You are on page 1of 9

DOI: 10.1111/j.1475-679X.2006.00201.

Journal of Accounting Research
Vol. 44 No. 2 May 2006
Printed in U.S.A.

Discussion of
Accounting Discretion in Fair Value
An Examination of SFAS 142
Goodwill Impairments


Beatty and Weber examine an accounting choice that managers made upon
adoption of Statement of Financial Accounting Standards 142: whether to
record a goodwill asset impairment as a cumulative effect of an accounting
change at the time of adoption or delay the recognition of such an impairment
to the future (perhaps indefinitely) when they would be recorded as expenses
in earnings from continuing operations. The authors consider several factors
that might influence managements reporting of transition effects, including
contracting, equity market incentives, and regulatory forces. Participants at the
2005 Journal of Accounting Research Conference questioned whether such a
complex accounting decision can be captured with simple linear models and
noisy proxy variables, while also speculating upon whether the results would
generalize to other settings. In this discussion, I summarize Beatty and Webers
research, highlight its contribution to the accounting literature, and provide
a record of the main issues raised by the conference participants.

1. Summary
Beatty and Weber (henceforth, BW) examine the accounting choice that
a subsample of U.S. companies made upon the adoption of Statement of
University of Arizona. I thank Richard Leftwich (the editor) for his helpful comments.
C , University of Chicago on behalf of the Institute of Professional Accounting, 2006



Financial Accounting Standards 142 (SFAS 142) Goodwill and Other Intangible Assets. In this discussion, I briefly summarize the content of the paper,
state its contribution to the literature, and provide a record of the relevant
issues raised by participants of the 2005 Journal of Accounting Research
BW pose the following question as their primary research motivation:
What factors affect managements decision to accelerate a goodwill impairment charge as an SFAS 142 below-the-line cumulative effect as opposed to
delaying the recognition of impairments into future above-the-line charges?
SFAS 142 changed the accounting for goodwill in three fundamental ways.
First, it eliminated the periodic amortization of these assets into expenses.
Second, it required that goodwill be assigned to a reporting unit for external reporting purposes that closely matched how management views the
corporations portfolio of businesses for internal purposes. Finally, it instituted a fair value test for determining goodwill impairments. Previously,
goodwill was treated similarly to other long-term assets for impairment testing: A comparison of undiscounted cash flows to the assets historical cost
was used. Upon adoption of the standard, firms were required to report any
reduction in the goodwill balance due to the stricter fair value test as a cumulative effect of an accounting change on the income statement, commonly
referred to as a below-the-line charge. Given that fair values are not readily
available for many of the reporting units to which goodwill balances were
assigned, managers enjoy a certain amount of discretion when applying this
standard. BW make several predictions regarding how managers use that
discretion during the transition period.
BW first identify a group of firms that are likely candidates for transition
impairment charges. These firms have goodwill outstanding at December
31, 2001, have a difference between the market and book values of their
equity that is less than their goodwill balance, and meet various other data
requirements for the cross-sectional tests.
Two different managerial choices are studied. First, using a probit model,
BW identify covariates of the decision to record a transition charge. Second,
using a censored regression, BW estimate how these same proxies are associated with the magnitude of the charge. These two models are each estimated
on two different samples, one limited in size due to lack of data availability
for some of the explanatory variables.
Three of the authors hypotheses focus on how contractual concerns
(debt, compensation, and employment) might affect the initial transition
charge. One hypothesis analyzes how the capitalization of earnings into stock
price (i.e., equity market incentives) might affect the transition charge. The
final hypothesis makes predictions regarding how regulatory constraints
might influence managements behavior.
The basis for BWs primary conclusions is reported in table 5 of the paper. There is weak evidence that the likelihood of taking a transition charge
is associated with less slack in a net worth debt covenant that includes the
effects of accounting changes. The evidence is somewhat stronger when




the dependent variable is the continuous measure of the magnitude of the

impairment. Firms are more likely to record transition charges when their
stock price is historically more highly associated with earnings from continuing operations, an earnings measure that would otherwise be subject
to impairment charges in the future. This relation holds when examining
the magnitude of the impairment as well. In the limited sample with bonus
and CEO tenure information available, BW document that, if the firms
proxy statement reveals a bonus plan that relies on earnings, then transition charges are less likely to be recorded and tend to be lower in magnitude.
Based on that same subsample, BW also conclude that the longer the CEOs
tenure, the less likely that a charge will be recorded, and its magnitude will
be lower. Finally, the authors conclude that if a firms stock is traded on
an exchange that uses objective financial statement measures to determine
trading eligibility, then such firms are less likely to record transition charges,
and such charges, if recorded, are lower in magnitude.

2. Contribution
Financial reporting rules require that managers make a significant number of subjective decisions when reporting accounting information to investors. As standard setters attempt (at least nominally) to move towards
principle-based standards, while embracing such concepts as fair value
accounting for assets and liabilities that are not actively traded, it seems
likely that the subjectivity in accounting reports will continue to rise.
The determinants of how these decisions are made by managers are still
not well understood, despite over three decades of academic research. This
ignorance is noted by Fields, Lys, and Vincent [2001, henceforth, FLV] in
their survey of the accounting choice literature. In particular, they claim that
one of the problems with the literature is that there have been few attempts
to take an integrated perspective (i.e., multiple goals) on accounting choice
(p. 258).
Most of the research surveyed by FLV analyzes a limited number of determinants of the accounting choice at hand, while ignoring many other
incentives that may be present. BW respond to FLVs call for more research that takes such an integrated approach, as they consider several
incentives that might affect a particular discretionary accounting choice.
In my opinion, this is a worthy goal, given the paucity of understanding
of the inputs to such decisions, and the ever increasing financial reporting
choices that managers are forced to make. I presume that many of the conference participants also believe that this is an interesting area of study, as
there was little discussion about why the authors chose to address this issue,
but rather, most of the comments focused on the research design and its
Some participants did question whether the result of this study would
generalize to other discretionary accounting choices. There was also some
discussion as to whether the paper reflected an overall critique of fair value



accounting, or whether it was a study of the determinants of how information

is presented in various components of the financial statements.
One relevant point not raised at the conference is the nature of the costs
borne (if any) by market participants and contracting parties when allowing
management this discretion. Presumably, BW view it as beyond the scope
of their paper to assess whether there are real costs that are not already
included in share prices and the various agency contracts. Yet this issue is
important. If these contracts, which arise endogenously, take into account
managements strategic use of accounting discretion, then studies of accounting choice lose some of their appeal, since these costs have already
been impounded into asset prices. BW are silent on this issue.

3. Issues
I group my remaining comments, because they relate to two critiques
regarding research on accounting choice, in general, and this paper, in
particular. First, BW, like many other researchers, face a daunting task in
empirically modeling the complex nature of the strategic reporting decision.
Second, given the trade-off that must be made when sacrificing realism for
tractability in the model, any measurement error in, or misinterpretation
of, the proxy variables used significantly limits the conclusions that can be
Conference participants noted that the simple empirical specifications
used by BW to test their hypotheses fail to capture the complex nature of
the transition impairment decision. For example, in complying with SFAS
142, management must first assign the goodwill balance to a reporting unit
prior to making an estimate of its fair value. As Watts [2003] notes, assigning
intangible assets to particular segments of the business is difficult, absent
a specific joint cost allocation model. Therefore, as noted by several participants, it would be interesting to study empirically how managers use
their discretion in this accounting choice. Unfortunately, as noted by the
presenter, the researcher faces the problem of developing an expectations
model of what is the proper allocation against which to study managements actual choice. Thus, BW are forced to ignore this important aspect
of the accounting choice decision.1
Another participant noted that accounting choices are often affected by
competitors actions, a force largely ignored in the BW model (though to
be fair, in sensitivity tests, BW use industry dummy variables as additional
controls). In addition, there are other players in this setting, such as auditors
and valuation consultants, that also behave strategically (and potentially

1 In their conclusion, BW state that they investigate the outcome of the managers goodwill
allocation and reporting unit decisions by examining the determinants of the SFAS 142 writeoff decisions. However, this is an indirect manner in which to study the allocation, and thus,
the paper does not shed much light on these accounting choices.




constraining some of managements discretion), yet their impact on the

decision is ignored.
While conference participants seemed willing to accept some sacrifice of
complexity for tractability, the remainder of their comments focused on
whether the proxy variables used in this simple model could accurately
capture the underlying forces. When combined with the criticisms above,
these issues make some of BWs conclusions problematic.
Participants generally had few objections to the debt contracting variables
used by BW. The authors carefully track down actual covenant information,
including whether net worth covenants included or excluded intangible
assets, as well as whether such covenants included or excluded the effects
of accounting changes. The results of the hypothesis test of the effect of
these covenants on the accounting choice can only weakly reject the null
hypothesis of no association in the probit regression (t-statistic of 1.42 for
the coefficient INWSlack). When examining this variables association with
the magnitude of the write-off in the censored regression, the statistical
significance is slightly higher (t-statistic of 1.93).
As a measure of regulatory constraints, BW include the dummy variable Nasdaq/Amex, which equals one if the firm is traded on an exchange
that incorporates objective financial information in the delisting decision.
BW predict and find that firms traded on such exchanges tend to record
lower transition impairments, presumably to maintain healthier ratios and
avoid delisting. Some participants noted that an exchange listing can capture other firm fundamental information. For example, NASDAQ firms
may be more intangible-asset intensive and have greater growth opportunities. However, this would work against the authors hypothesis, as many
of these growth options declined in value post-2000 and would suggest
that, if SFAS 142 were neutrally applied, these firms would record larger
transition charges, not the smaller ones observed. Similar to the debt contracting conclusion, I believe most participants find some comfort in the
conclusion that regulatory forces influence managements use of accounting
BW document that firms with a stronger association between share prices
and earnings in the pre-SFAS 142 period are likely to use accounting discretion to take larger transition impairment charges. The authors identify
the variable AsstPrc for each firm by estimating firm-specific regressions in
which share price is the dependent variable and earnings from continuing
operations serves as the independent variable. The coefficient from such a
regression serves as AsstPrc. The variable AsstPrc is interacted with a measure
of idiosyncratic firm risk, HRisk, a dummy variable equal to one if the firms
stock return volatility is above the sample median.
The hypothesis offered by BW is that the greater the capitalization rate
of operating earnings into share price, the more likely that firms will overstate the transition charge. This allows firms to avoid future impairments,
which would have a more negative effect on share prices as they would
be included in income from continuing operations, as opposed to an



impairment recorded as a cumulative effect of an accounting change, which

the market tends to discount (Bens and Heltzer [2005]).
Conference participants were skeptical of BWs interpretation of the
AsstPrc results. BW rely on a mechanistic view of how investors use earnings
information when setting share prices. The model assumes that investors will
correctly discount the overstated transition impairment charge, yet when a
future real economic impairment to the firms intangible assets occurs, investors will be unable to map this into current share price because a charge
has been excluded from earnings. I believe that investors would simply use
other financial and nonfinancial cues to determine when the actual goodwill impairment occurs in the future, and reduce their estimate of share
price accordingly.
The presenter noted that it is not necessary for investors to actually behave
as if earnings were the sole source of equity value relevant information, but
only that managers believe investors react this way. This is a plausible explanation.2 However, the spirit of the paper does not convey that this is largely
a story about nonrational behavior by the manager. As an alternative hypothesis, conference participants noted that the AsstPrc and HRisk variables
may simply be capturing companies that were classified as growth firms in
the late 1990s and 2000. Essentially, the firm-specific regressions are sorting
firms based on their price-earnings (P/E) ratios from the estimation period.
Many high P/E firms experienced a reversal of fortune prior to SFAS 142
adoption, as their share prices in the equity market dropped precipitously in
the early 2000s. It is quite possible that firms with higher AsstPrc and HRisk
were the most likely to record larger impairment charges when applying
the new SFAS 142 fair value tests, and these variables are correlated with
the measurement error of BWs expectation model of impairments. This is
especially problematic in the probit specifications of table 5 in which the
ExpectedImpair variable is not associated with the probability of a multisegment firm taking a charge.
Another example of an alternative explanation for a hypothesis variable
concerned Bonus, which is a dummy variable that is equal to one if the firm
has an earnings-based bonus plan that is affected by nonrecurring items,
and zero otherwise. The prediction is that a bonus based on an earnings
definition that includes the transition charge would dissuade the manager
from recording an impairment. While this is a reasonable hypothesis, measurement problems with the variable, and an endogeneity issue with Bonus
should prohibit strong conclusions about the strength of this incentive on
the accrual decision.
2 For example, Brav et al. [2005] find that a significant number of managers surveyed claim
to direct the company to repurchase its shares to offset earnings per share dilution attributed
to employee stock options, despite the fact that such a transaction may sacrifice shareholder
value if the funds could be used on projects where the return exceeds the cost of capital
(p. 515). In an empirical archival study, Bens et al. [2003] document similar evidence for
S&P 500 industrial firms. These results suggest that managements beliefs regarding investor
reliance on simple earnings statistics can have real effects on corporate behavior.




From a measurement error perspective, when using publicly available

data on bonus contracts, it is difficult to accurately determine how the
corporation defines earnings and thus whether the bonus calculation
would include or exclude the impairment. For example, Whirlpool notes in
both its 2002 and 2003 proxy statements that to determine incentive compensation for its top four executives a corporate performance target based
on return on equity was established.3 But it is not obvious whether this
measure includes or excludes the impairment charge. In 2002 Whirlpools
return on equity, including the effects of a $613 million transition impairment, was 36%, yet the executives earned their full bonus per the 2003
proxy statement, as this measure exceeded the performance target. Presumably, the contract excluded the impairment charge, but this can only be
inferred as it is not explicitly stated.
In addition to this measurement error, a participant noted a self-selection
problem associated with the Bonus variable. In this case, firms with earningsbased bonus plans may be less reliant on intellectual capital and intangible
assets to generate value. Hence, they are less likely to have impaired goodwill
by construction, and this might account for the observed negative association between Bonus and the dependent variables.
Finally, the Tenure variable, defined as the number of years the CEO has
held that position, is negatively associated with both the likelihood and
magnitude of a transition charge. BW interpret this to suggest that CEOs
with longer tenure are less likely to admit via a goodwill write-off that their
acquisitions were poor decisions. Yet this is an indirect proxy for whether
the CEO was actually in place when the goodwill being written off was first
recognized. An alternative interpretation of the Tenure variable is that it
also captures the life of the firm itself. As documented by Fama and French
[2001] firms that went public in the 1990s tended to be younger and less
profitable than previous generations of initial public offerings. Moreover,
with the equity price run-up of the late 1990s, many of these young firms were
more acquisitive (the young AOLs acquisition of the old Time Warner, for
example). By construction, these firms will have more goodwill and CEOs
with shorter tenure. A more convincing measure would capture the CEOs
tenure relative to the age of the firm, or better yet, whether the current CEO
is actually responsible for directing the acquisitions that led to the currently
recognized goodwill.
In summary, all empirical researchers face trade-offs when identifying
proxy variables in order to capture latent economic constructs. When the
inherently imperfect proxies are combined with simplified models of accounting choice, alternative explanations for observed relations abound.
Thus it can be difficult to reach unambiguous economic conclusions despite statistically significant associations.

3 Whirlpools Securities and Exchange Commission filings may be accessed via their web site



4. Conclusion
BW attack a problem that is of interest to accounting researchers: what
are the dominant forces that influence managements discretionary choices
when preparing financial reports? Although the choice set seems to be ever
expanding, as standard setters move away from objective measures such
as historical cost and towards subjective measures such as fair value, researchers still do not have a thorough understanding of how managers use
this discretion.
BW provide interesting insights by examining multiple forces that might
influence managements accounting choice, including contracting, equity
market incentives, and regulatory forces. While conference participants appreciated this multifaceted approach, they did not think it went far enough.
That is, accounting decisions can be quite complex, and a simple linear
framework may not capture many of the interesting subtleties involved.
Moreover, many of the proxy variables used in the BW framework were
difficult to interpret unambiguously. In addition, given an efficient stock
market and contracts that arise endogenously, it is unclear whether there
are any significant costs associated with managements strategic use of accounting discretion. That said, BWs results suggest that debt contracting
and regulatory forces appear to have an effect on accounting choice, and
this is a contribution to the literature. Future research should find these results useful, as it attempts to capture more of the complexity that academics
would like to see modeled.
BEATTY, A., AND J. WEBER. Accounting Discretion in Fair Value Estimates: An Examination of
SFAS 142 Goodwill Impairments. Journal of Accounting Research 44 ( 2006), this issue.
BENS, D., AND W. HELTZER. The Information Content and Timeliness of Fair Value Accounting: Goodwill Write-offs Before, During and After Implementation of SFAS 142. Working
paper, University of Chicago, 2005.
BENS, D.; V. NAGAR; D. SKINNER; AND M. WONG. Employee Stock Options, EPS Dilution, and
Stock Repurchases. Journal of Accounting and Economics 36 (2003): 5190.
BRAV, A.; J. GRAHAM; C. HARVEY; AND R. MICHAELY. Payout Policy in the 21st Century. Journal
of Financial Economics 77 (2005): 483527.
FAMA, E., AND K. FRENCH. Disappearing Dividends: Changing Firm Characteristics or Lower
Propensity to Pay. Journal of Financial Economics 60 (2001): 343.
FIELDS, T.; T. LYS; AND L. VINCENT. Empirical Research on Accounting Choice. Journal of
Accounting & Economics 31 (2001): 255307.
WATTS, R. Conservatism in Accounting Part I: Explanations and Implications. Accounting
Horizons 17 (2003): 20721.