THE THEORY OF STOCK

by Ray Ball,
MARKET EFFICIENCY:University ofRochester*
ACCOMPLISHMENT
S AND
LIMITATIONS

thirty years have passed since Eugene

ama introduced the idea of an
efficient stock market to the
financial economics literature, and it
continues to stimulate both insight and
controversy. Put simply, the idea is that
investors compete so fiercely in using
public information that they bid away its
value for earning additional returns. In so
doing, they quickly incorporate all
publicly available information into prices.
Harvard
economist
Benjamin
Friedman has dismissed the idea as a
credoa statement of faith, not
scientific research.1 Warren Buffett,
whose
legendary
investment
performance and philosophy have caused
some soul-searching among efficient
markets theorists, has described the
stock market as a slough of fear and
greed untethered to corporate realities.2
In my view, the concept of efficient stock
markets is one of the most important
ideas in economics.
In the theory of stock market
efficiency, public information that has not
been fully reflected in stock prices is like
gold lying in the streets; reports of either
are treated with equal skepticism. Take
the case of a company reporting a $3
increase in annual EPS when the
consensus forecast at the beginning of
that year was a $2 increase. How should
such information, which is essentially
free for investors to acquire the
moment it is placed in the public domain,
affect the companys stock price?

In an efficient market, the expected

part of the earnings increase ($2 of the
$3, in this case) should already be
reflected in the price, and investors
should trade on the new information (the
extra $1, or the earnings surprise) until
all the gains from so doing are competed
away. The stock price adjustment to the
information should be rapid (if not
instantaneous), and the new price
should make the stock a fair game
one which promises new investors a
normal rate of return.
In the mid-1960s, while working
together at the University of Chicago, Phil
Brown and I set out to test these
propositions by exploring how the stock
market
actually
responds
to
announcements of annual earnings. We
examined some 2300 annual earnings
reports by about 300 New York Stock
Exchange companies over the nine-year
period 1956-1964. After classifying each
of the 2300 reports as containing either
good news or bad news based on
past earnings, we calculated the stock
returns to an investor holding each of the
two groups of companies from one year
before
to
six
months
after
the
announcement.
What we found provided stronger
support for the theory than practitioners
in finance and accountingnot to
mention most of our academic colleagues
had expected. The results of our study,3
which are summarized graphically in
Figure 1, suggested that the market had

1
BANK OF AMERICA JOURNAL OF APPLIED CORPORATE FINANCE

already anticipated roughly 80% of the

surprise component in annual earnings
by
the
time
the
earnings
were
announced. There was some further
upward
drift
following
the

announcements of earnings increases

and
small
downward
movements
following decreases. But, over the entire
six-month period after the announce-

FIGURE 1

STOCK PRICE CHANGES IN

RELATION TO ANNUAL
EARNINGS ANNOUNCEMENTS*

EPS Increases
Market Index
EPS
Decreases

Month Relative to Earnings

Announcement
*Stock prices rise during the year before firms announce EPS increases as information flows into the market. Prices also fall in advance of EPS
decreases. There is some price reaction at the time of the announcement. The graph shows price movements for the average firm. Firms with
larger EPS changes tend to experience larger price movements. S( >urce: Ball and Brown (1968).

ments, investor returns from holding

each of the two groups of firms (those
with unexpected EPS increases and those
with decreases) would have been close to
zero. Thus, prices had incorporated the
information released in annual earnings
reports in a way that virtually eliminated
future opportunities to profit from that
news. (And, since publication of our study
in 1968, these results have been
replicated by many other studies of
different time periods in 15 different
countries.)
The impact of the simple idea of
market efficiency has been extensive and
enduring. Stimulated in large part by a
new research design known as the event

study (introduced in a 1969 study by

Fama, Fisher, Jensen, and Roll that I
discuss later), a large body of empirical
research in the 1970s provided consistent
evidence of the stock markets ability to
process information in a rational, in some
cases ingenious, fashion. This research so
transformed our view of stock markets
that contemporary observers cannot
begin to appreciate the general suspicion
of and ignorance about stock markets
that prevailed 30 years ago.
At the same time it was expanding
our knowledge of how the stock market
processes information, the early work on
market efficiency helped to establish a
receptive climate for three other major

2
BANK OF AMERICA JOURNAL OF APPLIED CORPORATE FINANCE

5
VOLUME 8 NUMBER 1 SPRING 1995

developments in financial theory: the

Miller- Modigliani theories of corporate
financial
policy;
the Sharpe-Lintner
Capital Asset Pricing Model (CAPM); and
the Black-Scholes option pricing model.
Each of these theoretical developments
has in turn affected financial practice in
important ways. The CAPM, for example,
is widely used today both by institutional
investors (in establishing their asset
allocation and performance measurement
criteria) and industrial corporations (in
setting their investment rate-of- return
targets). The Black-Scholes model is one
of the most popular, if not the dominant,
procedure for valuing stock options, as
well as many other derivatives and socalled contingent claims. And widespread practices such as indexing,
performance measurement, and asset
securitization are all unthinkable without
the idea of well-functioning security
markets.
Our current understanding of how
information is incorporated into stock
prices has also influenced corporate
disclosure policy. For example, it is now
common for corporations to release
between two and six different earnings
numbers (for example, those including
and excluding earnings from businesses
that were sold during the period, both
before and after the effects of accounting
method changes). Whereas the preefficiency mentality of accountants (and
the SEC) was to distrust investors
judgment and thus to choose the
earnings number for them, the dominant
corporate practice today, with the
blessing of the SEC, is to present and
then leave it to users to digest the full
range of information.
In
short,
the
theoretical
developments of the 60s and 70s have
helped transform the practice of finance.
And, in addition to these theoretical and
practical
achievements,
the
early
empirical testing of market efficiency
coincided
with
the
more
general
emergence of interest in and respect for
free markets that began in the 1970s,
first among economists and then later

among politicians. Indeed, the market

efficiency literature has helped pave the
way for what has proved to be a
worldwide liberalization of financial and
other markets.

3
BANK OF AMERICA JOURNAL OF APPLIED CORPORATE FINANCE

Despite
its
insights
and
accomplishments, however, the theory of
efficient markets has obvious limitations.
For example, it treats information as a
commodity that means the same thing to
all investors. It also assumeswith
potentially serious consequences, in my
viewthat information is costlessly
incorporated into prices. In reality, of
course,
investors
interpret
events
differently;
they
face
considerable
uncertainty about why others are trading;
and, especially in the case of smaller
firms, they face high costs (as a
percentage of firm value) in acquiring
and processing information. Given these
limitations, it is not surprising that
researchers soon began to accumulate
evidence that appeared to contradict the
theory. Whether the existence of such
anomalous" findings indicates a fatal
flaw in the theoryor, more likely, the
value of incorporating greater realism
only time and more research will tell.
My aim in this article is to provide an
admittedly
somewhat
personal
perspective on the accomplishments and
limitations of the theory, and to offer a
few suggestions about how the theory
might be modified to explain some of the
contradictory evidence. Briefly put, my
view is this: the theory of efficient
markets was an audacious and welcome
change from the comparative ignorance
about stock market behavior that
preceded it; and despite its now-obvious
theoretical and empirical flaws, it has
profoundly influenced both the theory
and practice of finance.

THE IDEA THAT MARKETS ARE

EFFICIENT
When Phil Brown and I started the
research for our 1968 paper, academics
in finance and economics conducted little
or no research on stock prices, and we
were strongly discouraged by highly
skeptical colleagues. The prevailing view
among academic economists, even at the
University of Chicago, was that the stock
market was not a proper subject of

serious study. While part of the

skepticism about our research stemmed
from economists' reservations about the
way accountants calculate earnings,
much more could be traced to a deep
suspicion of stock markets. Indeed, the
notion of modelling the stock market in
rational economic terms was academic
heresy at the time.
What little research had been
conducted on stock prices up to this point
had been performed for the most part by
statisticians.
The
scant
empirical
literature that preceded the efficient
market
hypothesis
modelled
price
behavior in statistical, not economic,
terms. Successive daily stock price
changes seemed for the most part
independent'7; that is, they displayed no
discernible trends or patterns that could
be exploited by investors. Prices on any
given day appeared likely to go up or
down with roughly equal probability.
Lacking an economic explanation for this
result, researchers described it in
statistical language as the random walk
hypothesis."
Moreover, it was a statistician at the
University of Chicago, Harry Roberts, who
first saw the potential import for
economics
of
the
random
walk"
literature.
In
an
important
paper
published in 1959, Roberts mused,
Perhaps
the
traditional
academic
suspicion about the stock market as an
object of scholarly research will be
overcome.' And, as he went on to say,
...there is a plausible rationale [for the
random walk model]. If the stock market
behaved like a mechan ically imperfect
roulette wheel, people would notice the
imperfections and, by acting on them,
remove them. This rationale is appealing,
if for no other reason than its value as
counterweight to the popular vieu of
stock market irrationality, but it is
obviously incomplete ,4
Among financial economists and
other students of the stock market,
statistical dependence in returns came1
to be viewed as valuable information just

4
BANK OF AMERICA JOURNAL OF APPLIED CORPORATE FINANCE

sitting there in the public domain, that is,

as so much gold in the streets. Any
predictable trends in prices over time
would mean that knowledge of past
returns (which is essentially costless to
acquire) could be used to predict future
returns. In short, dependence across time
1.

5.

was viewed as inconsistent with rational

behavior in competitive markets and thus
as evidence of what later became known
as market inefficiency.'*
The first use of the term efficient
market' appeared in a 1965 paper by
Eugene Fama, who defined it as:

11. V. Roberts. "Stock Market "Patterns" and Financial Analysis: Methodological Suggestions." Journal of l:iuance I t (1959). p. .
For a review of the random walk literature and precedents, see F.F. Fama. "Ffficient Capital Markets: A Review of Theory and Empirical
Work." Journal <>J' Finance!^ < 1970). 383-1 P.

Stimulated by a new research design known as the event study, a large body of empirical
research in the 1970s provided consistent evidence of the stock market's ability to process
information in a rational, in some cases ingenious, fashion.

a market where there are large numbers

of rational, profit-maximizers actively
competing, with each tlying to predict
future market values of individual
securities, and where important cu rrent
information is almost freely available to
all participants...
In an efficient market,'1 Fama argued,
on the average, competition will cause
the full effects of new information on
intrinsic values to be reflected
instantaneously in actual prices.6
The economics underlying this
model are very simple. Publicly-available
information by definition is accessible to
all investors at zero cost. While earnings
reports, for example, are costly for firms
to produce, once made public they are
nearly costless to obtain (though not
necessarily costless to interpret, as I
discuss later). And since revenue and
cost
are
equated
in
competitive
equilibrium, the implication of stock
market efficiency is that if the cost of
reproducing public information is zero,
then so are the expected gains. Security

6 F.F. Fama, "Random Walks

in Stoc k Market Prices.
h'inaucial Analysis Journal, September-October
( 196S). p.-i.

prices should therefore adjust to

information as soon as (if not before) it
becomes publicly available. Ideas don't
come much simpler in economics.

ACCOMPLISHMENTS OF THE
THEORY OF STOCK MARKET
EFFICIENCY
It is difficult to trace the influence of
ideas because their effects range from
the direct and concrete (such as the
effects on financial practice I noted
earlier) to the more subtle and abstract
(for example, their influence on how we
think about issues, and on the concepts
and language we use).
To gain some appreciation of how
thoroughly the research on market
efficiency has changed our thinking
about stock markets, imagine the
mindset of an observer of stock markets
prior to the research that began in the
late 60s. Ordinarily we view the markets
reaction to information through the lens
of chronological time. Reading the daily
financial press, for example, we observe
the market's response at a single point in
time to what is often a bewildering
variety of events and circumstances
affecting companies values. There are
announcements
of
earnings
and
dividend, new promotional campaigns,
labor disputes, staff retrenchments, new

5
BANK OF AMERICA JOURNAL OF APPLIED CORPORATE FINANCE

debt and equity issues, management

changes, proxy contests, asset write-offs,
bond rating changes, and changes in
interest rates, money supply figures, and
GDP data. Thus, what we see with the
unaided eye are price reactions to many
events at once, without seeing the
underlying pattern for any one of them.
In 1969, however, a study of the
stock market reaction to stock splits by
Eugena Fama, Lawrence Fisher, Michael
Jensen, and Richard Roll (hereafter FFJR)
introduced the concept of event time,
which may well have been the single
most important breakthrough in our
understanding of how stock prices
respond
to
new
information/
In
attempting to isolate the markets
reaction to stock splits (by eliminating the
surrounding noise of other events), FFJR
took many instances of the same event
occurring in many different companies at
different times, and standardized them
all in terms of a single event date
(designated in the literature as t=0). In
so doing, they were able to provide an
event-time view of the market's
reaction of the kind presented in Figure 2.
What does Figure 2 tell us?
Conventional wisdom held then, as it
does today, that stock splits are good
news for investors (and, although why
stock splits should be good news has long
been
something
of
a
puzzle
to
economists, FFJR reported a plausible
reason in their 1969 study: namely, 72%
of the splitting firms in their research
sample announced dividend increases
above the NYSE average in the year after
the split). What the conventional wisdom
may not have predicted, however, but
what emerges in Figure 2, is the speed
with which investors restore the market to
its equilibrium or fair game condition
after the announcement (as revealed by
the immediate flattening of the line after
t=0).
Thus, where we had previously seen
only the chaos of daily stock price
movements, FFJR s research design
enabled us to see order. And the large
empirical literature on market efficiency

that came after FFJR both refined their

event
study
technique
and
accumulated
impressive
evidence
unimaginable before the late '60sthat
stock prices respond in apparently
ingenious ways to information.7 One

S. For surveys with detailed analysis of individual studies and

their findings, see Fama ( 19^0). cited earlier: F.F. Fama. "Ffficient
Capital Markets: II," Journal of L'inance 16 (1991). Iri~5-I6l-; and S.F.
LeRoy. "Ffficient Capital Markets and Martingales." Journal <J
liconomic Literature 2~ (19H9), 1583*1021.

7to New Information," International

liconomic Rerieic 10 ( 1969). 1-2!.
6

BANK OF AMERICA JOURNAL OF APPLIED CORPORATE FINANCE

surprising and illuminating example: announcements of new public stock offerings are associated
with an immediate 3% average stock price
reduction. The prevailing explanation of this
negative market response is as follows: Investors
recognize that managers, as representatives of
existing shareholders, are more likely to issue new

one of the few areas in which a large amount of

data could be used to test a simple competitive
economic theory. And the data were unusually
plentiful. The University of Chicagos
Center for Research in Securities Prices (CRSP),
wrhich was established (with exceptional foresight)
by James H. Lorie in I960, provided comprehensive

FIGURE 2
STOCK PRICE CHANGES IN
RELATION TO STOCK
SPLITS*

-10-5

10

Month Relative to Split

Date

stock (rather than debt) when they think the

Stock prices rise during the 30 months before stock splits. Because all the information in the split is public knowledge bt the time it occurs, the
stock price reaction to the split is complete by "month 0." The graph shows price movements for the average firm. Firms with larger splits tend to
experience larger price movements. Source: FFJR (1969).

company is overvalued; hence announcements of

new stock offerings are interpreted, at least in the
average case, as conveying managers private
assessment of the firms prospects relative to its
current valuation.8 Another example worth noting
(in part for its touch of the bizarre): unexpected
CEO deaths are associated with price decreases in
the case of professional managers, as we might
expect in an economy where most companies are
well run. But, in the case of those CEOs who are
also the founders, death notices are accompanied
by price increases, with the implication that many
founders tend to hang on to their companies too
long.9
There was one other notable factor
contributing to the remarkable successes of
efficient markets theory in the late 1960s and
early 1970s. At that time, the stock market was

data on the universe of all NYSE stocks going back

to 1926. Such a rich source of datawhich, rare
for the time, came in machine-readable form and
were essentially error-freewas a luxury almost
unheard- of among economists.

Influencing the Climate for

Other Financial Economic
Theories
Together, then, the new theory and evidence
on market efficiency demonstrated to economists
for the first time that share price behavior could
be viewed as a rational economic phenomenon.
And this development in turn helped set the stage
for other breakthroughs.
Two other broad areas of financial-economic
enquiry were launched around the same time as
efficient market theory. First were the dividend
and capital structure irrelevance propositions
formulated by Merton Miller and Franco Modigliani
(first at Carnegie Mellon, later at the University of
Chicago) and presented in papers published in
1958 and 1961.10 Soon after came the capital
asset pricing model (CAPM), which was developed
separately by William Sharpe (at Stanford in a
1964 paper) and

8 For a nice summary of how the

market responds to new securities
issues, see C.W. Smith, Raising
Capital: Theory' and F'vidence,"
Midland Corporate Finance Journal, Vol. 4 No. 1
(Spring 1986), 6-22. This article
was based in turn on Investment
Banking and the Capital
11. F. Modigliani and M.H. Miller, The Cost of Capital, Corporation Finance, and The
Acquisition Process, "Journal of Financial
Theory of Investment," American Economic Review 48 (1958), 261-297; and
M.H. Miller and F. Modigliani, "Dividend Policy. Growth and the Valuation of
Economics Vol. 15 (1986), pp. 3-29.
Shares," Journal of Business 34 (1961), 411-4339 W.B. Johnson, R.P. Magee, N.J.
Nagarajan and H.A. Newman, An
Analysis of the Stock Price
Reaction to Sudden Executive
Deaths, Journal of Accounting
10and Economics 1 (1985), 151-174.
JOURNAL OF APPLIED CORPORATE
FINANCE

Whether the existence of such anomalous'1 findings indicates ? fatal Haw in

the theoryor, more likely, the value of incorporating greater realismonly
time and
more research will tell.

John Lintner (at Harvard in 1965). 11 The consistent

evidence of market efficiency offered strong
support for these concepts that shared similar
assumptions about the working of financial
markets. Indeed, without the impressive body of
empirical research on efficiency as background, it
is not clear that the equilibrium-based reasoning
of the Sharpe-Lintner CAPM (and later the BlackScholes option pricing model, introduced in 1973)
would have been so well and so quickly received.12
Development of the CAPM (as I discuss later) was
particularly important for subsequent research on
market efficiency because it provided researchers
with a method of estimating investors expected
returnsthe returns passive investors would
otherwise have earned in the absence of the
information being tested.
A considerable body of knowledge on corporate capital structure has also been built on the
work of Miller and Modigliani. The arbitrage-based
arguments
M&M
used
to
buttress
their
irrelevance propositions were a novelty in
corporate finance. They were developed over
approximately the same period that Roberts and
Fama were formulating their ideas on efficiency,
and they relied in part on the same basic
reasoning of competitive economic theory
underlying market efficiency. Among other
practical contributions, M&Ms then revolutionary
thinking on capital structure appears to have
supplied an important part of the logic for the
leveraged restructurings of the 1980s.13 As in the
case of the CAPM, finance scholars were more
receptive to the M&M propositions in large part
because of their consistency with the closelyrelated area of efficient markets.

Summing Up
Encouraged, then, by its initial accomplishments and reinforced by other theoretical

developments like the CAPM, research on market

efficiency spread rapidly across the globe, with
results that provided remarkably consistent
support for the theory. This research had an
enduring impact on our view of stock markets, of
financial markets generally, and perhaps of
markets of all kinds. Because this evidence tended
to contradict prior assumptions, which were based
on prejudice and casual observation rather than
on systematic research, academic attention was
quickly drawn to it. In a surprisingly short time,
academic attitudes toward stock markets had
shifted from one extreme to the other, from
suspicion to almost reverence.

LIMITATIONS IN THE THEORY

OF EFFICIENT MARKETS
It was not long, however, before researchers
began to report evidence that appeared to
contradict the theory. But, as I argue below,
finance scholars should not have been completely
surprised by such evidence. The theory of efficient
markets has not influenced theory and practice
because it is free of defects. Its influence has
been due to the insights it provides, not those it
fails to provide. The cup remains half full, even
though it is half empty.
I now discuss a number of limitations of the
theory that I have divided into three overlapping
categories: (1) the theorys failure to explain
certain aspects of share price behavior, referred
to in the literature as anomalies; (2) defects in
efficiency as a model of markets; and (3)
problems in testing the efficiency model.

Empirical Anomalies: Problems

in Fitting the Theory to the
Data

No theory can explain all the data writh which

it is confronted, especially when the data are as
abundant as stock prices. There is now a large
11W.F. Sharpe, Capital Asset
body of anomalous evidence that at least appears
Prices: A Theory of Market
to contradict market efficiency. The list of such
Equilibrium under
anomalies" includes:
12Conditions of Risk," Journal of Finance 19 Price Overreactions. There is evidence that the
prices
of
individual
stocks
overreact
to
(1964), 425-442; and J. Lintner,
information and then undergo corrections. The
"The
resulting negative correlations in prices appear to
13Valuation of Risk Assets and
create profit opportunities for contrarian trading
the Selection of Risky
strategies.H
Investments in Stock Portfolios Excess Volatility. Robert Shiller, among others,
and Capital Budgets,'' Review of Economics and has argued that the stock market in general
Statistics ^ (1965), 13-37.
overreacts to

15.

1 i. See M.H. Miller. The Modigliani-Miller Propositions Thirty Years After.

Journal of Applied Cotporate Finance \'o\. 2 No. 1 (Spring 1989).
W.E.M. l)e Bondt, and R.H. Thaler, Does the Stock Market Overreact?. Journal of
Finance. (1985). ^93-H05; and W.E.M. I)e Bondt and R.H. Thaler. Eurther
Evidence on Investor Overreaction and Stock Market Seasonal[ty." Journal of
Finance i2, 1987, 55--581.

9
VOLUME 8 NUMBER 1 SPRING 1995

events because of investors pursuit of fads and

other herd-like behavior. In support of his
argument, he presents evidence suggesting that
the volatility of stock prices is much too large to
be explained by the volatility of dividends.14
Price Underreactions to Earnings. In seeming
contradiction of the the tendency of prices to
overreact to information in general, research
indicates that prices underreact to quarterly
earnings reports, thus lending support for
earnings momentum strategies.15
The Failure of CAPM to Explain Returns. The
Capital Asset Pricing Model, as mentioned, has
provided the primary means for measuring
investors' expected returns; indeed, it was the
workhorse for calculating discount rates for an
entire generation of practitioners as well as
researchers. But evidence that high-beta stocks
do not earn higher returns than low-beta stocks
has led some to pronounce the horse dead. 16 (But
recent research suggests that announcements of
the death of beta may be premature.)17
The Explanatory Power of Non-CAPM
Factors. Stocks with (1) small capitalization, (2)
low market- to-book ratios, (3) high-dividend
yields, and (4) low P/E ratios tend to outperform
their expected returns (as estimated by the
CAPM), thus providing justification for yield-tilted,
small-cap,
and
other
popular
investment
strategies.
Seasonal Patterns. Researchers have provided
evidence of seasonal patterns of hourly, daily,
1.
monthly, and quarterly duration. Particularly puzzling is the weekend effect: the tendency of
stock returns to be negative over the period from
Fridays close to Mondays opening. Such
seasonals could lend support to technical
analysts claims to outperform market averages.
As I argue below, however, it is difficult to tell
whether such anomalies should be attributed to
defects in markets themselves, to flawrs in
market efficiency as a way of thinking about
2.
competitive markets, or to problems with the
research itself.

Defects in Efficiency as a
Model of Stock Markets

Failure to Incorporate Information Acquisition and Processing Costs. One of the most
important explanations of these anomalies is likely
to be the neglected role of information costs.
Information costs are neither new to economists
nor inconsistent with competitive markets.
Nevertheless, the costs of acquiring and
processing information have received scant
attention in the theory and empirical research on
stock market efficiency.
The first event studies such as the FFJR
study of stock splits and the Ball-Brown study of
annual earnings reports investigated relatively
simple cases in which information is wridely
disseminated in the financial press and on the
wire services to analysts and other investors. The
cost to individual investors of acquiring such
public information (as opposed to the cost of
producing that same information prior to
disclosure) w^as assumed to be negligible. The
information was also assumed to be simple to use,
with negligible processing (or interpretation)
costs.
But what about those academic research designs that simulate trading strategies that in fact
have substantial information acquisition and
processing costs? In such cases, the abnormal
returns reported by the research could be
significantly overstated. Consider the following:
If 1,000 professors of finance worldwide annually
incur costs of $50,000 each (including salary,
computer costs, and overhead) in searching for
anomalies, the expected return from their discovery of pricing errors in the market could be as
much as 5% per year (or $50 million) on a $1
billion portfolio. And, if they have been searching
for 20 years, the expected return from producing
such private information could be many times
larger.
But, if the expected returns from such strategies
are this large, why have some researchers
published their anomalous evidence? Why have
14 R.J. Shiller, Do Stock Prices they not instead used that information for private
gain?
Move Too Much to Bejustified by
3. More generally, what are the expected gains from
Subsequent
producing and trading on private information?
15Changes in Dividends?, American
4. What is the meaning of gains from a trading rule
Economic Review 71, (1981a), 421-436.
that is simulated using historical data together
with
modern
computing
and
statistical
167. V.I.. Bernard and J. Thomas.
techniques? (Does this differ from, say, the
Evidence That Stock Prices Do
simulated gains from owning a helicopter gunship
Not Fully
during the Middle Ages?)

17Reflect the Implications of

Current Earnings for Future
Earnings," Journal of Accounting and Economics 13
(1990), 305-340.

hS. For a survey, see M. Reinganum, The Collapse of the Efficient Market
Hypothesis: A Look at the Empirical Anomalies of the 1980s." Selected Readings
in Investment Management (Institutional Investor, 1988).

What is the predicted rate of return in a competitive market for an investor like
Warren Buffett who sits on the corporate board, is a close friend of the CEO,
and has a voice in major corporate decisions?
19.

For a defense of beta, see S.P. kothari and J. Shanken in this issue.

5. If an analyst forecasts the earnings of 25 stocks

each quarter at an annual cost of $200,000
(including salary and overhead), then the average
cost of a forecast is $2,000. This is one ten-thousandth of 1% of the market value of a $2 billion
firm. It is one tenth of 1% of the market value of
$2 million of stock in the firm held by clients. If we
further assume that the gains from forecasting are
approximately equal to the costs, can we reliably
detect such a low number in stock rates of return?
We have no well-developed answers to such
questions. To date, and for obvious reasons, researchers have tended to ignore information costs
and have relied on cases of publicly-disclosed
information (whose cost we somewhat arbitrarily
define as zero) in testing market efficiency.
The direction of the bias from ignoring
information costs is consistent with the anomalies
evidence. For example, information costs are a
likely candidate for explaining the small-firm
effect, the tendency of small-cap stocks to
produce higher returns. Such firms tend to have
small analyst followings, presumably because the
cost of acquiring and processing information is
large relative to the amounts invested in them.
Investors may rationally require higher returns
(than those predicted by the CAPM) to
compensate for their higher information costs (as
a percentage of value).
But little more can be said without knowing
more about the size of such information
processing and acquisition costs, and how such
costs are likely to affect actual (pre-cost) expected
returns. Moreover, there can be significant
differences among investors in the information
they possess, in the extent of their efforts to seek
information, and in their ability to process it.
Although the CAPM is silent on this issue, it would
be logical to expect more sophisticated, betterinformed, and more active investors to earn
higher returns. The question, however, is: How
much higher? For example, what is the predicted
rate of return in a competitive market for an
investor like Warren Buffett who sits on the
corporate board, is a close friend of the CEO, and
has a voice in major corporate decisions? Much of
Buffetts success may be attributed to his ability
to reduce the information (and "agency") costs
that confront ordinary passive investors. Lower
information costs could in turn be viewed as
reducing

Buffetts perceived risks and required rates of

return, thus making stocks more valuable to him
than to ordinary investors.
Heterogeneous Information and Beliefs.
This brings us to another, closely related
deficiency of efficient markets theoryone which
stems from its mechanical'' characterization of
investors as homogeneous, wholly objective,
information-processing
machines.
Investor
sentiment and confidence play no role, and
publicly available information is assumed to have
the same implications for all. In reality, of course,
individual investors can differ greatly in their
beliefs, and a single piece of informationsay,
IBMs latest earnings reportmay be interpreted
quite differently by different investors.
The assumption of homogeneous investors
was understandable at the time the theory was
developed. For one thing, it allowed researchers
to demonstrate to highly skeptical readers that
stock prices do respond to information, and in
rational wrays. But there were other grounds for
using such a clearly unrealistic assumption. As the
great Austrian economist F. A. Hayek argued in his
1945 classic, The Use of Knowledge in Society,20
market prices incorporate information that is not
and perhaps cannotbe fully know n by any
individual. For this reason, markets are likely to be
more rational in viewing information than the
individual investors who comprise them. Indeed,
that is the beauty of markets.
But this does not mean markets are infallible.
(To paraphrase Churchill on democracy, its not
that markets are perfect in allocating resources;
its just that all other systems are so much worse.)
For all their virtues, markets cannot fully
overcome the inevitable limits of individual
knowledgelimits
that,
under
certain
circumstances, can have pronounced effects on
market behavior.
To see how such limits might affect markets,
consider that the trading decisions of individual
investors are based not only on the individual's
information about a given stock or commodity,
but also on that individual's beliefs about the
information on w'hich other investors base their
trading. And because an investor possessing a
piece of information or holding a belief about the
future cannot know with certainty the extent to
which that information or belief is shared by
others, he does not know the

VOLUME 8 NUMBER11 SPRING 1995

extent to which the information is already

reflected in prices.
One plausible way of interpreting episodes of
high turnover and rapid price changes, such as
the stock market crash of 1987, is that investors
costs of becoming quickly informed about the
motives
for
others
trades
can
become
extraordinarily high. Under such conditions of
extreme uncertainty, the mechanistic model of
efficiency,
with
its
assumptions
of
zero
information costs and homogeneous investors,
clearly breaks down. In such cases, investors
required rates of return can be seen as rising
sharply in response to the sudden surge in price
volatility, and the general uncertainty is resolved
only gradually over time (in part through the
commentary of market analysts and others whose
economic role efficient markets theory has largely
ignored).
A central question for finance theory, then, is
whether events such as the crash of 87 are
unambiguous evidence of market irrationality. Are
we forced to conclude, along with Keynes (and the
new behaviorialist school of thought in finance),
that the market is a game of musical chairs in
which animal spirits regularly prevail over
reason? Or can such episodes plausibly be
described by Hayekian models in which rational
individuals with different information confront
extreme uncertainty?
Expanding efficient markets theory to accommodate information costs and differences among
investors may help us decide this issue. For, in
addition to the basic uncertainty that attends all
business activity, differences among investors
information and beliefs are an added source of
uncertainty and hence information costs; and, in
cases where such costs are high, investors may
expect to earn significantly higher returns.
A Brief Digression on the Role of
Security Analysts. As mentioned above,
efficient markets theory has virtually nothing to
say about the economic role of security analysts
and public commentary. The theorys silence on
security analysts has even led many academics to
view them as redundant.
What do stock analysts do? Besides talking to
management on a regular basis, they pay close
attention to economic events affecting the
companies they follow, and to the accompanying
market reactions. They also talk to large investors

about their reasons for buying or selling and so

attempt to gauge investor sentiment.' In so
doing, analysts can be viewed as reducing
investor uncertainty about the information and
trading motives of other investors.
For example, analysts earnings forecasts
function as estimates of the information about
near-term profitability that has already been built
into current prices. The existence of such
forecasts enables investors to compare their own
private forecasts against what amounts to a
public consensus. And by reducing one source of
investors uncertainty namely, about why others
are
tradingsecurities
analysts
reduce
information costs and raise the market value of
publicly traded companies. Media commentary on
markets can be viewed as playing a similar role.
Failure to Consider Transactions Costs.
Besides assuming zero information costs and
homogeneous investors, early versions of efficient
markets theory also assumed that the markets
themselves are costless to operate. But if stock
markets
are
largescale,
high-turnover
institutions whose primary purpose is to minimize
the transactions costs associated with helping
firms raise capital, they have not succeeded in
eliminating such costs.
Having said this, it is not clear what role
transactions costs play in setting stock prices. In a
paper published in 1978, Michael Jensen argued
that an efficient market can only be expected to
adjust prices within limits defined by the cost of
trading.21 According to this viewr, for example, if
transactions costs are 1%, an abnormal return to
a trading strategy of up to 1% would be
considered within the bounds of efficiency. It could
not be exploited for profit net of costs.
This view makes sense to traders, but it has
several shortcomings from a broader economic
perspective. First, there must exist some level of
transactions costs at which we are unwilling to
call a market efficient. If transactions costs were
extremely large, there would be few opportunities
to profit from price errors, net of costs.
Nevertheless, it makes little sense to describe a
market with extremely large costs of trading and
thus extremely large price errors as efficient.
Second, although transactions costs of x%
may be consistent with price errors of x%, they
are not consistent with price biases of that size.
Since most

Differences among investors' information ami hciiefs are an added Nource of uncertainty and hence

information costs: and, iwhere

such costs are
high,
investors may expect to earn signii
higherreturns.

event studies and related efficiency tests study

average returns, the pricing errors due to
transactions costs should net to zero over a broad
sample of trades; and thus transactions costs are
not obviously relevant to interpreting the results
of the studies.
Third, because transactions costs vary across
investors, defining efficiency in terms of
transactions costs can produce as many
definitions as there are investors. Whose
transactions costs are to be used in judging the
market to be efficient? A possible solution might
be to define efficiency in terms of the lowest-cost
trader. But, because some specialists and
institutional investors face transactions costs as
low as one tenth of one percent, this approach
can degenerate into ignoring transactions costs
entirely. Thus, the role of transactions costs in the
theory of market efficiency remains largely
unresolved.
Market Microstructure Effects. A closelyre- lated issue is the effect of the actual market
mechanism on transacted prices, known in the
literature as market microstructure effects.
Researchers typically assume that trades could be
executed at the closing prices recorded in data
files such as CRSPs. These price estimates,
however, have a margin of error that is as large as
their quoted spreads, which run about 3% for the
average NYSE-AMEX stock (and average close to
6% for NYSE-AMEX stocks with prices less than
$5). To compound the problem, investors cannot
always execute at quoted spreads because of
illiquidity.
Such market microstructure effects can be
quite large relative to the size of the short-term
abnormal returns reported by many event studies.
Consider the turn of the year or January effect
the widely-recognized tendency of stocks
(particularly low-price stocks) to produce large
returns in the first trading days of the calendar
year. A 1989 study by Donald Keim demonstrates
that this effect can be explained largely by a
pronounced (if unexplained) tendency of stocks to
trade near the bid at the end of the year and then

move toward the ask at the beginning.18

Similar microstructure effects may well also
be exaggerating the size of the short-term price
reversals, the seasonal patterns in daily or weekly
returns, and even the abnormal longer-term
returns reported in the anomalies literature. Such
effects are especially likely to influence the
findings of research on low-price stocks. In 1985,
for example, Werner DeBondt and Richard Thaler
published a study reporting very large gains from
a
contrarian
strategy
of
buying
stocks
immediately after large percentage declines and
holding them for five years.19 In a study published
this past year, however, two of my Rochester
colleagues and I discovered that DeBondt and
Thalers results were distorted by some very large
percentage gains on some very low-priced stocks.
One stock earned 3500% after trading at $ 1/16
(which is half of $1/8, the minimum increment in
which NYSE stocks are typically quoted). We also
found that adjustments of $1/8 in the prices of
their loser stocks would have reduced the fiveyear returns of this contrarian strategy by 2500
basis points.20
Like transactions costs, microstructure or
trading-mechanism
effects
present
the
researcher with a dilemma. Trading mechanisms
certainly are not costless to operate, and thus the
prices at which trades are transacted are likely to
be affected in systematic ways by institutional
arrangements. This seems particularly likely for

18 D.B. Keim, Trading Patterns,

Bid-ask Spreads and Estimated
Security Returns: The Case of
Common Stocks at Calendar Turning
Points," Journal of Financial Economics 25 (1989),
75-9B.
19 DeBondt and Thaler (1985),
cited earlier.
20 As a more general indication
of skewed returns, we found that
although the median returns of
winner" and loser stocks
differed by only 5%, the average
1

JOURNAL OF APPLIED
3 CORPORATE
FINANCE

small-capitalization and low-price stocks, and for

stocks and stock exchanges with low turnover.
But, unlike transactions costs, it seems
unreasonable to interpret price behavior caused
by the trading mechanism as evidence of market
inefficiency, because recorded prices are not
those at which the simulated trading strategies
could have been executed at the close of trading.
Yet taking this view to its limit leads here, as
in the case of transactions costs, to the perverse
outcome that the probability of a markets being
judged efficient increases with the size of bid/ask
spreads. Thus, as with transactions costs, the
precise role of market microstructure effects in
the theory of market efficiency is not apparent.

Problems in Testing Efficiency

as a Model of Stock Markets
As noted earlier, investors ability to exploit
public information for gain can be tested only by
returns differed by 57%. See R. Ball, S.P. Kothari and J. Shanken, Problems in
Measuring Portfolio Performance: An Application to Contrarian Investment
Strategies, Journal of Financial Economics 38 (1995), pp. 79-107.

comparing the returns earned when trading on

public information against the returns otherwise
expected from passive investing. In many
empirical tests of market efficiency, the Capital
Asset Pricing Model has been used to estimate
such expected returns. But, as also mentioned,
the CAPM has come under attack from empirical
studies showing at best weak correlations
between actual stock market returns and those
predicted by the CAPM. And because tests of
market efficiency are thus joint tests of both the
CAPMs success in measuring expectations and
the
markets
ability
to
incorporate
new
information in prices, flaws in the CAPM raise
doubts about the reliability of the existing
empirical researchboth the work that appears to
contradict market efficiency as well as that which
supports it.
Changes in Riskless Rates and Risk
Premiums. The CAPM states that a stocks
expected return is equal to the riskless rate (in
practice, the interest rates on Treasury notes or
bonds) plus a risk premium obtained by
multiplying the stock's beta times the market
equity risk premium (the amount a stock of
average risk is expected to earn above the
riskless rate). Expressed as an equation:
E(R) = Rf + B (Rm - Rf).

Unfortunately, the model is completely silent

on a number of important issues of practical as
well as theoretical import. For example, w hat
riskless rates, market risk premiums, and
individual betas would be consistent with an
efficient market? And how are such measures
expected to change over time, if at all? (Most
empirical tests of market efficiency simply assume
that riskless rates, market risk premiums, or betas
are constant.)
The second question is potentially troubling
because the more we allow both interest rates
and the market risk premium to vary over time,
the less ability we have to answer a question such
as: Are stock prices in general too volatile? In
1981, for example, Robert Shiller published a
study concluding that the historical variance of
stock prices over the 108-year period from 1871
to 1979 was much too large to be justified by the
variance of corporate dividends.2^ Using the
CAPM, however, its impossible to say whether
Shiller is right. Because of its assumption of
constant riskless rates and market risk premiums,
the CAPM provides no means of judging the
correct amount of variance in the market index.
Shillers test, for example, assumes in the
tradition of the CAPM that the market-wide
expected return is constant in nominal terms over
the entire sample period 1871-1979 in order to
place bounds on the price variance allowable in an
efficient market. Under this assumption, the
probability of rejecting market efficiency seems
almost 100%!21
But consider what happens when we relax
this assumption. Consider the S&P 500 as the
value of a claim on a perpetuity of expected
earnings produced by the 500 companies
comprising the index. If the real riskless rate is
assumed to be zero (close to its historical
average), then an increase in the risk premium
from 6% to 8% will reduce the value of the S&P
500 by 25%. Do we see too few or too many such
percentage declines in the S&P 500? It seems
quite plausible that market risk premiums could
change dramatically in response to sudden
changes in investors perceptions of political risk
and economic uncertainty. But the answer to this
question depends on how frequently the risk

21 This is a case of using an

excessive and misplaced faith in
the CAPM to

premium can be expected to change by as much

as 2%. (And, although the CAPM is silent on this
issue, researchers are now attempting to find
ways to address it.)
Trends in Real Rates and Market Risk
Premiums. Moreover, there is nothing in the
CAPM that would rule out serial correlation in
changes in either the riskless rate or the market
risk
premium.22
In
fact,
the
investment
opportunity set facing companies and investors
seems likely to exhibit positive serial correlation
that is, a tendency for increases to be followed by
further increasesdue to bursts of invention,
among other factors. A gradual aging of the
population seems likely to induce positive dependence in interest rates (as a growing number of
retirees seeks to live off income from investment
and a shrinking workforce operates those
investments).
In contrast, the market risk premium seems
particularly likely to have elements of negative
dependence. Consider, for example, how risk premiums would change if the level of general economic uncertainty perceived by investors varies
over time, but tends eventually to return to
normal levels. Increases in perceived risk will
cause the

27.

For some informal evidence of a sharp decline in market risk premiums over the
past 20 years, see J.R. Woolridge, "Do Stock Prices Reflect Fundamental
Values?. (in this issue).

22undermine what appears, in my

view at least, to be a more
reasonable belief in the
rationality of market pricing.
1
JOURNAL OF APPLIED
5 CORPORATE
FINANCE

Working along with such cycles in economic uncertainty and investor

confidence, changes in the supply of risky assets by corporations would reinforce
the pattern of negative serial correlation in investors required returns and
market risk premiums.

index to fall (even given the same level of

expected corporate profits) in order to give
investors a temporarily higher expected return
over the period of increased uncertainty. And, if
the uncertainties that trigger sharp price declines
are resolved over time, the index then tends to
rise.23
Working along with such cycles in economic
uncertainty and investor confidence, changes in
the supply of risky assets (due to the creation of
new assets and the obsolescence of old) by
corporations would reinforce this pattern of
negative serial correlation in investors required
returns and market risk premiums. For example,
suppose expectations of increased political and
economic stability are associated with rising stock
prices and P/E ratios (as has happened since the
recent U.S. House and Senate elections) in
response to declines in both interest rates and the
general
market
risk
premium.
For
U.S.
corporations, this means a reduction in their cost
of capital; and, in response to the lower cost of
capital, they can be expected to expand the
supply of new risky investments over the next few
years. As the aggregate supply of risky assets
increases, investors required rates of return will
tend to rise back toward normal levels over time,
causing prices and P/E ratios to fall.
Given the above assumptions, then (all of
which seem fairly plausible), an efficient market
would be expected to exhibit serial dependence in
both real interest rates and market risk
premiums. The point of these conjectures is to
illustrate that the cyclical patterns (or mean
reversion) in stock returns are not necessarily

evidence of the market irrationality asserted by

Shiller and the so-called behavioralists. Such
negative dependence is equally consistent with
rational responses by (1) investors to changes in
general political and economic conditions and (2)
corporations to the resulting changes in investor
demand (as reflected in interest rates and risk
premiums) for stocks.24
Changes in Betas. Finally, consider the
relation between beta and market efficiency.
Finance theory says nothing about how betas vary
both among companies and over time. It thus
offers no guidance as to what betas and what
patterns of changes in observed betas are
consistent with efficiency.
The current theory fails to account, for
example, for the fact that betas can be expected
to change in fairly predictable ways in response to
changes in stock price levels. To illustrate, if we
assume that a company has a given level of
overall risk (a fixed asset beta), the beta of the
companys stock (its equity beta) will rise along
with increases in its degree of financial leverage
(or debt-to-equity ratio). When expressed in
market value terms, moreover, a companys debtto-equity ratio will increase not only when it
issues more debt, but also when its stock price
falls. Loosely speaking, this means that when a
company's stock price drops sharply, investors
perception of the stocks risk and hence its
required return actually increases. Conversely,
when the stock price rises significantly, its
perceived risk and required retu rn fall
Such systematic changes in betas in
response to large stock price changes could

23 This point has been made by

F.F. Fama, Foundations of Finance (Basic
Books, New York), p. 149; K.C.
Chan. "On the Contrarian
Investment Strategy, Journal ofBusUiessGl
(1988), 147-163; K.F. Fama and
K.R. French, "Permanent and
Temporary Components of Stock
Prices," Journal of Political Fcouoni)96 (1988),
246-273; and R. Ball and S.P.
Kothari, "Nonstationary Expected
Returns: Implications for Serial
Correlation in Returns, Apparent
Price Reversals and Tests of
Market Efficiency, Journal of Financial Economics
25 (1989), 51-74.

24 The supply of risky investment

assets produced by corporate
investment decisions is typically
assumed by the CAPM to be fixed
and thus unaffected by changes in
prices and other financial
variables. But. without any
"supply-side" theory about how
corporations respond to changes
in real rates and risk premiums,
it is difficult to see how we
could be surprised by
c/wyobserved sequence of market
returns or use the sequence to
judge any market to be efficient
or inefficient. This

VOLUME 8 NUMBER I SPRING

1995

15

account for a number of the anomalies noted

earlier. For example, the longer-term mean
reversion of stock returnsthe tendency of high
stock returns to be followed by low ones, and vice
versadiscovered by Gene Fama and Ken French
could be explained largely by systematic shifts in
beta after price changes, as could the ability of
low P/E and low market-to- book stocks to
outperform market averages.25
Seasonal Patterns in Betas. To take
another example, betas could well exhibit
seasonals. There is no reason in principle why
securities relative risks, or even aggregate risk,
cannot vary with, say, the day-of-the-week.
Consider a firm that routinely makes its major
announcementsearnings,
dividends,
investments, acquisitionson Thursdays,

30.

theory including the DCF present value model, the option pricing model, and
efficient market theory itself. Each of these financial models is concerned
exclusively with investor demand for (and hence the price of) a certain asset in
relation to other assets with given prices. Indeed, I would argue that the failure
to consider the supply of investments is perhaps the greatest deficiency in
modern J'inancial economic theory.
Ball and Kothari (1989) provide evidence that endogenous risk variation explains
much of the serial correlation observed in Fama and French (1988). A related
issue is the evidence of apparent price reversals in De Bondt and Thaler (1985,
1987). Both Chan (1988) and Ball and Kothari (1989) report that the equity
betas of the most extreme losing" stocks can be expected to increase because
of the extreme increase in their market debt-equity ratios. Conversely, the
extreme "winning stocks are likely to have equity beta decreases due to
extreme leverage reductions.

25is a deficiency not only of the

CAPM, but of all major models in
modern finance

VOLUME 8 NUMBER I SPRING

1995

perhaps because it schedules board meetings on

that day. Thursday returns then will exhibit higher
variance and, possibly, covariance. To the extent
this is so, what appear to be seasonal patterns of
superior returns may turn out to be normal ones
when adjusted for the associated temporary increases in risk.

IS BEHAVIORAL FINANCE THE ANSWER?

Given such limitations in the concept of
market efficiency and the existing empirical tests
of the theory, a group of finance scholars known
as behavioralists has suggested that it is time to
abandon the premise of collectively rational
investors on which the theory rests. Behavioralists
argue that stock price corrections' and cycles
reflect systematic biases in how investors use
information. Investors are said to place too much
weight
on
current
information,
focusing
myopically on short-term earnings, for example,
while ignoring companies' long- run propects. And
stock prices accordingly are viewed as highly
unreliable guides to corporate resource allocation
and other important decisions. But, is behavioral
finance the answer to the limitations of efficient
markets theory?
I dont think so. First of all, the profit opportunities created by such alleged investor myopia
seem grossly inconsistent with competitive
markets. They are much too large to be credible,
thus casting doubt on the research methods.
Moreover, the failure of the vast majority of
professional
money
managers
(including
contrarians as well as earnings momentum types)
to outperform market averages with any degree
of consistency should reinforce our skepticism
about such claims.
Second, behavioral finance has anomalies of
its own. The extensive evidence of post-earningsan- nouncement drift cited earlier means that
investors place too little weight not too much, on
recent earnings information. Showing that a
theory has anomalies is one thing; replacing it
with an anomaly- free theory is another.
Finally, behavioralists have also criticized the
objectivity and openness of the research process
over the 1960s and 70s, arguing that the process
was slow to accept dissenting views. Benjamin
Friedman, for example, comments:
It is no coincidence that the research [in
support of efficient markets theory]...has emerged
almost entirely from the nations business
schools. From the perspective of the privateinterest orientation of these institutions, this
identification is both understandable and
appropriated
While Friedman is correct that the idea of

efficiency arose in business schools, his further

implication that the schools private-interest
orientation influenced the research findings
represents not just a distortion, but in fact a
complete inversion of the incentives of both
researchers and teachers. In my experience,
businessmen and MBA students have been more
receptive to the message that there is gold in the
streets than to the lesson that fiercely competitive
financial markets eliminate the easy opportunities
for gain.
Many critics of the literature on market efficiency have suggested that, it took too long to
recognize anomalies. Consider this recent
statement by Richard Thaler in his introduction to
a collection of behavioralist studies in finance:
The financial ivorld as described in Eugene
Fama's 1970 efficient markets survey was one
that had no urgent need for people. Markets were
efficient, prices were unpredictable, and financial
economists did not know how to spell the word
"anomaly. 52
This statement, perhaps precisely because it
is made from the secure vantage of hindsight, is
strongly at odds with my own view of how the
research evolved. In our 1968 paper cited at the
beginning, Phil Brown and I mentioned anomalous
evidence of post-earnings-announcement drift in
prices. Since this was the second event study
ever conducted (in the first, FFJR found no signs of
anomalies) and was published just three years
after the first time the words efficient market
appeared in print, it seems that the efficient
markets literature was not slow in introducing
anomalous evidence.
Moreover,
the
word
anomaly
was
introduced to the financial economics literature in
a paper I published in 1978 that surveyed a
number of puzzling findings related to earnings
and dividends.26 While that survey encountered
some resistance and

26 R. Ball, 'Anomalies in
Relationships Between Securities'
Yields and Yield- Surrogates, Journal
of Financial Economics 6, (1978), 103-126. I
borrowed the word anomaly from
Thomas Kuhns famous book, The Structure of
Scientific Revolutions (University of Chicago
Press, 1970), which introduced
the term to the

1
JOURNAL OF Mll.ll.l8> CORPORATE
FINANCE

The theoretical developments of the 60s and 70s have helped transform the
practice
of finance.

took two years to get published, other

anomalies were uncovered in the
meantime. Recognizing the pattern,
Michael Jensen quickly put together
an issue of the Journal of Financial
Economics devoted entirely to studies
of anomalies that was published in
the same year.
Overall, then, critics of the
literatures
responsiveness
to
anomalous results seem to have
overstated their case. In retrospect, I
am surprised by how quickly the
literature embraced the idea of
anomalies in the late 70s. We had
much more difficulty convincing
people that markets were efficient a
decade earlier.

CONCLUDING OBSERVATIONS
Are stock markets efficient? Yes
and no. On the one hand, the
research provides insights into stock
price behavior that were previously
unimaginable. On the other hand, as
research has come to show, the
theory of efficient markets is, like all
theories, an imperfect and limited
way of viewing stock markets. The
issue will be impossible to solve
conclusively while there are so many
binding limitations to the asset
pricing models that underlie empirical
tests of market efficiency.
Our models of asset pricing
began w^ith the CAPM and thus have
an accumulated history of only 30
years. With such a limited tradition in
asset pricing, we could hardly expect
to have a strong basis for concluding
that security prices do (or do not)
immediately restore equilibrium in
response
to
new
information,
particularly in the presence of
extreme uncertainty. Bearing such
constraints in mind, I believe that
much of the evidence on stock price
behavior does not and cannot reliably
address
the
factual
issue
of
efficiency at this point in time. Our

models of price equilibrium and our

data are not yet up to the task.
This is not to say that empirical
research cannot be informative under
these
circumstances.
We
have
learned a great deal from the
anomaly chasing of the last decade,
and there are enough puzzles and
clues to work on for some time.
Nevertheless, I do not believe that we
can reliably learn much about
whether (or to what extent) markets
are efficient from these results.
Do we have a better theory? I
think not. But, as researchers comb
through the evidence and seek more
realistic ways of modelling stock price
behavior, we are bound to improve
the theory over time.
Have we learned much from the
theory of stock market efficiency? I
think so. And I think this is a better
question. The economic theory of
competitive markets now seems
unlikely to be dislodged from its
central role in stock market research,
for several reasons. First, stock
markets must rank highly among
markets on a priori likelihood of being
competitive: there are no entry
barriers; there are many buyers and
sellers, who by and large appear to be
greedy and enterprising people; and
transaction costs are relatively low.
Second, as outlined above, the inevitability of binding limitations in both
our models of efficiency and in the
available data provides a credible
alternative explanation for the range
of
anomalies
that
have
been
documented over the last decade.
Third, I personally went through the
transformation from the pre-EMH view
of securities markets, and I am still
impressed by how well prices respond
to information, relative to what we
expected th i rty yea rs ago.
philosophy of science and scientific method. Kuhns book, I
might note, was published in the same year as Famas
review of the EMH literature, and only eight years before the
word appeared in the financial economics literature.

VOLUME 8 NUMBER 1 SPRING 1995

* RAY BALL
is Wesray Professor of Accounting at the University of Rochesters Simon School of Business.

17

VOLUME 8 NUMBER 1 SPRING 1995

7. K.F. Fama. L. Fisher. M.C.Jensen and R. Roll. "The Adjustment of Stock Prices
13- Teachers and researchers made frequent reference to the Sharpe-Lintner
model well before it was formally tested (by Black. Jensen, and Scholes in 19~2
and Eama and MacBeth in 1973), relying in part on the strong evidence of market
efficiency and on the perceiv ed overlapping of the two ideas.
25. Shiller (1981), cited earlier.