DIVIDEND POLICY: AN EMPIRICAL TEST USING TIlE
OCTOBER. 1987 CRASH

Wold Zemedkun

William Lewis Randolph
Norfolk State University

Norfolk, Virginia

The finance literature has not yet reached a consensus on the relevance of dividend
policy. The main consideration in the dividend policy debate is the impact of dividends
on the value of the finn's stock. Dividend policy issues have been investigated by
many. Miller and Modigliani [9] analyzed the special case of perfect markets and
concluded that dividend policy does not detennine or affect the value of the finn.
Others ([6], [8]) have argued· that investors prefer dividend to capital gains, and thus
dividend policy affects the valuation of the company stock. Most recent academic work
leans toward the argument that dividend policy is relevant Survey results reported
by researchers ([1], [11]) indicate that corporate chief financial officel8 believe that
dividend policy does affect stock prices. Baskin [2] analyzed dividend policy and found
an inverse correlation between dividend yield ~d stock price volatility.

The purpose of this study is to examine the relationship between dividend policy
and stock price volatility to detennine if dividend yield has an impact on either stock
price behavior, the rate of growth, or stock beta.I We have chosen to observe dividend
yield and stock price behavior in the period surrounding the crash of October 1987.
Such a period provides us with an excellent opportunity to examine factors that affect
the adjustment of prices of individual stocks to major market changes.

The primary hypothesis for this study is that stock prices of fInDS with high cash
dividend yields are less sensitive to a market crash and exhibit lower market risk
(beta) than stock prices of firms with low or no cash dividend yields. That is, the
market does care about dividend policy. Dividend policy is one important finn-specific
factor that could modify the effect of a crash on a finn's stock price. This is important
information for investors and has implications for portfolio strategies.

The concept of duration2 is the theoretical basis for the explanation of an inverse

1The beta of a stock is a measure of its price variability compared to the variability of the stock
market as indicated by an index such as the S&P 500 index. According to the Capital Asset
Pricing Model (CAPM), beta is a measure of the stock's nondiversifiable risk. As a statistical
measure beta is found by regressing a stock's price changes on the market index changes over
time. A stock with a high beta, greater than I, would be more volatile than the average stock
and therefore more risky.
2Unlike maturity, which looks only at the time to the last payment from an investment, dura-
tion gives weight to the time at which each cash payment occurs. The weight assigned to each
period is the present value of the cash flow for that period divided by the current price of the
security. Early payments are discounted less than those received later. Thus, duration is the
effective matmity of the investmenL



70 Journal ofBusiness Strategies Vol. 9, No.1

relationship between dividend yield and stock price volatility. As Reilly and Sidhu
[10] have shown, the stock with a high dividend yield will have a shorter duration
than a stock that pays no dividends. Duration considers the timing and pattern of
payments to the investor and indicates stock price sensitivity to changes in yield. Thus,
duration plays a vital theoretical role in the risk assessment of common stocks.

The important relationship between growth and dividend yield is also examined.
Firms that retain all of their earnings need to rely less on external sources of capital
during periods of high growth. The theoretical reason for finns to retain all of their
earnings is to finance growth. We expect to find that rums with low dividend or zero
dividend yields have a higher growth rate than finns with high dividend yields. This
is important because duration is a positive function of growth rate.

Data and Methodology

The data for this study is taken from the Standard and Poor's COMPUSTAT data
base. The PC PLUS version of the data base dated May 24, 1990, is used. The study
period is oriented around the crash of October 1987 and the study group of stocks is
drawn using three screening rules. The first screening rule is that the company's stock
be listed on the New York Stock Exchange (NYSE), the American Stock Exchange
(ASE), or trade in the over-the-counter market (OTC). This rule excludes Canadian
stocks and companies listed only on regional exchanges. The second screening rule
requires that enough data for the company be available to compute beta for September
1987. Beta is computed by using a minimum of 24 months of stock prices. The
third screening rule requires a stock price greater than three dollars in September 1987.
This rule is used in order to screen out "penny stocks." The National Association of
Security Dealers uses a three dollar stock price minimum to qualify newer companies
for inclusion in the National Market System. The application of these three screening
rules to the data base results in a sample of 3,039 companies available for the study.

We divide the resulting group of stocks into three categories based upon fiscal
1987 dividend yield. Companies with a dividend yield equal to or greater than three
percent are placed in the high dividend category. Companies that have a dividend
yield of less than three percent but greater than zero are in the medium dividend yield
category. Companies which paid no dividend in 1987 are in the zero dividend category.
The three percent boundary between the high and medium dividend categories divides
the companies which paid dividends into equal groups.

We fonn subsets of the overall study group by first separating the S&P 500 stocks
into one subset and then using listing criteria to fonn two more subsets. The first
subset consists of companies whose stock was part of the S&P 500 Index on September
1987. The second subset is those companies whose stock was listed on the NYSE or
ASE but are not in the S&P 500 Index. The third subset is made up of OTC
companies who are not in the S&P 500. This breakdown fonns subsets of companies
that are of different sizes.

We also study the monthly stock returns of the group using the period surrounding
the crash as the focal point Following the earlier empirical example of Fama, Fisher,



Spring 1992 Zemedkun &: Randolph: Dividend Policy 71

Jensen, and Roll (FFJR) [5], we use the market model to investigate the effect of the
crash on the stock prices of the various dividend groups. By adjusting for the market
effect or systematic risk, we can focus on the pattern of excess returns or prediction
errors surrounding the crash. While the FFJR methodology may have some theoretical
weaknesses, it has proven to be remarkably robust in detecting abnormal stock price
performance.

In order to test the hypothesis, we will employ the test statistic described in Dodd
and Warner [4]. To compute the statistic, the prediction error is standardized by its
estimated standard deviation. The standardized prediction error is accumulated over
the prediction intelVal and divided by the square root of the number of time periods
in this intelVal. The test statistic is then derived by summing the accumulated
standardized prediction error across finns and dividing by the square root of the number
of fInns.

Results

We show various characteristics of the sample group and it's subsets by exchange
listing in Tables 1 through 4. These Tables present the mean or median of a
characteristic for the sample and by high, medium, and zero dividend yield We use
the median for dividend yield and payout in order to avoid bias caused by outliers
which are present in these data elements. For the other characteristics we use the mean.

In Table 1, we see that the average beta is low for the high dividend group and
high for the zero dividend group. A test of the difference in the mean beta of the
three dividend subgroups indicates that there is less than a one percent probability that
the mean betas are equal.3 Our primary hypotheses, that risk and dividend yield are
inversely related, is supported by this evidence. The correlation between dividend yield

. and beta is -0.24, and, is highly signifIcant, further confmning the inverse relationship
between beta and dividend yield. In Tables 2 through 4, we can see that the
relationship between beta and dividend yield is consistent across the different listing
groups.

We measure the growth rate by computing the rate of change in earnings per share
(EPS), sales, and assets over the period 1983 to 1987. A company that retains all its
earnings can support a higher rate of growth without external financing. Thus, all
things being equal, we could expect an inverse relationship between growth and
dividend yield. This inverse relationship holds for the total sample and the subsets in
at least two of the growth factors measured.

From Table 1 we can see that the zero dividend group has the highest growth
rate in sales and assets but not EPS. We suspect this is because the zero dividend
group contains a large proportion of relatively new companies that are at the critical
point in the growth cycle and growth in earnings is taking a backseat to growth in
sales and assets to support new sales.

3See Iman and Conover [7, p. 282] for the procedures for inferences about the
difference in means.



72 Journal ofBusiness Strategies

Table 1: Characteristics or Total Sample

Vol. 9, No. 1

Number
Dividend Yield **
Dividend Payout**
Beta***
Growth Rates***

BPS
Sales
Assets

Market Value
(millions)***

Total
Sample

3039
0.89%
4.8%
0.99

6.7%
12.0%
12.2%
$1057

High
Dividend*

847
4.83%

54.0%
0.77

2.6%
6.2%
8.4%

$1957

Medium.
Dividend*

847
1.68%

22.2%
1.02

10.8%
10.8%
12.5%
$1571

Zero
Dividend*

1345
0.0%
0.0%
1.10

6.6%
16.6%
14.5%
$179

*High dividend stocks yield greater than 3%, medium. dividend stocks yield between
o and 3%, zero dividend stocks paid no dividend in 1987.
**The statistic computed for this characteristic is the median.
***The statistic computed for this characteristic is the average.

Table 2: Characteristics of S&P 500 Stocks

Number
Dividend Yield**
Dividend Payout**
Beta***
Growth Rates*··

BPS
Sales
Assets

Market Value
(millions)··*

Total
Sample

456
2.58%

32.6%
1.13

8.1%
7.2%
8.8%

$4771

High
Dividend*

195
4.44%

50.2%
1.00

2.5%
4.2%
6.5%

$6204

Medium.
Dividend*

211
1.88%

25.4%
1.17

12.3%
9.3%

11.1%
$4085

Zero
Dividend*

50
0.0%
0.0%
1.46

14.2%
9.5%
7.9%

$2168

*High dividend stocks yield greater than 3%, medium. dividend stocks yield between
o and 3%, zero dividend stocks paid no dividend in 1987.
**The statistic computed for this characteristic is the median.
*·*The statistic computed for this characteristic is the average.



Spring 1992 Zemedkun & Randolph: Dividend Policy

Table 3: Characteristics of Non-S&P SOO Listed Stocks

73

Total
Sample

Number
Dividend Yield**
Dividend Payout"'*
Beta**'"
Growth Rates***

BPS
Sales
Assets

Market Value
(millions)***

High
Dividend*

1284
1.52%
13.8%
0.97

7.3%
10.6%
11.9%
$602

Medium
Dividend*

417
5.33%
58.8%

0.74

3.0%
6.9%
9.5%
$889

Zero
Dividend*

388
1.65%
21.3%

1.02

10.6%
10.2%
12.2%
$861

479
0.0%
0.0%
1.13

8.4%
13.9%
13.3%
$147

*High dividend stocks yield greater than 3%, medium dividend stocks yield between
o and 3%, zero dividend stocks paid no dividend in 1987.
**The statistic computed for this characteristic is th~ median.
***The statistic computed for this characteristic is the average.

Table 4: Characteristics of OTC Stocks

Number
Dividend Yield **
Dividend Payout**
Beta***
Growth Rates"'*'"

BPS
Sales
Assets

Market Value
(millions)***

Total
Sample

1299
0.0%
0.0%
0.96

5.3%
15.3%
13.9%
$210

High
Dividend'"

235
4.38%

47.1%
0.63

1.9%
6.9%
8.2%
$323

Medium
Dividend'"

248
1.48%

20.5%
0.90

9.9%
13.1%
14.0%
$542

Zero
Dividend*

816
0.0%
0.0%
1.06

4.7%
18.8%
15.7%

$77

*High dividend stocks yield greater than 3%, medium dividend stocks yield between
o and 3%, zero dividend stocks paid no dividend in 1987.
*"'The statistic computed for this characteristic is the median.
**"'The statistic computed for this characteristic is the average.

Another characteristic that we measure is company size. The market value of the
common stock is the surrogate we use for size. When we examine the sample by
listing subset, Tables 2, 3, and 4, we see that there is a natural breakdown by size
due to the various listing requirements. The S&P 500 companies are the largest with
an average market value of $4,771 million. The other listed stocks are next in size



74 Journal ofBusiness Strategies Vol. 9, No.1

with an average of $602 million. And, the OTC companies are the smallest with an
average size of $210 million. The size difference between the high and medium
dividend groups in each subset is usually not significant, but the difference between
the medium and zero dividend groups is more evident In each subset the smallest
companies are the zero dividend companies, and the difference in size between the
zero dividend and the medium dividend companies is statistically significant at the 1%
level.

The monthly stock returns of the study group, separated by dividend category and
listing subset, are presented in Table 5. The returns for August 1987 to January 1988
are listed. The first set of returns is for the whole sample group and its categories by
dividend. In October 1987, the month of the stock market crash, the total sample went
down 28.2% (-0.282). The high dividend category went down 21.3%, the medium
dividend category down 27.4%, and the zero dividend category lost 33.1%. The
difference between the mean returns of the different dividend categories for October
1987 is significant at the 1% level. This is more evidence confmning that high
dividend yields are related to lower levels of risk.

Table 5: Monthly Stock Returns: August 1987 - January 1988

AUG SEP OCT NOV DEC JAN
Total Sample 0.023 -0.012 -0.282 -0.053 0.051 0.058

Hi Div 0.021 -0.027 -0.213 -0.038 0.032 0.067
Moo Div 0.033 -0.011 -0.274 -0.043 0.082 0.035
Zero Div 0.018 -0.003 -0.331 -0.068 0.043 0.068

S&P 500 0.027 -0.021 -0.254 -0.067 0.090 0.044
Hi Div 0.027 -0.033 -0.224 -0.069 0.060 0.071
Moo Div 0.025 -0.017 -0.269 -0.065 0.115 0.028
Zero Div 0.037 0.003 -0.302 -0.065 0.103 0.006

Listed Stocks 0.022 -0.013 -0.281 -0.049 0.045 0.058
Hi Div 0.018 -0.025 -0.209 -0.035 0.025 0.068
Med Div 0.035 -0.007 -0.286 -0.041 0.077 0.035
Zero Div 0.015 -0.007 -0.338 -0.068 0.036 0.070

OTC Stocks 0.023 -0.008 -0.294 -0.052 0.044 0.064
Hi Div 0.023 -0.028 -0.212 -0.018 0.024 0.061
Med Div 0.036 -0.011 -0.261 -0.029 0.063 0.039
Zero Div 0.018 -0.001 -0.329 -0.069 0.044 0.071

The October returns of each of the listing subsets in Table 5 follow the same
pattern with high yield stocks suffering the smallest losses in the crash and zero yield
stocks suffering the highest losses. The differences between the mean returns of the
various dividend categories within each listing subset are also statistically significant
at the 1% level. Thus, the behavior of stock prices in the crash was consistent with
their beta and the Capital Asset Pricing Model. This means that the prices of high
beta stocks lost more value during the crash than the value of low beta stocks.



Spring 1992 Zemedkun &: Randolph: Dividend Policy

Conclusion

75

Our analysis of companies with high, mediwn, and zero dividend yield in the
period sWTOWlding the stock marlret crash in. October 1987 shows a very strong inverse
relationship between risk and dividend yield. Stocks with high dividend yield have
lower betas and suffered the least in the crash. Stocks that pay no dividends have
significantly higher betas and suffered the greatest losses in the crash. Stocks with
dividend yields between zero and three percent, the mediwn dividend yield group, have
betas between the zero and high dividend paying stocks. The differences in the betas
of these three groups is statistically significant No significant excess returns were
evident in the crash time frame.

Our examination of growth rates by dividend yield group found that the zero
dividend group had the highest rate of growth in assets and sales. And the high
dividend yield group had the lowest rate of growth. Growth rates of the companies
under study are consistent with the theoretical view that companies that pay no
dividends can support higher levels of growth.

The concept of duration is the theoretical basis for the explanation of an inverse
relationship between dividend yield and stock price volatility. In general, stocks with
high dividend yields have lower growth rates and therefore will have shorter durations
than stocks that pay no dividends. Thus, price volatility is a direct function of duration.

The conclusion is that the market views companies that pay no dividends to be
more risky and companies that pay high dividends to be less risky than companies in
the mediwn dividend group. The market appears to be consistent (or perhaps efficient)
since beta explains the post crash returns fairly well for each of the three dividend
groups.

References

1. Baker, H. K., Furely, G. E., and Edelman, R. B. "A Survey of Management Views on
Dividend Policy." Financial Management (Autmnn 1985), pp. 78-84.

2. Baskin, J. "Dividend Policy and the Volatility of Common Stocks." Journal of Portfolio
Mana~t (Spring 1989), pp. 19-25.

3. Boquist, J., Racette, G., and Schlarbaum, G. "Duration and Risk Assessment for Bonds
and Common Stocks." Journal of Fmance, Vol. 30 (1975) pp. 1360-1365.

4. Dodd, P. and Warner, J.B. "On Corporate Governance: A Study of Proxy Contests."
Journal of Financial Economics, Vol. 11 (1983), pp. 401-38.

5. Fama, E., Fisher, L., Jensen, M., and Roll, R. "'The Adjustment of Stock Prices to
New Information," International Economic Review, Vol. 10 (1969), pp. 1-21.

6. Gordon, M. "Optimal Investment and Financing Policy," Journal of Fin!lllCft (1963),
pp. 264-272.



76 Journal ofBusiness Strategies Vol. 9, No. 1

7. Iman, R. L. and Conover, W. J. Modem Business Statistics, (1983).

8. litzenberger, R. and Ramaswamy, K. "The Effects of Dividends on Common Stock
Prices: Tax Effects or Infonnation Effects." Journal of Finance, Vol. 37 (May 1982),
pp. 429-443.

9. Miller, M. and Modigliani, F. "Dividend Policy, Growth and the Valuation of Shares."
Journal of Business, Vol. 34 (1961), pp. 411-433.

10. Reilly, F.K. and Sidhu, R.S. '''The Many Uses of Bond Duration." Financial Analysts
Journal (July-August 1980), pp. 58-72.

11. Sterk, W. and Vandenberg, P. "The Market Valuation of Cash Dividends and Tax
Differential Theory of Dividend Policy: A Case Revisited." The Financial Review, Vol.
25 (1990), pp. 441-455.


	Dividen Policy: An Empirical Test Using the October 1987 Crash