Taxes, Order Imbalance and Abnormal Returns around the ex‐Dividend day

• Author: David R. Gallagher,Graham Partington,Kingsley Y.L. Fong,Andrew B. Ainsworth
• Date: 01 September 2018
• Published date: 01 September 2018
• DOI: http://doi.org/10.1111/irfi.12155

Taxes, Order Imbalance

and Abnormal Returns around

the ex-Dividend day*

ANDREW B. AINSWORTH

†

,KINGSLEY Y.L. FONG

‡

,DAVID R. GALLAGHER

§,¶

AND GRAHAM PARTINGTON

†

†

University of Sydney Business School, University of Sydney, Sydney, New South

Wales, Australia

‡

UNSW Business School, University of South Wales, Sydney, New South Wales,

Australia

§

Macquarie Graduate School of Management, Sydney, New South Wales, Australiaand

¶

Capital Markets CRC Limited (CMCRC), Sydney, New South Wales, Australia

ABSTRACT

A costly arbitrage model, developed for the Australian imputation tax system,

shows that stocks paying dividends with a tax credit are likely targets for ex-

dividend arbitrage.We show that order imbalance, based on the direct observa-

tion of buyer and seller initiated trades, is a key factor in price movements

around the ex-dividend day. Buying pressure before the ex-dividend day aimed

at capturing the dividend and tax credit leads to an increase in prices that sub-

sequently reversein the ex-dividend period. This effect isconcentrated in those

stocks distributing a tax credit with theirdividend payments. The price pressure

resulting from order imbalance is substantially higher around the ex-dividend

day relative to the effect observed outside this period. Our results reject the

model of Frank and Jagannathan (1998) that bid-ask bounce is responsible for

the ex-day premium and provide support for explanations based on taxes,

transactioncosts, and incomplete price adjustment on the ex-day.

JEL Codes: G12; G14

Accepted: 17 August 2017

Inherent in tax and arbitrage-based explanations of ex-dividend price changes

is the notion that economic agents trade and prices adjust to reflect either one,

or more, of marginal tax rates, risk, or transaction costs (see Elton and Gruber

1970 Kalay 1982). The majority of the literature focuses on how taxes, risk and

* We thank an anonymousreferee, the Editor, Raymond Da Silva Rosa, Stephen Gray,Kartono Liano,

Terry Walter, Henk Berkman participants at the 2010 FMA Asian Conference, 14th INFINITI Confer-

ence on International Finance, and seminar participants at the University of New South Wales and

Macquarie University for helpful commentsand suggestions. We thank SIRCA forproviding data. This

research was funded through an ARC Linkage Grant (LP0561160) involving Vanguard Investments

Australiaand SIRCA.

© 2017 International Review of Finance Ltd. 2017

International Review of Finance, 18:3, 2018: pp. 379–409

DOI: 10.1111/irfi.12155

transaction costs directly affect the value that individual traders place on divi-

dends. However, there is a much less explored channel through which taxes,

risk, or transaction cost can affect prices. This second channel arises because

divergences in dividend valuations between traders can affect how aggressively

they price their orders. Differences in order aggression help to determine the

priority of the trader in the limit order book queue and the probability of trade

execution. Chordia and Subrahmanyam (2004), in a general context, show that

the aggressiveness of traders’orders can result in order imbalances that affect

prices. In the context of ex-dividend trading, for example, if there was a sub-

stantial surplus of cum-dividend buy orders from tax-advantaged dividend cap-

ture traders relative to sell orders from tax-disadvantaged dividend avoidance

traders, then cum-dividend prices would likely end up higher than if buy and

sell orders were more evenly balanced.

The idea that heterogeneity in valuations generates trading around the ex-

dividend date is well developed in the literature (e.g., Boyd and Jagannathan

1994; McDonald 2001). However, the nexus between this heterogeneity, the

aggressiveness of investor trading, any resulting order imbalance and the impact

on ex-dividend prices has received relatively little attention. One exception is the

ex-dividend microstructure model of Frank and Jagannathan (1998). In this

model it is hypothesized that order imbalances will exist around the ex-day due

to investors avoiding dividends. Consequently, a sell order imbalance is expected

on cum-dividend and a buy order imbalance is expected on the ex-dividend day

and prices are thus expected to bounce from the bid to the ask as the stock goes

ex-dividend. Jakob and Ma (2003) find partial support for this theory. For a small

sample of US stocks between November 1990 and January 1991 they show that

there is a buy order imbalance on the ex-day that is related to the ex-day price

drop. However, there is no evidence of any order imbalance before the ex-

dividend day. In contrast, Koski (1996) documents purchasing pressure on the

last cum-dividend day and selling pressure on the ex-day for high-yield stocks in

the US in 1988, but does not explore the link between order imbalance and

returns. Since the extant empirical work is based on restricted samples of stocks

and short time periods it is difficult to reach definitive conclusions.

1

A major

problem in expanding the sample size in such studies is the difficulty of accu-

rately identifying whether a trade was buyer or seller initiated. This is a problem

that we overcome in the current study.

The study extends the Boyd and Jagannathan (1994) theoretical model for

Australia and examines whether order imbalance influences prices on, and

around, the ex-dividend day. There are two reasons why we examine this ques-

tion using Australian data. First, the data allow us to directly observe whether

each trade made on the Australian Securities Exchange (ASX) was initiated by a

buy or sell order. Thus, we are able to undertake a much more comprehensive

analysis of ex-dividend price pressure than has previously been attempted. The

second benefit of studying Australian data is that the operation of the

1 For example, Jakob and Ma (2003) study 106 dividend events over a 63-day period.

© 2017 International Review of Finance Ltd. 2017380

International Review of Finance