Seasonality in ex dividend day returns
ABSTRACT It is documented that for both high- and low-yield stocks, ex day raw returns are systematically higher in January than for the other months of the year. Although such patterns are not predicted by any known tax-clienteles model, they are consistent with the price discreteness and spread models in the spirit of Bali and Hite (Journal of Financial Economics, 47, 127-59, 1998) and Bali (Journal of Economics and Finance, 27, 190-210, 2003). For high-yield stocks in January, the returns are about one-fourth those for low-yield stocks, and for the remaining months they are significantly negative. The rents that arbitrageurs earn for supplying liquidity are higher for low-yield stocks and are significantly higher in January.
- SourceAvailable from: umich.edu[show abstract] [hide abstract]
ABSTRACT: Ex-dividend day returns vary over time. The ex-day returns of high-yield stocks are persistently positive for some time periods and negative for others; in contrast, ex-day returns of low-yield stocks are always positive and less variable. The authors are unable to explain the variation with changes in the tax code but they do find a strong effect for the introduction of negotiated commissions. The authors find evidence that corporate dividend capturing is affecting ex-day returns and confirm the findings of R. H. Gordon and D. F. Bradford (1980) that the price of dividends is countercyclical. Copyright 1994 by American Finance Association.Journal of Finance. 02/1994; 49(5):1617-38.
Seasonality in ex dividend day returns
Finance and Economics Department, School of Business, Adelphi University,
1 South Avenue, Garden City, NY 11530, USA, E-mail: firstname.lastname@example.org
It is documented that for both high- and low-yield stocks, ex day raw returns are
systematically higher in January than for the other months of the year. Although
such patterns are not predicted by any known tax-clienteles model, they are consis-
tent with the price discreteness and spread models in the spirit of Bali and Hite
(Journal of Financial Economics, 47, 127–59, 1998) and Bali (Journal of Economics
and Finance, 27, 190–210, 2003). For high-yield stocks in January, the returns are
about one-fourth those for low-yield stocks, and for the remaining months they
are significantly negative. The rents that arbitrageurs earn for supplying liquidity
are higher for low-yield stocks and are significantly higher in January.
Perhaps the longest standing event study in economics is
the behaviour of the price of a stock on the day it goes
ex dividend. Beginning in the 1950s, studies examined the
cum-ex price adjustment and found that the price drop is
smaller than the dividend amount (Campbell and Beranek,
1955; Durand and May, 1960). Elton and Gruber (1970)
conclude that the differential taxation of dividends com-
pared to capital gains causes the incomplete price adjust-
ment, that tax-clienteles exist, that high (low) tax bracket
investors specialize in low (high) yield stocks, and that
their marginal tax rates can be deduced from the
Keloharju (1995), Sarig and Talkowsky (1997), and
Bartholdy and Brown (1999) document evidence favouring
tax-clienteles for stocks in Finland, Israel, and New
Brooks and Edwards (1980) argue that statements about
clientele effects can only be made if entrants, other than
long-term capital gains sellers, are excluded. Researchers
such as Kalay (1982) and Miller and Scholes (1982) criticize
the tax-clientele argument saying that the equilibrium is
inconsistent with no-arbitrage. If prices tend to fall by
less (greater) than the amount of the dividend then short-
term traders would buy (sell) before and sell (repurchase)
later to arbitrage the differential. This would occur until the
abnormal returns are driven to bounds of transaction costs.
Recently Bali and Hite (1998) (BH) propose a price discr-
eteness-based model where the dividend D is priced at D,
the tick multiple just short of the dividend amount. For
example, if the dividend equals 20¢ (37.5¢) then D equals
12.5¢ (25¢) if the tick size is 12.5¢. This occurs as a result of
the long-term investors’ aversion to dividends because
of taxes and the presence of transaction costs in the case
of arbitrageurs. This model can explain the tax-clientele
effects. BH also show that the price-drop-to-dividend
ratio declines between tick multiples. This is inconsistent
with tax-clienteles since the yield is rising as the ratio falls.
Bali (2003) extends BH by incorporating the effects of
bid-ask spreads as well on the cum-ex equilibrium.
He models the differential probability of observing a bid
(ask) price on cum (ex) days due to concentrated trading
patterns which also impact the price dynamics. The meth-
odology developed therein is used here to present the
This paper tests if there is seasonality in the behaviour of
ex dividend day returns and price changes. The study is
motivated by the observation in Rozeff and Kinney
(1976), Keim (1983), etc. that there exists a seasonal
January effect in stock returns and especially so for small
capitalization stocks. Bhardwaj and Brooks (1992) con-
clude that this effect is in fact a low-price effect. Ritter
(1988) documents an abnormally high buy/sell ratio for
individual investors in January compared to the rest of
the year; Ligon (1997) shows evidence that the January
Applied Economics Letters ISSN 1350–4851 print/ISSN 1466–4291 online # 2003 Taylor & Francis Ltd
Applied Economics Letters, 2003, 10, 929–932
effect is primarily related to the trading patterns of indi-
vidual investors. Gutner (2003) provides a recent discus-
sion in the popular press.
The present results are consistent with an equilibrium
where the liquidity providers around the ex day earn higher
returns in January than in the rest of the year. The mean
raw return for high-yield stocks in January (rest of the
year) equals 0.09% (?0.06%) with standard error equal
to 0.07% (0.02%) respectively. Using a t-test with unequal
variances produces a t-value of 1.98 which is significant
at the 5% level. For low-yield stocks these estimates
equal 0.36% (0.20%) and 0.03% (0.01%) respectively.
The t-test produces a value of 4.39. So returns in January
are higher for both yield groups while overall high-yield
returns are much smaller.
It appears that for high-yield stocks with high liquidity
the dividend is closer to being fully priced by the market
while the arbitrageur profits are larger for low-yield, low
dividend stocks with low liquidity. The results are not
consistent with a tax-clienteles based model but support
the transaction costs hypothesis.
The rest of the paper is as follows. The next section
details the data, model and empirical results and Section III
II. DATA, ESTIMATION AND EMPIRICAL
Eades et al. (1994) and Bali (2003) are followed in the data
selection process. The period from 2 July 1962 to 23 June
1997 is analysed which spans the tick size of one-eighth.
This procedure nets 84119 ex day events of which 16818
(67301) are high (low) dividend yields. The number of high-
yield events in January (non-January) equals 1054 (15764).
The corresponding number for low-yield events is 3227
(64074). The proportions of high- and low-yield events
occurring in January are quite close (6.3% versus 4.8%)
although they are below the uniform monthly average of
8.25%. This is a result of the tendency of a higher number
of dividend announcements to occur in the third month of
Table 1 shows that for January (non-January) months
the mean dividend for the high- and low-yield groups equal
US$0.42 (US$0.42) and US$0.33 (US$0.32) respectively.
The mean proportional (differential) price drop is signifi-
cantly lower (higher) in January versus other months
for both yield groups. The mean raw return is smaller for
high-yield stocks for both January and non-January
months compared to the corresponding months for low-
yield stocks. For high-yield group they are within discrete-
ness bounds. For low-yield group the mean raw return
exceeds the upper bound by a few basis points but the
differential is within estimates of bid-ask spreads.
The proportional price drop is used as the dependent
variable. The independent variables include rounded
down dividend scaled by cum price and the discreteness
term scaled by cum price. The following regression equa-
tion is estimated:
¼ ? þ ?jDUMjþ ?D
which is based on Equation 12 in Bali (2003). If D equals
the dividend then D equals the tick multiple just short of the
dividend D and ? equals the discreteness term (¼D?D).
?P equals the price drop (difference between cum and ex
prices) and P is cum price. The intercept equals the product
of two terms: (i) the difference in probability of observing
Table 1. Mean and standard errors of the high- and low-yield groups for the month of January and for February
through December. The data comprise NYSE taxable cash dividends from 2 July 1962 to 23 June 1997. The
relevant variables include dividend, dividend yield (DY), proportional price drop (PPD), differential price drop
(D??P), raw returns (RW) and the two discreteness bounds (DR and DR)
Dividend (¢)DY (%)PPD (%)D??P (¢)
DR (%) RW (%)DR (%)
High-yield sample for January (N¼1054)
High-yield sample for February–December (N¼15764)
Low-yield sample for January (N¼3227)
Low-yield sample for February–December (N¼64074)
Std error 0.07
ask price as cum price and that as ex price, and (ii) the
relative spread defined as the difference between ask and
bid prices on ex day divided by cum price. The dummy
variable DUMjequals 1 in January and 0 otherwise.
Table 2 presents the regression estimates for high- and
low-yield groups. Coefficients on the terms containing the
January dummy help to isolate the incremental effects
of January on the intercept, the dividend pricing term
and the discreteness term. ^ ? ? is significantly positive (1.74)
for high-yield stocks. Its magnitude is more than eight
times that of the intercept for low-yields (?0.21) which is
significantly negative. This could be partly due to the fact
that high-yield stocks have an average cum price of US$17
while for low-yield it equals US$36, and lower priced
stocks have higher relative spreads. The signs of the inter-
cepts indicate that high-yield stocks experience dividend
capture and low-yields experience selling cum and/or buy-
ing ex. The negatively significant values of ^ ? ?jfor high- and
low-yield groups (highly significant for the former (?0.60)
while marginally so for the latter (?0.17)) indicate that
high-yield group may have decreased dividend capture in
January; and low-yield group has increased selling cum
and/or buying ex in January. If, for high-yield stocks, it
is a result of increased selling cum and/or buying ex that ^ ? ?j
is negative, then the results support the research discussed
earlier. However if it is true that it occurs because of
reduced dividend capture then a new phenomenon has
been identified. This could be a result of dividend capturers
responding by trading less aggressively to the higher than
average bid-ask spreads posted by arbitrageurs in January.
For the high-yield group the slope coefficient on the
dividend pricing term ð^? ? ¼ ?0:04Þ and that on the discrete-
ness term ð^? ? ¼ ?0:50Þ are significantly negative, consistent
with arbitrageurs being short overnight. Since the inter-
cept and the slope coefficients are negatively correlated,
a negative ^ ? ?jis accompanied with positively significant^? ?j
(0.36) and insignificant^? ?j: Incidentally, the magnitudes of
^? ? and^? ?jare very close across the two yield groups further
emphasizing that discreteness is important and significant
for both the groups.
For the low-yield group^? ? is insignificantly different from
1 and^? ? is significantly positive (0.52). This is consistent
with the discreteness hypothesis. The estimates^? ?jand^? ?j
are not different from 0 implying that the incremental
intercept ð^ ? ?jÞ drives the higher January returns.
Models based on the assumption that the behaviour of
tax-clienteles governs the cum-ex price dynamics do not
predict a January seasonal in ex day returns. This effect
is documented for both high- and low-yield stocks. The
predictions of the proposed model which takes into
account the discreteness issues along with bid-ask spreads
are supported. High-yield stocks may have reduced divi-
dend capture in January while low-yields have increased
selling cum (and/or buying ex). This indicates that arbitra-
geurs extract higher returns around the ex day in January
compared to the rest of the year for both high- and low-
yield stocks. The results are consistent with other studies
of the January effect although they do not specifically study
ex dividend days.
Bali, R. (2003) Variation in ex day dividend pricing: myth or
reality?, Journal of Economics and Finance, 27, 190–210.
Table 2. Regression analysis of dividend pricing for high- and low-yield groups of NYSE taxable cash
dividends in the month of January and for the rest of the year from 2 July 1962 to 23 June 1997
The regression equation is:
¼ ? þ ?jDUMjþ ?D
where ?P is the price drop, P is cum price, D equals the tick multiple just short of the dividend amount,
? equals (D?D), DUMjis a dummy variable equal to 1 in January and 0 otherwise.
Note: Superscript a (b) denotes significant difference from 0 (1) at the 5% level.
Seasonality in ex dividend day returns931
Bali, R. and Hite, G. L. (1998) Ex dividend day stock price beha-
vior: discreteness or tax-induced clienteles?, Journal of
Financial Economics, 47, 127–59.
Bartholdy, J. and Brown, K. (1999) Ex-dividend day pricing in
New Zealand, Accounting and Finance, 39, 111–29.
Bhardwaj, R. and Brooks, L. (1992) The January anomaly: effects
of low share price, transaction costs, and bid-ask bias,
Journal of Finance, 47, 553–75.
Brooks, L. and Edwards, C. (1980) Marginal stockholders and
implied tax rates, Review of Economics and Statistics, 62,
Campbell, J. and Beranek, W. (1955) Stock price behavior on
ex-dividend dates, Journal of Finance, 10, 425–29.
Durand, D. and May, A. (1960) The ex-dividend behavior of
American Telephone and Telegraph stock, Journal of Finance,
Eades, K., Hess, P. and Kim, E. H. (1994) Time-series variation
in dividend pricing, Journal of Finance, 49, 1617–38.
Elton, E. and Gruber, M. (1970) Marginal stockholder tax rates
and the clientele effect, Review of Economics and Statistics,
Gutner, T. (2003) January effect?, Business Week, 24 February,
Hietala, P. and Keloharju, M. (1995) The ex-dividend day
behavior of Finnish restricted and unrestricted shares,
Applied Economics Letters, 2, 467–68.
Kalay, A. (1982) The ex-dividend day behavior of stock prices: a
re-examination of the clientele effect, Journal of Finance, 37,
Keim, D. (1983) Size-related anomalies and stock return season-
ality: further empirical evidence, Journal of Financial
Economics, 12, 13–32.
Miller, M. and Scholes, M. (1982) Dividends and taxes: some
empirical evidence, Journal of Political Economy, 90, 1118–
Ligon, J. (1997) A simultaneous test of competing theories regard-
ing the January effect, Journal of Financial Research, 20,
Ritter, J. (1988) The buying and selling behavior of individual
investors at the turn of the year, Journal of Finance, 43,
Rozeff, M. and Kinney, W. (1976) Capital market seasonality:
the case of stock returns, Journal of Financial Economics,
Sarig, O. and Talkowsky, E. (1997) Dividend effects in Israel:
a puzzle, Economics Letters, 54, 169–74.
932 R. Bali