Professional Documents
Culture Documents
To cite this article: Rakesh Bali & Jack C. Francis (2012) Further evidence from ex-dividend days, Applied Economics Letters,
19:6, 537-540, DOI: 10.1080/13504851.2011.587763
Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Applied Economics Letters, 2012, 19, 537540
Applied Economics Letters ISSN 13504851 print/ISSN 14664291 online # 2012 Taylor & Francis 537
http://www.informaworld.com
DOI: 10.1080/13504851.2011.587763
538 R. Bali and J. C. Francis
II. The Model III. Analysis
The proportional price drop (ex-day price drop, P, Bali (2003a) provides the data selection procedure.
divided by cum price, P) is as follows (Bali, 2003a): Table 1 reports the results of regression equations
(23). For the full sample the intercept is significantly
P D negative (-0.25) and increases in the post-event period
qc qx :spr b: l: e 1
P P P (0.13). This implies that overall the arbitrageurs are
long the stocks from cum to ex and there is a tendency
where spr is the relative spread, qc (qx) is the probabil- for more sales in the post-event period. The coefficient
^ is significantly larger than 1 (1.14) and falls in the
b
ity that the closing transaction price on the cum (ex)
post-event period b ^ 0:15 consistent with
day is the ask price and = D D. ev
If investors tend to sell cum and buy ex due to decline in dividend capture post-event. The estimate
^l is positively significant (0.63) and is unchanged in
dividend aversion, then qc , qx, that is, a negative
intercept in Equation 1. The liquidity providers then the post-event period.
inventory the dividend overnight. If dividend capturing Other estimates are consistent with the model.
Downloaded by [The University of British Columbia] at 19:00 29 October 2014
investors (e.g. corporations) trade due to regulatory Both high and low yields indicate selling cum and
reasons, they may buy cum and sell ex which leads to buying ex with significant intercepts (-0.33 and
qc . qx and implies a positive intercept in Equation 1. -0.20, respectively). They are higher in the post-
The liquidity providers are then short the stock event period consistent with more sales. The magni-
tude of the intercept is smaller for low yields con-
overnight.
sistent with them having lower relative spreads. For
We estimate the following regression: ^ is significantly larger than 1 (1.20) and
high yields b
P D D ^l is significantly positive (0.57). For low yields b ^ is
a aev :DUMev b: bev : :DUMev not significantly different from 1 (1.05) and l is ^
P P P
positively significant (0.65). Negatively significant
l: lev : :DUMev e 2 ^ (-0.26 for both groups) indicates reduced value
P P b ev
placed on dividend post-event.
where a is the intercept term from Equation 1; b and l The GARCH terms are significant for the full sam-
are coefficients on the dividend pricing and discreteness ple and for the two yield groups. The coefficient on the
terms, respectively; DUMev = 1 in the post-event per- dummy term in the variance equation is positively
iod and 0 otherwise; and aev, bev and lev measure the significant for all groups. The error variance is signifi-
respective changes caused by the events. The error term cantly higher post-event consistent with a decline in
follows a Generalized Autoregressive Conditional market quality. GARCH (2,2) did not converge.
Heteroscedasticity (GARCH (1,1)) process which is These results complement the conclusions in Bali and
tested for change in the post-event period. Specifically, Francis (forthcoming), who show a post-event decline
in volume for high yields.
ht b0 b1 e2t1 b2 ht1 b3 DUMev 3 The DurbinWatson statistics indicate positive
autocorrelation for all groups. The Chow tests do
where ht is the conditional variance and b indicates not detect a structural break post-event for any
standard GARCH (1,1) terms and the dummy term. group. To test for the effects of outliers we form a
The null hypothesis states that the slope coefficients subsample by omitting bottom and top 1% of obser-
of the terms with the event dummy (aev, bev and lev) vations with extreme (D P) values. The dividend
should equal 0 in Equation 2. Any significance is only pricing estimates are unaffected but for high yields the
necessary but not sufficient condition for rejecting the GARCH term assumes more prominence at the
no-arbitrage conditions. expense of ARCH terms.
We test fairly weak arbitrage conditions and utilize Figure 1(a) and (b) provides the time-series varia-
tion in monthly raw returns along with the discrete-
the observable discreteness in dividends alone and
ness bounds for high and low yields over the period
compute bounds assuming the price drop equals the
1994 to 2000. For the whole period and for both
tick multiple just below the dividend (D) and D plus 1
groups RWs are generally constrained by the weak
tick, that is,
no-arbitrage bounds. Overall, for the full sample,
the mean raw return is 0.22 (bounds -0.21 and 0.15)
DR RW DR 4 per cent. For high yields the mean raw return is
0.12 (bounds -0.33 and 0.28) per cent. For low
where RW is the ex-day raw return. yields the mean raw return is 0.24 (bounds -0.18 and
Downloaded by [The University of British Columbia] at 19:00 29 October 2014
Table 1. Regression analysis for the full sample, high- and low-yield groups of NYSE taxable cash dividends around the moves to reduction in capital gains tax rate from 28% to
20% and to trading in 6.25 (dividends . 12.5)
Panel A: Pricing dynamics
Full sample
Estimate -0.25a 0.13a 1.14b -0.15a 0.63a 0.15 0.18
SE 0.03 0.04 0.02 0.03 0.07 0.21
High-yield group
Estimate -0.33a 0.49a 1.20b -0.26a 0.57a -0.11 0.17
SE 0.12 0.15 0.04 0.05 0.12 0.26
Low-yield group
Further evidence from ex-dividend days
b0 b1 b2 b3
Full sample
Estimate 0.004a 0.005a 0.993a 0.003a
SE 0.001 0.000 0.000 0.000
High-yield group
Estimate 0.956a 0.031a 0.550a 0.583a
SE 0.286 0.009 0.130 0.178
Low-yield group
Estimate 0.007a 0.007a 0.990a 0.005a
SE 0.001 0.001 0.001 0.001
Notes: The error term is modelled as ht b0 b1 e2t1 b2 ht1 b3 DUMev where ht is the conditional variance.
a
Denotes significant difference from 0 at the 5% level.
539
540 R. Bali and J. C. Francis
(a) fall by about one-half in the second period. Gene-
1.5 rally, the raw returns are within trading costs for
both yield groups in the two periods. For neither
1.0 group are returns statistically different across the
two periods (tst = 0.9 and 1.6 for high and low
Returns (%)
0.0
Feb-95
Mar-96
Apr-97
May-98
Jun-99
Jul-00
events analysed by Bali and Francis (forthcoming).
We find that taxes may affect investor behaviour but
Month
our results do not support the contention that tax
(b)
Downloaded by [The University of British Columbia] at 19:00 29 October 2014
0.0
0.5
References
Bali, R. (2003a) Variation in ex day dividend pricing: Myth
1.0 or reality?, Journal of Economics and Finance, 27,
190210.
Jan-94
Feb-95
Mar-96
Apr-97
May-98
Jun-99
Jul-00