Recent deliberations by both the International
Accounting Standards Board (IASB) and the
Financial Accounting Standard Board (FASB) in
the United States have focused on how fair values
of assets and liabilities should be measured. The
issue of when, rather than how, fair value measure-
ment should be applied is still far from resolved,
however. Fair values have been mandated for
some assets and liabilities under both IASB and
FASB standards, but it is fair to say that principles
governing the applicability of fair values have yet
to be articulated: when is fair value accounting ap-
propriate and when is it not? Or, in terms of my
charge for this paper, under what circumstances is
fair value a plus or a minus?

To prepare for my task, I made a survey of pub-
lic statements made for and against fair value ac-
counting by a variety of standard setters,
regulators, analysts, and preparers. The stated ‘mi-
nuses’ typically point to the dangers of fair value
estimates from marking to model rather than mark-
ing to market, concerns about introducing ‘excess
volatility’ into earnings, and feedback effects (on
banks’ lending practices, for example) that could
damage a business and, indeed, heighten systemat-
ic risk. A few antagonists question whether fair
values (for bank assets and liabilities, for example)
really capture the economics of a business (in fos-
tering core deposits and making loans). In counter-
point, the proponents of fair value argue that fair
value is a superior economic measure to historical
cost. Consider the following arguments, often ad-
vanced as ‘pluses’:
• Investors are concerned with value, not costs, so

report fair values.

• With the passage of time, historical prices be-
come irrelevant in assessing an entity’s current
financial position. Prices provide up-to-date in-
formation about the value of assets.

• Fair value accounting reports assets and liabili-
ties in the way that an economist would look at
them; fair values reflect true economic sub-
stance.

• Fair value accounting reports economic income:
in accordance with the widely accepted Hicksian
definition of income as a change in wealth, the
change in fair value of net assets on the balance
sheet yields income. Fair value accounting is a
solution to the accountant’s problem of income
measurement, and is to be preferred to the hun-
dreds of rules underlying historical cost income.

• Fair value is a market-based measure that is not
affected by factors specific to a particular entity;
accordingly it represents an unbiased measure-
ment that is consistent from period to period and
across entities.
So self-evident do these points seem to be that

fair value accounting is often just presumed to be
‘more relevant’. The words, ‘fair value’ sound
good (who could be against ice-cream and fair
value?!) while ‘historical cost’ sounds, well, passé.
As it turns, out, however, each of these statements
becomes qualified under scrutiny. Can economic
argument lead to constructive arguments for im-
plementing fair value accounting?

1. Some preliminaries
Pluses and minuses can only be evaluated against
an alternative, so I will take the approach of asking
if (or under what conditions) fair value accounting
is an improvement over historical cost accounting.
In discussions about fair value, people often pro-
ceed at cross-purposes, so a few points need to be
clear before we proceed.

1.1. What is fair value?
Three notions of fair value accounting enter the

discussion, and one must be clear which is being
entertained.

Accounting and Business Research Special Issue: International Accounting Policy Forum. pp. 33-44. 2007 33

Financial reporting quality: is fair value
a plus or a minus?
Stephen H. Penman*

*The author is at the Graduate School of Business,
Columbia University, shp38@columbia.edu This paper draws
on some of the themes in a White Paper prepared for the
Center for Excellence in Accounting and Security Analysis
(CEASA) at Columbia Business School. See D. Nissim and S.
Penman, The Boundaries of Fair Value Accounting, White
Paper No. 2 (Center for Excellence in Accounting and
Security Analysis, Columbia University, 2007). Comments re-
ceived at the Information for Better Markets Conference have
been helpful as has a close reading of the manuscript by
Martin Walker and Pauline Weetman.

1. Fair value variously applied in a ‘mixed attrib-
ute model’:
In this treatment, fair value is used alternative-
ly with historical cost for the same asset or lia-
bility but at different times; the accounting is
primarily historical cost accounting, but fair
values are applied under certain conditions.
Examples are fair values applied in fresh-start
accounting (that then proceeds under historical
cost accounting), impairment from historical
cost to fair value (really a form of fresh-start
accounting), using fair values to establish his-
torical cost (for barter transactions and dona-
tions, for example) or in the allocation of
purchase price (between goodwill and tangible
assets, for example), and reference to fair value
to discipline estimates under historical cost ac-
counting.

2. Fair value continually applied as entry value:
Assets are revalued at their replacement cost,
with current costs then recorded in the income
statement, with unrealised (holding) gains and
losses also recognised. Revenue recognition
and matching is maintained but income, based
on current costs, is said to be a better indicator
of the future and not path-dependent.

3. Fair value continually applied as exit value:
Assets and liabilities are remarked each period
to current exit price, with unrealised gains and
losses from the remarking recorded as part of
(comprehensive) income.

The pluses and minuses of fair value in applica-

tions (1) and (2) can be debated, but note that both
are really modified cost accounting; both maintain
standard revenue recognition – applying exit
prices to recognise value from business activity
only on actual exit of the product or service to the
market – but with modifications to the expense
matching.1 Application 3 applies exit values to
continually remark assets and liabilities but with-
out actual exit (realisation).

The FASB, in its recent Statement 157, Fair
Value Measurements endorses fair value as exit
value, with a seeming nod from the IASB subject
to some minor reservations:2

‘Fair value is the price that would be received to
sell an asset or paid to transfer a liability in an
orderly transaction between market participants
at the measurement date.’
While the IASB and FASB presumably envision

exit values being applied to determine fair value in
the mixed attribute model (1), I will limit my com-
ments to fair value applied in (3).3 It is the recog-
nition of exit values without an historical exit
transaction that places this fair value accounting in
such contrast to historical cost accounting. The
top-line notion of revenue disappears, and income
is simply the change in fair values on the balance
sheet. Accordingly, the accounting issues are quite
different. Continually remarking equity invest-
ments to fair value rather than using the equity
method involves different issues from impairing
equity method investments for a permanent loss
under mixed attribute accounting. And so with
marking inventories, core deposits, bank loans, in-
surance contracts, debt, and so on to fair value on
a continual basis. ‘Fair value accounting’ as envi-
sioned in application (3) is a potential shift in par-
adigm.4

1.2. Fair value to whom?
As with any policy issue, prescriptions cannot be

made without an understanding of the objectives
of the exercise. To whom are we reporting? Whose
pluses and whose minuses? Different users may
demand different accounting reports, and confu-
sion reigns if issues are discussed at cross purpos-
es. A shareholder might recognise a gain from a
fall in the market value of debt as creditworthiness
deteriorates, but not the creditor. Bank sharehold-
ers might wish to see bank deposits at fair value,
but not the depositors. A bank regulator would also
be concerned about reporting deposits at less than
face value if such reporting affected depositors’
confidence in the banking system. While an in-
vestor might welcome the information about
volatility that fair value accounting reveals, not so
a central banker who might be concerned about
feedback effects on systematic risk. A bank regula-
tor might be concerned about marking up banks’

34 ACCOUNTING AND BUSINESS RESEARCH

1 For example, impairment to fair value under application
(1) fresh-starts the matching of expenses to future revenues
when there is a downward revision in future revenues antici-
pated, that is, cost have expired. Application (2) matches cur-
rent costs rather than historical costs to (current) revenues.
FASB Statement 33 (now suspended) was an experiment with
application (2), but those issues are not part of the current de-
bate. See Statement of Financial Accounting Standards No.
33, Financial Reporting and Changing Prices (Norwalk,
Conn.: FASB, September 1979).

2 See Statement of Financial Accounting Standards No. 157,
Fair Value Measurements (Norwalk, Conn.: FASB, September
2006), paras 5–15 and Discussion paper, Fair Value
Measurements Part 1: Invitation to Comment (London: IASB,
November 2006).

3 Statement 157 is explicit in stating that the standard deals
with the measurement of fair value (when fair value measure-
ments are applicable), not with the issue of when fair value
measurements are applicable. IASB discussion documents
have the same flavour. However, the application question is
very much open and (presumably) part of the conceptual
framework agenda.

4 Under application (3), some assets or liabilities might be
carried at fair value (continually) while others are carried at
historical cost (continually). So, marketable securities might
be marked to market, with inventories at historical cost. This
form of a ‘mixed attribute model’ differs from moving be-
tween fair value and historical cost for the same asset and lia-
bility.

capital during speculative times with the resulting
incentive for profligate lending.5

In this talk, I take a shareholder perspective:
what are the pluses and minuses of using fair value
accounting (rather than historical cost accounting)
for reporting to shareholders? This, I submit, is
hardly controversial; the shareholders are the own-
ers to whom management and auditors report. But
it does mean that, if standard setters have a broad-
er set of constituents in mind, with an objective of
general purpose financial reporting, then they may
see the issues differently.

1.3. My approach
Normative statements about accounting issues

are often statements of the author’s received wis-
dom combined with some a priori thinking: here is
what I think about the matter, says the author, sup-
ported by some inductive and deductive logic. This
approach, applied in the ‘accounting theory’ era of
the 1950s to the 1970s, gave us numerous prescrip-
tions but little resolution. It would be helpful to
refer to concrete research results for answers, but
theoretical and empirical research has not delivered
a definite resolution either. Recent accounting-
based valuation theory has given us some insight to
which I will refer later. Empirical research (of the
type discussed by Wayne Landsman) documents
correlations between fair value measurements and
stock prices that are useful for understanding
whether fair values are ‘relevant to investors’. But
it does not give us much of a handle on the policy
question of whether fair values should be reported
in place of historical cost accounting (which, re-
search shows, is also relevant to investors).6

My approach, I must confess, is largely a priori.
But I hope to get some bite by taking what might
be referred to as a demand approach. Accounting,
as I see it, is a product and products are a matter of
design. The design – and the quality of the product
– should be judged on how well it serves the cus-
tomer. So, with the customer identified as the
shareholder (above), I ask which product features
– fair value or historical cost – help (or frustrate)
the customer. Unfortunately, inferring demand
from statements made in the current regulatory en-
vironment is difficult, given that regulation affects
behaviour. We do observe the voluntary applica-
tion of fair value accounting (without the coercion
of regulation) in some situations – unregulated
hedge funds use fair value accounting, for example
– and so we can defer to ‘the market’ for lessons.
Such observations are limited, however, so I resort
to a priori analysis. But I do so with an eye to the
shareholder; I presume that shareholders require
accounting information for two purposes:
1. Valuation. Shareholders use accounting infor-

mation to inform them about the (fair) value of
the equity: What is the equity worth?

2. Stewardship. Shareholders use accounting in-
formation to assess the stewardship of manage-
ment, the owners’ employees: How efficient
have managers been in making investments
and conducting operations to add value for
shareholders?

More concretely, I force an orientation to practi-
cal tasks for which information is demanded: To
what extent does fair value accounting aid or frus-
trate the tasks of equity valuation and monitoring
managers’ stewardship? This focus, also, is hardly
controversial. The first task is that of the equity an-
alyst, the second the pursuit of those involved in
corporate governance on behalf of shareholders.

In view of the above, many of the points I make
below are not particularly original. I want to be a
little more analytical than simply listing the stan-
dard litany of complaints about and statements in
favour of fair value accounting. But, in doing so,
some well-worn points come to the surface. By
presenting them in a more organised framework,
my hope is that they will be more imperative.

1.4. Information for better markets
It is often said that financial reporting should have

the objective of providing all relevant information
to capital markets. So (it follows), if both historical
cost information and fair values are relevant, both
should be reported. Nothing here subtracts from that
position (if one wants to adopt it). The issue is
which measurement basis should go through the
discipline of the accounting system to determine the
summary, bottom-line numbers, earnings and book
value on which investors and analysts focus (for
whatever bounded rationality reason). Alternatives
to the accounting information (within the system)
can, of course, be supplied in footnotes, much like
some fair value information is now disclosed.

2. The conceptual merits of fair
value accounting versus historical
cost accounting
As with most accounting issues, it is important to
distinguish conceptual issues from those that have

Special Issue: International Accounting Policy Forum. 2007 35

5 Papers that deal with fair value from the view of the cen-
tral banker and bank regulator include A. Enria et al., Fair
Value and Financial Stability Occasional Paper Series No. 13,
European Central Bank, April 2004; G. Plantin, H. Sapra, and
H. Shin, ‘Marking to market, liquidity, and financial stability’,
Monetary and Economic Studies (Special Edition), October
2005; K. Burkhardt and R. Strausz, ‘The effect of fair vs. book
value accounting on banks’, unpublished paper, Free
University of Berlin, April 2004; and ‘Fair value accounting
for financial instruments: some implications for bank regula-
tion’, BIS Working paper No.209, August 2006.

6 Indeed, inferences from the empirical research are limited
because stock prices, from which ‘relevance’ is inferred, are de-
termined from information under current accounting practices,
and those prices might be different under alternative practices.

to do with measurement. Here I ‘conceptualise’
how both fair value accounting and historical cost
accounting would satisfy the valuation and stew-
ardship goals of shareholder reporting, in principle
(if measurement were no problem). I then overlay
the concepts with measurement in Section 3.

2.1. The concepts behind fair value accounting
Putting aside measurement issues, fair value ac-

counting conveys information about equity value
and managements’ stewardship by stating all as-
sets and liabilities on the balance sheet as their
value to shareholders:7

• the balance sheet becomes the primary vehicle
for conveying information to shareholders;

• with all assets and liabilities recorded on the
balance sheet at fair value, the book value of eq-
uity reports the value of equity (the Price/Book
ratio = 1.0);

• the income (profit and loss) statement reports
‘economic income’ because it is simply the
change in value over a period;

• following the economic principle that current
changes in value do not predict future changes in
value, earnings cannot forecast future earnings.
But this is of no concern for valuation, because
the balance sheet provides the valuation;

• (unexpected) earnings, being a shock to value,
reports on the risk of the equity investment.
Volatility in earnings is informative for value at
risk;

• the P/E ratio is Price/Shock-to-value, that is, a
realisation of value at risk (with a very different
interpretation to that under historical cost);

• income reports the stewardship of management
in adding value for shareholders.
In short, the balance sheet satisfies the valuation

objective and the income statement provides infor-
mation about risk exposure and management per-
formance.

The accounting for investment funds – mutual
funds and hedge funds – applies this strict fair
value accounting, and investors are willing to trade
in and out of these funds at book value (‘net asset
value’) with the presumption that book value
equals value (with no gains and losses between
shareholders). Further, the income (returns) for
these funds is accepted as a comprehensive meas-
ure of the fund managers’ investment perform-
ance, both the investment success and the
volatility to which investors have been subjected.
The accounting is sufficient; one does not require
a balanced scorecard.

2.2. The concepts behind historical cost
accounting

Historical cost accounting is often misinterpret-
ed in the debate, with the criticism that it reports a
balance sheet with old, historical costs rather than
current values. This statement is correct, but belies
an understanding about how historical cost works
for valuation and performance assessment. Under
historical cost accounting,
• the income statement is the primary vehicle for

conveying information about value to sharehold-
ers, not the balance sheet;

• earnings report how well the firm has performed
in arbitraging prices in input (supplier) markets
and output (customer) markets; that is, historical
cost earnings reports the value-added buying in-
puts at one price, transforming them according
to a business model, and selling them at another
price;

• in contrast to fair value accounting, current in-
come forecasts future income on which a valua-
tion can be made;

• the P/B ratio is typically not equal to 1.0 and the
P/E ratio takes current earnings as a base and
multiplies it according to the forecast of future
earnings;

• earnings do not report shocks to value, but
shocks to trading in input and output markets;

• earnings measure the stewardship of manage-
ment in arbitraging input and output markets,
that is, in adding value in markets.
Historical cost accounting views value as gener-

ated in business by purchasing inputs (from suppli-
ers), transforming them according to a business
plan, and selling the consequent product (to cus-
tomers) over cost; in short, value is added by arbi-
traging (entry and exit) prices in input and output
markets for goods and services according to a
business plan. Historical cost accounting does not
report the (present) value of expected outcomes
from the business plan. Rather, it reports on
progress that has been made in executing the plan,

36 ACCOUNTING AND BUSINESS RESEARCH

7 This idea is close to that of ‘value in use’ but with a focus
on the shareholder rather than on the entity. The value-in-use
concept (or its variant, ‘deprival value’) appears (for example)
in Accounting Standards Board, Statement of Principles for
Financial Reporting (London: ASB, 1999), Australian
Accounting Research Foundation, Accounting Theory
Monograph No. 10, Measurement in Financial Accounting
(AARF, 1998) and has long been part of the discussion, for ex-
ample in J. Horton and R. Macve, ‘ ‘Fair value’ for financial
instruments: how erasing theory is leading to unworkable
global accounting standards for performance reporting’,
Australian Accounting Review 10 (July 2000): 26–39 and R.
Macve and G. Serafeim, ‘ “Deprival value” vs “fair value”
measurement for contract liabilities in resolving the “revenue
recognition” conundrum: towards a general solution’.
Unpublished paper, London School of Economics, June 2006.

recognising value added (earnings) from actual
transactions in the input and output markets being
arbitraged. The income statement comes to the
fore with a matching of revenues (value received
from transactional exit prices) with costs (value
surrendered in transactional input prices). The bal-
ance sheet is not a statement of values (for the
large part), by design, but rather a by-product of
this matching, with liabilities such as accrued ex-
penses, deferred revenues, and deferred taxes gain-
ing their legitimacy from the matching process
rather than as representations of the value of obli-
gations.

The term, ‘historical cost’ is unfortunately pejo-
rative. A better term, one that captures the essence,
is ‘historical transactions accounting’, for the ac-
counting reports a history of transactions, and it is
that history of engaging with markets from which
valuations are made and management performance
assessed.

3.3. The demand for fair values
A demand for fair values could be imputed if

historical cost information is shown to be deficient
for valuation and performance evaluation, with
fair values providing the remedy. Here I compare
the two for purposes of valuation.

To separate concepts from measurement issues,
it is helpful to compare fair value accounting and
historical cost accounting implemented in their
ideal form. Ideal fair value accounting reports a
book value that is sufficient to value a firm but
earnings that are useless for the purpose. Ideal his-
torical cost accounting produces a balance sheet
that does not report value, but earnings that are
sufficient to value a firm. Consider the following
equity valuation model based on expected earn-
ings (that is a legitimate one in valuation theory in
the sense that it gives the same value as that based
on expected dividends):

Valuet = Expected Earningst+1 (A)
r

Here r is the required return for the equity holders.
Under ideal fair value accounting, earnings are

forecasted from the current book value:
Expected Earningst+1 = r x Book Valuet (B)
That is, book value (ideally measured at fair

value) is sufficient for forecasting earnings and for
valuation. Under ideal historical cost accounting
earnings are forecasted from current earnings:

Expected Earningst+1 = Earningst (C)
That is, current earnings (ideally measured) are

sufficient for forecasting earnings and for valua-
tion. In the parlance, current earnings indicate per-
manent earnings.8 Accordingly, under historical
cost accounting equity value is determined by cap-
italising current earnings:

Valuet = Earningst (D)
r

The lessons are:
1. It is not necessary to state the balance sheet at

fair value to satisfy the valuation objective.
Valuations can be made from the historical cost
income statement.

2. Assuming that one knows the required equity
return, there is no reason, in principle, to say
that fair value accounting is better than histori-
cal cost accounting. The resolution must turn
on how measurement strays from the ideal.
Historical cost comes with considerable meas-
urement issues; does fair value measurement
provide a solution?

3. If one does not know the required return (and
we don’t!), fair value accounting has a distinct
advantage. Valuation under historical cost ac-
counting requires a required return (to convert
earnings, a value flow, to a stock of value). Fair
value accounting delivers the value directly
from the balance sheet without relying on a re-
quired return (as with the mark-to-market in-
vestment fund). That is, the forecast of earnings
in (B) is not necessary, for book value already
reports the value. As a bonus, realisations on
value at risk are reported in the income state-
ment to give an indication of what the required
return should be.

In short, fair value accounting is a plus, imple-
mentation issues aside. However, historical cost
accounting has features that provide an alternative
should ideal fair value accounting not be attain-
able. Many of the statements about fair value ac-
counting in the bullet points in the introduction to
this paper are misdirected, at least at the conceptu-
al level.

As an illustration of this last statement – and to
focus on the practical valuation task — the appen-
dix carries out a valuation of the Coca-Cola
Company using historical cost numbers. Coke has
a lot of value missing from the balance sheet – its
price-to-book ratio is currently 6.3 – mainly be-
cause US GAAP does not allow its brand asset to
be carried on the balance sheet. Those who com-
plain that accounting is poor because intangible as-
sets are missing from the balance sheet might
argue that brands should be booked (as in the UK

Special Issue: International Accounting Policy Forum. 2007 37

8 One must accommodate retention that yields additional
earnings; the forecast here is for the case of full payout.
Valuation under ideal fair value accounting and ideal histori-
cal cost accounting is modelled in J. Ohlson and X. Zhang,
‘Accrual accounting and equity valuation,’ Journal of
Accounting Research, 36 (Supplement 1998): 85–111.

before IFRS). After reading the Appendix, I hope
you will be impressed by how readily Coke can be
valued without getting the balance sheet straight.
To point (1) above: missing (intangible) assets in
the balance sheet are no problem (for valuation) if
the earnings from those assets are reported in the
income statement. Note that the Coke case is not
one where valuation model (D) with ideal histori-
cal cost accounting applies. That model implies
a forward P/E of 10 (for a 10% required return,
say), but Coke’s P/E is 19.3. Nor is it a case

where

the forecast (C) strictly applies. But the imperfec-
tions of historical cost accounting can be accom-
modated.

A core accounting concept underlies the use of
historical cost accounting in valuation: the can-
celling error property. Provided that earnings are
comprehensive (clean-surplus) earnings, it is al-
ways true that

Stock returnt = Earningst + (Pt – Bt) – (Pt–1 – Bt–1)

where P is equity price and B is the book value of
equity.9 With fair value accounting, P = B at all
points in time, so earnings always equal the stock
return – just like earnings for the mark-to-market
investment fund always equals the market return
on the assets (cum-dividend). However, P = B is
not necessary; provided that the error in the bal-
ance sheet, P – B is the same at the end of the pe-
riod as at the beginning, capitalising earnings still
works. Historical cost reports a balance sheet with
error, but the focus is on earnings. We teach the
cancelling error property to our first-year account-
ing students by pointing out that earnings is the
same whether one expenses R&D immediately or
capitalises it and amortises, provided there is no
growth; that is, balance sheet errors cancel.
Growth changes this (and therefore growth intro-
duces a change in price premium over book value).

But growth can be accommodated in valuation, as
the Coca-Cola example shows.

3. Fair value measurements
Concepts are the place to start, but the rubber hits
the road with measurement. If ideal fair value ac-
counting can be implemented, all is OK, for noth-
ing is lost by abandoning historical cost
accounting, and something is gained; we have a
net plus. However, if fair value measurements do
not achieve the ideal and at the same time we lose
the information provided by historical cost ac-
counting, damage can be done.

After defining fair value as market exit price,
the recent FASB Standard 157, Fair Value
Measurements, then identifies three levels of ‘in-
puts’ to determine market price, distinguished by
increasing levels of subjectivity.10 Levels 2 and 3
refer to estimates of hypothetical market prices.
The criticisms of subjective measurement are well
known, and the FASB’s Level 3, in particular, rais-
es concerns. To sort out the pluses and minuses, it
is worthwhile to focus on Level 1 measurement –
where market prices for identical assets and liabil-
ities are observed in active markets – for, if fair
value accounting is not appropriate in that case,
concerns are just magnified when subjective esti-
mates are made.

3.1. Pluses and minuses of Level 1 fair value
measurements

Implementation of fair value accounting (as pro-
posed by the FASB and IASB) involves two ques-
tions. The first is whether exit value measures
value to shareholders. The second is whether fair
values can be applied at the level of aggregate as-
sets and liabilities that jointly produce value for
shareholders – an issue of matching.
Fair value as exit value

One could envision the implementation of ideal
fair value accounting with subjective estimates of
fair values of assets and liabilities (for sharehold-
ers), but that flies in the face of the idea that ac-
counting information should be based on
objective, reliable evidence. The FASB and IASB
commendably maintain the ‘reliability’ criterion
by requiring that fair value be backed up with an
observed market price (at least in their Level 1 im-
plementation). However, equating fair value to
market value is quite constraining, for it equates
value (to shareholders) to market (exit) price.

Plus: Fair (market) values are a plus when value
to the shareholders is determined solely by expo-
sure to market price; that is, shareholder value is
one-to-one with market prices.

A marketable bond in which a firm invests its
‘excess cash’ is exposed to changes in market price
that determines the amount of cash on liquidation,

38 ACCOUNTING AND BUSINESS RESEARCH

9 I believe this equation first appears in P. Easton, T. Harris,
and J. Ohlson, ‘Accounting earnings can explain most of secu-
rity returns: the case of long event windows,’ Journal of
Accounting and Economics, 15 (June–September 1992):
119–142, but textbooks of old used to discuss the cancelling
error property.

10 Statement 157 defines the three levels as follows:
Level 1inputs are quoted prices (unadjusted) observed in ac-
tive markets for identical assets and liabilities.
Level 2 inputs are inputs other than Level 1 quoted prices that
are observable, directly or indirectly; examples include quot-
ed prices for similar assets or liabilities in active markets,
quoted prices for identical or similar assets or liabilities in
markets that are not active, inputs such as observed interest
rates, credit risks, volatilities, and default rates, and inputs cor-
roborated by observable market data by correlation or other
means.
Level 3 inputs are unobservable inputs for the asset or liabili-
ty, reflecting the firm’s own assumptions about the assump-
tions that market participants would use in pricing the assets
or liability.

and shareholder welfare is tied to the market price,
one-for-one. Accordingly, fair value is appropriate.
It is similarly appropriate for shares held in a trad-
ing portfolio where the investor gets the return,
one-for-one, from the change in market price.

Minus: Fair (market) values are a minus when
the firm arbitrages market prices. That is, fair
value is not appropriate when the firm adds value
(for shareholders) by buying at (input) market
prices and selling at (output) market prices.

Raw material used in manufacturing does not get
its value from a change in its exit market price, but
as an input into a process that adds value to its
market price by producing a product and selling it
to customers; change in shareholder value is not
one-to-one with the change in the market price of
the input. With respect to stewardship, the manag-
er should not be rewarded on the basis of changes
in the market price of the raw material, but for
adding value (earnings) from buying the input
favourably and selling it, transformed, to cus-
tomers with a mark-up.

The one-to-one condition says that fair value is a
minus where firms are involved in (expectational)
arbitrage (of input and output) prices in their busi-
ness model; that is, the business model adds value
to market prices. Or stated differently, fair value is
not appropriate when there is a top-line notion of a
customer from whom value is received in an exit
price, with value added over an input price. Fair
value is appropriate when value comes from prop-
erty rights and obligations, and value is added or
lost (solely) from fluctuations in the market values
of those rights and obligations.11

Here are some cases where the one-to-one con-
dition for fair value applies:
1. Investments in securities in a trading portfolio

and derivative instruments on such securities.
2. Pension assets: The firm has performed by con-

tributing to the fund and has no influence on
the performance of the fund, but shareholder
welfare is affected directly by changes in the
market value of the fund.

3. Investments by an insurance company. In the
business model these securities are value in re-
serve and that value depends on market price.

4. Real estate held for speculation with no plan
for developing or utilising the real estate.

5. Options that give the counter party (but not the

firm) the call rights; the firm is a passive count-
er party. Warrants and call and put options on
the firm’s own stock are an example. Freddie
Mac and Fannie Mae mortgages are of this
type. These are essentially traded put options
on real estate – the right of property owners to
sell property back to these institutions.
Shareholders’ welfare is determined by the
counter party’s call, not the firm’s. The market
value of the instrument reflects the probability
of this call and changes in the market value re-
flect changes in shareholders’ welfare as this
probability changes.

Below are some cases where the one-to-one con-
dition for fair value does not apply:
1. Inventory: the firm adds value by finding a cus-

tomer.
2. Investment in a subsidiary where the firm has

influence.
3. Assets and liabilities whose value changes as

interest rates change but there is also a numer-
ator effect (on future earnings) as well as a de-
nominator effect from change in interest rates.
These typically are instruments that involve
customer relationships. Examples: commercial
loans, mortgages held by originating banks,
and core deposits. Historical cost accounting
allows one to observe the numerator effects.

4. Performance obligations. Fair value accounting
books the liability to perform at the price that
someone else would charge to satisfy the obli-
gation, not at the cost at which the company
can perform (possibly with comparative advan-
tage).

5. Receivable allowances and warranty liabilities.
Value to shareholders is based on firm perform-
ance in servicing these items (through its
credit department and customer service depart-
ment), not what the market would charge for
non-recourse relief from the obligation. (Note:
market values can be information for disciplin-
ing estimates, but as an exercise in improved
historical cost accounting, not as an application
of fair value accounting.)

6. Insurance assets and liabilities, other than in-
vestment assets.

7. Real estate held as input to business enterprise
(for example, real estate development, real es-
tate rentals). For real estate rentals, historical
cost accounting recognises value through rental
income in the income statement.

8. Environmental clean-up liabilities. Fair value
is the amount that someone would charge for
the clean up, not the anticipated cost to the firm
in managing the problem.

Special Issue: International Accounting Policy Forum. 2007 39

11 The perspective is similar to that under Coase’s transac-
tions cost theory of the firm. Firms exist because markets are
not perfect and thus prices do not measure value under all con-
ditions. Firms and their hierarchies are more efficient than
markets in some respect, entrepreneurs exploit those efficien-
cies, and historical cost accounting reports the efficiency of
firms is dealing with imperfect prices.

Fair values are particularly inappropriate when
they replace historical cost accounting from which
(fair) value is assessed:

Minus: Fair market prices are a minus if they
substitute for historical cost information and (effi-
cient) prices depend on historical cost information.

Carrying investments in a subsidiary at market
prices (rather than under the equity method or pro-
portionate consolidation) obscures the profitability
of the subsidiary and the value of the parent which
is based on that profitability. The value of an inter-
mediary function – adding value from the spread
between borrowing and lending rates – is obscured
by fair valuing loans and borrowings if those exit
values do not incorporate the firm-specific ‘intan-
gibles’ in customer and depositor relationships. So
with the insurance business that involves customer
relationships in managing premiums and losses,
along with operating costs.12

The loss of historical cost information can lead
to inefficient prices. The spectre of inefficient
prices raises another issue:

Minus: Fair values bring price bubbles into fi-
nancial statements.

Fair values (as exit prices) come with a caveat.
Provided the one-to-one condition is satisfied, fair
values are value to shareholder if market prices are
‘efficient’. In a price bubble, however, inefficient
prices are booked on the balance sheet, with bub-
ble gains flowing through to the income statement.
For trading portfolios where investments are held
short-term, this may not be a large problem. But
where the portfolio is held for the long-term, it is a
problem. For instance, pension assets marked to
bubble prices may give the appearance of satisfac-
tory funding of future pension obligations and in-
surance assets may give the appearance of
adequate or even excess reserves against future in-
surance losses.
Fair value matching

Fair value accounting is often promoted as a way
to avoid the ‘myriad of rules’ involved in imple-
menting revenue and expense matching in the in-
come statement. But fair value accounting has its
own matching concept that is difficult to implement.

Minus: Fair value accounting fails without asset
and liability matching.

Under a business model, assets and liabilities are
used jointly to generate value for shareholders. If
so, the stand-alone fair value of an asset has little
meaning. To capture value added (from exposure
to market prices), one matches fair values of all the
assets and liabilities that generate the value togeth-
er, leaving none out. A particular danger lies in fair
valuing an asset and not matching the fair value of
an associated liability whose price changes are
negatively correlated with those of the asset.

In the income statement, such fair value mis-
matching results in recognition of gains and not
offsetting losses (or vice versa). The case of nega-
tive correlation produces ‘excess volatility’ in
earnings about which one can indeed complain.
Note further that, even if fair values of individual
assets and liabilities are matched and summed, the
total may not capture the value of the group if they
are used synergistically. (In this case it is difficult
to see that the one-to-one condition would hold
anyway, so the point may be mute.).

Cases where mismatching can occur:
1. Core deposits. These are intangible assets for

banks which represent the ability to obtain rel-
atively inexpensive funds from demand, sav-
ings and small denomination time deposits.
Their value is difficult to estimate but it is neg-
atively related to the value of the loan portfo-
lio: When interest rates rise, the value of the
loan portfolio typically declines but the value
of the core deposits intangible asset typically
increases. If the loan portfolio is marked-to-
market but the value of the core deposits intan-
gible is not recognised, earnings and book
value will be artificially depressed. But the one-
to-one condition says that customer deposits
should not be fair valued. So, if loan portfolios
are fair valued, a mismatching occurs.

2. Borrowings. The decline in the value of a
firm’s assets (due to deteriorating profitability)
is accompanied by an offsetting decline in the
value of its debt obligations (due to deteriorat-
ing credit quality). If the decrease in the value
of liabilities is recognised as a gain in the in-
come statement, but the decrease in asset value
is not recognised (for example, due to difficul-
ties in measuring the value of some intangible
assets such as customer relationships), earnings
will be overstated at times when high quality
information is especially important.

The matching issue is particularly difficult when
an instrument whose value varies with price is
used in conjunction with assets and liabilities
whose value is tied to customer relationships. A
fair value option under IASB standards13 (and pro-
posed in a current FASB exposure draft14), at-

40 ACCOUNTING AND BUSINESS RESEARCH

12 This is not to say that information about the sensitivity of
earnings to changes in interest rates (for banks) or embedded
values (for insurance companies) are not relevant footnote in-
formation.

13 The IASB included a fair value option in International
Accounting Standard No. 39, Financial Instruments:
Recognition and Measurement (London: IASB, December
2003). An IASB amendment restricting the fair value option
was published in June 2005 under the title, The Fair Value
Option. The fair value option must be applied to a group of as-
sets and/or liabilities that is both managed, and it performance
evaluated, on a fair value basis.

14 FASB Proposed Statement of Financial Accounting
Standards, The Fair Value Option for Financial Assets and
Liabilities (Norwalk, Conn.: FASB, January 2006).

tempts to address the matching problem, but not so
if the asset or liability fair valued under the option
does not satisfy the one-to-one condition.

3.2. Pluses and minuses of Level 2 and Level 3
fair value measurements

Levels 2 and 3 admit estimates of hypothetical
market prices. Level 3, while insisting that the fair
value is based on an estimate of market price
(rather than value-in-use), permits ‘unobservable
inputs’ that ‘reflect the reporting entity’s own as-
sumptions about assumptions that market partici-
pants would use in pricing the asset or liability.’

The objections to using subjective estimates are
well understood. However, any accounting beyond
mere cash accounting involves estimates. The
question of where to draw the line on estimates
(Level 2 but not Level 3?) is difficult to handle a
priori, for resolution rides largely on one’s assess-
ment, not only of the integrity of managers but
also of their (honest) subjective biases. The com-
petence and independence of monitors – auditors,
assessors, and corporate boards – must also be
evaluated, along with the effectiveness of controls.
(Honest) managers are naturally optimistic, for it
is their business plan. Accounting, however, serves
as a counterweight to managements’ optimism, so
raising their estimates to the level of accounting
information contaminates. Some argue that such
estimates elicit information from management that
might not otherwise surface. The stewardship per-
spective underscores the downside; rewarding
managers based on their estimates exposes the
shareholder to moral hazard.

Here are a few points to consider when enter-
taining the use of estimated fair values.

First, the restriction in Section 3.1 that fair value
accounting applies only when shareholder value is
solely determined by exposure to market prices
means that, in most cases, there will be an active
market where Level 1 measurements are available.
If a firm has to execute by finding a customer in an
illiquid market, value is usually determined by that
ability to execute, not solely by market prices. So
situations where estimation is required may be
limited (if the one-to-one condition is honoured).

Second, one must question whether Level 3 real-
ly enforces a discipline in estimating market
prices. Using one’s own assumptions could yield
estimates resulting in upfront (day one) estimated
profits.

Third, fair value estimation errors introduce
error into the balance sheet but also the income
statement (which reports the change in fair value).
Indeed, with random errors in both the opening

and closing balance sheet – bias aside – the errors
are compounded in the income statement.15 If one
has a fuzzy balance sheet, fair value is less inform-
ative, but if one also loses the informative histori-
cal cost earnings, mark up a definite minus. In the
extreme, estimated fair values could produce an
uninformative balance sheet and a less uninforma-
tive income statement.

Fourth, it is sometimes said that historical cost
involves estimates and estimated fair values are no
different. But estimates to effect matching under
historical cost are based on, and audited against,
the historical transaction record – like the histori-
cal experience with credit losses, useful lives and
warranty service costs. Level 2, with ‘observable
inputs’, could be interpreted as invoking this no-
tion. But the notion is quite different from specu-
lating about the present value of the cash flows
when marking to model.

Fifth, historical cost estimates true up against the
actual transaction record, and usually fairly quick-
ly. Fair value estimates usually do so but, without
an associated ‘historical transaction accounting’,
estimated fair values settle up against estimated
fair values. So, the fair value of a long-term con-
tract on the output of an energy plant might be re-
estimated each year but, without a reporting on the
actual sales and expenses of running the plant each
year, the subsequent estimated fair values become
elusive. The same can be said about insurance con-
tracts: the fair value of an insurance contract is in-
formed by the historical experience reported in
premiums matched to losses and expenses. Note
that the FASB’s fair value accounting for employ-
ee stock options (at grant date) does not settle up
at all (so, if the option is not exercised, the record-
ed expense is not reversed).

Sixth, I suspect that an analyst will have difficul-
ty in carrying out a quality analysis on fair value
accounting. How would estimation errors, biased
or random, be discovered? Disclosures about the
valuation methodology are required under FASB
Statement 157 and these presumably will help. But
I am not clear on how earnings quality diagnostics
of the type applied under historical cost accounting
(again, with reference with what is normal in the
transaction history) would be developed.

Seventh, observed market behaviour is instruc-
tive. Hedge funds (largely unregulated) apply fair
value accounting and estimate fair values for illiq-
uid assets. They do so under the rigour of formal
valuation committees with oversight of their
boards and auditors. But there is a danger in share-
holders trading in and out of the fund at values
based on estimates. So funds typically limit the
percentage of illiquid assets held or require lock-
ups or transfers to side pockets until realisation.
Private equity funds typically require realisation
before distribution. In short, the tolerance for esti-

Special Issue: International Accounting Policy Forum. 2007 41

15 This effect is demonstrated formally in K. Peasnell,
‘Institution-specific Value’, BIS Working paper No. 210,
August 2006.

mated fair value (by shareholders) is limited.
Eighth, the research reported by Wayne

Landsman at this conference indicates that the in-
formativeness of fair values declines as estimates
are introduced.

4. Conclusion: pluses and minuses
In this paper I have taken a demand approach in
considering the pluses and minuses of fair value
accounting: Do fair values enhance the task of eq-
uity valuation and stewardship assessment? Surely
I have not exhausted this exploration, but some
points have been made. At a conceptual level, fair
value accounting is a plus; equity value is read
from the balance sheet, with no further analysis
needed, and the income statement reports realisa-
tions for determining value at risk. But concepts
are one thing and implementation another. With fair
value defined as exit price, the minuses add up (!).

Fair value accounting works well, for both valu-
ation and stewardship, with investment funds
(where shareholders trade in and out of the fund at
net asset value). This case is instructive for it is the
situation where the one-to-one relationship be-
tween exit prices and fair value to shareholders
holds. That one-to-one condition fails, however,
when a firm holds net assets whose value comes
from execution of a business plan rather than fluc-
tuations in market prices, even when exit prices
are observed in active markets. Asset and liability
matching problems confound the problem further.
Overlay the minuses of estimated fair values when
actual prices are not observed, and the minuses do
add up.

I have spent some time laying out the valuation
properties of historical cost accounting – better re-
ferred to as ‘historical transaction accounting’ –

because I sense that those properties are not al-
ways appreciated in the discussion of fair values.
A balance sheet focus is not necessary for valua-
tion, for we also have an income statement.
Historical cost income statements report earnings
from assets and valuation can be made from earn-
ings even though their fair value is not on the bal-
ance sheet. The Coca-Cola valuation in the
appendix makes the point. I don’t see a plus from
booking an estimate of the market price of Coke’s
brand asset to the balance sheet and then running
the revaluations through the income statement.
This scenario is not on the horizon, one would
think, but there is little difference in principle in
applying exit values to banks’ loans and deposits
that also get their value from brands and other cus-
tomer intangibles.

This having been said, the implementation of
historical cost accounting is not without its prob-
lems, and many criticisms of historical cost ac-
counting under current GAAP are well taken. The
analyst is frustrated by a number of features of
GAAP. I really have not engaged in tallying up the
pluses and minuses of fair values against the plus-
es and minuses of historical cost accounting with
all its measurement issues. But, it is difficult to see
how fair value accounting (with exit prices) solves
the problems with historical cost accounting when
the one-to-one condition is not satisfied. That con-
dition is a necessary condition for fair value ac-
counting. Further, while the implementation
problems with historical cost accounting are due to
difficulties of revenue and expense matching, fair
value accounting also has its own (asset and liabil-
ity) matching problems and these appear to be se-
rious ones.

42 ACCOUNTING AND BUSINESS RESEARCH

Special Issue: International Accounting Policy Forum. 2007 43

Appendix. A valuation of the Coca-Cola Company based on historical cost information

At the close of trading on 8 December 2006, the Coca-Cola Company’s shares traded at $48.91 each. The price-
to-book ratio was 6.3, indicating a lot of value missing from the balance sheet, largely because US GAAP does
not allow Coke’s intangible (brand) assets to be booked to the balance sheet. The forward P/E was 19.3, based
on analysts’ consensus EPS forecast for 2007.
The following valuation yields a value of $49.09 per share using only information available in the historical
cost financial statements. The valuation is crude (and can be refined), but the point is that we get close to the
market price by using historical cost information and, indeed, with three line items.

The historical cost numbers
Here are the relevant line items for years 2002–2005 ($m):

2005 2004 2003 2002
Sales (1) 21,962 21,044 19,656 17,545
Operating income, after tax (2) 5,065 4,427 4,192 3,841
Net operating assets (average) (3) 16,985 16,006 15,220 14,526

The financial statement analysis
From these line items, the following valuation inputs can be calculated:

2005 2004 2003 2002
Operating profit margin (2 ÷ 1) 23.1% 21.0% 21.3% 21.9%
Asset turnover (1 ÷ 3) 1.29 1.31 1.29 1.21
Average operating profit margin 21.8%
Average asset turnover 1.28
Average sales growth rate, on a base
of 2001 sales of $17,354m) 6.6%

The valuation model
We employ a standard residual income valuation model that calculates missing value in the balance sheet from
a forecast of forward (2006) operating income:

Value of Equity2005 = Book Value of Equity2005 + Residual Income from Operations2006
Required Return – Growth Rate

where

Residual Income from Operations2006 = Forecasted Operating Income2006 – (Required
Return x Net Operating Assets2005)

Only the residual income from operations is forecasted because residual earnings from interest on net debt are
usually close to zero.

The forecast
As the book value of equity and net operating assets for 2005 are in the 2005 financial statements, we need
only a forecast of operating income for 2006, the required return, and the growth rate for residual income.

• For the required return, we will use 10% which is approximately the current Treasury rate of 4.6% plus a risk
premium of 5.4%.

• If both the profit margin and the asset turnover are constant, then residual operating income grows at the sales
growth rate.16 The condition is approximately satisfied for Coke, so we set the growth rate at the sales growth
rate of 6.6%.

• The historical financial statements supply a forecast of operating income and residual operating income:

Forecasted sales for 2006 = Sales for 2005 x (1 + Average sales growth rate)
= $21,962 x 1.066
= $23,411

Forecasted operating income for 2006 = Sales for 2006 x Average profit margin
= $23,411 x 0.218
= $5,104

Forecasted residual operating income for 2006 = $5,104 – (0.10 x 17,113)
= $3,392

44 ACCOUNTING AND BUSINESS RESEARCH

Appendix. A valuation of the Coca-Cola Company based on historical cost information (continued)

The valuation
With a 2005 book value of equity of $15,935, the calculated value with these inputs is

Value of Equity2005 = $15,935 + 0003,392

0.10 – 0.066

= $115,700m, or $49.09 per share

The valuation is crude, by design, to make a point. It uses only information in the historical financial statements
(plus as assumed required return). Yet is comes quite close to the market price of $48.91. Adding more infor-
mation (about sales growth rates) and a different required return will change the valuation, but the historical
cost financial statements yield considerable insights. Most importantly, it challenges the notion that one needs
to have fair values on the balance sheet to value equity claims. Indeed, it is hard to see how fair value estimates
of assets and liabilities would enhance the valuation.

In choosing Coca-Cola, I am of course being selective; not all firms are as easy to value as Coke. The histor-
ical cost information for a bio-tech start-up with no product yet out of R&D is not much use for valuation, for
example. The financial reports would report losses and, possibly, negative book values. However, again, it is dif-
ficult to see how exit prices would redeem the accounting. Better for the analyst to get a biochemistry degree.

16 See S. Penman, Financial Statement Analysis and Security Valuation. 3rd ed. (New York: The McGraw-Hill Companies,
2007), p. 523.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.