Professional Documents
Culture Documents
Market Effiency
Market Effiency
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
American Accounting Association is collaborating with JSTOR to digitize, preserve and extend
access to The Accounting Review
Market Efficiency
William H. Beaver
ABSTRACT: Efficient market research arose in response to allegations from the profes-
sional investment community and critics of financial accounting, and it preceded a formal,
conceptual development of market efficiency. Several ambiguities exist with respect to
previous definitions of market efficiency. Market efficiency is defined here in terms of the
equality of security prices under two information configurations (i.e., with and without
universal access to the information system of interest). Casually, a securities market is
efficient with respect to an information system if and only if security prices act as if
everyone knows that information system. If this condition holds, prices are said to "fully
reflect" the.information system. Several attributes of this definition are advantageous,
relative to those of earlier definitions.
The purpose here is to discuss some This research is partially supported by the Stanford
important ambiguities associated with Program in Professional Accounting, major contributors
the concept of market efficiency and to to which are: Arthur Andersen & Co.; Arthur Young &
Company; Coopers & Lybrand; Deloitte Haskins &
offer a precise definition of the concept. Sells; Ernst & Whinney; and Peat, Marwick, Mitchell &
Some of these issues have been discussed Co. I am indebted to the members of the Stanford
elsewhere, although mostly outside of the Accounting Workshop for their many helpful sugges-
tions on an earlier version of this paper.
accounting literature, narrowly defined.
These ambiguities make the widely William H. Beaver is Thomas D. Dee II
cited definitions of market efficiency Professor of Accounting, Stanford Uni-
conceptually incomplete and deficient in versity.
a fundamental sense. This can be some-
Manuscript received April, 1979.
what discomforting, given the sizable Revisions received September, 1979 and February, 1980.
empirical literature cited in favor of Accepted March, 1980.
23
conditions under which market efficiency or process by which security prices are
would or would not be expected to formed. Investors come to the securities
obtain. Potentially, theory can provide market with a diverse set of endowments
predictions regarding differences in mar- of current and future consumption, pref-
ket efficiency among different securities erences for current and future consump-
or differences in market efficiency with tion, and beliefs about future states of the
respect to different types of information world. The security prices arise in a
systems. In other words, theory can serve series of ' spot" markets that periodically
as a guide to future empirical research reopen (e.g., daily). The reopening of
and can complement the predominantly these markets implies that these markets
empirical tradition of efficient market are incomplete at least in a nominal
research. One of the premises of our sense. Whether they permit the same
research culture is that the "generaliza- allocation that would be attained in
bility" (i.e., information content) of complete markets is another matter
empirical evidence is enhanced by the [Arrow, 1964]. In incomplete markets,
presence of a conceptual, theoretical expectations are formed about future
underpinning. Until recently, the re- prices, and equilibrium can be charac-
search has existed with little or no formal terized as dependent on those expecta-
foundation. The empirical findings have tions. One of the earliest discussions of
largely preceded a formal, conceptual the relationship between current prices
development of market efficiency. and individuals' expectations appears in
It is difficult to interpret empirical Muth's [1961] work on rational expec-
evidence as a test of market efficiency in tations.
the absence of a definition of what the Under uncertainty, capital market
concept means. A definition permits equilibrium can be characterized as a
researchers who offer their research as mapping from endowments, preferences,
evidence regarding market efficiency to and beliefs into prices. Individuals' be-
show how their evidence is a test of liefs will be conditioned upon the infor-
market efficiency. In other words, there mation which each received. Hence,
can be a link between the conceptual and equilibrium prices at the time t will in
the empirical levels, and the researchers part depend upon the signals received at
can state precisely what additional as- time t by each individual.3
sumptions are involved in making that
transition.2 The issues will be illustrated 2 This is not to say that a precise conceptual definition
is a sufficient condition for an unambiguous interpreta-
throughout by the research concerning
tion of empirical work. Other aspects of the research
changes in depreciation methods. design may introduce ambiguities.
The analysis will consist of five sec- In general, individuals may not agree on the prob-
ability of future states, but typically equilibrium models
tions: (I) discussion of equilibrium se-
assume agreement on the prices and other relevant
curity prices, (II) definition of market attributes associated with each state [Radner, 1974].
efficiency, (III) attributes of the defini- Under information asymmetries among individuals, the
partitioning of states over which there is conditional
tion, (IV) theories of market efficiency,
price agreement is a nontrivial issue. Presumably, the
and (V) implications for empirical re- attributes that partition the states must be publicly
search. observable in order to posit conditional price agreement
[Radner, 1968, 1972; Demski, 1974]. Information
asymmetries also raise nontrivial issues about the mean-
I. DISCUSSION OF EQUILIBRIUM
ing, existence, nature, and stability of an equilibrium.
SECURITY PRICES For example, the concept of what is meant by equilibrium
must be extended to incorporate the information asym-
Market efficiency is viewed here as a metry. These issues have been explored by Wilson
property of an equilibrium mechanism [1977], Holmstrom [1977], Kobayashi [1977], Gjesdal
For convenience and precision, some Even apart from the influence of endow-
minor notation is introduced. The in- ments and preferences, prices are also a
formation system of individual i at time t
function of beliefs. If probabilistic assess-
is denoted Titi The configuration of in- ments are treated from a subjectivist,
formation systems across individuals at personalistic viewpoint, the meaning of
time t is it, * Ihit}X intrinsic value is ambiguous in a world
where I is the number of of individuals. The
heterogeneous beliefs, even if individ-
signal received by individual i at time t uals were identical with respect to endow-
from Tit is denoted yit, and the configura- ments and preferences. Notwithstanding
tion of signals at time t is {Ylt, , its ambiguity, notions of intrinsic value
Yit, , Yit} {Ylt}. The beliefs of in- continue to play a role in discussions of
dividual i at time t about future security the price formation process. For example,
prices that will prevail at time t + k the Financial Accounting Standards
(for k > 0) is denoted Board's Tentative Conclusions [1976, p.
32] provides a discussion of intrinsic
fit(Plt + k, * * * X, Pit +ken .. * Pit +k|17it, Yit)3
value, which, in turn, draws upon the
where J is the number of securities. The definition provided by Lorie and Hamil-
configuration of beliefs at time t is ton [1973, p. 114]:
{.ftt* ) ** *9 it( f **X t(t * )}-{Pitt * )}-Intrinsic value is the value that the security
ought to have and will have when other inves-
The price of security j at time t is Pjt, andtors have the same insight and know14-dge as
the configuration of prices is {P1,..., the analyst.
market to be inefficient. Yet, by the tions can be contrived that make the
above definition, the market would be definition of market efficiency circular.
said to be inefficient because of such For example, define information available
individuals. as that which is fully reflected in prices. In
3. Moreover, heterogeneity of beliefs this case, all securities markets are effi-
may prevail among individuals even if cient by definition. However, the defini-
they possess identical information with tion was intended more as an intuitive
respect to firm-specific and economy- description of the concept rather than as
wide data. For example, they may have a rigorous definition. Fama [1970] pro-
purchased different education and may vided the more formal "fair game"
interpret signals in a different fashion model:
because of different educational back- zj, = + rj,, + E(Pjt + 1 | st) (2)
grounds. Efficient market research is
typically concerned with only a subset of E(zj, + 1kt) = 0
the total information in an economy. rj,+1 = the realized return on
security j in period t + 1
Educational differences are not of im-
(where return is defined
mediate concern. Again, it would seem
as the percentage
helpful if the definition of market effi-
change in security price
ciency could admit to heterogeneous
adjusted for dividends
beliefs, which arise because of informa-
received),
tion differences beyond the scope of
interest. E(Tjt+1 Dt) = the expected return on
The purpose of this section has been to
security in period t + 1,
provide a brief characterization of se-
conditional upon Jt,
curity prices. In this setting, the role of Ot = the information set as-
sumed to be fully re-
heterogeneity was explored with the in-
flected in prices in pe-
tention of providing a presumption that
riod t,
it might be helpful if the definition of
market efficiency were able to accommo- zjt+ 1 = abnormal return on se-
curity j in period t + 1.
date such heterogeneity. The analysis
now turns to the issue of defining market Expression (2) states that the abnor-
efficiency. mal return Zjp+ 1 on security j in period
t?+1 is the actual, realized return rj1+ 1
II. DEFINITION OF MARKET EFFICIENCY minus the expected return E(Pijt +P1|t).
Subsequent notions of market effi- E(Pj t + 1 jDt) is conditioned upon Ft, which
ciency did not rely explicitly on intrinsic is the information "set" assumed to be
value and are illustrated by the following fully reflected in prices at time t. Accord-
definition, which appears in Fama [1970, ing to Fama, in an efficient market, the
p. 384]. expected value of the abnormal return
is zero.
... in an efficient market prices "fully
There are several ambiguities asso-
reflect" the information available .. . (1)
ciated with Expression (2). First, Le Roy
This definition has been extensively cited [1976], among others, has criticized this
and criticized, by Fama [1970] among model as being tautological in that it
others, on the grounds that the terms merely implies the expected deviation of a
fully reflect and information available arerealization from its expected value is zero.
vague and nonoperational. In fact, defini- The model is open to this interpretation
because it is not explicit about what in- erally, the market efficiency condition
formation conditions the distribution of states that prices act as if it holds.
the rj+ 1 term in Expression (2). Fama More precisely, a distinction will be
[1976] denies the tautology but admits made between information system effi-
that it is not empirically testable unless ciency (called j-efficiency) and signal
some equilibrium model of security re- efficiency (called y-efficiency) as follows:
turns is specified.'0 Second, It is not
y-efficiency
well-defined. Notwithstanding the am-
biguity of the phrase assumed to be fully A securities market is efficient with
reflected in prices, the meaning of the respect to a signal yt if and only if
term information set is unclear. It is the configuration of security prices
unclear whether it reflects to every possi- {P4jt is the same as it would be in
ble signal (i.e., with respect to the entire an otherwise identical economy
information system) or merely to the sig- (i.e., with an identical configura-
nal that is empirically observed. Third, tion of preferences and endow-
Expression (2) is belief-oriented, yet ments) except that every indivi-
dual receives yt as well as yt. (3)
E(ijj,+l j 1 F) is not well specified under
heterogeneous beliefs. q-efficiency
Subsequent attempts to repair the
A securities market is efficient with
definitional problems have defined mar-
respect to qt, if and only if y-effi-
ket efficiency as the equality of some
ciency holds for every signal (yt)
aspect of equilibrium under two different
from C. (4)
information configurations. The three
aspects selected have included equality of Under Expression (3) (or (4)), prices will
beliefs, equality of actions (i.e., portfolio be said to fully reflect y (or rt). The signal
choice), or equality of prices. Examples which includes both yt and yit will be
appear in Le Roy [1976], Fama [1976], denoted y*. The configuration of ex-
Rubinstein [1975], and Beja [1976]. The tended signals is {y*}. The configuration
following subsection will offer a defini- of the extended information systems is
tion and section III will discuss its at- {lt 1}-
tributes. This distinction removes the ambiguity
A. Information System Efficiency and
introduced by the use of the term infor-
Signal Efficiency mation set. This ambiguity arises because
As defined here, market efficiency with the term was used to describe previous
respect to an information item means empirical studies. These studies deal only
that prices act as if everyone knows that with observed signals. Presumably, such
information." For example, market effi- signals were viewed as sampling from a
ciency with respect to changes in the process, where information system effi-
depreciation method for annual report
purposes (see Comiskey [1971], Archi- 10 The "fair game" model as stated in Expression (2)
has also been criticized for concentrating solely on the
bald [1972], Ball [1972], and Kaplan and expected values of the return distribution and ignoring
Roll [1972], among others) means the higher moments. However, Fama subsequently [1976]
market prices act as if there is universal repaired this particular problem by defining market
efficiency in terms of the entire return distribution.
knowledge of the change in accounting
l This definition originated with William Sharpe of
methods. Even though the universal Stanford University and is also referred to in Beja [1976].
knowledge condition may not hold lit- It is similar to one offered by Rubinstein [1975].
tions (i.e., portfolio choice). For example, not be altered). If the market is already
the portfolio positions could differ under efficient with respect to that information
{,it} vs. {it} even though prices are the as defined in Expression (4), the policy
same. As indicated earlier, the empirical maker can be assured that the release of
research was responding to assertions by such data will not alter security prices.
analysts, accountants, and others that 4. The definition defines market effi-
prices did not fully reflect certain infor- ciency in a specific informational setting
mation. Hence, price-oriented empirical (e.g., Ct). As such, it permits as fine a
research and a price-oriented definition partitioning of the market efficiency con-
are most directly responsive to such dition (e.g., according to each con-
allegations. ceivable Ct) as the researcher may wish.
However, even if prices are one impor- In this context, the distinction between
tant consequence, this does not deny strong, semi-strong, and weak form
that there may also be other concerns efficiency can be viewed as a coarse par-
which may require action-oriented defini- titioning of information systems of in-
tions of market efficiency. For example, terest. This three-fold distinction is con-
Beaver [1973] cites "improperly" diversi- venient for many purposes such as the
fied portfolios as one of the consequences classification of previous empirical
or "costs" that may be incurred by an research as in Fama [1970]. However, in
investor even in an efficient market. other cases, it may not be precise enough,
Consider the "myopic" individual of if the market is efficient with respect to
Section 1. This is merely a particular some ij within a given category, but not
illustration of a broader set of conse- others. Expression (4) focuses upon
quences that may not be reflected in equilibrium price conditions with and
price-oriented definitions. Investors may without universal access to Ct. As such, it
not be indifferent between two informa- avoids severe definitions of market effi-
tion settings (with and without universal ciency and the use of terms such as the
access to i'). They may not choose the information available, the information the
same portfolios, even though the prices market uses, or the true probability distri-
are efficient with respect to q'. To address bution, which have appeared in other
this issue fully requires a specification of definitions and whose meanings have not
the purpose or intent of studying market been well defined.
efficiency and more explicit assumptions
about the equilibrium process. Defini- IV. THEORIES OF MARKET EFFICIENCY
tions which focus on price effects have the It is one matter to define market effi-
property of being descriptive of a class ofciency, but it is another to construct an
extant empirical research, which in turn equilibrium model where market effi-
focuses on one (albeit only one) aspect of ciency with respect to q' (or yt) would
the equilibrium. obtain. A theory of market efficiency
Even where the policy maker is con- involves a specification of the process by
cerned with the effects of information on which information becomes reflected in
portfolio holdings, a price-oriented defi- prices. Under what conditions would
nition of market efficiency can still be of
market efficiency with respect to qt be
interest. For example, consider a policy expected? Unless individuals are charac-
decision to release some information terized as throwing away something of
publicly in an exchange setting (i.e., value, information in not used because it
financing and production decisions will
is costly (e.g., costly to purchase qj or
costly to observe yt). Yet much of the efficiency with respect to more informa-
empirical research has examined market tion than merely the knowledge that a
efficiency with respect to publicly avail- change took place.
able information, such as changes in At an institutional level, some (e.g.,
depreciation methods for annual report Bernstein [1975]) appear willing to give
purposes. Why would one ever expect the analyst community credit for being
prices not to "fully reflect" publicly the mechanism by which market effi-
available information? Won't market ciency is attained. Such information
efficiency hold trivially? The answer, of intermediaries have not yet been ex-
course, lies in the possibility that such plicitly modeled. However, work by
data are not universally available at zero Kihlstrom and Mirman [1975], Gross-
cost to all individuals. man [1976], and Grossman and Stiglitz
Knowledge of the change may not be [1975] provide the beginnings of an
universal. Moreover, there may not be analysis of the conditions under which
universal (costless) access to other infor- market efficiency would or would not
mation regarding the implications of the obtain. In these models, individuals
change. In other words, it may be costly "extract" information from prices. In the
to obtain the accounting training (e.g., Grossman and Stiglitz model, individ-
knowledge of depreciation methods, un- uals choose to become informed or
derstanding a firm can have a different uninformed, and at equilibrium each
set of books for tax and for annual report individual is indifferent, because either
purposes, etc.) and hence such knowledge action (after deducting information costs)
is not universal. However, interpretation offers the same expected utility. In order
of the change goes beyond this and in- for there to be incentives to purchase
volves the assessment of managements' information, prices cannot "fully reflect"
motivations for changing depreciation the information obtained. In the Gross-
methods (e.g., reflecting managements' man model, every individual is equally
expectations about future earnings, or uninformed in that each receives a
plans for additional asset acquisitions). garbled signal, but prices act as an aggre-
Such analysis (i.e., information) is pro- gation of everyone's information, such
vided by the financial and accounting that the prices "reflect" information that
community, but perhaps not costlessly. is superior to that held by each and every
Hence, market prices might not reflect individual. However, individuals extract
this potentially costly information. This this superior information from prices,
constitutes a simplified explanation of and the prices "fully reflect" that superior
why a nonzero probability of market information.
inefficiency with respect to "publicly In a related vein, Verrecchia [1979] has
available" information might be as- constructed a model where price acts as
sessed. However, the cost of the informa- an aggregator of beliefs (as distinct from
tion might be such that security prices act an aggregator of information). As the
as if the information were costlessly number of individuals increases, prices
available to all investors. This "as if" behave as if everyone observed the un-
behavior is one interpretation of the garbled signal. The major difference is
empirical research on semi-strong form that there is no explicit learning from
efficiency. From this perspective, the prices involved. Hence, while prices re-
empirical studies of change in accounting flect the ungarbled signal, individual
methods are viewed as testing market beliefs or portfolios may not reflect this
information. Neither Grossman nor Ver- whether the firm has changed deprecia-
recchia requires the existence of a subset tion methods or, alternatively, is it a
of "initially" more informed individuals broader information system, as discussed
(e.g., superior analysts). earlier?
The models are partial equilibrium, 2. It permits the researcher to specify
where assumptions regarding the con- whether system efficiency or signal effi-
tracting opportunities permitted are cen- ciency is being tested. If the former, what
tral. For example, these analyses are theory leads the researcher to assume
essentially models of private information generalizability to unobserved and un-
production. Had contracts which per- tested signals from the system? If the
mitted collective production (or non- latter, why is the more restricted view
production) of information been con- adopted? Is it merely because of a lack
sidered, the information production of justification for a broader view or
incentives could have been dramatically because of some theoretical reason?
altered. One such collective choice would 3. It permits the researcher to define
be the centralized production of a signal, the information system as broadly or
thereby avoiding the costs of redundant narrowly, as desired. Again, presumably
production. Many of the issues related to theoretical considerations could at least
the institutional structure of information partially motivate the classifications of
production are discussed in Gonedes information systems.
[1976]. 4. It permits the researcher to dis-
tinguish the concept from the manner in
V. IMPLICATIONS FOR EMPIRICAL which the concept is tested. The advan-
RESEARCH tages of doing so were discussed earlier.
5. It permits the researcher to move
The previous discussion has several
rigorously from the conceptual level to
potential implications for empirical re-
the empirical level. The additional as-
search in market efficiency. A precise
sumptions in such a movement can be
definition will facilitate theoretical de-
specified, and the researcher can be more
velopment. As such, it can provide a
precise about the reasons why the empiri-
specification of the conditions under
cal evidence constitutes an indirect test. 16
which market efficiency would or would
Needless to say, the empirical studies
not obtain. For example, would differ-
do not directly test the conditions out-
ences be expected across securities or
lined in Expressions (3) or (4). The prices
across information systems? The em-
that would prevail in the hypothetical
pirical evidence would follow in response
to empirically testable implications of the economy of universal possession of qt are
not observed. Instead, the empirical
theory rather than in response to allega-
studies infer the security price behavior
tions by analysts, critics, and regulators.
In other words, the theory can serve as a
that would be observed if such a condition
held. A prominent example is the change
guide to future empirical research.
in depreciation methods. In this case, the
An explicit adoption of this formal
early studies held that if the market were
definition enables the researcher to be
more precise in several respects:
16 These five implications for empirical research are
1. It permits the researcher to specify not unique to the particular formal definition offered
here. They presumably would flow from any formal
precisely what is the information system
definition which did not suffer from the ambiguities
under study. For example, is it only cited earlier.
efficient with respect to this information, game" model, which suffers from several
the changes in equilibrium prices, ceteris ambiguities, as discussed previously.
paribus, would be zero. By way of residualMoreover, the ability to earn abnormal
analysis and other aspects of the research returns does not distinguish between
design, an approximation of the ceteris information "fully reflected" in prices
paribus condition was attempted. As and information "never reflected" in
Gonedes and Dopuch [1974] point out, prices." Both situations can imply an
in testing for market efficiency such inability to earn abnormal returns, yet it
studies have assumed the information is potentially important to distinguish
effect would be zero in a market which between them. For example, from a
fully reflected the information. A con- policy perspective, requiring public dis-
textual argument (i.e., no direct impact closure of such data for exchange pur-
on cash flows) was typically offered to poses would have no effect on stock
justify such an assumption. prices in the "fully reflect" case but would
Historically, there has been a consider-in general have price effects in the "never
able emphasis upon the ability to earn reflect" case. Moreover, critics such as
abnormal expected returns as a test of Briloff [1972] contend that certain mis-
market efficiency. Its prominence is un- pricing was "permanent" (i.e., for an
derstandable, given the heritage of mar- indeterminate length of time), and, if so,
ket efficiency. As indicated earlier, the it would not be detected by an abnormal
origin of market efficiency is the profes- returns test. Finally, market efficiency
sional investment community which sells studies have not been exclusively con-
services to clients on the basis that cerned with an ability to earn abnormal
inefficiencies will lead to abnormal re- returns. Most prominently, consider the
turns. From this perspective, a market studies of the change in depreciation
inefficiency that cannot be converted into methods. The studies were concerned
an abnormal return is "academic." As a with the "announcement effect" of the
result, it is natural that much of the earnings reported under the new depreci-
research on weak form efficiency and the ation method. In other words, did equi-
early work on semi-strong form efficiency librium prices change when depreciation
adopted an abnormal return testing per- methods were changed? Addressing the
spective. Under this perspective, a trading issue in this manner is consistent with
strategy is adopted based upon some
publicly available signal. 1 The ability to earn abnormal returns rests in part
Notwithstanding this tradition, Ex- upon what information is "fully reflected" in prices at
twi'o points in time (the beginning and end of the period
pressions (3) and (4) define market
over which the return is defined). Hence, it refers t6 a
efficiency in terms of an equality of dynamic property of prices over time. It may be intui-
equilibrium prices under two informa- tively appealing to conjecture that market efficiency at
both points in time (with respect to a') implies the in-
tion configurations. By contrast, they do
ability to earn abnormal returns (from knowledge of '')
not describe market efficiency in terms over that interval. However, it is difficult to argue that
of the ability to earn abnormal expected the "ability to earn abnormal returns" is an implication
returns (i.e., as in the "fair game" model).of a definition without specifying the definition. A
definition is a prerequisite to the derivation of implica-
The ability to earn abnormal returns is a tions.
potential implication of market efficiency 18 This feature underscores the fact that the ability to
but not a definition of it. 17 earn abnormal returns is an implication, not a definition
of market efficiency. A definition is an if and only if
A formal statement of "the ability to statement, while an implication goes in one direction but
earn abnormal returns" is the "fair the converse is not necessarily implied.
REFERENCES
Radner, R., "Competitive Equilibrium under Uncertainty," Econometrica (January 1968), pp. 31-58.
,"Existence of Plans, Prices, and Price Expectations in a Sequence of Markets," Econometrica
(1972), pp. 289-303.
,"Market Equilibrium and Uncertainty: Concepts and Problems," Frontiers of Quantitative
Economics (Vol. II) (Editors: M. Intriligator and D. Kendrick (North-Holland Publishing Company,
1974), pp. 43-105.
Roll, R., "A Critique of the Asset Pricing Theory's Tests; Part I: On Past and Potential Testability of the
Theory," Journal of Financial Economics (March 1977), pp. 129-176.
,"Ambiguity When Performance is Measured by the Securities Market Line," Journal of Financ
(September 1978), pp. 1051-1069.
Ross, S., "The Current Status of the Capital Asset Pricing Model (CAPM)," Journal of Finance (June
1978), pp. 885-901.
Rubinstein, M., "Securities Market Efficiency in an Arrow-Debreu Economy," American Economic
Review (December 1975), pp. 812-824.
Verrecchia, R., "On the Theory of Market Information Efficiency," Journal of Accounting and Econo
(March 1979), pp. 77-90.
Wilson, R., "The Core of an Economy with Differential Information," (Technical Report No. 220,
IMSSS, Stanford University, 1977).