Ebook Copeland''s financial theory and corporate policy: Part 2

PART

II

Corporate
Policy: Theory,
Evidence, and
Applications
T

HE FIRST PART OF THE text covers most of what
has come to be recognized as a unified theory of decision
making under uncertainty as applied to the field of finance.
The theory of finance, as presented in the first half of the text, is applicable to a wide
range of finance topics. The theoretical foundations are prerequisite to almost any
of the traditional subject areas in finance curricula; e.g., portfolio management,
corporation finance, commercial banking, money and capital markets, financial institutions, security analysis, international finance, investment banking, speculative
markets, insurance, and case studies in finance. Since all these topics require a thorough
understanding of decision making under uncertainty, all use the theory of finance.
The second half of this text focuses, for the most part, on applications of the
theory of finance to a corporate setting. The fundamental issues are: Does financing
matter? Does the type of financing (debt or equity) have any real effect on the value
of the firm? Does the form of financial payment (dividends or capital gains) have any
effect on the value of claims held by various classes of security holders?
Because these issues are usually discussed in the context of corporate finance
they may seem to be narrow. This is not the case. First of all, the definition of a
corporation is very broad. The class of corporations includes not only manufacturing
firms but also commercial banks, savings and loan associations, many brokerage
houses, some investment banks, and even the major security exchanges. Second, the
debt equity decision applies to all individuals as well as all corporations. Therefore
although the language is narrow, the issues are very broad indeed. They affect almost

every economic entity in the private sector of the economy.

357


358

CORPORATE POLICY: THEORY, EVIDENCE, AND APPLICATIONS

As we shall see, the theoretical answer to the question "Does financing matter?"
is often a loud and resounding "Maybe." Often the answer depends on the assumptions
of the model employed to study the problem. Under different sets of assumptions,
different and even opposite answers are possible. This is extremely disquieting to the
student of finance. Therefore we have presented empirical evidence related to each
of the theoretical hypotheses. Frequently, but not always, the preponderance of evidence supports a single conclusion.
It is important to keep in mind that hypotheses cannot be tested by the realism
of the assumptions used to derive them. What counts for a positive science is the
development of theories that yield valid and meaningful predictions about observed
phenomena. On the first pass, what counts is whether or not the hypothesis is consistent with the evidence at hand. Further testing involves deducing new facts capable
of being observed but not previously known, then checking those deduced facts
against additional empirical evidence. As students of finance, which seeks to be a
positive science, we must not only understand the theory, but also study the empirical
evidence in order to determine which hypothesis is validated.
Chapter 11 is devoted to various empirical studies related to the efficient market
hypothesis. Most of the evidence is consistent with the weak and semistrong forms
of market efficiency but inconsistent with the strong form. In certain situations,
individuals with inside information appear to be able to earn abnormal returns. In
particular, corporate insiders can beat the market when trading in the securities of
their firm. Also, block traders can earn abnormal returns when they trade at the
block price, as can purchasers of new equity issues. The last two situations will surely

lead to further research because current theory cannot explain why, in the absence
of barriers to entry, there appear to be inexplicable abnormal rates of return. Chapter
12 returns to the theoretical problem of how to evaluate multiperiod investments in
a world with uncertainty. It shows the set of assumptions necessary in order to extend
the simple one-period CAPM rules into a multiperiod world. It also discusses two
interesting applied issues: the abandonment problem, and the technique for discounting uncertain costs.
Chapter 13 explores the theory of capital structure and the cost of capital. This
is the first of the corporate policy questions that relate to whether or not the value
of the firm is affected by the type of financing it chooses. Also, we define a cost of
capital that is consistent with the objective of maximizing the wealth of the current
shareholders of the firm. This helps to complete, in a consistent fashion, the theory
of project selection. Capital budgeting decisions that are consistent with shareholder
wealth maximization require use of the correct technique (the NPV criterion), the
correct definition of cash flows (operating cash flows after taxes), and the correct cost
of capital definition.
Chapter 14 discusses empirical evidence on whether or not the debt-to-equity
ratio (i.e., the type of financing) affects the value of the firm. This is one of the most
difficult empirical issues in finance. Although not conclusive, the evidence is consistent
with increases in the value of the firm resulting from increasing debt (up to some
range) in the capital structure. However, much work remains to be done in this area.
Chapter 14 also provides a short example of how to actually compute the cost of
capital.


CORPORATE POLICY: THEORY, EVIDENCE, AND APPLICATIONS

359

Chapter 15 looks at the relationship between dividend policy and the value of
the firm. There are several competing theories. However, the dominant argument

seems to be that the value of an all-equity firm depends on the expected returns from
current and future investment and not on the form in which the returns are paid out.
If investment is held constant, it makes no difference whether the firm pays out high
or low dividends. On the other hand, a firm's announcement of increase in dividend
payout may be interpreted as a signal by shareholders that the firm anticipates permanently higher levels of return from investment, and of course, higher returns on
investment will result in higher share prices.
Chapter 16 presents empirical evidence on the relationship between dividend
policy and the value of the firm that, for the most part, seems to be consistent with
the theory—namely, that dividend policy does not affect shareholders' wealth. The
chapter also applies the valuation models (presented in Chapter 15) to an example.
Chapter 17 uses the subject of leasing to bring together a number of further
applications of capital structure and cost of capital issues. We also illustrate how
option pricing can help clarify the nature of an operating lease under which the lessee
may exercise a contractual right to cancel (with some notice and with moderate
penalties). Chapter 18 discusses several applied topics of interest to chief financial
officers pension-fund management, executive compensation, leveraged buyouts,
ESOP's and interest rate swaps.
Chapters 19 and 20 consider the widespread phenomenon of mergers. They begin
with the proposition that without synergy, value additivity holds in mergers as it
does in other types of capital budgeting analysis. Mergers do not affect value unless
the underlying determinants of value—the patterns of future cash flows or the applicable capitalization factors are changed by combining firms. Empirical tests of
mergers indicate that the shareholders of acquired firms benefit, on the average, but
the shareholders of acquiring firms experience neither significant benefit nor harm.
Chapters 21 and 22 conclude the book by placing finance in its increasingly
important international setting. A framework for analyzing the international financial
decisions of business firms is developed by summarizing the applicable fundamental propositions. The Fisher effect, which states that nominal interest rates reflect
anticipated rates of inflation, is carried over to its international implications. This
leads to the Interest Rate Parity Theorem, which states that the current forward
exchange rate for a country's currency in relation to the currency of another country
will reflect the present interest rate differentials between the two countries. The

Purchasing Power Parity Theorem states that the difference between the current spot
exchange rate and the future spot exchange rate of a country's currency in relation
to the currency of another country will reflect the ratio of the rates of price changes
of their internationally traded goods. We point out that exchange risk is a "myth"
in the sense that departures from fundamental parity theorems reflect changes in
underlying demand and supply conditions that would cause business risks even if
international markets were not involved. The fundamental relations provide the
principles to guide firms in adjusting their policies to the fluctuations in the exchange
rate values of the currencies in which their business is conducted.



The only valid statement is that the current price embodies all
knowledge, all expectations and all discounts that infringe upon the
market.
C. W. J. Granger and 0. Morgenstern, Predictability of Stock Market
Prices, Heath Lexington Books, Lexington, Mass., 1970, 20

Efficient Capital Markets:
Evidence

Empirical evidence for or against the hypothesis that capital markets are efficient
takes many forms. This chapter is arranged in topical order rather than chronological
order, degree of sophistication, or type of market efficiency being tested. Not all the
articles mentioned completely support the efficient market hypothesis. However, most
agree that capital markets are efficient in the weak and semistrong forms but not in
the strong form. The majority of the studies are very recent, dating from the late 1960s
and continuing up to the most recently published papers. Usually capital market
efficiency has been tested in the large and sophisticated capital markets of developed
countries. Therefore one must be careful to limit any conclusions to the appropriate

arena from which they are drawn. Research into the efficiency of capital markets is
an ongoing process, and the work is being extended to include assets other than common stock as well as smaller and less sophisticated marketplaces.

A. EMPIRICAL MODELS USED FOR
RESIDUAL ANALYSIS
Before discussing the empirical tests of market efficiency it is useful to review the three
basic types of empirical models that are frequently employed. The differences between
them are important. The simplest model, called the market model, simply argues that

361


362

EFFICIENT CAPITAL MARKETS: EVIDENCE

returns on security j are linearly related to returns on a "market" portfolio. Mathematically, the market model is described by
Rit = a; + ki Rint + Cit .

(11.1)

The market model is not supported by any theory. It assumes that the slope and
intercept terms are constant over the time period during which the model is fit to the
available data. This is a strong assumption, particularly if the time series is long.
The second model uses the capital asset pricing theory. It requires the intercept
term to be equal to the risk-free rate, or the rate of return on the minimum variance
zero-beta portfolio, both of which change over time. This CAPM-based methodology
is written
[Rnz, — R ft ][1j + E ft .

Rit = R ft

(7.32)

Note, however, that systematic risk is assumed to remain constant over the interval
of estimation. The use of the CAPM for residual analysis was explained at the end
of Chapter 10.
Finally, we sometimes see the empirical market line, which was explained in Chapter 7 and is written as
Rjt

j1

) 0t

Lfljt

8

jt•

(7.36)

Although related to the CAPM, it does not require the intercept term to equal the
risk-free rate. Instead, both the intercept, ')i ot , and the slope, j)s,„ are the best linear
estimates taken from cross-section data each time period (typically each month).
urthermore, it has the advantage that no parameters are assumed to be constant
over time.
All three models use the residual term, cit , as a measure of risk-adjusted abnormal
performance. However, only one of the models, the second, relies exactly on the theoretical specification of the Sharpe-Lintner capital asset pricing model.
In each of the empirical studies discussed, we shall mention the empirical technique by name because the market model is not subject to Roll's critique (discussed

in Chapter 7), whereas the CAPM and the empirical market line are. Thus residual
analysis that employs the CAPM or the empirical market line may be subject to
criticism.

B. ACCOUNTING INFORMATION
Market efficiency requires that security prices instantaneously and fully reflect all
available relevant information. But what information is relevant? And how fast do
security prices really react to new information? The answers to these questions are
of particular interest to corporate officers who report the performance of their firm
to the public; to the accounting profession, which audits these reports; and to the
Securities and Exchange Commission, which regulates securities information.
The market value of assets is the present value of their cash flows discounted at
the appropriate risk-adjusted rate. Investors should care only about the cash flow


ACCOUNTING INFORMATION

363

Table 11.1 FIFO versus LIFO

Revenue
Cost of goods sold
Operating income
Taxes at 40%
Net income
eps (100 shares)
Cash flow per share

LIFO

FIFO

100
90
10
4
6
.06
.96

100
25
75
30
45
.45
.70

Inventory at Cost

Fourth item 90 —> LIFO
Third item
60
Second item 40
First item
25 —> FIFO

implications of various corporate decisions. However, corporations report accounting
definitions of earnings, not cash flow, and frequently the two are not related. Does

an efficient market look at the effect of managerial decisions on earnings per share
(eps) or cash flow? This is not an unimportant question, because frequently managers
are observed to maximize eps rather than cash flow because they believe that the
market value of the company depends on reported eps, when in fact (as we shall see)
it does not.
Inventory accounting provides a good example of a situation where managerial
decisions have opposite effects on eps and cash flow. During an inflationary economy
the cost of producing the most recent inventory continues to rise. On the books, inventory is recorded at cost so that in the example given in Table 11.1 the fourth item
added to the inventory costs more to produce than the first. If management elects
to use first-in-first-out (FIFO) accounting, it will record a cost of goods sold of $25
against a revenue of $100 when an item is sold from inventory. This results in eps of
$.45. On the other hand, if LIFO (last-in-first-out) is used, eps is $.06. The impact
of the two accounting treatments on cash flow is in exactly the opposite direction.
Because the goods were manufactured in past time periods, the actual costs of production are sunk costs and irrelevant to current decision making. Therefore current
cash flows are revenues less taxes. The cost of goods sold is a noncash charge. Therefore, with FIFO, cash flow per share is $.70, whereas with LIFO it is $.96. LIFO
provides more cash flow because taxes are lower.
If investors really value cash flow and not eps, we should expect to see stock prices
rise when firms announce a switch from FIFO to LIFO accounting during inflationary periods. Sunder [1973, 1975] collected a sample of 110 firms that switched
from FIFO to LIFO between 1946 and 1966 and 22 firms that switched from LIFO
to FIFO. His procedure was to look at the pattern of cumulative average residuals
from the CAPM. A residual return is the difference between the actual return and the
return estimated by the model:

e j, = R j, — E(R

sit ).

The usual technique is to estimate a ft over an interval surrounding the economic event
of interest. Taking monthly data, Sunder used all observations of returns except for

364

EFFICIENT CAPITAL MARKETS: EVIDENCE

those occurring plus or minus 12 months around the announcement of the inventoryaccounting change. He then used the estimated ,6";„ the actual risk-free rate, and the
actual market return during the 24-month period around the announcement date to
predict the expected return.' Differences between estimated and actual returns were
then averaged across all companies for each month. The average abnormal return in
a given month is
1

N

AR E = —
e.,
N 1=1

where

N = the number of companies.

The cumulative average return (CAR) is the sum of average abnormal returns over
all months from the start of the data up to and including the current month, T:
CAR =

E=t

AR E ,

where
T = the number of months being summed (T = 1, 2, . . , M),
M = the total number of months in the sample.
If there were no abnormal change in the value of the firm associated with the switch
from FIFO to LIFO, we should observe no pattern in the residuals. They would
fluctuate around zero and on the average would equal zero. In other words, we would
have a fair game. Figure 11.1 shows Sunder's results. Assuming that risk does not
change during the 24-month period, the cumulative average residuals for the firms
switching to LIFO rise by 5.3% during the 12 months prior to the announcement of
the accounting change. This is consistent with the fact that shareholders actually value
cash flow, not eps. However, it does not necessarily mean that a switch to LIFO
causes higher value. Almost all studies of this type, which focus on a particular phenomenon, suffer from what has come to be known as postselection bias. In this case,
firms may decide to switch to LIFO because they are already doing well and their
value may have risen for that reason, not because of the switch in accounting method.
Either way, Sunder's results are inconsistent with the fact that shareholders look only
at changes in eps in order to value common stock. He finds no evidence that the
switch to LIFO lowered value even though it did lower eps.
More recently Ricks [1982] studied a set of 354 NYSE- and AMEX-listed firms
that switched to LIFO in 1974. He computed their earnings "as if" they never
switched and found that the firms that switched to LIFO had an average 47% increase
in their as-if earnings, whereas a matched sample of no-change firms had an average
2% decrease. Ricks also found that the abnormal returns of the switching firms were
significantly lower than the matched sample of no-change firms. These results are
inconsistent with those reported by Sunder.
The studies above indicate that investors in efficient markets attempt to evaluate
news about the effect of managerial decisions on cash flows not on eps. This fact has
Sunder used a moving-average beta technique in his second study [1975]. However, it did not substantially change his results.


ACCOUNTING INFORMATION

.125

.125

110 firms switching
to LIFO

.075

.075

22 firms' switching
from LIFO
•
•
•

••••
••
..•• •••••••••
0

— ...
••••

••
72
0

.025 - vir

•

—.025

—.025

—.075

—

.075

365

4D.
0.• •

•...
•
••

L
I
—.125
—12 —7 —2 3 8 13

I
I

—.0125
—12 —7 —2 3 8 13

Months from the date of change

Months from the date of change

Figure 11.1
Cumulative average residuals for 24 months around the
accounting change. (From S. Sunder, "Relationship between
Accounting Changes and Stock Prices: Problems of
Measurement and Some Empirical Evidence," reprinted from
Empirical Research in Accounting: Selected Studies, 1973, 18.)

direct implications for the accounting treatment of mergers and acquisitions. Two
types of accounting treatment are possible: pooling or purchase. In a pooling arrangement the income statements and balance sheets of the merging firms are simply added
together. On the other hand, when one company purchases another, the assets of
the acquired company are added to the acquiring company's balance sheet along with
an item called goodwill. Goodwill is the difference between the purchase price and
the book value of the acquired company's assets. Regulations require that goodwill
be written off as a charge against earnings after taxes in a period not to exceed 40
years. Because the writeoff is after taxes, there is no effect on cash flows, but reported
eps decline. The fact that there is no difference in cash flows between pooling and
purchase and the fact that cash flows, not eps, are the relevant information used by
investors to value the firm should convey to management the message that the accounting treatment of mergers and acquisitions is a matter of indifference.' Yet many
managements prefer pooling, presumably because they do not like to see eps decline
owing to the writeoff of goodwill. No economically rational basis for this type of
behavior can be cited.
In a recent empirical study tiong, Kaplan, and Mandelker [1978] tested the effect
of pooling and purchase techniques on stock prices of acquiring firms. Using monthly

Prior to the 1986 Tax Reform Act, the Internal Revenue Service (IRS) allowed the book value of the
assets of the acquired firm to be written up upon purchase. This reduced the amount of goodwill created,
but even more important, it created a depreciation tax shield that did not exist in a pooling arrangement.
Therefore cash flows for purchase were often higher than pooling. In these cases purchase was actually
preferable to pooling, at least from the point of view of the acquiring firm.


366

EFFICIENT CAPITAL MARKETS: EVIDENCE

data between 1954 and 1964, they compared a sample of 122 firms that used pooling
and 37 that used purchase. The acquired firm had to be at least 3% of the net asset
value of the acquiring firm. Mergers were excluded from the sample if another merger
took place within 18 months, if the acquiring firm was not NYSE listed, or if the
merger terms were not based on an exchange of shares. (This last criterion rules out
taxable mergers.)
Using the simple time-series market model given below, they calculated cumulative abnormal residuals:
ln

= oci + ln

+ ui„

where
= return on the jth security in time period t,
o ; = an intercept term assumed to be constant over the entire time period,
16'; = systematic risk assumed to be constant over the entire time period,

Rmt = market return in time period t,

nit = abnormal return for the jth security in time period t.
When the cumulative average residuals were centered around the month of the actual
merger, the patterns revealed no evidence of abnormal performance for the sample
of 122 poolings. This is shown in Fig. 11.2. Therefore there is no evidence that "dirty
pooling" raises the stock prices of acquiring firms. Investors are not fooled by the
accounting convention.
These results are just as important for acquiring firms that had to write off goodwill against their after-tax earnings because they used the purchase technique. As
shown in Fig. 11.3, there is no evidence of negative abnormal returns, which is what
we would expect if investors looked at eps. Instead, there is weak evidence that shareholders of acquiring firms experienced positive abnormal returns when the purchase
technique was used. This is consistent with the hypothesis that investors value cash
flows and that they disregard reported eps.
The empirical studies of Sunder [1973, 1975], and Hong, Kaplan, and Mandelker [1978] provide evidence on what is meant by relevant accounting information.
Cumulative
abnormal
residual
.04—
.02 —
0••
•••
• • • • °••
v.:.
—.02
4
—.0 -- —50 —40 —30
—20 —10

• •

•

•

• •

• •

•:•••
10 20 30 40 50 60

Figure 11.2
Cumulative abnormal residuals for 122 poolings with market value greater than book
value in the month relative to merger. (From H. Hong, R. S. Kaplan, and G. Mandelker,
"Pooling vs. Purchase: The Effects of Accounting for Mergers on Stock Prices," reprinted
with permission of Accounting Review, January 1978, 42.)


367

ACCOUNTING INFORMATION

Cumulative
abnormal
•
residual
• • e•
.04—
• • ••
• •
•
.02

•
••
• •1
I
I
II.
1.•
0
•
•••
•••
••
—.02
• •
•„. •
—.04—
•S• ••••
—50 —40 —30 —20 —10
0
•

•

• •
•
• .

•
••

•

I

••• . •
•r•
••••

10

20

.0

o„
i•

30

•
•

.

%

•••• .

•

• •

.

..
•-• -0

41.
0.

.

%

i

I

I

40

50

60

Figure 11.3
Thirty-seven purchases with market value greater than book value in the month relative
to merger. (From H. Hong, R. S. Kaplan, and G. Mandelker, "Pooling vs. Purchase:
The Effects of Accounting for Mergers on Stock Prices," reprinted with permission of

Accounting Review, January 1978, 42.)

By relevant we mean any information about the expected distribution of future cash
flows. Next, a study by Ball and Brown [1968] provides some evidence about the
speed of adjustment of efficient markets to new information.
Earnings data and cash flows are usually highly correlated. The examples discussed above merely serve to point out some situations where they are not related
and therefore allow empiricists to distinguish between the two. Ball and Brown used
monthly data for a sample of 261 firms between 1946 and 1965 to evaluate the usefulness of information in annual reports. First, they separated the sample into companies
that had earnings that were either higher or lower than those predicted by a naive
time series model. Their model for the change in earnings was
AN/i, = a

+ 1), Arn, + e ft ,

(11.2)

where
ANIii = the change in earnings per share for the jth firm,
Am, = the change in the average eps for all firms (other than firm j) in the market.
Next, this regression was used to predict next year's change in earnings, AN/ j,,,,:
ANI;,, +1 = a + b1 Amt +1,

(11.3)

where

a, b = coefficients estimated from time series fits of Eq. (11.2) to the data,
Am t +i = the actual change in market average eps during the (t + 1)th time period.
Finally, estimated earnings changes were compared with actual earnings changes. If
the actual change was greater than estimated, the company was put into a portfolio

where returns were expected to be positive, and vice versa.
Figure 11.4 plots an abnormal performance index (API) that represents the value
of $1 invested in a portfolio 12 months before an annual report and held for T


368

EFFICIENT CAPITAL MARKETS: EVIDENCE

Abnormal
performance
index
1.12—
1.10
Variable 2

1.08
1.06
1.04
1.02
1.00
Total sample
0.98
0.96
0.94
0.92
0.90
0.88

I

I
12 10 8
6 4
2 0 2 4 6
Month relative to Annual Report announcement date

Figure 11.4
Abnormal performance index of portfolios chosen
on the basis of differences between actual and
predicted accounting income. (From R. Ball and
P. Brown, "An Empirical Evaluation of Accounting
Income Numbers," reprinted with permission of
Journal of Accounting Research, Autumn 1968, 169.)

months (where T = 1, 2, . . . , 12). It is computed as follows:
N T

API = —

n (1 + e ft),

N1=1 t=i

where
N=

the number of companies in a portfolio,

T= 1, 2, . . . , 12,
sit = abnormal performance measured by deviations from the market model.

A quick look at Fig. 11.4 shows that when earnings are higher than predicted, returns
are abnormally high. Furthermore, returns appear to adjust gradually until, by the
time of the annual report, almost all the adjustment has occurred. Most of the infor-


ACCOUNTING INFORMATION

369

mation contained in the annual report is anticipated by the market before the annual
report is released. In fact, anticipation is so accurate that the actual income number
does not appear to cause any unusual jumps in the API in the announcement month.
Most of the content of the annual report (about 85% to 90%) is captured by more
timely sources of information. Apparently market prices adjust continuously to new
information as it becomes publicly available throughout the year. The annual report
has little new information to add.
These results suggest that prices in the marketplace continuously adjust in an
unbiased manner to new information. Two implications for the corporate treasurers
are: (1) significant new information, which will affect the future cash flows of the firm,
should be announced as soon as it becomes available so that shareholders can use
it without the (presumably greater) expense of discovering it from alternative sources;
and (2) it probably does not make any difference whether cash flow effects are reported
in the balance sheet, the income statement, or footnotes—the market can evaluate
the news as long as it is publicly available, whatever form it may take.
The Ball and Brown study raised the question of whether or not annual reports
contain any new information. More recent studies by Aharony and Swary [1980],
Joy, Litzenberger and McEnally [1977], and Watts [1978] have focused on quarterly
earnings reports where information revealed to the market is (perhaps) more timely
than annual reports.' They typically use a time series model to predict quarterly
earnings, then form two portfolios of equal risk, one consisting of firms with earnings

higher than predicted and the other of firms with lower than expected earnings. The
combined portfolio, which is long in the higher than expected earnings firms and
short in the lower than expected earnings firms, is a zero-beta portfolio that (in perfect markets) requires no investment. It is an arbitrage portfolio and should have
zero expected return. Watts [1978] finds a statistically significant return in the quarter
of the announcement a clear indication that quarterly earnings reports contain new
information. However, he also finds a statistically significant return in the following
quarter and concludes that "the existence of those abnormal returns is evidence that
the market is inefficient."
Quarterly earnings reports are sometimes followed by announcements of dividend changes, which also affect the stock price. To study this problem, Aharony and
Swary [1980] examine all dividend and earnings announcements within the same
quarter that are at least 11 trading days apart. They conclude that both quarterly
earnings announcements and dividend change announcements have statistically significant effects on the stock price. But more important they find no evidence of market
inefficiency when the two types of announcement effects are separated. They used
daily data and Watts [1978] used quarterly data, so we cannot be sure that the conclusions of the two studies regarding market inefficiency are inconsistent. All we can
say is that unexpected changes in quarterly dividends and in quarterly earnings both
have significant effects on the value of the firm and that more research needs to be
done on possible market inefficiencies following the announcement of unexpected
earnings changes.
See also articles by Brown [1978], Griffin [1977], and Foster [1977].


370

EFFICIENT CAPITAL MARKETS: EVIDENCE

Using intraday records of all transactions for the common stock returns of 96
(large) firms, Paten and Wolfson [1984] were able to estimate the speed of market
reaction to disclosures of dividend and earnings information. In a simple trading
rule, they bought (sold short) stocks whose dividend or earnings announcements exceeded (fell below) what had been forecast by Value Line Investor Service. The initial
price reactions to earnings and dividend change announcements begin with the first

pair of price changes following the appearance of the news release on the Broad Tape
monitors. Although there was a hint of some activity in the hour or two preceding
the Broad Tape news release, by far the largest portion of the price response occurs
in the first 5 to 15 minutes after the disclosure. Thus, according to Patell and Wolfson,
the market reacts to unexpected changes in earnings and dividends, and it reacts very
quickly.

C. BLOCK TRADES
During a typical day for an actively traded security on a major stock exchange,
thousands of shares will be traded, usually in round lots ranging between one hundred
and several hundred shares. However, occasionally a large block, say 10,000 shares
or more, is brought to the floor for trading. The behavior of the marketplace during
the time interval around the trading of a large block provides a "laboratory" where
the following questions can be investigated: (1) Does the block trade disrupt the
market? (2) If the stock price falls when the block is sold, is the fall a liquidity effect,
an information effect, or both? (3) Can anyone earn abnormal returns from the fall in
price? (4) How fast does the market adjust to the effects of a block trade?
In perfect (rather than efficient) capital markets all securities of equal risk are
perfect substitutes for each other. Because all individuals are assumed to possess the
same information and because markets are assumed to be frictionless, the number of
shares traded in a given security should have no effect on its price. If markets are less
than perfect, the sale of a large block may have two effects (see Fig. 11.5). First, if it
is believed to carry with it some new information about the security, the price will
change (permanently) to reflect the new information. As illustrated in parts (c) and (d)
of Fig. 11.5, the closing price is lower than the opening price and it remains low
permanently. 4 Second, if buyers must incur extra costs when they accept the block,
there may be a (temporary) decline in price to reflect what has been in various articles
described as a price pressure, or distribution effect, or liquidity premium, as shown in
parts (a) and (c). Figure 11.5 depicts how hypothesized information or price pressure
effects can be expected to show up in continuous transactions data. For example, if

the sale of a large block has both effects [Fig. 11.5(c)], we may expect the price to
fall from the price before the trade ( — T) to the block price (BP) at the time of the
block trade (BT), then to recover quickly from any price pressure effect by the time of

4
The permanent decline in price is also tested by looking at the pattern of day-to-day closing prices. Ev
dence on this is reported in Fig. 11.6.


BLOCK TRADES

No price pressure

No new information

Price pressure
Price

Price

BP
Time

Time
Open

—T BT +T

Close

Open

(a)
New information

371

Price

---fi
I

Open

Close

(b)

Price

BP

—T BT +T

I

I

I

—T BT +T
(c)

Close

Time

Open

—T BT +T

Close

Time

(d)

Figure 11.5
Competing hypotheses of price behavior around the sale of a large block.

the next trade ( + T) but to remain at a permanently lower level, which reflects the
impact of new information on the value of the security.
Scholes [1972] and Kraus and Stoll [1972] provided the first empirical evidence
about the price effects of block trading. Scholes used daily returns data to analyze
345 secondary distributions between July 1961 and December 1965. Secondary distributions, unlike primary distributions, are not initiated by the company but by
shareholders who will receive the proceeds of the sale. The distributions are usually
underwritten by an investment banking group that buys the entire block from the
seller. The shares are then sold on a subscription basis after normal trading hours.
The subscriber pays only the subscription price and not stock exchange or brokerage
commissions. Figure 11.6 shows an abnormal performance index based on the market

model and calculated for 40 trading days around the date of a secondary distribution.
The abnormal performance index falls from an initial level of 1.0 to a final value of
.977 just 14 days after the sale, a decline of 2.2%. On the day of the secondary distribution, the average abnormal performance was —.57.. Because this study uses only
close-to-close daily returns data, it focuses only on permanent price changes. We have
characterized these as information effects [Fig. 11.5(c) and (d)]. Further evidence that
the permanent decline in price is an information effect is revealed when the API is
partitioned by vendor classification. These results appear in Table 11.2.
On the day of the offering the vendor is not usually known, but we may presume
that the news becomes available soon thereafter. One may expect that an estate liquidation or portfolio rebalancing by a bank or insurance company would not be motivated by information about the performance of the firm. On the other hand, corporate


372

EFFICIENT CAPITAL MARKETS: EVIDENCE

Performance index
1.05 —
1.04 —
Market adjustments
1.03 —
1.02 1.01—
1.00 — .........................
.99 —
•
. ••
• ..........
.98—
.97 .96 I
.95 I
—25 —20 —15 —10 —5 0

5
10 15
Day relative to distribution day
Figure 11.6

Abnormal performance index on days around a
secondary distribution. (From M. Scholes, "The Market
for Securities: Substitution vs. Price Pressure and the
Effects of Information on Share Prices," reprinted with
permission of Journal of Business, April 1972, 193.)
Copyright © 1972, The University of Chicago Press.
Table 11.2 Abnormal Performance Index for Secondary
Distributions Partitioned by Vendor Category

API

No. of
Observations
in Sample

Category

—10 to
+10 Days

0 to
+ 10 Days

192
31

36
23

Investment companies and mutual funds
Banks and insurance companies
Individuals
Corporations and officers

—2.57.
—.3
—1.1
—2.9

—1.47.
—0.0
—.7
— 2.1

— .7

—.5

50

Estates and trusts

From M. Scholes, "The Market for Securities: Substitution vs. Price Pressure and the Effects of Information on Share Prices," reprinted with permission of Journal of Business, April 1972, 202. Copyright
© 1972, The University of Chicago Press.

insiders as well as investment companies and mutual funds (with large research staffs)

may be selling on the basis of adverse information. The data seem to support these
suppositions. Greater price changes after the distribution are observed when the seller
is presumed to have a knowledgeable reason for trading.'
Mikkelson and Partch [1985] studied a sample of 146 registered and 321 nonregistered secondary offerings between 1972 and 1981. Using daily data, they find an
A second test performed by Scholes showed that there was no relationship between the size of the distribution (as a percentage of the firm) and changes in the API on the distribution date. This would lead
us to reject the hypothesis that investment companies and mutual funds may have had an impact because
they sold larger blocks.
5


BLOCK TRADES

373

average statistically significant initial announcement price decline of —2.87% for
registered secondaries and —1.96% for nonregistered secondaries. There was no significant price change at the actual offering date for registered distributions. The announcement date price declines are permanent, they are positively related to the size
of the offering, and they are related to the identity of the seller (with the largest
declines occurring when the vendors are directors or officers). These results are consistent with a permanent information effect. Mikkelson and Partch also find that the
underwriting spread of secondaries is positively related to the relative size of the
offering. This is consistent with the argument that the underwriting spread reflects
compensation for the underwriter's selling effort or liquidity services. Therefore even
though Mikkelson and Partch find no rebound in market prices following secondary
offerings, they cannot conclude that the costs of liquidity are unimportant.
The data available to Kraus and Stoll [1972] pertain to open market block
trades. They examined price effects for all block trades of 10,000 shares or more
carried out on the NYSE between July 1, 1968, and September 30, 1969. They had
prices for the close the day before the block trade, the price immediately prior to the
transaction, the block price, and the closing price the day of the block trade. Abnormal performance indices based on daily data were consistent with Scholes' results.
More interesting were intraday price effects, which are shown in Fig. 11.7. There is
clear evidence of a price pressure or distribution effect. The stock price recovers substantially from the block price by the end of the trading day. The recovery averages

.713%. For example, a stock that sold for $50.00 before the block transaction would

Price (logarithmic)
•
Closing price (P_ 1)

\

1.8610% (Et)

T

1.1380% (E3)

•Price prior to block (PPB)
Closing price (Po)
• --t-.7130% (E2)

✓ i

End of day —1

1
•
Block Price (BP)

End of day 0

• Time

Figure 11.7
Intraday price impacts of block trading. (From A. Kraus
and H. R. Stoll, "Price Impacts of Block Trading on the
New York Stock Exchange," reprinted with permission
of Journal of Finance, June 1972, 575.)


374

EFFICIENT CAPITAL MARKETS: EVIDENCE

have a block price of $49.43, but by the end of the day the price would have recovered
to $49.79.
The Scholes and Kraus-Stoll studies find evidence of a permanent price decline
that is measured by price drops from the closing price the day before the block trade
to the closing price the day of the block transaction. These negative returns seem to
persist for at least a month after the block trade. In addition, Kraus and Stoll found
evidence of temporary intraday price pressure effects. The implications of these findings are discussed by Dann, Mayers, and Raab [1977], who collected continuous
transactions data during the day of a block trade for a sample of 298 blocks between
July 1968 and December 1969. The open-to-block price decline was at least 4.56% for
each block in the sample. The reason for restricting the sample to blocks with large
price declines was to provide the strongest test of market efficiency. If an individual
or a group of investors can establish a trading rule that allows them to buy a block
whose open-to-block price change is at least 4.56%, then sell at the end of the day,
they may be able to earn abnormal profits. This would be evidence of capital market
inefficiency.
Testing a trading rule of this type takes great care. Normally, a block trade is not
made publicly available until the trade has already been consummated and the transaction is recorded on the ticker. The semistrong form of market efficiency is based
on the set of publicly available information. Therefore a critical issue is: Exactly how
fast must we react after we observe that our —4.56% trading rule has been activated

by the first publicly available announcement that occurs on the ticker tape? Figure
11.8 shows annualized rates of return using the —4.56% rule with the purchase made
x minutes after the block and the stock then sold at the close. Returns are net of
actual commissions and New York State transfer taxes. For both time periods that
are reported, we would have to react in less than five minutes in order to earn a
positive return. Such a rapid reaction is, for all practical purposes, impossible. It
seems that no abnormal returns are available to individuals who trade on publicly
available information about block trades because prices react so quickly. Fifteen
minutes after the block trade, transaction prices have completely adjusted to unbiased
estimates of closing prices. This gives some idea of how fast the market adjusts to new,
unexpected information like a block trade.
What about people who can transact at the block price? Who are they and do
they not earn an abnormal return? Usually, the specialist, the floor trader (a member
of the NYSE), brokerage houses, and favored customers of the brokerage houses can
participate at the block price. Dann, Mayers, and Raab show that with a —4.56%
trading rule, an individual participating in every block with purchases of $100,000 or
more could have earned a net annualized rate of return of 203% for the 173 blocks
that activated the filter rule. Of course, this represents the maximum realizable rate
of return. Nevertheless, it is clear that even after adjusting for risk, transactions costs.
and taxes, it is possible to earn rates of return in excess of what any existing theory
would call "normal." This may be interpreted as evidence that capital markets are
inefficient in their strong form. Individuals who are notified of the pending block
trade and who can participate at the block price before the information becomes
publicly available do in fact appear to earn excess profits.
However, Dann, Mayers, and Raab caution us that we may not properly under-


BLOCK TRADES
480 B
4701

120-1 A
100

10/1/68 — 3/31/69

,L) 20

% Rate of return

2 40

100 K 50 K — —
10 K ........

320
310

80
60-

375

40 —

7/1/69 — 12/31/69

20

Il - - - - - - - - 0 I ---0 .I

1
—20—20 — ,:\
-40—
—40
• •
—60— k
—60—
..■
U•4
—80—
—80—
—100—
—100—
I
I
I
I
5
0
15
5
10
Time of purchase (minutes after block)
(b)
(a)

I
10

I

15

Annualized' rates of return on initial wealth, —4.56 percent rule;
purchase at first price at least x minutes after block, sell at close" (using
only first block per day). Gross returns less actual commissions and
NY State transfer taxes (curves represent levels of initial wealth).
Annualized rates of return are calculated by squaring the quantity
one plus the respective six-month return.
Blocks occurring within x minutes of the close were assumed not
to have been acted upon.
b

Figure 11.8

Annualized rates of return on the —4.56% rule. (From L.
Dann, D. Mayers, and R. Raab, "Trading Rules, Large
Blocks, and the Speed of Adjustment," reprinted from Journal
of Financial Economics, January 1977, 18.)

stand all the costs that a buyer faces in a block trade. One possibility is that the
specialist (or anyone else) normally holds an optimal utility-maximizing portfolio. In
order to accept part of a block trade, which forces the specialist away from that portfolio, he or she will charge a premium rate of return. In this way, what appear to be
abnormal returns may actually be fair, competitively determined fees for a service
rendered—the service of providing liquidity to a seller.
To date, the empirical research into the phenomenon of price changes around a
block trade shows that block trades do not disrupt markets, that markets are efficient
in the sense that they very quickly (less than 15 minutes) fully reflect all publicly
available information. There is evidence of both a permanent effect and a (very)
temporary liquidity or price pressure effect as illustrated in Fig. 11.5(c). The market
is efficient in its semistrong form, but the fact that abnormal returns are earned by

individuals who participate at the block price may indicate strong-form inefficiency.


376

EFFICIENT CAPITAL MARKETS: EVIDENCE

D. INSIDER TRADING
A direct test of strong-form efficiency is whether or not insiders with access to information that is not publicly available can outperform the market.6 Jaffe [1974]
collected data on insider trading from the Official Summary of Security Transactions
and Holdings published by the Securities and Exchange Commission. He then defined
an intensive trading month as one during which there were at least three more
insiders selling than buying, or vice versa. If a stock was intensively traded during
a given month, it was included in an intensive-trading portfolio. Using the empirical
market line, Jaffe then calculated cumulative average residuals. If the stock had
intensive selling, its residual (which would presumably be negative) was multiplied
by —1 and added to the portfolio returns, and conversely for intensive buying. For
861 observations during the 1960s, the residuals rose approximately 5% in eight
months following the intensive-trading event, with 3% of the rise occurring in the
last six months. These returns are statistically significant and are greater than transactions costs. A sample of insider trading during the 1950s produces similar results.
These findings suggest that insiders do earn abnormal returns and that the strongform hypothesis of market efficiency does not hold.
Jaffe also investigated the effect of regulation changes on insider trading. Two of
the most significant changes in security regulation resulted from (1) the Cady-Roberts
decision in November 1961, when the SEC first exercised its power to punish insider
trading and thus established the precedent that corporate officials trading on insider
information were liable for civil prosecution; and (2) the Texas Gulf Sulphur case in
August 1966, when the courts upheld the earlier (April 1965) SEC indictment of
company officials who had suppressed and traded on news about a vast mineral
strike. After examining abnormal returns from intensive insider-trading samples
around dates of these historic decisions, Jaffe was forced to the following conclusion:

The data could not reject the null hypothesis that the enforcement of SEC regulations
in these two cases had no effect on insider trading in general. At best the regulations
prohibit only the most flagrant examples of speculation based on inside information.
A study by Finnerty [1976] corroborates Jaffe's conclusions. The major difference is that the Finnerty data sample was not restricted to an intensive trading
group. By testing the entire population of insiders, the empirical findings allow an
evaluation of the "average" insider returns. The data include over 30,000 individual
insider transactions between January 1969 and December 1972. Abnormal returns
computed from the market model indicate that insiders are able to "beat the market"
on a risk-adjusted basis, both when selling and when buying.
A study by Givoly and Palmon [1985] correlates insider trading with subsequent news announcements to see if insiders trade in anticipation of news releases.
The surprising result is that there is no relationship between insider trading and
news events. Although insiders' transactions are associated with a strong price moveThe Securities and Exchange Commission defines insiders as members of the board of directors, corporate officers, and any beneficial owner of more than 10% of any class of stock. They must disclose, on a
monthly basis, any changes in their stock holdings.


NEW ISSUES

377

ment in the direction of the trade during the month following the trade, these price
movements occur independent of subsequent publication of news. This leads to the
conjecture that outside investors accept (blindly) the superior knowledge and follow
in the footsteps of insiders.
One of the interesting implications of the empirical work on insider trading is
that it is consistent with the point of view that markets do not aggregate information.
In Chapter 10, fully aggregating markets were described as those that reflect all
available information even though it is not known to all market participants. In a
fully aggregating market an insider should not be able to make abnormal returns
because his trading activity would reveal his information to the market. The evidence
on profitable insider trading shows that this is clearly not the case.

E. NEW ISSUES
There has been a long history of articles that have studied the pricing of the common
stock of companies that is issued to the public for the first time. To mention a few, the
list includes papers by the Securities and Exchange Commission [1963], Reilly and
Hatfield [1969], Stickney [1970], McDonald and Fisher [1972], Logue [1973],
Stigler [1964], and Shaw [1971]. They all faced a seemingly insoluble problem: How
could returns on unseasoned issues be adjusted for risk if time series data on preissue
prices were nonexistent? Any estimate of systematic risk, e.g., requires the computation of the covariance between time series returns for a given security and returns
on a market portfolio. But new issues are not priced until they become public. An
ingenious way around this problem was employed by Ibbotson [1975]. Portfolios
of new issues with identical seasoning (defined as the number of months since issue)
were formed. The monthly return on the XYZ Company in March 1964, say two
months after its issue, was matched with the market return that month, resulting
in one pair of returns for a portfolio of two months seasoning. By collecting a large
number of return pairs for new issues that went public in different calendar months
but that all had two months seasoning, it was possible to form a vector of returns
of issues of two months seasoning for which Ibbotson could compute a covariance
with the market. In this manner, he estimated the systematic risk of issues with
various seasoning. Using the empirical market line, he was able to estimate abnormal
performance indices in the month of initial issue (initial performance from the offering
date price to the end of the first month) and in the aftermarket (months following
the initial issue). From 2650 new issues between 1960 and 1969, Ibbotson randomly
selected one new issue for each of the 120 calendar months.
The estimated systematic risk (beta) in the month of issue was 2.26, and the
abnormal return was estimated to be 11.4%. Even after transactions costs, this
represents a statistically significant positive abnormal return. Therefore either the
offering price is set too low or investors systematically overvalue new issues at the
end of the first month of seasoning. Later evidence shows that the aftermarket is
efficient; therefore Ibbotson focused his attention on the possibility that offering

prices determined by the investment banking firm are systematically set below the


378

EFFICIENT CAPITAL MARKETS: EVIDENCE

Table 11.3 Gain and Loss Situations for a New Issue

I
II
III
IV

Situation

Investors

Investment
Banker

Maximum offering price > market price > offering price
Maximum offering price > offering price > market price
Maximum offering price = offering price > market price
Market price > maximum offering price = offering price

Gain
Parity
Parity
Gain

Parity
Loss
Loss
Parity

fair market value of the security. Regulations of the SEC require a maximum offering
price for a new issue, which is usually filed two weeks in advance of the actual offering,
although it can be adjusted in some cases.' The actual offering price is set immediately before the offering. The existence of a regulation that requires the actual
offering price to be fixed creates the possibility of a "Heads I lose, tails you win"
situation for the underwriter. Table 11.3 shows the four possibilities that can occur
in a firm commitment offering (the underwriting syndicate buys the issue from the
firm for the offering price less an underwriting spread, then sells the issue to the
public at the fixed offering price). The best the underwriter can do is achieve a parity
situation with no gain or loss. This happens whenever the market price turns out to
be above the offering price (situations I and IV). Obviously, the investment banker
does not want the market price to equal or exceed the maximum offering price
(situations III and IV). This would infuriate the issuing firm and lead to a loss of
future underwriting business. Therefore we usually observe situations I and II. But
if the investment banking firm receives adequate compensation from its underwriting
spread for the risk it undertakes, and if it cannot gain by setting the offer price lower
than the market price, then why do we not observe offer prices (which, after all, are
established only moments before the issues are sold to the public) set equal to the
market value? Why can investors systematically earn an abnormal return of 11.4%
during the first month of issue? This conundrum, like the difference between the
block price and the closing price on the day of the block, cannot easily be explained
by existing finance theory.
What about new issue performance in the aftermarket, i.e., for prices from the
first market price onward? Figure 11.9 shows abnormal returns (based on the empirical market line) in the aftermarket for six-month holding periods and the significance
tests (t-tests). The 9 periods other than the initial offering period include only 2

periods with results that are statistically different from zero (and returns in these 2
periods are negative). Ibbotson concludes that the evidence cannot allow us to reject
the null hypothesis that aftermarkets are efficient, although it is interesting to note
that returns in 7 out of 10 periods show negative returns.
Figure 11.10 shows plots of changes in systematic risk in the aftermarket; note
the decline. The results show that the systematic risk of new issues is greater than
In most cases the maximum offering price is set high enough to cause little concern that it may actually constrain the actual offering price.


NEW ISSUES

379

4.000 —

.150

t - test

Abnormal return
.100

2.000

.050

0.000
=MINIM=

NOMMEN

.000

■

—2.000

1111111•1

■

=1•I

I
I
I
I
I
I
—4.000
I
I
I
I
—.050
60
48
36
12
24

0
60
36
48
24
0
12
aMonths of seasonin g (n), n = 1, 7, 13, 19, 25, 31, 37, 43, 49, 55 ; r = 6 ; reg ressions 1-6, 7-12, ..., 55-60.

Figure 11.9
Abnormal returns for issues of different seasoning. (From R. Ibbotson, "Price Performance
of Common Stock New Issues," reprinted from Journal of Financial Economics, September
1975, 254.)

6.000

3.000

t - test

Beta
4.000

2.000

111•111■
MIII=111

1.000

2.000

.000

.000

—1.000
0

—2.000
60
48
36
24
12
0
60
48
aMonths of seasonin g (n), n = 1, 7, 13, 19, 25, 31, 37, 43, 49, 55 ; r = 6 ; re gressions 1-6,
7-12, ..., 55-60. These t-statistics are designed to test if p„, r, 0 > 1.
12

24

36

Figure 11.10
Systematic risk of issues with different seasoning. (From R. Ibbotson, "Price Performance
of Common Stock New Issues," reprinted from Journal of Financial Economics, September
1975, 260.)

380

EFFICIENT CAPITAL MARKETS: EVIDENCE

the systematic risk of the market (which always has a beta equal to one) and that
their systematic risk is not stable in that it drops as the new issues become seasoned.
Weinstein [1978] studied the price behavior of newly issued corporate bonds
by measuring their excess holding period returns. Excess returns were defined as the
difference between the return on the ith newly issued bond and a portfolio of seasoned
bonds with the same (Moody's) bond rating. Data were collected for 179 new issues
between June 1962 and July 1974. Weinstein's conclusions for newly issued bonds are
similar to those of Ibbotson [1975] for newly issued stock, namely, that the offering
price is below the market equilibrium price but that the aftermarket is efficient.
Weinstein found a .383% rate of return during the first month and only a .06% rate
of return over the next six months.

F. STOCK SPLITS
Why do stocks split, and what effect, if any, do splits have on shareholder wealth?
The best known study of stock splits was conducted by Fama, Fisher, Jensen, and
Roll [1969]. Cumulative average residuals were calculated from the simple market
model, using monthly data for an interval of 60 months around the split ex date for
940 splits between January 1927 and December 1959. Figure 11.11 shows the results.
It plots the cumulative average return for the stock split sample. Positive abnormal
returns are observed before the split but not afterward. This would seem to indicate
that splits are the cause of the abnormal returns. But such a conclusion has no
economic logic to it. The run-up in the cumulative average returns prior to the stock
split in Fig. 11.11 can be explained by selection bias. Stocks split because their price
has increased prior to the split date. Consequently, it should hardly be surprising

that when we select a sample of split-up stocks, we observe that they have positive

Cumulative average
residual, Um
0.44

Figure 11.11
Cumulative average residuals for 60
months around stock splits. (From E. F.
Fama, L. Fisher, M. Jensen, and R. Roll,
"The Adjustment of Stock Prices to
New Information," reprinted with
permission of International Economic
Review, February 1969, 13. Copyright
â International Economic Review.)

0.33

0.22

0.11

.ã ã

ãã
.
0 . 1 1 1 1 1
—29 —20 —10

1

0

1 I
10

I I 1
20
30

Month relative to split, m


STOCK

Cumulative average
+
residual, U,„
0.44

SPLITS

381

Cumulative average
residual, Urn
0.44
••••.....

..•
0.33

0.33

0.22

0.22
.....••••

0.11—

0.11
r•

..*
0 1.-'1 • 1 1 1 1
-29
-20
-10

1
0

1 1
10

1 1
20

1
30

0 1•:1 I 1 1 1
-29
-20
-10

1_
0

1

I I
10

Month relative to split, m

Month relative to split, m

(a)

(b)

1 1
20

1
30

Figure 11.12
Cumulative average residuals for splits with (a) dividend increases and (b) decreases. (From

E. F. Fama, L. Fisher, M. Jensen, and R. Roll, "The Adjustment of Stock Prices to New
Information," reprinted with permission of International Economic Review, February 1969,
15. Copyright © International Economic Review.)

abnormal performance prior to the split date. Selection bias occurs because we are
studying a selected data set of stocks that have been observed to split.
Farna et al. [1969] speculated that stock splits might be interpreted by investors
as a message about future changes in the firm's expected cash flows. They hypothesized that stock splits might be interpreted as a message about dividend increases,
which in turn imply that the managers of the firm feel confident that it can maintain
a permanently higher level of cash flows. To test this hypothesis the sample was divided into those firms that increased their dividends beyond the average for the
market in the interval following the split and those that paid out lower dividends.
The results, shown in Fig. 11.12, reveal that stocks in the dividend "increased" class
have slightly positive returns following the split. This is consistent with the hypothesis
that splits are interpreted as messages about dividend increases.' Of course, a dividend increase does not always follow a split. Hence the slightly positive abnormal
return for the dividend-increase group reflects small price adjustments that occur
when the market is absolutely sure of the increase. On the other hand, the cumulative
average residuals of split-up stocks with poor dividend performance decline until
about a year after the split, by which time it must be very clear that the anticipated

8 This does not imply that higher dividend payout per se causes an increase in the value of the firm. In
Chapter 15 "Dividend Policy" we shall see that higher dividends are interpreted as signals that the future
cash flows from the firm will increase.