previous (expected) and current (unexpected) valuation effects, and
hence, ignores the timeliness of deferred tax information. Nevertheless, we
believe that a price-level model is more suitable for an investigation of the
deferred tax components for the following reasons. First, many of the deferred
tax components can sometimes be anticipated.'* A price-level model does not
require a specification of the unexpected change in deferred taxes. Because net
deferred taxes include many different components resulting from many differ-
ent transactions, a return earnings (flows) model would require a specification
of an expectation model for each component. Second, return earnings associa-
tions do not explicitly consider balance sheet information, whereas a price-
The Valuation of Deferred Taxes 603

level model allows us to examine the valuation of different types of balance


sheet assets and liabilities, including net deferred taxes and their components.5

Model development
F&O (1995) present a model in which the market value of equity equals the
recorded book value of shareholders' equity (the sum of net operating assets,
net financial assets, and net deferred taxes) plus any unrecorded goodwill. They
show that, under the assumption of "clean surplus accounting" (i.e., the change
in book value of shareholders' equity is equal to net income minus dividends),
unrecorded goodwill is equal to the present value of expected future abnormal
earnings — defined as actual earnings minus lagged book value of sharehold-
ers' equity times the firm's cost of capital. They further argue that if only net
operating assets generate abnormal earnings, unrecorded goodwill must equal
the present value of expected abnormal operating earnings, defined as actual
operating income minus the cost of capital times the beginning-of-period net
operating assets.
Net deferred taxes play an important role in the calculation of unrecorded
goodwill. Even though net deferred taxes may not generate explicit abnormal
eamings, their presence may alter the goodwill calculation, with the adjustment
depending crucially on whether and when the deferred taxes reverse (settle). In
the extreme case where deferred tax liabilities (assets) never reverse, these
deferred taxes should be added to shareholders' equity (written off). In the
other extreme, where deferred taxes settle in the next period, operating assets
(and therefore expected operating eamings) must be adjusted to reflect the pre-
sent value of the deferral. In most other cases, the classification of deferred tax
components to either operating assets, financial assets, or shareholders' equity
may alter investors' assessment of future abnormal operating eamings.
Consistent with F&O (1995), we relate equity value (P,) to current abnor-
mal operating earnings {AE, ) , net operating assets (NOAf ), net financial
assets (NFAf) and net deferred taxes {DT,), obtaining tbe following valuation
equation:

P, = 70 + JyNOA, + y2NFA, + y:^E, + 74 Dr, + e, (1)

The coefficient on current abnormal operating income (73) depends pri-


marily on the persistence of abnormal operating eamings over time. No persis-
tence implies a 73 of zero, whereas full persistence implies a coefficient equal
to one over the weighted average cost of capital used to value the firm. In the
case of unbiased accounting, the coefficients on both net operating and net
financial assets should equal one. However, conservative accounting implies a
coefficient larger than one for operating and financial assets. If accounting for
operating assets is more conservative than that of financial assets, we would
expect 7i to be larger than 72 .The effect of conservative accounting on valua-
tion is exacerbated by an expected growth in operating and financial assets.
604 Contemporary Accounting Research

F&O do not consider conservatism in accounting for financial assets, because


these assets are assumed to be marked to market and to earn zero abnormal
earnings.6
Consequently, growth in net operating assets is expected to be larger than
growth in net financial assets, implying that the coefficient on net operating
assets will be even larger compared with the coefficient on net financial assets.
Finally, the coefficient on deferred taxes (74) depends on the timing and likeli-
hood of settlement. The more that deferred taxes are expected to reverse in the
next period, the larger the valuation coefficient of this variable. If net deferred
taxes are valued the way net financial assets are, the coefficient on DT, would
be one. However, if deferred taxes are valued in the manner of net operating
assets, we would expect their coefficients {y^ and 74, respectively) to be equal.
We estimate equation (1) using cross-sectional data for 1992-94. However,
cross-sectional estimation of equation (1) may result in biased estimates for
either of the following two reasons. First, the valuation coefficients on net
operating assets {NOA,) and current abnormal operating earnings {AE,) depend
on the firm's cost of capital. Because each firm may have a different cost of
capital, measurement error may be introduced resulting in biased estimates.
This problem seems less severe for net financial assets (NEA,), because the
coefficient on these assets is predicted to be one for all firms. Second, the val-
uation of current abnormal operating income depends on the persistence of
these eamings over time. Again, to the extent that firms differ in terms of
abnormal earnings persistence, and those differences are correlated with
observed variables, cross-sectional estimation of equation (1) may result in
biased coefficients.
To address these problems, we modify equation (1) in the following ways.
First, we add lagged abnormal operating eamings (LAE,) to the model.
Including this variable may capture cross-sectional variation in earnings per-
sistence. Second, we allow each coefficient in the model, except the one on
NEA to vary by industry.7 This will allow us to control for systematic differ-
ences in cost of capital and abnormal earnings persistence across industries.
Third, to reduce any size effects, we deflate all variables by the number of com-
mon shares. Finally, to check that our results are not artifacts of intertemporal
differences, we also estimate equation (1) separately for each year t (t = 1992,
1993, 1994).
Consequently, we obtain the following empirical model:

e, (2)
The Valuation of Deferred Taxes 605

where Ij is an indicator variable that equals one if the observation belongs to


industry 7 (/ = 1, 2 ... 6, according to Sharpe's industry classification without
financial institutions and utilities) and firm subscripts, i, are understood. Similar
to the other types of asset classes in the model, we coded deferred tax assets as
positive numbers and deferred tax liabilities as negative numbers. Consequently,
we expect all coefficients in equation (2) to be positive. The foregoing discus-
sion also leads us to predict that aj is, on average, larger than a2.
To examine the effect of expected reversal on the valuation of deferred
taxes, we classified deferred tax components into seven categories (a compre-
hensive list of the components is provided in the Appendix): (1) depreciation
and amortization; (2) losses, credits, and alternative minimum taxes carried for-
ward; (3) restructuring charges; (4) environmental charges; (5) employee ben-
efits, including deferred compensation, pensions, and tax assets resulting from
the adoption of SFAS Nos. 106 and 112; (6) valuation allowance required by
SFAS No. 109; and (7) all other components. We then estimate the following
equation:

i + 0L2tNFA, +

0)

where DTCj^f represents the kxh component of deferred taxes in year t (k- 1,2
... 7; f = 1992, 1993, 1994), and S^t represents the regression coefficient on the
kth deferred tax component in year t. As in equation (2), firm subscripts are
understood. We restricted the valuation coefficients on all deferred tax compo-
nents to be equal across industries. This restriction, as Table 4 will show, is not
rejected by the data.
The valuation coefficient of each component depends on the likelihood and
the expected time to reversal. We consider timing differences from depreciation
and amortization as having a relatively low probability of reversing, because
under the going-concern assumption, firms are likely to continue to invest in
new depreciable assets, replacing old assets. In addition, the expected settle-
ment period for deferred taxes from depreciation and amortization depends on
the useful life of the depreciable assets. Because many long-term assets (e.g.,
property, intangibles) depreciate over a long period of time, we predict the val-
uation coefficient on depreciation and amortization timing differences to be
close to zero. As for losses carried forweu^d, legal restrictions prevent the firms
from delaying the settlement of these deferred tax assets, hence, these tax loss-
es may expire. Moreover, because most firms cannot sustain losses for a long
period of time, deferred tax assets from losses are not expected to persist.
606 Contemporary Accounting Research

Consequently, the valuation coefficient on losses is also expected to be rela-


tively small.
Deferred taxes from restructuring charges are likely to reverse over a rela-
tively short period of time as the firm completes the restmcturing plan. On the
other hand, deferred taxes from environmental liabilities are expected to settle
over a longer period of time, which suggests that the valuation coefficient on
deferred taxes from restructuring charges should be larger than the one on
deferred taxes from environmental liabilities. However, the valuation coeffi-
cients on these deferred tax components also depend on whether the firm
accrues an unbiased estimate of environmental liabilities and restructuring
charges. If, for example, the recognized environmental liability is understated,
investors would expect environmental charges to recur, which, in turn, would
raise the valuation coefficient on deferred taxes from environmental charges.8
Moreover, if environmental charges are more likely to recur than restructuring
charges, our prediction regarding the relation between their valuation coeffi-
cients may weaken or even reverse.
Deferred taxes from employee benefits are primarily due to nonrecurring
pensions and other post-retirement benefit liabilities (e.g., the catch-up charge
reflecting the adoption of SFAS Nos. 106 and 112). These charges are also
expected to reverse, however, reversals are expected to occur over a long peri-
od of time, equal to the average remaining life of employees and retirees.
Consequently, the valuation coefficient on this item should also be smaller than
the one on restructuring charges.
Finally, SFAS No. 109 (paragraph 17) requires that if, on the basis of avail-
able evidence, "it is more likely than not" that all or a portion of tax assets will
not be realized in the future, the tax assets must be written down by creating a
valuation allowance. Many criticize the valuation allowance as being subjec-
tively determined and difficult to verify (e.g., Deloitte & Touche 1992). Some
even argue that the valuation allowance is used to manage reported income. To
the extent that firms' information on future realizations of tax assets is not ver-
ifiable, the valuation allowance provides a noisy signal of the asset value;
hence, the valuation coefficient on the valuation allowance component of
deferred taxes is expected to be close to zero.

Data and variables


Our sample includes all public Fortune 500 companies, available on the
Compact Disclosure data base and covered by the COMPUSTAT industrial
files. Given that deferred tax information must be collected from the financial
statements, we had to restrict our sample to a manageable size. We decided to
concentrate on Fortune 500 companies because these firms are more likely to
have significant deferred taxes. We collected data on the separate components
of each company's deferred taxes for the years 1992, 1993, and 1994, provid-
ing the company had adopted SFAS No. 109. We decided to concentrate on the
The Valuation of Deferred Taxes 607

period 1992-94 because deferred tax components were available only for firms
that adopted SEAS No. 109, which became effective for fiscal years starting
after December 15, 1992.
We exclude financial institutions (single-digit Standard Industrial
Classification (SIC) code 6) and electric utilities (two-digit SIC code 49) from
our sample. Financial institutions were excluded because of the difficulty of
separating financial assets from operating assets. Electric utilities were exclud-
ed because of the different structure of their financial statements, especially the
effect of regulatory accounting on the classification and measurement of bal-
ance sheet items.
We obtained deferred tax data on 1,336 firm year observations over fiscals
1992-94. We deleted 215 observations with missing price or other accounting
data, resulting in a sample of 1,121 firm year observations. To minimize the
effect of outliers on our inferences, we deleted 7 observations (0.6 percent) for
which the absolute value of the /^-Student (regression residual divided by the
residuals' standard error) statistic was greater than 3. Our final sample includes
243 observations for 1992, 459 observations for 1993, and 412 observations for
1994, a total of 1,114 firm years. Table 1 presents information on the sample
selection procedure and on the sample's distribution by year and industry. We
use an industry classification similar to the one used in Sharpe (1982).
We use share price at fiscal year-end (COMPUSTAT item 199) to calculate
the market value of equity per share, the dependent variable in our regressions.
Net financial assets per share {NEA) were calculated as cash and cash equiva-
lents (item 1) plus short-term investments (item 193), minus long-term debt
(item 9), current portion of long-term debt (item 34), and preferred stock (item
130), all divided by the number of shares outstanding at fiscal year-end (item
25). Net operating assets per share {NOA) were calculated as book value of
shareholders' equity (item 60) minus net financial assets plus net deferred tax
liabilities, and divided by the number of shares outstanding. We coded all assets
as positive numbers and all liabilities as negative numbers.
Operating income is calculated as primary earnings per share before extra-
ordinary items (item 58) plus the tax-adjusted interest expense per share (item
15) divided by the average number of shares used to calculate earnings per
share (item 54), and minus the tax-adjusted interest income and other nonoper-
ating income (items 190 and 62, respectively) per share. All tax adjustments are
made by multiplying by: one minus the statutory federal tax rate (34 percent for
1992, and 35 percent thereafter).^ Abnormal after-tax operating income in year
t is calculated as operating income in year t minus expected normal earnings
(cost of capital times last period's net operating assets). Similar to Penman
(1996), we use 10 percent as a proxy for the expected rate of return on operat-
ing assets. 10
608 Contemporary Accounting Research

TABLE 1
Sample selection*

Panel A: Pooled sample


Observations witb complete deferred tax data 1,336
Price or accounting data are not available on COMPUSTAT (215)
Observations with I /f-Student'l > 3 (7)
Total observations 1,114

Panel B: Sample by year


1992 observations 243
1993 observations 459
1994 observations 412
Total observations 1,114

Panel C: Sample by industry classificationst


(1) Basic industries 163
(2) Capital goods 204
(3) Construction 42
(4) Consumer goods 565
(5) Energy 78
(6) Transportation 62
Total observations 1,114

Notes:
* Fortune 500 firms, excluding financial institutions (single-digit SIC code = 6) and public
utilities (double-digit SIC code = 49), that adopted SFAS No. 109 during 1992-94.
t Industry classifications are based on Sharpe (1982). The following four-digit SIC codes are
assigned to each group:
(a) Basic Industries: 1000-1299, 1400-1499, 2600-2699, 2800-2829, 2870-2899,
3300-3399;
(b) Capital Goods: 3400-3419, 3440-3599, 3670-3699, 3800-3849, 5080-5089, 5100-5129,
7300-7399;
(c) Construction: 1500-1599, 2400-2499, 3220-3299, 3430-3439, 5160-5219;
(d) Consumer Goods: 0000-0999, 2000-2399, 2500-2599, 2700-2799, 2830-2869,
3000-3219, 3420-3429, 3600-3669, 3700-3719, 3850-3879, 3880-3999, 4830-4899,
5000-5079, 5090-5099, 5130-5159, 5220-5999, 7000-7299, 7400-9999;
(e) Energy: 1300-1399, 2900-2999;
(0 Transportation: 3720-3799, 4000-4799.

Results
Table 2 presents the distribution of the regression variables. The average (medi-
an) market value of equity per share is $31.39 ($29.00). The average (median)
net operating assets (NOA) is $25.17 ($19.58), about 80 percent of the market
value of equity. Financial assets are, on average, 36.7 percent (median 24.4 per-
cent) of market value of equity per share. The negative sign indicates that most
The Valuation of Deferred Taxes 609

TABLE 2
Descriptive statistics (all in dollars per share)*

Distribution of regression variablest Mean Median Std.Err. 1-Qrt 3-Qrt


Market value of equity (P) 31.39 29.00 15.32 19.75 40.81
Net operating assets (NOA) 25.17 19.58 21.99 10.69 32.26
Abnormal operating eamings (A£) -1.19 0.02 3.90 -1.44 0.82
Lagged abnormal oper. eamings (LAE) -1.91 -0.14 4.67 -2.04 0.64
Net financial assets (ATM) -11.51 -7.07 16.17 -16.03 -1.92
Net deferred taxes ( o r ) -0.37 -0.06 3.57 -1.30 0.61

Notes:
* Fortune 500 firms (excluding financial institutions and public utilities) with complete
COMPUSTAT data over 1992-94 and complete SFAS No. 109 disclosures.
t Market value of equity per share (P) is share price at fiscal year end. Net financial assets
(NFA) are calculated as cash and cash equivalents plus short-term investments, minus short-
term debt, long-term debt, and preferred stock. Deferred taxes (DT) are based on information
collected from financial statements (income tax footnote). Net operating assets (NOA) are
calculated as book value of equity minus net financial assets plus net deferred tax liabilities.
Operating income is defined as earnings per share net of interest expense, interest income,
and other nonoperating income, all tax adjusted and per share. Abnormal operating income
(AE) is defined as operating income minus 10 percent of lagged net operating assets. Lagged
abnormal operating income (LAE) is similarly defined and is equal to AE,_,. All variables are
per share and adjusted for stock splits and stock dividends.

firms use tax deductible debt to finance operations. Net deferred tax assets are,
on average, about 1 percent (median 0.2 percent) of the market value of equity
per share. The negative sign indicates that more firms have net deferred tax lia-
bilities, rather than net deferred tax assets. Finally, the median abnormal oper-
ating income {AE) and lagged abnormal operating income {LAE) are close to
zero. However, the negative mean of these variables indicates that their distri-
bution is skewed to the left.

Table 3 presents the variables' correlation matrix (Pearson correlations are


presented above the diagonal, and Spearman correlations are presented below
the diagonal). As expected, positive correlations exist between market value of
equity and net operating assets (Pearson = 0.36; Spearman = 0.33), and
between market value of equity and abnormal operating income (Pearson =
0.13, Spearman = 0.27). The correlations between net financial assets and mar-
ket value of equity are negative (Pearson = -0.16, Spearman ,= -0.08), consis-
tent with a lower expected return on financial assets relative to operating assets.
Finally, the correlations between market value of equity and net deferred taxes
are close to zero.
High negative correlations exist between net operating assets and all the
other explanatory variables. In particular, the correlations between net operat-
ing assets and net financial assets are a Pearson of -0.88 and a Spearman of
610 Contemporary Accounting Research

TABLE 3
Variables' correlation matrix*
Pearson (Spearman) correlations are above (below) the diagonal t

Variable P NOA AE LAE NFA DT


1> 1.00 0.36 0.13 0.10 -0.16 -0.03
NOA 0.33 1.00 -0.38 -0.28 -0.88 -0.32
AE 0.27 -0.45 1.00 0.65 0.43 0.02
LAE 0.25 -0.36 0.66 1.00 0.35 -0.10
NFA -0.08 -0.83 0.51 0.44 1.00 0.12
DT -0.00 -0.37 0.17 0.05 0.26 1.00
Notes:
* See Table 2 for variable definitions
t Correlations with absolute values above 0.10 are significant at the 0.01 level.

-0.83. However, current and lagged abnormal operating income are positively
correlated (Pearson = 0.65, Spearman = 0.66). These high correlations suggest
a multicollinearity problem in the estimation procedures.

Table 4 presents estimation results of equation (2) for a pooled sample of


1,114 firm year observations and for three yearly samples (1992, 1993, and
1994). Panel A of Table 4 presents results for a restricted version of equation
(2) — without industry controls. Panel B contains results for equation (2) with
industry controls. We therefore report the average coefficients and average
White (1980) f-statistics across the seven industry classifications. We also
report results of F-tests (and their corresponding p-values) that all coefficients
for a certain variable are equal across industries.
Results in panel A indicate that net operating assets, net financial assets,
current abnormal operating income, and net deferred taxes are significantly
larger than zero (at the 0.005 level) in the pooled model and in the three year-
ly models (the 1992 AE is significant only at the 0.025 level)." Lagged abnor-
mal operating earnings obtain a positive coefficient. However, this variable is
significantly larger than zero only at the 0.05 level in the pooled model
(/ = 1.85), and is not significant at the yearly models. We also find that — after
controlling for current and lagged abnormal income, net financial assets, and
net deferred taxes — a dollar of net operating assets is valued at $0.89 (rang-
ing between $0.81 and $0.93 in the yearly models), significantly less than one
at the 0.01 level. The marginal value of a dollar of net financial assets in the
pooled model is $0.79 (ranging between $0.70 and $0.80 in the yearly models),
also significantly less than one at the 0.01 level. Finally, the marginal value of
a dollar of net deferred taxes is $1.20 (ranging between $1.01 and $1.30 in the
yearly models), and is significantly above one at the 0.05 level.
The Valuation of Deferred Taxes 611

TABLE 4
Regressions of share price (P) on net operating assets per share {NOA), net financial
assets per share {NFA), current and lagged abnormal operating earnings per share {AE,
LAE), and deferred taxes per share {DT)*

Panel A: OLS regressions (industry controls are not included)t


Year INT. NOA NFA AE LAE DT R^ #Obs.
Pooled 19.75 0.89 0.79 0.83 0.20 1.20 0.36 1,114
t-stat.t 30.66 22.26 14.75 6.39 1.85 10.15

1992 18.90 0.81 0.70 0.59 0.35 1.24 0.32 243


t-stat. 14.28 9.27 5.95 2.02 1.21 5.00

1993 20.90 0.90 0.78 0.71 0.21 1.30 0.35 459


t-stat. 20.66 13.95 9.16 3.61 1.26 6.77

1994 18.46 0.93 0.80 1.46 0.20 1.01 0.40 412


t-stat. 17.01 14.84 9.75 5.76 1.26 5.29

Panel B: Regressions with industry intercepts and slopes


Year INT. NOA NFA AE LAE DT R^ #Obs.
Pooled 16.83 1.01 0.80 1.74 0.35 1.12 0.39 1,114
t-stat 8.54 13.05 13.92 2.74 0.77 3.58
F-stat. 6.76 3.82 — 2.31 3.26 1.07
P(E) 0.00 0.00 — 0.04 0.01 0.38

1992 14.82 1.03 0.85 0.18 1.95 1.55 0.36 243


t-stat 3.24 5.18 6.20 0.19 1.30 2.76
F-stat 1.92 0.89 — 1.32 1.62 0.52
P(F) 0.09 0.49 — 0.25 0.16 0.76

1993 14.79 1.17 0.89 2.60 0.10 1.42 0.43 459


t-stat. 5.67 9.10 9.81 2.38 0.35 2.72
F-stat. 3.66 2.70 — 6.71 4.50 1.45
P(F) 0.00 0.02 — 0.00 0.00 0.20

1994 14.92 1.02 0.78 3.17 0.31 0.78 0.46 412


t-stat. 3.25 8.08 9.00 2.26 0.82 1.90
F-stat. 1.58 1.55 — 8.59 2.51 0.82
P(F) 0.16 0.17 — 0.00 0.03 0.54
Notes:
* See Table 2 for variable definitions and Table 1 for sample selection and industry
classifications.
t Panel A contains ordinary least squares (OLS) regressions without industry controls. Panel B
contains industry intercepts and slopes for all variables except net financial assets {NFA). In
panel B, we report average coefficients and average /-statistics across industries. We also
report the F-value and p-value of a test for a specific coefficient is equal across industries.
612 Contemporary Accounting Research

t Using one-tailed tests, /-statistics above 1.3 are significant at the 0.10 level, r-statistics above
1.7 are significant at the 0.05 level, r-statistics above 2.4 are significant at the 0.01 level, and
r-statistics above 2.6 are significant at the 0.005 level. White (1980) corrected t statistics are
presented below the coefficients.

We conducted F-tests to examine whether different classes of assets are


valued differently. Consistent with our prediction, these tests indicate that net
operating assets are valued more than net financial assets (F-statistics are
16.46, 3.86, 8.31, 8.34 for the pooled, 1992, 1993, and 1994 samples, respec-
tively; significant at the 0.05 level or better). We also find that net deferred
taxes are valued more than net financial assets (F-statistics are 13.39, 5.05,
8.03, and 4.18 for the same periods, respectively; significant at the 0.02 level
or better). Finally, we find that net deferred taxes are valued more than net
operating assets in the pooled model and in the 1993 sample (F-statistics are
8.16 and 5.20, respectively; significant at the 0.02 level or better). However,
the difference between deferred taxes and net operating assets is not significant
at the 0.05 level in the 1992 and 1994 samples (F-statistics are 3.60 and 0.19
with p-values of 0.07 and 0.66, respectively).
Results in panel B of Table 4 (controlling for industry in all variables
except net financial assets) are slightly different than those reported iti panel A.
First, the valuation coefficient of net operating assets is 1.01 in the pooled
model (ranging from 1.02 in 1994 to 1.17 in 1993). This coefficient varies sig-
nificantly by industry, as indicated by the F-test presented below the coefficient
(F = 3.82, p-value = 0.00). The valuation coefficient of net financial assets is
virtually unaffected by the inclusion of industry controls (0.80, t = 13.92). We
repeated our tests allowing the coefficient on net financial assets to vary by
industry (not reported). We could not reject that the valuation of net financial
assets is equal across industries (F-statistic = 0.88; p-value = 0.49).
The valuation coefficient on abnormal operating income (1.74, t = 2.1 A) is
larger relative to the one presented in panel A. This coefficient also varies by
year. It is 0.18 in 1992, 2.60 in 1993, and 3.17 in 1994. The coefficient on
lagged abnormal operating income is not significant at the 0.10 level. The val-
uation coefficient on deferred taxes is 1.12 (/ = 3.58). This coefficient varies by
year (1-55, 1.42, and 0.78 in 1992, 1993, and 1994, respectively). However, the
coefficient does not vary significantly by industry, as indicated by the corre-
sponding F-test.
Table 5 provides descriptive statistics (panel A) and a correlation matrix
(panel B) of the deferred tax components used in estimating equation (3). The
net deferred tax liability is, on average, $0.37 per share (1.5 percent of net
operating assets). There is, however, much variation in the magnitude of the
various components. Deferred tax liabilities resulting from depreciation and
amortization of long-term assets are, on average, $2.34 per share (9.2 percent
of net operating assets), and almost all firms with deferred taxes (1,081 out of
1,114 firm years) report this item separately. Six hundred and eighty-nine of
1,114 observations (62 percent) report deferred tax assets from losses, credits,
The Valuation of Deferred Taxes 613

TABLE 5
Analysis of deferred tax components*

Panel A: Descriptive statistics


Components (per share)t Mean Median Std.err. 1-qrt 3-qrt #5^0

Depreciation & amort. -2.34 -1.07 3.40 -2.70 -0.35 1,081


Losses & credits 0.83 0.18 1.59 0 0.87 689
Restructuring charges 0.16 0 0.48 0 0 274
Environmental charges 0.07 0 0.30 0 0 113
Employee benefits 0.88 0.29 1.59 0 1.03 886
Valuation allowance -0.44 0 1.05 -0.33 0 525
Other deferred taxes 0.47 0.24 2.45 -0.13 0.79 1,109
Net deferred taxes -0.37 -0.06 3.57 -1.30 0.61 1,114

Panel B: Pearson (Spearman) correlations above (below) the diagonal*


Components Depr.& Losses& Restruc- Environ- Employee Valuation Other
(per share) t amort. credits turing mental benefits allowance
Depreciation & amort. 1.00 -0.44 -0.09 -0.17 -0.33 0.11 -0.33
Losses & credits -0.43 1.00 0.23 0.05 0.41 -0.62 0.11
Restructuring charges -0.02 0.12 1.00 0.02 0.30 -0.21 0.16
Environmental charges -0.21 0.12 0.01 1.00 0.10 -0.05 0.06
Employee benefits -0.37 0.24 0.14 0.13 1.00 -0.30 0.29
Valuation allowance 0.11 -0.57 -0.23 -0.08 -0.25 1.00 -0.18
Other deferred taxes -0.04 0.05 -0.04 -0.02 0.08 -0.23 1.00
Notes:
* See Table 1 for sample selection.
t Total deferred taxes are partitioned into seven components: deferred taxes related to
depreciation and amortizations; tax losses carded forward, including alternative minimum tax
(credits and losses) carried forward; deferred tax assets from restructuring charges; tax assets
from environmental charges; employee benefits, including deferred compensation, pensions,
and other post-employment benefits; valuation allowance as required by SFAS No. 109; and
other deferred taxes, defined as total deferred taxes minus the six identified components.
* Correlations with absolute values above 0.10 are generally significant at the 0.01 level.

or alternative minimum tax carried forward (mean $0.83 per share, median
$0.18 per share). The number of observations witb deferred tax assets related
to restructuring (mean $0.16 per share) and environmental charges (mean $0.07
per share) are 274 and 113, respectively. The number of firm years with
deferred tax assets related to employee benefits and deferred compensation
(mean $0.88 per share, median $0.29 per share) is 886. Five hundred and twen-
ty-five firm years (47.1 percent) report valuation allowances, reducing tax
assets by $0.44 a share on average. Because many of the valuation allowances
are related to losses carried forward, firms estimate that, on average, half of
their losses will not be utilized (0.44 divided by 0.83). All other components
614 Contemporary Accounting Research

are $0.47 per share (1.9 percent of net operating assets). Overall, deferred tax
asset components are, on average, 9.6 percent of net operating assets, whereas
depreciation and amortizations create, on average, a deferred tax liability of 9.2
percent of net operating assets.
Panel B of Table 5 presents the components' correlation matrix (Pearson
above the diagonal, and Spearman rank correlations below the diagonal). As
would be expected under SEAS No. 109, high negative correlations exist
between deferred tax assets from credits and losses carried forward and the val-
uation allowance (Pearson = -0.62; Spearman = -0.57). These high correlations
suggest that many firms believe that all or a portion of their tax losses will not
be utilized against future taxable profits during the time allowed by law. Also,
deferred tax assets from restructuring, environmental, and employee benefits
are also negatively correlated with the valuation allowance. Consistent with the
requirement of SEAS No. 109, the correlation between deferred tax liabilities
from depreciation and amortization and the valuation allowance is relatively
small. Positive correlations exist between deferred taxes from restructuring
charges and employee benefits and deferred taxes from losses; the higher the
restructuring and employee benefit charges, the higher the probability of loss-
es carried forward. Finally, a positive correlation exists between tax assets from
employee benefits and tax assets from restructuring charges (Pearson = 0.30,
Spearman = 0.14), perhaps because many firms took a big restructuring charge
in the same fiscal year they adopted SEAS No. 106.
Regression results for equation (3) are shown in Table 6. We report results
for a pooled model and for the 1992, 1993, and 1994 samples. For reasons dis-
cussed subsequently, we also report regression results for 1992, using share
prices at the end of the first quarter of 1993 as the dependent variable.
Furthermore, the regression intercepts, the coefficients on net operating assets,
and the coefficients on current and lagged abnormal operating earnings repre-
sent the average across industries. For these variables, we report the results of
an F-test that all coefficients are equal across industries. Given that net
deferred taxes do not vary by industry (as reported in Table 4, panel B), we
excluded industry controls from the deferred tax components.
Consistent with the results reported in Table 4, the coefficient on net oper-
ating assets is close to one (ranging between 0.96 in the pooled and 1.08 in the
1993 model). The coefficient on net financial assets remains smaller than one
(ranging between 0.75 in 1994 and 0.85 in 1993).12 The average coefficient on
abnormal operating earnings per share is 1.59, and it is significant at the 0.005
level. The coefficients on the deferred tax components vary by type. In all mod-
els but 1992's, we can reject the hypothesis that all deferred tax components
have the same valuation coefficient. The F-statistics and the corresponding p-
values of these tests were 12.06 (0.00) in the pooled model, and 0.81 (0.56),
6.65 (0.00), 4.59 (0.00) in the 1992, 1993, and 1994 ones, respectively.
Except for the coefficient on tax losses {DT-losses), all coefficients on the
deferred tax components (pooled model) are positive, as expected. The coeffi-
The Valuation of Deferred Taxes 615

cients on deferred taxes from depreciation and amortization (DT-depr.) are pos-
itive and significantly larger than zero (0.50; t = 2.86) at the 0.005 level.
However, results from the yearly models suggest tbat the variable's signifi-
cance is driven mainly by tbe 1992 data. The magnitude of the coefficient and
its low significance level indicate that many equity investors do not expect
deferred tax liabilities from depreciation and amortization to settle in tbe near
future. Instead, investors believe tbat, due to lack of discounting future rever-
sals and the long-term nature of these deferred tax liabilities, these liabilities
are overstated causing book value of equity to be understated.
The coefficients on deferred taxes from restructuring charges and employ-
ee benefits are positive and generally significantly larger than zero at the 0.01
level for both the pooled model and the yearly models. However, tbe magnitude
of the coefficients seems high relative to the valuation of other deferred tax
components. One possibility is that our model captures information on restruc-
turing and employee benefits that is not captured by net operating assets and
abnormal operating income. The coefficients on environmental charges are
positive. However, they are significantly larger tban zero at only the 0.10 level
and only in the pooled model. The magnitude of the environmental coefficients
is smaller than those on restructuring charges, reflecting tbe longer expected
time to settlement.
Contrary to our expectation, the coefficient on deferred taxes from losses
and credits carried forward (DT-losses) is negative in the pooled model (-0.43,
t = -1.13) and in the 1994 model (-0.98, t - -1.70); it is positive in the 1992
and 1993 ones. However, none of tbe coefficients is significant at tbe 0.05 level
(one-tailed test). This result suggests that investors do not expect these losses
and credits carried forward to be utilized. Alternatively, this result may be dri-
ven by a valuation model misspecification. If current and lagged abnormal
earnings (as calculated here) fail to capture differences between firms with and
witbout losses carried forward, and if firms witb losses carried forward are
expected to generate lower future abnormal earnings, we would expect a nega-
tive valuation coefficient.
The coefficient on the valuation allowance in the pooled model is positive,
as expected, and significantly larger than zero at tbe 0.03 level. This signifi-
cance level is driven by the 1992 data, tbe year SFAS No. 109 was first imple-
mented. Because investors had little access to these disclosures prior to their
first release during the first quarter of 1993, fiscal year-end prices may not
reflect the actual amounts disclosed. To examine tbis conjecture, we estimated
tbe 1992 regression using share prices at the end of first quarter of 1993 as the
dependent variable. Results of this regression, which are reported at the bottom
of Table 6, indicate that the coefficient on the valuation allowance is similar to
the one reported earlier, however, its significance level decreases substantially.
In addition, the coefficient on deferred taxes from environmental charges
decreases from 2.08 to 0.04. The remaining coefficients are similar in magni-
tude and significance levels.
616 Contemporary Accounting Research

cn Ov CN cn
CN

CN VO p- Ov ov
CO
d d d d d

CN VO u-i VO o cn 00
00
u-i OS
00 CN u-1 cn VO
^^ CO -^ cn d cn d

.72
.23
.76
CN O

.09
CN
VO
OS §
CN CN in
r- VO 00
-^ CN
d d d o cn
Ov
Ov
o ,^ m 1 CN u-1 VO '^ (N
OO OS oo CN O OS
1
Q g CN CN CN cn
i CN CN

o o Ov
1oo

.04
.01
00 OO CN ;^
CN Ov
CN 00 q
CN d
00
d d 'o
B

.45
Ov .97 oo
O 00 U 00
cn CO CN •3 cn '•^

rices
o
sd d
.92
cn O VO
,98

cn cn
O Ov
T d d ex o d
?
-0.

o VO VO u^ 1 oo o •s r- 00
u-i OO VO >n CN Ov 00 Ov *—;
1
sin

'-' CN
d CN CN d d d d
•a
CN
CN CN
OS no O
CN in oo
u^
Ov ^ [..»
CN w-i • ^
CO
o CN VO r -
cn
^^ 2- OS
o
.-H
00

o o CN o O o O o CN o O o —1
o cd CN O

E
c/:
Ov CN CO J^ OS ,^
o CN CO cn OS o u^ VO VO o oo cn
'—\
>n VO o o o o CN VO UI o r- CN VO
o
CN CN VO CN CN VO
'^ CN CN o o ^^ O o o o o
iriabli

1 o
.88

cn VO
VO
1 cn t--
oo OO
d cn d vd d OS d 00 > o vd
c
0.
o u-1 _ m
.04

VO o
OS Ov VO
o 00
Ov CN
00 VO
VO
o 00 VO OS
OS CN CN
c VO
o o o cn o o
o CO «-)
'"^
o o VO CO o Ov m o o 00 CN o fl. —' VO
•a

CN CN
OO
;o o o JO Ov o cn
VO en
00 § cn 00
VO §:Ov §
u-1
14.

ision
15.

VO o u-1 CN o n- CO o 2; cn o

u
1992 R
Pooledi
f-stat. t

/--stat.
/•-stat.
^-stat.
F-stat.

f-stat.
f-stat.
f-stat.

1992
1994
1993
1992

1/
The Valuation of Deferred Taxes 617

Notes:
* All firms are Fortune 500 firms (excluding financial institutions and electric utilities) covered
by the Industrial COMPUSTAT, with available deferred tax disclosures in the financial
statements.
t See Table 1 for sample selection and Table 2 for variable definitions. Net deferred taxes (DT)
are partitioned into seven components: deferred taxes due to depreciation and amortizations
(DT-depr.y, alternative minimum tax, credits, and losses carried forward (DT-losses)\ deferred
tax assets from restructuring charges (DT-restruc.y, deferred tax assets from environmental
charges (DT-envir.); deferred compensation, pensions, and other post-employment benefits
(DT-benefit); valuation allowance under SFAS No. 109 (DT-vatuat.y, and other remaining
items {DT-other).
i White (1980) corrected t statistics are presented below the coefficients. When applicable, we
also present F-statistics and p-values for tests of coefficient equality across industries. Using
one-tailed tests, f-statistics above 1.3 are significant at the 0.10 level, f-statistics above 1.7 are
significant at the 0.05 level, r-statistics above 2.4 are significant at the 0.01 level, and t-
statistics above 2.6 are significant at the 0.005 level.

The coefficients on the valuation allowance {DT-valuat.) should be inter-


preted together with the one on deferred taxes from carryforwards {DT-losses).
Firms are required to determine the valuation allowance based on the probabil-
ity of generating future taxable profits against wjliich carryforwards can be uti-
lized. That is, the valuation allowance reduces deferred tax assets from losses
and credits carried forward to their net realizable value. Because the results
indicate that investors do not assign a positive valuation coefficient to these
assets, it is difficult to interpret the coefficients on the valuation allowance. 13
To enhance our understanding of the valuation of losses carried forward,
we calculated the net realizable value of deferred tax assets from losses and
credits carried forward as the sum of DT-losses and DT-valuat., and included it
in equation (3) instead of as two separate components.l'* The coefficients on
this variable (not reported) are generally negative but not significantly differ-
ent from zero (at the 0.05 level, one-tailed test). There is no visible effect on
the valuation of the remaining variables. This result indicates that net tax loss-
es are not valued as real assets, although we cannot rule out the possibility of
correlated omitted variables in our valuation equation. We leave this question
open for further research.
Finally, looking at the pooled model, all other deferred taxes {DT-other)
are valued in a manner similar to net operating assets, that is, their valuation
coefficient is similar in magnitude to the valuation coefficient on net operating
assets.
To check whether our results are sensitive to different model specifica-
tions, we repeated the analysis using three additional models. First, instead of
using lagged abnormal operating earnings, we used abnormal operating earn-
ings in the year following year t (first lead). This specification change had little
effect on our results (not reported in a table). The average coefficient on net
operating assets is 1.11 (average r-statistic is 12.21), the coefficient on net finan-
cial assets is 0.81 {t - 12.24), and the coefficients on the deferred tax compo-
nents are similar to those reported in Table 6. However, the average coefficient
618 Contemporary Accounting Research

on lead abnormal operating income is higher than the one reported in Table 6
(4.26, t = 5.48), consistent with a lower degree of measurement error in abnor-
mal operating income.'5 We also repeated our tests after adding the second lag
of abnormal operating earnings to our model. Including the second lag has no
visible effect on our results, and its coefficient is not statistically significant.
Second, we deflated equations (2) and (3) by net operating assets (NOA).
This model explains the market-to-//OA ratio, with the explanatory variables
mentioned earlier deflated by NOA. Finally, we estimated a market-to-book
specification that includes earnings over book value of equity, and the deferred
tax components divided by book value of equity as explanatory variables for
the firm's market-to-book ratio. Book value of equity was adjusted by sub-
tracting (adding) net deferred tax assets (liabilities). Results of these specifica-
tions were similar to those reported in Tables 4 and 6.

Summary and conclusions


This study examines the market valuation of deferred taxes by using a valua-
tion model based on the theoretical framework of Feltham and Ohlson (1995).
We introduce net deferred taxes as a distinct category of assets to the valuation
equation and present the market value of equity as a function of current and
lagged abnormal operating income, net operating assets, net financial assets,
and net deferred taxes, allowing each asset category to obtain a different valu-
ation coefficient. Our empirical model also allows each valuation coefficient
(except net financial assets) to vary by industry. Tbe results of the empirical
analysis indicate that operating, financial, and deferred taxes are value relevant
in explaining the cross-sectional variation in market values of equity.
We find that a dollar of net operating assets is, on average, valued as a dol-
lar, a dollar of net financial assets contributes less than one dollar to equity
value, and that a dollar of net deferred taxes is valued slightly more than a dol-
lar. We suspect that the valuation of net financial assets below one is due to the
correlation between net borrowing and the present value of expected future
abnormal operating eamings. A decrease in net financial assets (i.e., increase in
net borrowing) may indicate lower future abnormal earnings because of unsuc-
cessful operations. Consequently, a misspecified measure of the present value
of future abnormal operating earnings may result in a valuation coefficient on
net financial assets that is lower than one.
We further investigate the value relevance of the deferred tax components.
These components, which were obtained from the financial statements of
Fortune 500 firms over 1992-94, were classified into seven categories:
deferred taxes from depreciation and amortization, losses and credits carried
forward, restructuring charges, environmental charges, employee benefits,
SFAS No. 109 valuation allowances, and other net deferred taxes. We find that
separating deferred taxes into components provides value-relevant information.
In particular, the valuation coefficient on deferred tax liabilities from depreci-
ation and amortization is close to zero. This result reflects investors' expecta-
The Valuation of Deferred Taxes 619

tions that the firm will continue to invest in depreciable assets, thus reducing
the probability of future reversal. Consequently, investors believe that these lia-
bilities are overstated because of a lack of discounting of future reversals, caus-
ing book value of equity to be understated.
We also find that deferred taxes from restructuring charges have valuation
coefficients larger than other deferred tax components. This result reflects, at
least partially, the higher likelihood of reversal in the short run. Finally, we find
that the net realizable value of deferred taxes from losses and credits carried
forward (deferred tax assets from carryforwards net of the SEAS No. 109 valu-
ation allowance) is negatively associated with stock prices. This result suggests
that investors, on average, do not expect these carryforwards to be utilized,
although we cannot rule out the possibility that model misspecification is dri-
ving this result.
Overall, these results, which are not sensitive to changes in model specifi-
cation, are consistent with investors' valuation of deferred taxes depending on
when these deferred taxes reverse. As such, these results are consistent with the
FASB's claim that net deferred taxes should be accounted for in a way similar
to any other balance sheet assets and liabilities.

Appendix
Deferred tax database: List of deferred tax items obtained from notes to tbe financial
statements
1. Alternative minimum tax
2. Credits carried forward
3. Loss provisions and losses carried forward
4. Amortizations
5. Depreciation and long-term assets
6. Intangible assets and development costs
7. Start-up costs
8. Bad debts and receivables
9. Warranty reserves
10. Inventory
11. Long-term contracts
12. Capital leases
13. Research and development costs
14. Capitalized interest
15. Employee benefits
16. Restructuring reserves
17. Environmental reserves
18. Investments (including consolidations)
19. Financial reserves
20. Prepaid expenses
21. Deferred income
22. Accrued liabilities
23. Valuation allowance
24. Other deferred taxes
25. Total net deferred taxes
620 Contemporary Accounting Research

Data were collected for Fortune 500 firms, excluding fmancial institutions and public
utilities (one-digit SIC code of 6, and two-digit SIC code of 49), that adopted SFAS No.
109 between 1992-94. The classification to different components depends primarily on
firms' disclosures in the fmancial statements and the terminology used therein.

Endnotes
1 The terms net deferred taxes and net deferred tax liabilities will be used
interchangeably to describe the net effect of interperiod tax allocation on the
firm's balance sheet. Most firms in our analysis have both deferred tax assets and
deferred tax liabilities resulting from temporary differences. However, for most
firms, the net amount is a deferred tax liability.
2 Recently, the Intemational Accounting Standards Committee (IASC) issued a
revised version of International Accounting Standard No. 12, "Accounting for
Taxes on Income" (IASC 1996), prohibiting discounting of deferred taxes.
3 See, for example, Huckel and Livnat (1992) and White, Sondhi, and Fried (1994,
chapter seven). Foster (1986, p. 76) says that "a bank loan officer may classify
deferred taxes as part of debt, whereas an investor may classify deferred taxes as
part of equity.
4 For example, using estimates of future capital expenditures, one could anticipate
a company's change in deferred taxes due to depreciation.
5 Alternatively, one could conduct an event study around the announcement of
deferred tax infonnation. However, this approach may not be feasible because of
the problem to identify the time during which deferred taxes were announced.
Also, infonnation on deferred taxes is disclosed together with other financial
information. Finally, as a portion of defened taxes may be anticipated, an event
study is expected to be of low power due to the lack of expectation models for
deferred tax components.
6 For further details, see F&O (1995, equation 12).
7 We do not expect the valuation coefficient on net financial assets (NFA) to vary
by industry, as these assets are usually marked to market or presented at or close
to fair value.
8 SFAS No. 5 requires that if a loss is probable and estimable, the "best" (unbiased)
loss estimate should be accrued. SFAS Interpretation No. 14 adds that if all loss
estimates are equally likely, the lowest loss estimate should be accrued (liability
understatement). According to Securities and Exchange Commission (SEC) Stajf
Accounting Bulletin No. 92 it is inappropriate to offset environmental liabilities
by recoveries from insurers and other Potentially Responsible Parties (liability
overstatement). Barth and McNichols (1994, p. 201) show that the median
unrecognized environmental liability according to investors valuation is only 1.4
percent of market value of equity.
9 We repeated our tests excluding short-term investments (COMPUSTAT item 193)
from financial assets. Our results were not sensitive to this different
classification. In particular, the regression coefficients had the same significance
levels as those reported.
10 To the extent that the "true" cost of capital is conelated with the model variables,
a mismeasured cost of capital may bias the regression coefficients. To examine
the sensitivity of our results to different cost of capital definitions, we repeated
our analysis with an industry-adjusted return on equity (using one-digit SIC
codes) as a proxy for the discount rate obtaining similar results. Frankel and Lee
(1996) use a country-specific discount rate, however, the discount rate, in their
study is fixed for all firms within a country.
11 All reported significance levels are of a one-tailed test.
The Valuation of Deferred Taxes 621

12 The five-year yield on corporate bonds was 6.05 percent, 5.2 percent, and 7.7
percent at December of 1992, 1993, and 1994, respectively. One would expect a
negative correlation between the valuation of net financial assets and the yield on
bonds. The valuation coefficients on net financial assets are 0.83, 0.85, and 0.75,
in the 1992, 1993, and 1994 models, respectively, which is consistent with the
above claim.
13 Given the high correlations between DT-losses and DT-valuat., we, estimated the
valuation model omitting one of them. When DT-valuat. is omitted, the
coefficients on DT-losses become negative in all periods and significantly lower
than zero at the 0.01 level in the pooled model and in 1994. When DT-losses is
omitted the coefficients on DT-valuat. are positive in all periods and significantly
larger than zero at the 0.01 level in all cases but 1993.
14 Recall that assets (liabilities) were coded as positive (negative) numbers.
Therefore, the valuation allowance {DT-valuat.), which is coded as a negative
number, is added to DT-losses.
15 We could obtain lead eamings only for our 1992 and 1993 samples. 1995
eamings data were not available at the time this article was written.

References
Accounting Principles Board. 1967. Accounting for income taxes. Opinion (No. 11)
New York, NY: American Institute of Certified Public Accountants.
Accounting Standards Committee. 1985. Accounting for deferred taxation. Statement
of Standard Accounting Practice (No. 15) London, UK: Institute of Chartered
Accountants.
Barth, M. E., and M. F. McNichols. 1994. Estimation and market valuation of
environmental liabilities relating to superfund sites. Journal of Accounting
Research 34 (Supplement): 177-209.
Beaver, W. H., and R. E. Dukes. 1972. Interperiod tax allocation, earnings
expectations and the behavior of security prices. Accounting Review 47 (April):
320-32.
Chaney, P. K., and D. C. Jetter. 1992. The effect of deferred taxes on security prices.
Journal of Accounting, Auditing and Finance 7: 91-116.
Deloitte and Touche. 1992. Accounting for income taxes: An analysis of FASB
statement no. 109. New York, NY: Deloitte & Touche.
Eeltham, G. A., and J. A. Ohlson. 1995. Valuation and clean surplus accounting for
operating and financial activities. Contemporary Accounting Research 11
(Spring): 689-732.
Financial Accounting Standards Board (FASB). Statement of Financial Accounting
Standards {SFAS)
1975 No. 5: Accounting for contingencies.
1990 No. 106: Employers' accounting for post-retirement benefits other than
pensions.
1992 No. 109: Accounting for income taxes.
1993 No. 112: Employers' accounting for post-employment benefits.
Financial Accounting Standards Board. 1976. Reasonable estimation of the amount of
a loss. FASB interpretation no. 14.
Foster, G. 1986. Financial statement analysis. Englewood Cliffs, NJ: Prentice Hall.
Frankel, R., and C. M. C. Lee. 1996. Accounting diversity and international valuation.
Working paper. University of Michigan and Comell University.
Givoly, D., and C. Hayn. 1992. The valuation of the deferred tax liability: Evidence
from the stock market. Accounting Review 67 (April): 394-410.
622 Contemporary Accounting Research

Huckel, K. S., and J. Livnat. 1992. Cashflow and security analysis. Homewood, IL:
Business One-Irwin.
International Accounting Standards Committee (IASC). 1996. Income taxes (revised).
International Accounting Standard J^o. 12: London, UK: IASC.
Ohlson, J. A., and S. H. Penman. 1992. Disaggregated accounting data as explanatory
variables for returns. Journal of Accounting, Auditing and Finance 1 (Fall):
553-73.
Penman, S.H. 1996. The articulation of price-earnings ratios and the market-to-book
ratios and the evaluation of growth. Journal of Accounting Research 34
(Autumn): 235-60.
Securities and Exchange Commission (SEC). 1993. Accounting and disclosure relating
to loss contingencies. Staff Accounting Bulletin No. 92 Washington, D.C.: SEC
White, G. I., A. C. Sondhi, and D. Fried. 1994. The analysis and use of financial
statements. New York, NY: John Wiley and Sons.
White, H. 1980. A heteroscedasticity-consistent covariance matrix estimator and a
direct test for heteroscedasticity. Econometrica: 817-38