ORE Open Research Exeter

Transcription

1 ORE Open Research Exeer TITLE vdend valuaon radng and ransacons coss: he 997 paral abolon of dvdend ax cred repaymens AUTHORS Holland Kevn; Hodgknson Lynn; Jackson Rchard H.G. JOURNAL Accounng and Busness Research EOSITE IN ORE 04 July 2008 Ths verson avalable a hp://hdl.handle.ne/0036/30956 COYRIGHT AN REUSE Open Research Exeer makes hs work avalable n accordance wh publsher polces. A NOTE ON VERSIONS The verson presened here may dffer from he publshed verson. If cng you are advsed o consul he publshed verson for pagnaon volume/ssue and dae of publcaon

2 vdend Valuaon Tradng and Transacons Coss: he 997 aral Abolon of vdend Tax Cred Repaymens aper number 04/08 Lynn Hodgknson Unversy of Wales Bangor Kevn Holland Unversy of Wales Aberyswyh Rchard H G Jackson* Unversy of Exeer Absrac Alhough n July 997 UK resden ax-exemp shareholders los he rgh o repaymen of ax creds on dvdends pad by UK resden companes hey could connue o receve ax cred repaymens n respec of dvdends receved from Irsh resden companes - many of whch are lsed on he London Sock Exchange. In July 997 he rae of ax cred on Irsh companes dvdends was 2%. Ths was reduced o % n ecember 997 and he rgh o ax cred repaymens was abolshed n January 999. We oban nsghs no he ncenves and behavour of ax-exemp nvesors n response o hese changes n he relave ax aracveness of nvesmens n Irsh resden companes over UK resden companes. In parcular we examne he rade off beween he avalably of dvdend ax cred repaymen and he sum of ransacon coss and porfolo resrucurng coss. In addon we examne he behavour of sock prce and radng volume of UK resden companes socks around ex-dvdend days n order o provde a bass of comparson o and o exend he work of Bell and Jenknson (2002). We conclude as follows: () only a s hghes rae 2% was he level of dvdend ax cred on Irsh companes dvdends suffcen o nduce changes n UK ax-exemp shareholders nvesmen sraeges; () mpled oal expeced nvesor coss n dvdend capure are n he order of 2.6% conssen wh he cosng of uncerany over ex-dvdend prce and he effcacy of an-avodance legslaon; () ax-exemp shareholders were he margnal nvesors n UK resden companes pror o he July 997 conssen wh he fndngs of Bell and Jenknson (2002); and (v) ax nduced dvdend capure of UK companes dvdends s evdenced n radng volume paerns. Acknowledgemen The auhors would lke o hank semnar parcpans a he Unverses of Ason Lverpool and Sydney and conference delegaes a he Brsh Accounng Assocaon Annual Conference 2003 Mancheser for her helpful commens and suggesons and a former nernaonal ax parner of a Bg 4 accounancy frm for asssance n nerpreng relevan legslaon. *Conac Auhor: Rchard Jackson Senor Lecurer n Accounng and Fnance School of Busness and Economcs Sreaham Cour Rennes rve Unversy of Exeer Exeer EX4 4U. Tel Secreary Fax E-mal: rchard.jackson@exeer.ac.uk

3 ISSN

4 . Inroducon The paymen of dvdends represens an mporan predcable neracon beween companes and shareholders. Numerous effecs or non-effecs have been arbued o dvdend polcy. Modglan and Mller (963) demonsrae ha under ceran marke assumpons frm value s ndependen of dvdend polcy hough unancpaed changes can mpose ax and adjusmen coss on shareholders. Conversely n a seng characersed by nformaon asymmery beween mangers and shareholders dvdend polcy can ncrease frm value by reducng agency coss (Rozeff 982) and revealng managers nsde nformaon o he capal markes (Ross 977). Oher value relevan effecs have been arbued o dfferenal nvesor level margnal ax raes as beween dvdend ncome and realsed gans and o ck sze selemen coss uncerany n ex-dvdend prces and marke mcrosrucure. These effecs are no wholly muually exclusve and herefore one or more of hem can nerfere n and confound aemps o examne any oher n solaon. For example a hypohessed ax-nduced effec may no occur because of he exsences of ransacon coss or he cumulaon of a number of ax changes whch are ogeher suffcen o overcome ransacons coss may gve rse o spurous emprcal deecon and value oversaemen n respec of one of he changes. In a marke valuaon conex nvesor level ax consderaons are only mporan when hey mpose coss or benefs on he prce seng or margnal shareholder and herefore nfluence her prcng decsons. In he UK nsuonal nvesors domnae he London Sock Exchange (LSE) n erms of he volume and value of shares owned and raded; and more mporanly ax-exemp nsuonal nvesors domnae n erms of prce seng (Bell and Jenknson 2002). Ths paper nvesgaes a change n he dvdend-relaed cash flows accrung o ax-exemp nsuonal nvesors. Alhough he UK Fnance (No.2) Ac 997 abolshed he rgh of UK resden ax exemp shareholders o clam repaymen of ax creds on dvdends receved from UK resden companes on or afer 2 nd July 997 such shareholders could connue o receve ax cred repaymens on a sub-group of UK lsed companes. Under he erms of he UK - Republc of Ireland ouble Taxaon Convenon n operaon as a July 997 ax exemp UK penson funds and UK nsurance company ax exemp penson busnesses (ogeher hereafer referred o as TEIC nvesors ) connued o oban ax cred repaymens on dvdends receved from companes resden n he Republc of Ireland many of whch were (and ndeed sll are) lsed on he London Sock Exchange. 2 Ths sae of affars connued only unl 3 s ecember 998 he double axaon Smlarly under Irsh domesc legslaon Irsh resden penson funds could also oban repaymens of ax creds from Irsh resden companes. The general repaymen of dvdend ax creds o ax exemp nvesors was abolshed from Irsh domesc legslaon n Aprl The Irsh Sock Exchange Ld was esablshed followng he ecember 995 demerger of The Inernaonal Sock Exchange of he Uned Kngdom and he Republc of Ireland Ld. Unl ha pon mos Irsh companes had dual lsng on he ubln and London Sock Exchanges (Mullgan 996). In order o allevae he marke 3

5 convenon havng been revsed by an amendng proocol daed 4 h November 998. Consequenly hree ax regmes can be denfed as follows: () pror o 2 nd July 997 TEIC nvesors n boh UK resden and Irsh resden companes were enled o he repaymen of ax creds; () beween 2 nd July 997 and 3 s ecember 998 TEIC nvesors were enled o he repaymen of ax creds on dvdends receved only from Irsh resden companes (an Irsh dvdend ax cred rae change durng hs perod allows o be furher sub-dvded no he perods 2 nd July 997 o 2 nd ecember 997 nclusve and 3 rd ecember 997 o 3 s ecember 998 nclusve); and () afer he s January 999 TEIC nvesors were no enled o a repaymen of ax cred upon dvdend from eher UK resden or Irsh resden companes. Observng changes n nvesor behavour as beween he above perods can provde nsghs no nvesor ncenves expecaons and audes concernng dvdends. In order o absrac from sgnallng and agency consderaons we examne ex-dvdend day prce and volume changes developng and esng a number of hypoheses concernng sock prce and radng volume changes n response o varaons n he avalably and level of dvdend ax cred repaymens. These hypoheses lnk he relave ax aracveness of nvesmen n Irsh resden (as compared o UK resden) companes over he perods se ou above and he acons of TEIC nvesors and are desgned o oban nsghs no he ncenves o and behavours of such nvesors. In parcular we examne he rade off beween he avalably of dvdend ax cred repaymen and ransacon coss. Furher we examne he behavour of sock prce and radng volume of UK resden companes socks around ex-dvdend days n order o provde a bass of comparson and o exend he work of Bell and Jenknson. The paper s man conrbuons can be summarsed as follows: () he developmen and esng of axbased hypoheses over a perod conanng several axaon changes; () he examnaon of radng volume across a number of perods where here are clear varaons n he expeced profably of dvdend capure radng; and () he developmen of a model whch can explan he propensy of dvdend capure radng acvy n erms of ransacon coss. The poenal for mpac of presence concerns of some Irsh companes a specal assocaon agreemen allowed dual resden companes he opon o manan her dual lsng saus - and many of he larger Irsh companes connue o hold such dual lsng saus (Lucey 200). Irsh frms seekng an nal quoaon pos ecember 995 however had o choose beween he wo markes. 4

6 confoundng unrelaed changes o marke and radng arrangemens s mgaed n ha he perod over whch he dvdend axaon changes consdered n he paper occurred and herefore he sample perod are relavely shor. Our resuls can be summarsed as follows. Frs TEIC nvesors appear o be he margnal nvesors n Irsh resden shares only durng he perod when Irsh dvdend ax cred repaymens were avalable a he rae of 2% (as compared o zero per cen for dvdends from UK resden companes). Ths resul s conssen wh changes n he relave raes of ax cred repaymen beween UK and Irsh resden companes nducng a change n TEIC nvesors sraeges.e. evdence of a dvdend clenele formng and subsequenly dsspang. Second he absence of evdence suggesng TEIC nvesors where he margnal nvesors n Irsh resden shares when Irsh dvdend ax cred repaymens were avalable a he rae of only % (agan as compared o zero per cen for UK resden companes) s conssen wh TEIC nvesors ransacon coss beng of suffcen magnude o ouwegh he now reduced poenal ax benefs of holdng shares of Irsh resden companes and wh such nvesors seekng o maxmse afer ax reurns raher han ax benefs. Thrd our model of ransacon and porfolo resrucurng coss mply expeced ransacon coss n he order of 2.6% whch s hgher han he round rp ransacons coss suggesed by James (2000). Ths fndng s conssen wh he expeced coss of devaon from an oherwse opmal porfolo for he purposes of dvdend (and relaed ax cred) capure beng sgnfcan uncerany over exdvdend prce and he effcacy of an-avodance legslaon whch mposes addonal coss by requrng a mnmum holdng perod f dvdend ax cred capure o be fully effecve. Fourh for UK resden companes TEIC nvesors appear o be he margnal nvesor pror o he abolon of ax cred repaymens - conssen wh he fndngs of Bell and Jenknson. Ffh we observe paerns n radng volume conssen wh ax nduced dvdend capure for he UK resden companes (bu daa consderaons preven a smlar analyss for Irsh resden companes). Ths remander of hs paper proceeds as follows: he nex secon dscusses he relevan legslaon and llusraes he mpac of he changes on he axaon of dvdends; prevous work examnng boh prce changes and volumes raded around he ex-dvdend day are dscussed n he hrd secon; he fourh secon ses ou mehodology hypoheses daa source defnon of varables; he resuls are presened and dscussed n he ffh secon; and he fnal secon concludes. 5

7 2. Relevan legslaon Ths secon summarses he relevan ax legslaon coverng he repaymen of dvdend ax cred and relaed an-avodance legslaon n he conex of he UK - Republc of Ireland ouble Taxaon Convenon. urng he early par of he perod under revew he Republc of Ireland and he Uned Kngdom boh operaed a paral mpuaon sysem regulang he neracon of he corporae and personal ax sysems wh respec o dvdends. The key feaures of boh counres sysems were dencal. 3 A company payng a (ne) dvdend was requred o pay advance corporaon ax (ACT) o he Inland Revenue/Revenue equal o he produc of he lower rae of ncome ax and he gross dvdend where he gross dvdend s equal o he ne dvdend plus he ACT. Shareholders receved he ne dvdend plus an mpued ax cred ( ax cred ) whch ogeher equalled he gross dvdend. If he shareholder s margnal ax rae on dvdend ncome ( margnal rae ) was equal o he lower or basc rae of ncome ax no furher acon was requred. If he shareholder s margnal rae was greaer han he lower rae an addonal paymen by he shareholder was due o he ax auhores beng equal o he amoun of he gross dvdend mulpled by he excess of he hgher rae over he basc rae. If on he oher hand he shareholder s margnal rae was less han he lower rae of ax hey could oban a repaymen equal o he amoun of he gross dvdend mulpled by he dfference beween her margnal rae and he lower rae of ax. Hence a ax-exemp nvesor could reclam an amoun equvalen o he dvdend ax cred n full. Afer s July 997 however he rgh of TEIC nvesors o oban repaymen of he ax cred on dvdends receved from UK resden companes was abolshed. The ably of ax-exemp nvesors resden n Ireland o oban repaymen of he ax cred on dvdends receved from companes resden n he Republc of Ireland connued unl he abolon of dvdend ax creds n Aprl 999. Inrgungly hough durng he perod 2 nd July 997 o 3 s ecember 998 TEIC nvesors could connue o oban repaymen of he ax creds on dvdends pad by companes resden n he Republc of Ireland under he erms of he UK - Republc of Ireland ouble Taxaon Convenon hen n operaon (alhough no smlar rgh exsed under UK domesc legslaon). One objecve of axaon convenons s he avodance of dscrmnaon beween axpayers of he wo conracng saes. os July 997 Irsh domesc legslaon connued o allow repaymen of ax creds on dvdends from companes resden n he Republc of Ireland o ax exemp nvesors smlarly 3 Ths verson of he UK mpuaon sysem was abolshed by he Fnance Ac of 999 wh effec from 6 Aprl

8 resden. The convenon exended ha rgh o TEIC nvesng n hose same Republc of Ireland companes - hereby ensurng comparable reamen. 4 The avalably of ax cred repaymens o ax exemp shareholders may generae a preference for dvdend ncome over capal growh. 5 Indeed one of he jusfcaons of he July 997 change n he UK was o releve he pressure on companes o pay dvdends n order o sasfy he preferences of ax-exemp shareholders: The presen sysem of ax creds encourages companes o pay ou dvdends raher han renves her profs (Chancellor of he Exchequer 997). If non-ax facors lead ax-exemp shareholders o hold low dvdend yeld porfolos such shareholders may aemp o ncrease dvdend ncome by emporarly alerng her porfolos o nclude hgher dvdend yeldng shares. For example purchasng and hen sellng a hgh dvdend yeld share around s ex-dvdend dae would provde addonal dvdend ncome whls mnmsng he coss of emporary porfolo mbalance. An-avodance legslaon however exss n boh he Republc of Ireland and he Uned Kngdom desgned o hwar he ax benef of such bond washng acves. 67 When a ax-exemp nvesor purchases a share and whn a monh and sells (boh ransacons beng a curren marke prce ) and durng he perod of ownershp he share urns ex-dvdend hen he amoun of avalable ax cred repaymen s reduced by an approprae amoun. 8 Ths legslaon however was n place n boh he Republc of Ireland and he Uned Kngdom hroughou he perod under examnaon n hs paper and herefore does no affec he relave ax aracveness of dvdend ncome durng he hree phases of he denfed naural expermen. 4 The enlemen o repaymen was gven by paragraph 2(b) whch orgnally read as follows: A resden of a Conracng Sae who receves dvdends from a company whch s a resden of he oher Conracng Sae shall subjec o he provsons of sub-paragraph (c) of hs paragraph and provded he s he benefcal owner of he dvdends be enled o he ax cred n respec hereof o whch an ndvdual resden n ha oher Conracng Sae would have been enled had he receved hose dvdends and o he paymen of any excess of ha ax cred over hs lably o ax n ha oher Conracng Sae. Sub-paragraph (c) referred o above provdes ha he enlemen o repaymen s no avalable where he benefcal owner conrols 0% or more of he vong power n he company payng he dvdend. In he amended reay he orgnal paragraph 2 was deleed. 5 There may be non-ax reasons for such a preference. For example Secon 34 of he ensons Ac 995 empowers rusees o make nvesmens bu he Ac does no defne he erm nvesmen. Case law provdes some clarfcaon alhough here s sll uncerany. In Re Wragg (99) 2 Ch. 55 p.64 nvesmen s defned narrowly as he purchase of propery n order o be held for he sake of he ncome wll yeld (presumably n conras o radng where he move may be o purchase an asse wh he expecaon of a subsequen resale a a hgher prce). Alhough he cours have recognsed new approaches o nvesmen pung he emphass n nvesmen on he makng of capal prof a he expense of nvesmen yeld (Marson v. Moron [986] WLR 343 a page 350) he cauous approach mus be o adop he src nerpreaon se ou n Re Wragg n he absence of a more defne defnon (Nabarro Nahanson 2000). 6 The erm bond washng s generc o all acves ha aemp o wash an underlyng secury free of he relaed coupon or dvdend paymen. 7 In he Republc of Ireland he legslaon s conaned n Secon 750 of he Taxes Consoldaon Ac 997. The correspondng provsons n he Uned Kngdom are found n Secon 733 of he Income and Corporaon Taxes Ac The reducon beng gven by: max. poenal ax cred perod from prevous ex - dvdend dae o one day before dae of purchase dae perod from prevous ex - dvdend dae o 'curren' ex - dvdend 7

9 The ax cred raes applcable o dvdends from boh Republc of Ireland resden and Uned Kngdom resden companes vared over he perod under revew and are se ou n anel A of Table. anel B of Table summarses he perods of avalably of ax cred repaymen o ax-exemp nvesors as dscussed above. 3. revous leraure 3.. Ex-dvdend prce behavour In addon o he amoun of dvdend a number of oher facors have been denfed n he leraure as poenally nfluencng he level of he ex-dvdend day fall n share prce. In a rsk-neural world whou axes and absracng from ransacon coss selemen coss and oher marke mperfecons he drop n share prce when a share urns ex-dvdend should equal he dvdend pad o shareholders. In oher words he rao of he prce drop o he dvdend pad ofen referred o as he drop off rao (OR) should be equal o one. Followng he semnal work of Campbell and Beranek (955) many emprcal sudes have examned wheher or no hs hypohess holds. Elon and Gruber (970) repor ha he prce drop s more ofen han no less han he dvdend pad and argue ha he OR wll only be equal o one f he effecve ax raes on dvdends and capal gans are he same. They argue ha shareholders can choose o sell a share when s cum-dvdend and forego he dvdend paymen generang he followng cash flows: T G ( ) p () where he ex-dvdend dae of he share s a me ; s he share prce a close on he radng day before ; p s he prce orgnally pad for he share; and T G s he rae of axaon on capal gans whch he nvesor faces. Alernavely he shareholder may choose o sell he share when s ex-dvdend and generae cash flows under an mpuaon sysem as follows: T G ( ) p + ( TI ) ( T ) ACT (2) 8

10 where s he share prce on he ex-dvdend day; s he amoun of dvdend pad T I s he rae of axaon on dvdend ncome and T ACT s he rae of mpuaon ax (equal o he basc rae of ax durng he perod analysed). Invesors wll be ndfferen beween sellng cum- or ex-dvdend provded expressons () and (2) are equal. Assumng hs condon and rearrangng he resulng equaon he OR s gven by: ( ) ( TI ) = ( T )( T ) G ACT (3) and hence he OR wll equal one f and only f T I = T + T T T. Under a classcal ax G ACT G ACT sysem such as ha n he US he OR wll equal one f and only f T I = T G. Lakonshok and Vermaelen (983) Eades e al. (984) oerba and Summers (984) Kaplans (986) Barclay (987) Fedena and Grammakos (993) and Lasfer (995) amongs ohers also sugges ha dfferenal axaon may explan a OR of less han one. If nvesors ax raes are no homogenous however arbrage may occur as nvesors wh relavely favourable ncome ax raes purchase shares cum-dvdend from nvesors who face heaver ncome ax raes. If hs arbrage occurs he OR wll n par reflec he ransacon coss of hese shor-erm nvesors. Oher posed explanaons for ORs dfferen from one nclude ck sze selemen coss uncerany n ex-dvdend prces and marke mcrosrucure effecs. Bal and He (998) argue ha f he dvdend paymen s no a mulple of he curren ck sze hen he share prce canno change by he exac amoun of he dvdend and conend ha he ex-dvdend prce fall wll be rounded down o he neares ck sze endng o deflae he OR. Lasfer (995) pons ou ha pror o July 994 he London Sock Exchange operaed a fxed-selemen sysem under whch nvesors f hey purchased shares a he sar of he selemen perod dd no have o sele for wo weeks. A large proporon of shares ended o go ex-dvdend on he frs day of he selemen perod hence f an nvesor purchased a share cum-dvdend (a he end of he prevous selemen perod) hey would sele wo weeks earler han f hey were o purchase he share ex-dvdend. Frank and Jaganahan (998) examne ex-dvdend day prcng n Hong Kong where neher dvdends nor capal gans are axed and sll fnd ha he OR rao ends o be less han one posng marke mcrosrucure as an explanaon. Under he mpuaon sysem ax-exemp nvesors able o reclam mpued ax creds upon dvdends should prefer dvdends o capal gans - and hence we argue ha he OR (based on he ne 9

11 dvdend) f such ax-exemp nvesors are he margnal nvesors and absracng from all non-ax mperfecons should be greaer han one Tax-exemp nvesors The promnence of ax-exemp nvesors and parcularly penson funds s evdenced by he sgnfcance of her aggregae holdngs (22.% of oal UK raded equy was held by penson funds as a 3 s ecember 997). The resuls of Bell and Jenknson suppor he hypohess ha ax-exemp nvesors are he margnal nvesors n UK socks especally socks wh hgh dvdend yeld. Absen ransacons coss and porfolo adjusmen consderaons we would expec ncreased acvy by TEIC nvesors swchng from UK resden o Irsh resden socks durng he perod 2 nd July 997 o he 3 s ecember 998 snce o hs nvesor group over hs perod dvdends pad by Irsh resden companes would be worh more han dvdends pad by UK resden companes. Furher f TEIC nvesors were he margnal nvesors n Irsh resden socks hroughou he perod a sgnfcan decrease n he value of he dvdend.e. a fall n he OR should be observed followng he abolon of he ax cred on dvdends from Irsh resden companes. Swches n nvesor holdngs performed n order o capure he axaon-based benefs need only have been emporary snce TEIC nvesors would only need o hold shares n he Irsh resden company for a monh ncludng he dae when hey wen ex-dvdend n order o capure dfferenal benef n full (.e. no reduced by an-avodance an-bond washng legslaon). A change n porfolo o capure hese ax effecs may however have requred nvesors o devae from her opmal porfolo hence poenally ncreasng her exposure o rsk ceers parbus. Ths swch may only have been emporary bu snce he prce a whch he nvesor could sell ex-dvdend (a a pon a monh or more snce purchase n order o enjoy he full dfferenal benef) was unceran he rsk nvolved was furher ncreased. Tradng also as always would have ncurred ransacons coss. Hence raonal nvesors would only have raded o capure dvdends f he expeced dfferenal axaon-based benefs exceeded he expeced coss of addonal rsk and ransacon coss nvolved Ex-dvdend radng volume Mchaely and Vla (995) pos ha examnng ex-dvdend prcng alone may be nsuffcen o denfy nvesors ax preferences and sugges ha changes n radng volume around he ex-dvdend day may provde furher nformaon beyond ha conaned n prces. They argue ha s no possble o denfy from prces wheher nvesors face homogenous or heerogenous effecve ax raes and sugges ha radng volume ncreases sgnfcanly around he ex-dvdend day reflecng dvdend capure and ha wll ncrease furher as nvesors face ncreasngly dfferen preferences for 0

12 dvdends or capal gans. Boh ransacon coss and poenal ncreases n rsk (due o emporarly movng from an opmal porfolo) are unavodable coss nvolved n swchng shares o capure ax creds. As dscussed above f he expeced coss of swchng ouwegh he poenal benefs nvesors wll no rade and hence he hgher he ransacons coss and he ncrease n sysemac rsk he less aracve dvdend capure becomes. Mchaely and Vla (995) fnd ha boh ransacon coss and rsk sgnfcanly mpac on he volume of shares raded around he ex-dvdend day whereas her mpac on prces s found by hese auhors o be less sgnfcan. L (2002) repors ha excess radng volume ncreases wh dvdend yeld. L also examnes he mpac of dfferen ax regmes and argues ha ax regme changes should have an mpac on exdvdend excess radng volume because hey cause changes n he relave axaon of dvdends and capal gans and goes on o say he gans o exchangng ax burdens among nvesors wh dfferen ax saus ncreases [decreases] and hs ncreases [decreases] radng volume around he ex-dvdend days. The dfference beween he effecve ax raes for dvdends and capal gans ncreased n he US followng an ncrease n ncome ax raes n 993 and a decrease n capal gans ax raes n 997. L expeced o fnd an accompanyng ncrease n radng volume reflecng he ncreased poenal for ax arbrage bu nsead found a decrease. He conends ha hs may be due o a decrease n dvdend yeld suggesng frms may have changed her dvdend polcy followng he ax regme changes. The removal of he ax creds frs n he Uned Kngdom and hen n he Republc of Ireland decreased he exen of dfferenal axaon beween capal gans and dvdends over me. The decrease n poenal benefs was lkely o dscourage he exen of dvdend capure acvy followng he removal of he ax creds and hence s expeced ha he exen of excess radng volume around he ex-dvdend day has on he whole decreased over he perods examned. Foregong dscusson concernng radng volume around ex-dvdend days mus be compounded by consderaon of he effecs of applcable an-avodance legslaon. As se ou n Secon 2 such legslaon was n operaon n boh he Republc of Ireland and he UK hroughou he perod of hs sudy denyng full repaymen of dvdend ax cred (where avalable) o nvesors f hey dd no hold he underlyng share for one monh or more. For example purchase of a share on s las day cum-dvdend and sale on he ex-dvdend dae would resul assumng successful operaon of he an avodance legslaon n a very small proporon of he assocaed dvdend ax cred o be reclamed.

13 4. Mehodology hypoheses daa source varables and models We examne he marke valuaon of dvdends and radng volumes around ex-dvdend regmes for UK and Irsh companes quoed on he London Sock Exchange over he followng four ax regmes: ) ror o he 2 nd July 997 when TEIC nvesors n boh UK and Irsh resden companes were able o reclam he ax cred assocaed wh hose companes dvdends. In he followng hs perod s desgnaed perod Boh. ) From 2 nd July 997 o 2 nd ecember 997 (nclusve) when TEIC nvesors were able o reclam he ax cred assocaed wh dvdends from Irsh resden companes a a rae of 2% on he gross dvdend bu were unable o make any such clam n respec of dvdends from UK resden companes. Ths perod s desgnaed Irsh2. ) From 3 rd ecember 997 o 3 s ecember 998 (nclusve) when TEIC nvesors were able o reclam he ax cred assocaed wh dvdends from Irsh resden companes a a rae of % on he gross dvdend bu were unable o make any such clam n respec of dvdends from UK resden companes. Ths perod s desgnaed Irsh. v) Followng 3 s ecember 998 when TEIC nvesors were no able o reclam any ax cred n respec of dvdends pad by eher UK or Irsh resden companes. Ths perod s desgnaed Neher. 4.. rce behavour Ths sub-secon commences wh he formulaon of regresson models o explan ex-dvdend prce behavour. We formulae wo prmary regresson models n respec of each of Irsh and UK resden companes dvdend evens - an unscaled and a scaled model - and n each case perform he esmaons usng boh unadjused and marke-adjused prces. The secon goes on o formulae hypoheses concernng he models coeffcens. In lne wh Elon and Gruber (970) we compare he drop n share prce on he ex-dvdend day wh he ne dvdend pad as follows: (4) 2

14 where for company gong ex-dvdend a me - and are respecvely he closng prces per share on he cum-dvdend and ex-dvdend days and s he ne dvdend per share pad. If he drop n share prce s equal o he ne dvdend pad he OR as per expresson (4) wll equal. The OR n effec ndcaes he value of he dvdend o he margnal nvesor n he sock. If dvdends are ax-advanaged o capal gans whch ndeed hey were for TEIC nvesors up o 2 nd July 997 n respec of boh UK and Irsh resden socks and furher up o 3 s ecember 998 n respec of Irsh resden socks and TEIC nvesors are he margnal nvesors we would expec he OR o be greaer han one durng he perods of ax-advanage. 9 Alhough we repor he ORs here are mehodologcal problems wh he sasc as s no normally dsrbued and s dsored by very small ne dvdend paymens. An alernave procedure o denfy he relaonshp beween he prce drop and he dvdend s o examne he slope coeffcen n he followng regresson: = + ε (5) where ε s a sochasc error erm. Followng Frank and Jaganahan (998) who argue ha a sgnfcanly negave nercep s evdence of a marke-mcrosrucure effec we allow for an nercep as follows: = α + + ε (6) Fnally n order o examne wheher or no he marke value of dvdends has changed over he hree perods of our sudy (perods Boh Irsh and Neher) we nclude perod slope dummes and esmae va OLS he followng augmened model: + = α + IRISH 2 + ( ) ( IRISH 2 ) + IRISH( IRISH ) + ε (7) where s a zero-one dummy equal o one when he ex-dvdend day falls n he perod pror o 2 nd July 997; IRISH2 s a zero-one dummy equal o one when he ex-dvdend day falls n he perod from 2 nd July 997 o 2 nd ecember 997 (nclusve); and IRISH s a zero-one dummy equal 9 Absracng from oher facors as dscussed earler whch may affec he OR. 3

15 4 o one when he ex-dvdend day falls n he perod from 3 rd ecember 997 o 3 s ecember 998 (nclusve). The base case of he model s herefore for ex-dvdend days fallng n he perod followng he 3 s ecember 998. Ths esmaon s performed separaely for dvdend evens assocaed wh Irsh resden frms ( he unscaled Irsh model ) and hose assocaed wh UK resden frms ( he unscaled UK model ). In he UK model here s no effec from changng raes of Irsh dvdend ax cred so he esmaed model s smplfed o: ( ) ( ) IRISH IRISH ε α = (8) where IRISH s a zero-one dummy equal o one when he ex-dvdend day falls n he perod from 2 nd July 997 o 3 s ecember 998 (nclusve). In dscussng he UK daa and he UK model herefore we desgnae perod Irsh o be from 2 nd July 997 o 3 s ecember 998 when TEIC nvesors were able o reclam he ax cred assocaed wh dvdends from Irsh resden companes a a rae of eher 2% or % on he gross dvdend bu were unable o make any such clam n respec of dvdends from UK resden companes. We also esmaed scaled forms of he above models o mgae any heeroscedascy concerns. IRISH IRISH IRISH IRISH ε α = 2 2 (9) for dvdend evens assocaed wh Irsh resden frms ( he scaled Irsh model ) and IRISH IRISH ε α = (0)

16 for dvdend evens assocaed wh UK resden frms ( he scaled UK model ) where n each he varables are as prevously defned. We esmae each of hese models usng boh unadjused and marke-adjused prces 0 makng he marke adjusmen as follows: = Rm () where R m s he reurn on he marke on he ex-dvdend day as proxed by he reurn on he Fnancal Tmes All Shares Index and s he equy bea of company as esmaed usng he marke model and daly reurns daa over he -05 o 6 days pror o ex-dvdend day. Bell and Jenknson use monhly reurns n calculang equy beas n order o overcome problems wh hn radng whereas we choose o adop he Scholes-Wllams (977) adjusmen. In order o use monhly daa o compue he equy bea s necessary o use reurns from a consderably longer perod han s he case f daly reurns are used over whch longer perod here s greaer chance ha he rsk characerscs of he company may have changed and hence a greaer chance ha he equy bea may be based. If TEIC nvesors n Irsh resden socks were he margnal nvesors durng perods Boh Irsh2 and Irsh hen we would expec each of IRISH 2 and IRISH o be sgnfcanly greaer han zero n he Irsh models (snce he dvdend would be worh sgnfcanly more o he ax-exemp penson fund nvesor pror o he removal of he ax-cred). If such nvesors were no normally he margnal nvesors n Irsh resden frms bu were he margnal nvesors durng he second and hrd perods (perods Irsh2 and Irsh ) - when TEIC nvesors could have been araced o Irsh resden companes n order o capure he ax-cred whch was no longer avalable n respec of UK resden companes we would expec IRISH 2 and IRISH o be sgnfcanly greaer han zero bu no. If neher IRISH 2 nor IRISH are sgnfcanly dfferen from zero we would conclude ha TEIC nvesors dd no ake he opporuny o capure he ax cred by nvesng n Irsh resden socks and/or hey were no he margnal nvesors a any me durng he perod under analyss. Fnally wheher or no s sgnfcanly greaer han zero f IRISH 2 s sgnfcanly greaer han zero whls IRISH s no we would conclude ha TEIC nvesors aced o capure dvdends and assocaed ax creds durng perod Irsh2 bu no durng Irsh; and furher ha 0 We repor he resuls based on marke-adjused prces n he nex secon. Resuls based on unadjused prces are no qualavely dfferen for he Irsh models. For he UK models slgh dfferences n resuls based on unadjused prces n comparson o he man repored resuls are dsclosed by way of foonoe n he nex secon. Full resuls based on unadjused prces are avalable from he auhors. 5

17 dvdend capure was perceved o be worhwhle wh a dvdend ax cred rae of 2% on gross dvdend bu no when he ax cred rae was % -.e. we may deduce some bounds for oal ransacons and porfolo resrucurng coss as expeced by hese nvesors. If our resuls for he UK model are o suppor hose of Bell and Jenknson we would deduce ha axexemp nvesors were he margnal nvesors n UK resden socks and ha should be sgnfcanly greaer han zero (snce he dvdend would be worh sgnfcanly more o he TEIC nvesor pror o he removal of he ax-cred). Furher he coeffcen IRISH should no be sgnfcanly greaer han zero as TEIC nvesors n UK resden socks could no reclam he axcred durng perod Irsh. Our hypoheses are herefore as follows: Hypohess. TEIC nvesors were he margnal nvesors n Irsh resden socks quoed on he LSE pror o he removal of he ax cred for such socks on he 3 s ecember 998. As a resul IRISH 2 and IRISH are all sgnfcanly greaer han zero n he Irsh models. Hypohess 2. TEIC nvesors were he margnal nvesors n Irsh resden socks quoed on he LSE pror o he reducon n dvdend ax cred rae for such socks from 2% o % on 3 rd ecember 997. As a resul and IRISH 2 are sgnfcanly greaer han zero and IRISH s no sgnfcanly greaer han zeros. Hypohess 3. TEIC nvesors were no normally he margnal nvesors n Irsh resden socks quoed on he LSE pror o 2 nd July 997 bu were he margnal nvesors durng he perod 2 nd July 997 o 3 s ecember 998 n order o capure he preferenal reamen wh respec o he ax cred afforded o nvesmen n Irsh resden companes. As a resul IRISH 2 and IRISH are sgnfcanly greaer han zero n he Irsh models bu s no. Hypohess 4. TEIC nvesors were no normally he margnal nvesors n Irsh resden socks quoed on he LSE pror o 2 nd July 997or afer 2 nd ecember 997 bu were he margnal nvesors durng he perod 2 nd July 997 o 2 nd ecember 997 n order o capure he preferenal reamen wh respec o he ax cred a 2% afforded 6

18 o nvesmen n Irsh resden companes. As a resul IRISH 2 s sgnfcanly greaer han zero n he Irsh models bu and IRISH are no. Hypohess 5. TEIC nvesors were he margnal nvesors n UK resden socks quoed on he LSE pror o he removal of he ax cred for such nvesmen on he 2 nd July 997. As a resul s sgnfcanly greaer han zero n he UK models bu IRISH s no Cluserng To conrol for poenal effecs of UK resden companes shares gong ex-dvdend on he same days.e. concern ha resduals may no be ndependenly and dencally dsrbued he UK model esmaons were repeaed usng porfolos of socks grouped on he bass of her ex-dvdend day. The resuls do no dffer subsanally from hose se ou n he nex secon. Therefore n he neress of economy hey are no separaely repored n full bu dfferences from he man repored resuls are dsclosed by way of foonoe. Concernng dvdend evens of Irsh regsered companes 70 of our sample of 98 evens (7%) occurred on days absen any oher sample dvdend evens; and only 28 of our sample of evens (29%) occurred on days on whch here was one or more oher sample dvdend even. Therefore re-esmaon of he Irsh models o allow for cluserng was no consdered necessary Tradng behavour As dscussed above radng volumes are expeced o be hgher around he ex-dvdend day as axprvleged nvesors purchase he shares cum-dvdend and sell ex-dvdend. Smlar o Mchaely and Vla (995) and L (2002) we compue he excess radng volume on days 5 o +5 around he exdvdend day EV as follows: number of shares n sock raded on day average no. of shares n sock raded per day durng esmaon perod (2) where [ 5 + 5] n erms of days relave o he ex-dvdend day. We adop an esmaon perod of 05 o 6 days relave o he ex-dvdend day. aronng he sample by Irsh versus UK resdence and he Irsh sample by four ax regme perods and he UK sample by hree we calculae average excess volumes as follows: These resuls are avalable from he auhors. 7

19 EV = n n = EV (3) where [ 5 + 5] relave o he ex-dvdend day and n s he number of ex-dvdend day observaons durng he ax regme concerned durng relave day durng he es perod. We also calculae he sgnfcance wh whch we may nfer ha he mean radng volume durng he es perod exceeds ha durng he esmaon perod (-es on dfference n means one aled es) for each dvdend even and agan paronng he sample as above calculae he proporons of each sub-sample n whch such sgnfcance s below he 50% 0% 5% 2.5% and % levels. We hen compare proporons over ax regme of separaely Irsh and UK resden companes. As he exen of ax-heerogeney was greaer for boh Irsh and UK resden socks pror o he removal of he ax cred we would expec excess radng o be greaer durng ha perod. In parcular durng he perod 2 nd July 997 o 3 s ecember 998 no only was here ax-heerogeney based on nvesors ax posons bu he ably of nvesors n Irsh resden companes sll o reclam dvdend ax creds mean here was ax-heerogeney based on he resdency of he companes. We would expec as a resul ha he volume of shares raded n Irsh resden socks o be greaer durng hs perod; and greaes durng he perod 2 nd July 997 o 2 nd ecember 997 whle he rae of Irsh dvdend ax cred remaned a 2%. Hence he followng hypoheses: Hypohess 6. The number of shares raded n Irsh resden socks quoed on he LSE around dvdend evens was hgher durng he perod pror o 2 nd July 997 han was afer he 3 s ecember 998 and hghes durng he perod 2 nd July 997 o he 3 s ecember 998 conssen wh TEIC nvesors changng her porfolos n order o capure he preferenal reamen wh respec o he ax cred afforded o nvesors n Irsh resden companes. Hypohess 7. The number of shares raded n Irsh resden socks quoed on he LSE around dvdend evens was hgher durng he perod pror o 2 nd July 997 han was afer 2nd ecember 997 and hghes durng he perod 2 nd July 997 o 2 nd ecember 997 conssen wh TEIC nvesors changng her porfolos n order o capure he preferenal reamen wh respec o he ax cred a 2% afforded o nvesors n Irsh resden companes. 8

20 Hypohess 8. The number of shares raded n UK resden socks quoed on he LSE around dvdend evens was hgher pror o 2 nd July 997 han was afer hs dae - conssen wh hgher ax heerogeney aa The daa source for hs sudy was aasream. aasream ls FTAL99 (ls of FTSE All Share Index consuens n 999) yelded 8 UK frms lsed on he LSE n 999; and aasream ls FLON (all non-uk-regsered frms lsed on he LSE) ogeher wh ls EAIR ( dead Irsh frms) yelded 57 Irsh regsered LSE lsed frms. Sarng wh he populaon of all hese frms dvdend evens durng he perod 996 o 999 (nclusve) reducons were made for unusable evens as per Table 2 o gve a useable sample of 4028 UK resden company dvdend evens and 93 Irsh resden company dvdend evens wh ex-dvdend dae durng he perod January 996 o 3 ecember 999 for whch share prce daa were avalable from 05 days pror o he ex-dvdend dae hrough o he ex-dvdend dae. To avod problems assocaed wh hn radng of company shares and followng Bell and Jenknson who pon ou ha alhough no prce change s observed f here s no rade on he ex-dvdend day he prce mgh change when he share s raded laer s approprae o nclude n he OR and regresson analyses only dvdend evens where he sock was raded on he ex-dvdend day. Therefore we exclude dvdend evens where here was no share prce change on he ex-dvdend day. Fnally we exclude from he samples dvdend evens where he declaraon and paymen daes sraddle changes n avalably n dvdend ax cred repaymen o TEIC nvesors.e. dvdends declared by Irsh or UK resden frms before 2 nd July 997 and pad on or afer ha dae and dvdends declared by Irsh resden frms up o and ncludng 3 s ecember 998 and pad afer ha dae. As per Table 2 hs resuls n samples of 3596 UK resden company dvdend evens and 98 Irsh resden company dvdend evens for whch analyses of basc sascs OR and regresson esmaon are performed. Tradng volumes daa were also colleced for he perod 05 days pror o and 5 days followng he exdvdend day where such daa were avalable from our aasream source yeldng samples of 2955 UK resden company dvdend evens and 70 Irsh resden company dvdend evens for whch analyss of radng volume s performed (see Table 2). The number of Irsh resden company dvdend evens n our sample s herefore modes parcularly n perod Irsh2 wh poenal mplcaons for he level of sgnfcance of he relaonshps ha we hypohesse and seek. I s mporan o noe however ha our samples are based on he populaon of 9

21 all LSE-lsed UK and Irsh resden frms dvdend evens wh an ex-dvdend dae beween January 996 and 3 ecember 999 wh sample reducon owng only o mehodologcal necessy and daa avalably. Snce ax reamen of dvdend and n parcular avalably and rae of assocaed ax cred reclam depend upon he dae he dvdend was pad (raher han for example he dae was declared or approved) our sample dvdend evens are assgned o perods (Boh Irsh2 Irsh or Neher for Irsh resden companes dvdends and Boh Irsh or Neher for UK resden companes dvdend evens) based on paymen dae. 5. Emprcal resuls 5.. Basc sascs The basc sascs for he samples analysed are presened n Table 3. The mean OR for Irsh resden socks (Table 3 anel A) rses from 0.52 durng perod Boh o durng perod Irsh2; hen falls durng perod Irsh o 0.77 a level smlar o ha seen n perod Boh; and fnally rses n perod Neher o Ths s conssen wh our fourh hypohess ha TEIC nvesors were no normally he margnal nvesors n Irsh resden socks quoed on he UK Sock Exchange pror o 2 nd July 997or afer 2 nd ecember 997 bu were he margnal nvesors durng he perod 2 nd July 997 o 2 nd ecember 997 n order o capure he preferenal reamen wh respec o he ax cred a 2% afforded o nvesmen n Irsh resden companes. Tesng he changes n OR usng boh a paramerc es (analyss of dfferences n means) and a non-paramerc es (Wlcoxon Mann Whney es) (Table 4 anel A) we fnd srong suppor for hs hypohess. Under dfference of means esng he change beween perods Boh and Irsh2 s sgnfcan a he 5% level; ha beween Irsh2 2 and Irsh a %; and he change beween perods Irsh and Neher s no sgnfcan. The respecve sgnfcances va Wlcoxon Mann Whney esng are % 5% and nsgnfcan. The mean OR s greaes for UK resden companes (Table 3 anel B) when he ax-cred s avalable fallng slghly from n perod Boh o n perod Irsh and fallng agan o n perod Neher. Ths lends some suppor o our ffh hypohess (above) alhough he fall n OR beween perods Boh and Irsh s lower han expeced. Our resuls as beween perods Boh and Neher are n agreemen wh hose of Bell and Jenknson who argue ha snce he only group of nvesors affeced by he removal of he ax cred were ax-exemp nvesors a sgnfcan change n 2 To deermne wheher he companes payng a dvdend durng he perod Irsh 2 acceleraed her dvdend paymen o ake advanage of he hgher ax cred we compared he monh n whch he dvdend was pad o he 20

22 he OR suggess ha he margnal nvesors n UK resden companes were ax-exemp. The level of OR durng perod Irsh under hs analyss however prevens a clear-cu concluson. The mehodologcal ssues n usng he raw OR are however as dscussed above. We move on below o regresson analyss as used o denfy an mpled OR from slope coeffcens Regresson analyss The resuls of he OLS esmaons along wh dagnosc sascs are summarsed n Tables 5 and 6. The resduals of he unscaled models (Table 5) and scaled models (Table 6) exhb hgh levels of nonnormaly. Non-normaly n he dsrbuon of he resduals can be as a resul of non-normaly n he dependen varable (Rawlngs 988). In our daa he man form of deparure from non-normaly s kuross specfcally he resduals are lepyokurc or peaked. 3 Whls can be argued ha wh a suffcenly large sample sze hs should no be a cause for concern we assess he sensvy of he repored resuls by employng a ransformaon of he dependen varable. In order o broaden he dsrbuon we employ a cube roo ransformaon of he dependen varable (Rawlngs 988). Though he Jarque-Bera sasc s sll sgnfcan n wo of he models he ransformaon has he desred effec of sgnfcanly reducng he level of he dagnosc sasc n all four models. 4 The repored Breusch-agan es sascs ndcae sgnfcan deparures from he assumpon of homoscedasc varances n boh he UK unscaled and scaled models consequenly he repored - sascs ncorporae he Whe (980) adjusmen. The sensvy of he models o he presences of nfluenal observaons has been assessed usng he EFITS sasc (Belsley Kuh and Welsch 980) whch suggess ha he repored resuls are no sensve o he ncluson of nfluenal observaons. 5 The resuls of esmaon of he unscaled Irsh and UK models are summarsed n Table 5. The coeffcen a s sgnfcanly greaer han 0 and s no sgnfcanly less han one. Ths coeffcen represens he mpled OR n respec of UK resden companes dvdends n he model s monh he dvdend was pad n he subsequen year. The same monh was used for he paymen of boh dvdends. 3 For he sample of UK frms kuross sascs of and occurred n he unscaled and scaled measures of he dependen varable respecvely. The ssue was less problemac n he Irsh sample where he correspondng fgures were and respecvely. 4 The wo sgnfcan es sascs occur: () n he unscaled Irsh model (6.68 compared wh n he absence of he ransformaon); and () n he scaled UK model (34.76 compared wh >77363 n he absence of he ransformaon). Snce each of he four ransformed models s qualavely dencal o s correspondng unransformed model n he neress of economy we repor only he unransformed models. Full resuls are avalable for he auhors. 5 Observaons wh a EFITS sascs n excess of he crcal level were excluded and he model reesmaed. In each of he four models he resuls are qualavely dencal o hose repored n Tables 5 and 6 and are no herefore repored separaely. Full resuls are avalable from he auhors. The crcal value s calculaed as 2 p n where p s he number of ndependen varables plus one and n s he number of observaons (Belsley e al. 980). 2

23 base case.e. he perod followng removal of enlemen by TEIC nvesors o ax cred repaymen. Coeffcen.200 (beng he sum of coeffcens and a s posve and sgnfcan leadng o an mpled OR of ) n respec of UK resden companes dvdend pror o he removal of he dvdend ax cred. Coeffcen IRISH n hs esmaed model s posve bu nsgnfcan. Alhough he sgn of he nercep s negave as hypohessed s no sascally sgnfcan a acceped levels. The mpled OR a.200 greaer han he.069 repored by Bell and Jenknson for he large companes n her sample s sll below he heorecal fgure of.25 whch would be expeced absen ransacons coss durng a perod when he margnal nvesor was able o reclam he ax cred and ha rae of ax cred on gross dvdend was 20% - as was he case durng perod Boh. 6 Overall herefore we deduce suppor for Hypohess 5.e. ha TEIC nvesors were he margnal nvesors n UK resden socks quoed on he LSE pror o he removal of he ax cred for such nvesmen on he 2 nd July 997 The resuls of he unscaled Irsh model can be summarsed as follows: he coeffcen s sgnfcanly greaer han zero bu no sgnfcanly less han one and mples a OR n respec of Irsh resden companes dvdends of followng removal of enlemen by TEIC nvesors o ax cred repaymen. Coeffcens and IRISH are boh negave bu nsgnfcan. Coeffcen IRISH 2 a s posve and sgnfcan provdng furher suppor of Hypohess 4 n confrmaon of our resuls from analyss of basc sascs above; and lendng no suppor o alernave Hypoheses 2 or 3 each of whch requres a leas one of and IRISH o be posve and sgnfcan. The esmaed nercep for he unscaled Irsh model s of he expeced (negave) sgn bu s nsgnfcan. As repored n Table 6 esmaon of he scaled UK model confrms he qualave resuls from esmaon of he unscaled UK model (above): he coeffcen on ne dvdend a s boh sgnfcanly greaer han zero and sgnfcanly less han one and s and s close o he OR of repored by Bell and Jenknson for he larges 250 companes n her sample. Coeffcen s posve and sgnfcan. In hs model he mpled OR s (agan beng he sum of coeffcens and ) n respec of UK resden companes dvdend pror o he removal of he a 6 Re-esmaon of he unscaled UK model based on unadjused (.e. no marke adjused) prces gave qualavely smlar resuls. The only dfference was coeffcen IRISH beng sgnfcan a he % level (cf. no sgnfcan above). Re-esmaon of he unscaled UK model o allow for cluserng also gave qualavely smlar resuls. The only dfferences was coeffcen beng sgnfcan a he 5% level (cf. no sgnfcan above). IRISH 22

24 dvdend ax cred. Coeffcen IRISH s no sascally sgnfcan a normally acceped levels; and he nercep s of he hypohessed (negave) sgn and sascally sgnfcan mplyng he presence of marke mcrosrucure effecs on ex-dvdend prces. 7 The scaled Irsh model provdes no explanaory power (havng an adjused R 2 sasc of zero) however he lack of power s sensve o he excluson of nfluenal observaons Tradng volume Mchaely and Vla (995) argue ha radng volume s posvely relaed o ax-nduced heerogeney n sock ownershp. ror o he removal of he ax cred here was a hgher degree of ax heerogeney as dvdends were of a subsanally dfferen value o ax-exemp nvesors han o ohers due o he ax cred. Therefore we would expec a hgher radng volume pror o he removal of he ax cred. If TEIC nvesors n order o capure a ax cred were emporarly purchasng Irsh resden socks cum-dvdend and hen sellng he socks ex-dvdend durng he perod 2 nd July 997 o 3 s ecember 998 (nclusve) we would expec o observe a hgh radng volume n such socks over hs perod. Alernavely f such dvdend capure acvy were resrced o he perod when he dvdend ax cred assocaed wh Irsh resden socks dvdends was 2% we would expec o observe a hgh radng volume n he perod 2 nd July 997 o 2 nd ecember 997 (nclusve) only. urng hese perods here was no only a hgh degree of ax heerogeney among nvesors bu also a hgh degree of ax heerogeney among socks. Hypoheses 6 and 7 (above) peran. Smlarly n respec of UK resden socks Hypohess 8 (above) predcs a hgher radng volume before removal of he rgh of TEIC nvesors o re-clam dvdend ax cred. For Irsh resden socks Table 7 anel A suggess ha conrary o expecaons alhough posve abnormal radng volume occurred durng each of he perods was hghes durng perod Neher and lowes durng perod Irsh2. I was expeced ha he hghes abnormal radng volume would occur pror o he removal of he ax cred (when here was a hgher degree of ax-heerogeney) whereas he hghes abnormal radng volume acually occurred afer he removal of he ax cred. Foregong dscusson of he ne value of he dfferenal ax reamen subsumng ransacons and porfolo adjusmen coss perans. L (2002) also repors resuls for he US whch conradc expecaons snce hghes excess radng volume occurred when ax heerogeney was a s lowes. Also n lne 7 Re-esmaon of he scaled UK model based on unadjused (.e. no marke adjused) prces gave qualavely smlar resuls. The only dfferences were: () coeffcen was no sgnfcan (cf. 5% above); and () coeffcen IRISH was sgnfcan a he 5% level (cf. no sgnfcan above). Re-esmaon of he scaled UK model o allow for cluserng by usng porfolos of socks grouped on he bass of her ex-dvdend day also gave qualavely smlar resuls. The only dfference was he nercep no beng sgnfcan (cf. sgnfcan above). 8 The scaled Irsh model was margnally sgnfcan when esmaed afer excludng nfluenal observaons (refer Foonoe 5 above) n parcular he coeffcen on he Irsh2 varable was sgnfcan a he 2.5% level. 23