Copyright
Paul M Healy.

Earnings and stock splits online

. (page 1 of 2)
Online LibraryPaul M HealyEarnings and stock splits → online text (page 1 of 2)
Font size
QR-code for this ebook


* /






^' o/ T^'"



v-






1988



t, ' ^-< n '. n



• f^.'^ .-■'



WORKING PAPER
ALFRED P. SLOAN SCHOOL OF MANAGEMENT



FARKTNGS ANri 9.T0CK SPLITS

by

Paul M. Healy

Paul Asquith

Krishna G. Falepu



October, 1987



WP #1944-87



MASSACHUSETTS

INSTITUTE OF TECHNOLOGY

50 MEMORIAL DRIVE

CAMBRIDGE, MASSACHUSETTS 02139



EARlv'INGS AND STOCK SPLITS

by

Paul M. Healy

Paul Asquith

Krishna G. Palepu

October, 1987 WP #1944-87



EARNINGS AND STOCK SPLITS



Paul Asquith^^^

Paul Healy ^
Krishna Palepu



October 1987



Harvard Business School
MIT Sloan School of Management



We wish to thank Robert Holthausen, Robert Kaplan, Richard Leftwich,
Robert Merton, Richard Ruback and Karen Wruck for their comments on an
earlier draft of this paper. We also wish to thank the participants at
seminars at Harvard University, New York University and the University of
Chicago. Finally we wish to thank Lydia Magliozzi and Eric Wolff for their
assistance in data collection.



FEB 81988
RECc



Earnings and Stock Splits



This paper examines whether stock splits convey information about
earnings. It finds that firms split their shares after a significant
Increase in stock prices and earnings. Furthermore, the market's reaction
to a split announcement is significantly related to prior earnings
increases. Thus it appears as though the market interprets stock splits as
confirmation that past earnings increases are permanent. The evidence also
suggests that the market's reaction to split announcements is not due
solely to expectations of near-term cash dividend increases.



- 2 -
1 . INTRODUCTION

Stock splits have long been puzzling phenomena to financial
economists. Stock splits usually occur after an increase in stock prices
and usually elicit a positive stock price reaction upon announcement. The
reasons for these stock price increases, however, have not been fully
understood. Most theorizing has focused on the market's positive reaction
to the split announcement. Since stock splits themselves do not directly
affect a company's cash flows, the increase in a company's stock price at
the time of these announcements must, assuming market efficiency, reflect
the release of new information. An explanation first proposed by Fama,
Fisher, Jensen and Roll (1969) hereafter FFJR, and repeated in subsequent

studies is that the market interprets stock split announcements as

2
improving the probability of near-term dividend increases.

FFJR report data consistent with the dividend explanation: 71.5% of
-their sample firms experience a percentage dividend increase in the year
after the split which is larger than the average increase for all
securities on the New York Stock Exchange. Thus, FFJR argue that a large
price increase at the time of a stock split is due to altered expectations
concerning future dividends rather than due to any intrinsic effects of the
splits themselves.

This 'dividend hypothesis' however, does not appear to fu]]y explain
the observed market reaction to stock split announcements. For example,
Grinblatt, Masulis and Titman (1984), hereafter GMT, report a significant
stock price reaction to the announcement of stock splits by firms that do
not pay cash dividends in the three years prior to the split. Since only
11% of their sample firms initiate a cash dividend during the year
following the split, GMT conclude that the valuation changes associated



with stock split announcements cannot be attributed totally to revised

expectations about near-term dividend increases.

Fama (1976) suggests that it is likely that the information revealed

by stock splits concerns earnings rather than dividends. He states:

"in their (FFJR's) view, the unusual behavior of the returns on a
splitting security in the months immediately preceding a split
reflects the information content of the dividend change that usually
accompanies a split... (A)n alternative view, completely consistent
with their empirical results, is that dividends are a passive variable
in the whole process. That is, companies tend to increase dividends
when earnings increase and to decrease dividends when earnings
decrease. In this view, the FFJR data suggest that splits tend to
occur when firms have experienced unusual increases in earnings, which
accounts for the positive average residuals of splitting shares in the
months preceding the split.

Lakonishok and Lev (1987), hereafter LL, examine both dividends and
earnings growth surrounding stock split announcements and conclude that the
evidence is consistent with "either stabilization of earnings growth
subsequent to the abnormal presplit growth, improved cash dividends
prospects, or both."

The objective of this paper is to examine whether stock splits convey
information about firms' earnings in the period surrounding the split
announcements. In order to mitigate any confounding effects of
simultaneous dividend changes, only firms that do not pay cash dividends at
the time of the stock split are included in the sample. Our tests, based
on a sample of 121 stock split announcements from the period 1970-1980,
lead to several conclusions.

First, stock splitting firms experience significant unexpected
Increases in their earnings for four years prior to the stock split. Even
when the earnings of the splitting firms are adjusted for the performance
of their industries, splitting firms experience significant earnings
increases in the year prior to the split. These earnings increases are not



reversed for at least four years subsequent to the split, indicating that
they are permanent. Stock splitting firms also exhibit significant
unexpected increases in their stock prices in the four years prior to and
on the split announcement date but not in the subsequent period.

Second, there is a significant cross-sectional relationship between
the stock price reaction to the split announcements and unexpected earnings
changes in the two years prior to the split. That is, the higher the
pre-split earnings changes, the larger the market reaction to the split
announcement. In contrast, there is no significant relationship between
the split announcement return and earnings changes after the split date.

Together, these conclusions indicate that firms split their shares
after they experience significant increases in their earnings and stock
prices. Furthermore, it appears as though the market interprets stock
splits as confirmation that past earnings increases are permanent, not
transitory. The results are also consistent with a "trading range" motive
for stock splits. This motive, often cited by practitioners, is that
splits are used to keep share prices within an "optimal" trading range.

The rest of the paper is organized as follows. The data and sample
used in our analysis are described in the next section. Section three
discusses the hypotheses, tests and results. The paper's conclusions are
discussed in section four.

2 . DATA

The initial sample of stock splits is identified from the CRSP Daily
Master file. To be included in the sample, a firm has to: (1) not pay cash
dividends prior to or at the time of the stock distribution announcement;



5 -



(2) announce a stock distribution (designated by the firm as either a stock
split or stock dividend) of at least 25% during the time period 1970 -
1980; (3) not announce any other stock distribution of 2.5/o or more in the
five years prior to the event; (4) be listed on the New York or American
Stock Exchange at the time of the event; (5) have stock price and return
data available for the announcement date and two days prior to the event;
and (6) have the stock distribution announcement date available in the Wall
Street Journal , or on the CRSP Daily Master file.

Cash dividend paying firms that split their stock often make
simultaneous announcements concerning dividend clianges. The sample is
therefore restricted to non-cash dividend paying firms in order to separate
the market reaction to stock splits from the reaction to simultaneous cash
dividend announcements. The sample only includes stock distributions over
25% because the accounting treatment for them does not affect retained
earnings. If the distribution is below 20%, the amount of the distribution
must be transferred from retained earnings to capital stock. The transfer
is at the firm's discretion for distributions between 20% and 25%. The
restriction of no other stock distributions in n five year period is to
insure independence of observations in the sample since five years of prior
earnings are analyzed with each split. Listing on the ASF. or the NYSE is
required to guarantee data availability. Announcement dates are necessary
to properly determine the market's reaction.

For each firm in the initial sample the following additional data are
collected: (1) the stock price two days prior to the split announcement;
(2) the return on the firm's common stock, and on the CRSP equally weighted
market index, for the day before and the day of the split announcement; (3)



- 6



the first quarterly earnings announcement date preceding the split
announcement; (4) earnings per share before extraordinary items and
discontinued operations from the twenty-four quarterly announcements
immediately preceding the split date (to construct earnings for years -6 to
-1), and the twenty quarterly announcement following the split date (to

o

construct earnings for years to 4).

Annual earnings are defined so that year earnings are created from
the four quarterly announcements subsequent to the split (quarters to 3);
year -1 earnings are constructed from the four quarterly earnings reports
prior to that event; and the earnings for years -6 to -2 and 1 to 4 are
constructed using the quarterly earnings data from quarters -24 to -5, and
from quarters 4 to 19. Defined this way, the earnings for years -6 to -1
are announced strictly before the split announcement, and the earnings for
years to 4 are announced strictly after the split date. This is done
because stock split announcements occur throughout a fiscal year and fiscal
year earnings include some quarters before the split and some quarters
after the split. By using quarterly data to create annual event earnings
the earnings can be aligned with the split announcement and earnings
performance before and after the announcement can be strictly separated.

After this data collection and earnings definition process, to be
included in the final sample a firm has to satisfy the following
conditions: (1) the date of the quarterly earnings announcement immediately
preceding the split announcement is reported in the Wall Street Journal
Index and (2) earnings for two or more years can be constructed from the
Quarterly Compustat Industrial tapes.

The final sample consists of 121 firms which do not pay cash dividends
and which announce a stock distribution of at least 25% during the period



7 -



1970 - 1980. Table 1 reports the number of stock splits by calendar year.
About 55% of the sample stock splits are from the three year period 1970 -
1972; another 20% is from the year 1980; the remaining 25% are from the
years 1974 - 1979. There is, thus, evidence that the splits are clustered
in time.

To test whether any observed earnings patterns of the sample firms are
due to industry and time-related trends, industry data on earnings per
share before extraordinary items and discontinued operations are also
collected for each final sample firm. The industry for each sample firm is

defined as the set of firms listed on Standard and Poor's Compustat

9
Industrial tapes with the same four digit SIC Code.

3 . HYPOTHESES

Since stock split announcements follow and are contemporaneous with

stock price increases, there must be (assuming market efficiency) new

positive information associated with them. This paper investigates whether

stock splits communicate information about firms' earnings. The following

hypotheses are tested:

HI: Firms that split stock experience significant earnings increases
in the years prior to the split announcement.

H2: Firms that split stock experience significant earnings increases
in the years subsequent to the split announcement.

H3: The market reaction to the announcement of a stock split is

positively related to the earnings increases prior or subsequent
to the split.

The earnings pattern of the splitting firms suggested by HI, H2, and
H3 differs in two respects from the time-series behavior of annual earnings
reported in the literature for a random sample of firms. First, studies by



8 -



Ball and Brown (1968), Ball and Watts (1972), and Watts and Leftwich (1977)
suggest that annual earnings follow a random walk. Thus the average
earnings change for a random sample of firms is expected to be zero. In
contrast, HI and H2 predict that earnings changes for splitting firms are
positive.

Second, Brooks and Buckmaster (1980), Beaver, Lambert, and Morse
(1980), and Freeman, Ohlson and Penman (1982) report that firms with large
earnings increases in one year experience earnings declines in the
following year. In other words, large earnings changes are usually
transitory and followed by earnings decreases. In contrast to this usual
earnings behavior, when firms split their stock the market may expect the
earnings increases prior to or subsequent to the split to be permanent.

Beaver, Lambert and Morse (1980) also report that, since large
earnings increases are partially transitory, investors do not fully
Incorporate them into a firm's future expected earnings. Hence, if there
are earnings increases prior to the stock split announcement, investors may
attach a low probability to these earnings increases being permanent. H3
predicts that positive stock price reactions to split announcements are due
to investors revising their probability that the past earnings increases
are not transitory.

In addition to these earnings hypotheses, HI to H3, this paper also

provides some evidence on the "dividend hypothesis" originally proposed by

FFJR (1969) and subsequently discussed by GMT (198A) and LL (1987). This

is:

H4: The stock market responds positively to stock splits because they
usually mean increased near-term cash dividends.



- 9



Tests of hypotheses HI and H2 are presented in section 4.2. Tests of
hypothesis H3 is presented in section 4.3. Hypothesis H4 is discussed in
section 4.1.

4. RESULTS

4. 1 Market Reaction To Stock Split Announcements

Table 2 reports risk-adjusted returns around the announcement of stock
splits. These are computed using the two parameter market model. The
parameters are estimated over the period -630 days to -481 days relative to
the stock split announcement day (AD) using the CRSP equal -weighted market
index. The announcement day is the date the Wal l Stree t Jo urnal reports
the stock split.

The mean risk-adjusted returns are insignificant for the period AD-480
to AD-251 days and significant for days AD-240 to AD-2. The holding period
abnormal returns for days AD-240 to AD-121, AD-120 to AD-61, AD-60 to
AD-21, AD-20 to AD-11 and AD-10 to AD-2 are all positive and significant.
The cumulative abnormal returns for this entire period are 50.5%. Thus
stock split announcements are preceded by large increases in the announcing
firms' stock price.

During the two days surrounding the split announcement (AD-1 and AD),
the mean risk-adjusted return is 3.9%, which is significantly different
from zero at the 1% level. As reported in Panel B of Table 2, the median
abnormal return is similar (3.5%). Further, 76% of the sample firms have
positive abnormal announcement period returns. Finally, the abnormal
returns for the twenty days subsequent to the stock split announcement are
insignificant.



- 10 -



The abnormal returns in Table 2 are similar to those reported by
earlier studies. For example, FFJR (1969) report a cumulative abnormal
return of 35.8% during the 29 months preceding the stock split
announcements in their sample. GMT (1984) report a 3.41% mean two day
announcement return for a large sample of stock split and stock dividends.

Hypothesis H4 suggests that the market's reaction to split
announcements is due to an implicit signal about increased near-term cash
dividend payments. Since our sample consists of firms that do not pay cash
dividends at the time of the split, an increase in cash dividends implies
dividend initiation. In order to investigate this possibility, we follow
our sample for five years subsequent to the split and record cash dividend
initiations. Of the 121 firms in the sample, 81 (67%) do not initiate a
cash dividend payment during the five years after the split announcement.
Of the remaining 40, the mean (median) lag between the stock split
announcement and the payment of the first cash dividend was 32 (34) months.

Only 11 (9%) sample companies pay cash dividends within one year from the

12
date of the stock split.

The percentages of splitting firms that initiate cash dividends are
similar to those for the population of all firms. For the sample period
1970-1980, 10% of the non-cash dividend paying firms listed on Compustat
initiate dividends in any given year. Furthermore, 70% of the firms which
do not pay cash dividends do not initiate dividends during the next five
years .
4.2 Earnings Performance Before and After Stock Splits

To examine whether there is a systematic earnings pattern for firms
that split stocks, earnings changes are estimated for five years before
(years -5 to -1), the year of the stock split announcement (year 0), and



11 -



four years subsequent (years 1 to A) . As mentioned above all earnings are

calculated anjiually for the four quarters relative to the split announce-

13
ment. To aggregate results across firms, we standardize earnings changes

for each firm j in years -5 to 4 by its stock price two days before the

announcement of the stock split. P.. The standardized earnings change in

year t, 6E^ . , is therefore defined as:
tj'



^^tj = (^j -^-lj)/Pj t = -5, ..., 4 (1)



where A is earnings before extraordinary items and discontinued
^ J

operations for firm j in year t.

Student t statistics are estimated to test whether the mean standard-
ized earnings changes of the stock split sample firms are significantly
different from zero in each year. Panel A in Table 3 reports the mean and

quartile earnings changes and the number of firms for which data are

14
available for the stock split sample in years -5 to 4. The mean

standardized earnings changes and probabilities for years -4 to are 1.05%

(p=0.02), 0.66% (p=0.12), 1.24% (p=0.01), 2.55% (p=0.01), and 2.03%

(p=0.01). The mean earnings changes are insignificant in all the

subsequent years.

The above results indicate that firms that announce stock splits

experience a significant earnings increase for the four years before the

stock split announcement. The largest of these increases occurs in the

year immediately before the stock split announcement. The earnings

continue to increase during the year of the split. While earnings

increases cease in the years subsequent to stock splits, past increases



- 12



are not reversed in these years. Therefore, the split firms' earnings
increases prior to and contemporaneous with stock split announcements
appear to be permanent. These conclusions are consistent with hypothesis
HI but not with H2.

Since the splits in our sample are clustered in time, it is possible
that the earnings performance of the split firms is due to industry- and
time-related factors rather than firm-specific factors. To test this, the
split firms' performance is compared to the performance of their
industries. For each sample firm, industry earnings changes for year - 5
to 4 are estimated using all firms that have (a) the same four digit SIC
code on Standard and Poor's Annual Compustat and Research Industrial tapes;
and (b) earnings data available for two consecutivo years of the eleven
years surrounding a split announcement. The earnings changes for these
comparison firms are then standardized by their stock prices two days prior
to the test firm's split announcement and an industry median is
constructed. Industry adjusted earnings in each year are calculated for
each splitting firm by subtracting its industry median earnings change from
the firm's earnings change.

Panel B of Table 3 reports the mean and the quartile industry adjusted
earnings changes. The mean industry adjusted earnings changes are
insignificant in all years except -1 and 3. The mean adjusted earnings
change is 2.82% (p=0.01) in year -1 and 2.94% (p=0.05) in year 3.

These results indicate that the strong earnings performances of the
splitting firms are matched in general by their industries' performances.
The split firms are in industries that have been experiencing good times.
In the pre-split period, split firms perform significantly better than
their industries only in year -1. This superior earnings performance by



13



splitting firms relative to their industries is permanently incorporated
into their earnings levels: in the years subsequent to the split
announcements, the split firms do not experience significantly lower
earnings growth than their industries. These results are similar to LL
(1987) who examine earnings growth rates around stock splits and find that
splitting firms have significantly higher earnings growth than a control
sample for the period before and during the year of the split. Their
sample of stock splits includes firms that pay cash dividends.

To summarize, firms that announce stock splits experience significant
unexpected earnings increases for several years prior to and in the year
of the stock split announcements. In the year immediately prior to the
split announcement, the split firms outperform their industries. This
increase in earnings appears to be permanent and is not reversed for at
least four years after the split. These conclusions are consistent with
hypothesis HI.

4.3 Relation Between Stock Split Announcement Returns and Earnings Changes

This section next tests whether the earnings changes documented above
are related to the market's reaction at the announcement of the stock
splits. Table 4 reports the Pearson and Spearman correlation coefficients
between the split announcement returns and the standardized earnings
changes in each of the years -5 to 4. The announcement return for each
firm is the risk-adjusted return for one day prior to and the day of the
stock split announcement. Standardized earnings changes in each year are
computed as described above.

The Pearson correlation coefficients are insignificant in years -5 to
-3 and to 4. In years -2 and -1, the correlations are 0.31 and 0.19,



14



both significantly different from zero. The Spearman rank correlation
results are similar, indicating that the Pearson correlations are not
driven by a few outliers. These correlations indicate that the split
announcement returns are positively related to earnings increases for the
two years prior to the split. There is no relationship between the
announcement returns and the earnings changes before these years or after
the split date even though years -4, -3, and have significant earnings
increases.

The split announcement returns are also regressed on earnings changes
in the two years before the split (years -2 and -1) and the two years after
the split (years and 1). We estimate the following regression model:



SRET. = a + B,6E „ . + 3-6 E , . + B-6E- . + B,6E, . + t. (2)
J , 1 -2, J 2 -l,j 3 0,j 4 l,j J



where SRET. is the risk-adjusted return for firm j for one day prior to and

the day of the stock split announcement and 6E is the earnings change in

^ J

year t standardized by the stock price two days prior to the stock split
announcement .

Table 5 reports the regression estimates. The estimated coefficients
for the earnings changes in years -2 and -1 are 0.306 and 0.241, both
significantly different from zero at the 5% level in a two-tailed test.
These estimates indicate that, all else being constant, a 10% increase in
standardized earnings in year -2 (year -1) is likely to be accompanied by a
3.06% (2.41%) risk-adjusted return in the two days surrounding the stock
split announcement. The coefficients of the earnings changes in years
and 1 are insignificant.



15 -



The correlations and the multiple regression results are both
consistent with the hypothesis H3. The information released by the stock
split announcements is related to the earnings increases in the two prior


1

Online LibraryPaul M HealyEarnings and stock splits → online text (page 1 of 2)