NBER WORKING PAPER SERIES CAPITAL GAINS TAXES AND ASSET PRICES: CAPITALIZATION OR LOCK-IN?

Transcription

1 NBER WORKING PAPER SERIES APITAL GAINS TAXES AND ASSET PRIES: APITALIZATION OR LOK-IN? Zhonlan Da Edward Maydew Doulas A. Shackelford Harold H. Zhan Workn Paper NATIONAL BUREAU OF EONOMI RESEARH 1050 Massachusetts Avenue ambrde, MA June 2006 We thank Ashq Al, Robert Keschnck, Suresh Radhakrshnan, Scott Wesbenner, Yexao Xu, the assocate edtor, an anonymous referee, and semnar partcpants at the 2006 NBER Behavoral Response to Taxaton/Publc Economcs Proram Meetn, the 2006 UN Tax Symposum, and the Unversty of Texas at Dallas for helpful comments. Zhonlan Da s at the School of Manaement, Unversty of Texas at Dallas, Rchardson, TX 75083, zda@utdallas.edu, Edward Maydew s at the Kenan-Flaler Busness School, Unversty of North arolna, hapel Hll, N 27599, Edward_Maydew@unc.edu, Doulas A. Shackelford s at the Kenan-Flaler Busness School, Unversty of North arolna, hapel Hll, N 27599, Doulas_Shackelford@kenan-flaler.unc.edu, Harold H. Zhan s at the School of Manaement, Unversty of Texas at Dallas, Rchardson, TX 75083, harold.zhan@utdallas.edu. All errors are our own. The vews expressed heren are those of the author(s) and do not necessarly reflect the vews of the Natonal Bureau of Economc Research by Zhonlan Da, Edward Maydew, Doulas A. Shackelford and Harold H. Zhan. All rhts reserved. Short sectons of text, not to exceed two pararaphs, may be quoted wthout explct permsson provded that full credt, ncludn notce, s ven to the source.

2 aptal Gans Taxes and Asset Prces: aptalzaton or Lock-In? Zhonlan Da, Edward Maydew, Doulas A. Shackelford and Harold H. Zhan NBER Workn Paper No June 2006 JEL No. H2, G1, D4, M4 ABSTRAT Ths paper examnes the mpact on asset prces from a reducton n the lon-term captal ans tax rate usn an equlbrum approach that consders both demand and supply responses. We demonstrate that the equlbrum mpact of captal ans taxes reflects both the captalzaton effect (.e., captal ans taxes decrease demand) and the lock-n effect (.e., captal ans taxes decrease supply). Dependn on tme perods and stock characterstcs, ether effect may domnate. Usn the Taxpayer Relef Act of 1997 as our event, we fnd evdence supportn a domnant captalzaton effect n the week follown news that sharply ncreased the probablty of a reducton n the captal ans tax rate and a domnant lock-n effect n the week after the rate reducton became effectve. Nondvdend payn stocks (whose shareholders only face captal ans taxes) experence hher averae returns durn the week the captalzaton effect domnates and stocks wth lare embedded captal ans and hh tax senstve nvestor ownershp exhbt lower averae returns durn the week the lock-n effect domnates. We also fnd that the tax cut ncreases the tradn volume durn the week mmedately before and after the tax cut becomes effectve and n stocks wth lare embedded captal ans and hh tax senstve ownershp durn the domnant lock-n week. Zhonlan Da School of Manaement Unversty of Texas at Dallas Rchardson, TX zda@utdallas.edu Edward Maydew Kenan-Flaler Busness School Unversty of North arolna hapel Hll, N edward_maydew@unc.edu Doulas A. Shackelford Kenan-Flaler Busness School Unversty of North arolna hapel Hll, N and NBER dou_shack@unc.edu Harold H. Zhan School of Manaement Unversty of Texas at Dallas Rchardson, TX harold.zhan@utdallas.edu

3 aptal Gans Taxes and Asset Prces: aptalzaton or Lock-n? I. Introducton Ths paper jontly tests two effects of captal ans taxaton on equty tradn: a demand-sde captalzaton effect and a supply-sde lock-n effect. Prevous studes have tested these effects separately, but, to our knowlede, ths s the frst study to evaluate them jontly and emprcally document the relatve domnance of each effect surroundn an event of a tax rate chane. Employn an equlbrum approach, we show that n eneral ther net tax effect on asset prces s ambuous. Evaluatn returns and tradn volume around the 1997 reducton n the captal ans tax rate, we fnd evdence of the captalzaton and the lock-n effects jontly affectn tradn. In partcular, the captalzaton effect domnates the lock-n effect n the week follown an ncrease n the probablty of a reducton n the captal ans tax rate, as buyers respond to nformaton that future captal ans tax rates wll be lower. The lock-n effect, on the other hand, domnates the captalzaton effect after the rate reducton actually became effectve. Taxaton s one of the most prevalent market frctons n fnancal markets. It affects nvestors decsons and dstorts the valuaton of assets. aptal ans taxes, n partcular, play an mportant role n determnn an nvestor s tradn stratees and ultmately can affect asset prces. Because nvestors endoenously respond to the mposton of captal ans taxes, the tax effect on asset prces can be complcated and dffcult to measure. In hs revew of taxes n the fnance lterature, Graham (2003) concludes that Thouh ntrun n theory, the professon has made only modest proress n documentn whether nvestor taxes affect asset prces we need more evdence about the mportance of personal taxes affectn asset prces To date, research on the effects of nvestor level captal ans taxes on asset prces has produced conflctn results. Several studes report that the presence of captal ans tax reduces stock prce and current stock return (see Guenther and Wllenbor (1999), Lan and Shackelford (2000), Ayers, Lefanowcz, and Robnson, (2003), amon others), whle other studes document that mposn captal ans tax ncreases stock prce and current stock return (see Feldsten, Slemrod, and Ytzhak, (1980), Landsman and Shackelford (1995), Reese (1998), Poterba and Wesbenner (2001), Klen (2001), Bloun, Raedy, and 2

4 Shackelford (2003), Jn (2006), Ells, L, and Robnson (2006), amon others). The former s referred to as the captalzaton effect of taxes and s often justfed by the arument that nvestors would demand a lower prce to buy the assets f they have to pay captal ans taxes n the future. The latter s referred to as the lock-n effect and s attrbuted to nvestors requrn hher prces to sell assets f they have to pay taxes on selln them. Reconzn that the two effects work n opposte drectons, the purpose of ths paper s to understand the nteracton of the two effects and the crcumstances under whch one effect domnates the other surroundn a tax rate chane. Theoretcal studes on taxes and asset prcn have been scarce and often focus on tradn stratees for nvestors to avod payn captal ans taxes and ther mpact on asset prces when nvestors face embedded captal ans on ther asset holdns. For example, onstantndes (1983) shows that nvestors can rebalance ther portfolos wthout trern captal ans taxes f they are allowed to sell short assets n whch they have embedded ans. Ths allows nvestors to separate ther optmal lqudaton of assets from ther optmal consumpton and nvestment polces. Klen (1999) ntroduces a eneral equlbrum model of asset prcn wth captal ans taxes when nvestors face short sale constrants so that they cannot rebalance ther portfolo wthout trern captal ans taxes lablty. He makes predctons about the effects of captal ans taxes on asset prces wthout explctly solvn for the equlbrum prce. Vard (2000) analyzes the dynamc asset prcn effects and ncdence of realzaton-based captal ans taxes. Under the assumpton of small realzaton taxes, he derves the frst-order condtons for equlbrum asset prces. To obtan the frst-order effects, he lnearzes the frst-order condtons around the no-tax equlbrum. He fnds that asset prces are ncreased by the current realzaton tax, to partly offset the sale dsncentve assocated wth the tax, consstent wth the lock-n effect. Shackelford and Verreccha (2002) develop a tradn model where the lon-term and short-term captal ans tax rates dfferental creates a trade-off between optmal rsk-sharn and optmal tax-related tradn stratey. They show that sellers are reluctant to sell apprecated assets sooner because they are subject to hher captal ans taxes. To entce sellers, buyers must provde compensaton n the form of hher sales prces. 3

5 In ths paper, we analyze the effects of captal ans taxaton on prces, whle jontly consdern the captalzaton effect and the lock-n effect. Intutvely, the captalzaton arument approaches the tax effect from buyers perspectve (demand sde), whle the lock-n effect vews the tax mpact from sellers perspectve (supply sde). A more complete analyss of captal ans tax effects must smultaneously allow for demand and supply to nteract. In equlbrum, the net effect on stock markets of the captal ans tax wll be the combnaton of both effects. Our study provdes such a unfed framework and offers predctons for the captal ans tax effect on securty markets. Our analyss suests that a chane n captal ans taxes nfluences asset prces by shftn both the demand for assets and the supply of assets. Specfcally, when the captal ans tax s ncreased, the demand curve for assets s shfted down, reflectn the declne n prces requred to attract buyers. An ncrease n the captal ans tax also shfts the supply curve up, reflectn the boost n prces requred to entce current owners to sell. The equlbrum net tax effect on asset prces s ambuous, dependn on whch effect domnates. An ncrease n captal ans taxes unambuously reduces the float of assets (number of shares actvely traded). In the event of a captal ans tax cut, the demand curve for the assets shfts up and the supply curve shfts down. The equlbrum net tax effect on asset prce s stll ambuous, but the float of assets s unambuously ncreased. To detal the predctons of our analyss, suppose the captal ans tax rate s reduced. If the captalzaton effect domnates the lock-n effect, stock prces wll ncrease leadn to hher current stock returns. onversely, f the lock-n effect domnates the captalzaton effect, we predct that stock prces wll decrease and lower current stock returns. These effects are lkely to apply to all stocks traded on the stock market and consttute the market wde captal ans tax effect on stock prces. Furthermore, the captal ans tax effect wll vary dependn upon the characterstcs of stocks because of ther dfferences n tax costs. For nstance, rowth stocks (.e., stock whose valuaton depends larely on future dvdend rowth) are more lkely to face captal ans taxes than ncome stocks (.e., those stocks currently dstrbutn dvdends). onsequently, n the event of a captal ans tax cut, rowth 4

6 stocks should experence even hher returns than ncome stocks when the captalzaton effect domnates the lock-n effect. For stocks wth lare prce apprecaton and a hh percentae of tax senstve nvestor ownershp (such as ndvdual nvestors and mutual funds), a captal ans tax cut wll reduce nvestors tax cost of selln these stocks for portfolo rebalancn when the lock-n effect wll domnate, leadn to even lower current returns on these stocks. These consttute the cross-sectonal effect of a captal ans tax chane on asset prces. Althouh the captalzaton effect and the lock-n effect co-exst, the relatve mportance of the two effects should vary around the tmn of a captal ans tax rate chane. Specfcally, n the event of a captal ans tax cut, the captalzaton effect (prce ncrease caused by demand shft upward) wll be stroner than the lock-n effect before the tax cut becomes effectve and the lock-n effect (prce decrease caused by supply shft downward) wll domnate the captalzaton effect after the tax rate cut effectve date. The reason for the tmn dfference s that nvestors react to chanes n the probablty of a captal ans tax rate cut before the rates actually fall. In other words, buyers ncrease ther demand for stocks n response to the news of future tax cut. onversely, because captal ans are taxed upon realzaton, tax senstve stockholders lkely wll refran from selln shares wth embedded ans untl the captal ans tax rate cut becomes effectve. onsequently, we select dfferent event wndows for a domnant captalzaton effect and a domnant lock-n effect n our emprcal nvestaton. Dfferent event wndows are crtcal for dentfyn the relatve domnance of captalzaton and lock-n. We perform the emprcal tests of these predctons by examnn return and volume responses to the 1997 captal ans tax cut on stocks ncluded n the RSP dataset for the perods between January 1, 1995 and December 31, Our emprcal analyss confrms that whle both the captalzaton and the lock-n effects jontly nfluence asset prces, the mantude of each effect dffers across the tmn of the tax cut and stocks wth dfferent characterstcs. The 1997 captal ans tax rate reducton provdes a rare opportunty to jontly nvestate the effects of captalzaton and lock-n on asset prces. In late Aprl, 1997, nformaton leaked that the Democratc Whte House and the Republcan onressonal leadershp had reached an accord to reduce the captal ans tax rate. Ths news preceded 5

7 the actual effectve tax rate by about one week. Durn that nterm week, we fnd that the captalzaton effect domnated the lock-n effect. Ths s consstent wth ndvduals, the only shareholders beneftn from reduced rates, buyn shares (throuh both personal accounts and mutual funds) as the probablty of lower captal ans tax rates when they sell n the future ncreased. onversely, we fnd that the lock-n effect domnated the captalzaton effect durn the week follown the effectve date of the tax cut. Ths s consstent wth the tax senstve ndvdual nvestors selln stocks (throuh both personal accounts and mutual funds) wth lare embedded ans after the tax cut became effectve. Althouh consstent wth our predcton, broad market movements surroundn the effectve date may reflect other factors movn the markets durn those two weeks. Our cross-sectonal analyses, however, do provde compelln evdence about the effects of captalzaton and lock-n. Specfcally, we fnd that: Non-dvdend payn stocks experenced a stroner captalzaton effect than dvdend-payn stocks durn the week the captalzaton effect domnated. Stocks wth lare prce apprecaton n the past and hh ndvdual percentae ownershp experenced stroner lock-n effect and earned lower mmedate returns durn the week the lock-n effect domnated. Tradn volume was reater durn the week mmedately before and after the tax cut becomes effectve Tradn volume was reater for stocks wth lare embedded captal ans and a hh percentae of ndvdual and mutual fund ownershp durn the week the lock-n effect domnates. Snce constructn alternatve explanatons for these cross-sectonal fndns s dffcult, we nfer from these results that captalzaton and lock-n effects jontly affect market returns n the predcted manner. 6

8 The paper s oranzed as follows. Secton 2 dscusses the effect of captal ans taxes on stock prces and ts emprcal mplcatons usn a smple demand and supply framework. Secton 3 lays out the emprcal methodoloy and secton 4 provdes emprcal analyss and dscussons. Fnally, secton 5 concludes. II. aptal ans taxes and asset prces n an equlbrum framework onsder an economy n whch tax senstve nvestors are requred to pay taxes on apprecaton n stock value upon selln. The overall tax effect on stock prces wll be affected by both stock buyers and sellers. In the presence of the captal ans taxes, stock buyers wll requre a lower prce to acqure the stock to compensate them for ther future tax lablty (the captalzaton effect); whle stock sellers wll requre a hher prce to sell the stock to recover ther current tax cost (the lock-n effect). The former wll shft the demand for the stock and the latter wll move the supply for the stock at all prce levels. To demonstrate the effect of captal ans tax on stock prce, we use an equlbrum approach based on the demand and supply framework. We assume that the demand curve for the stock s downward-slopn so that nvestors are wlln to buy more shares of the stock at lower prces and fewer shares at hher prces, and on the other hand, the supply curve s upward-slopn so that nvestors are wlln to sell more shares at hher prces and fewer shares at lower prces, both before and after the captal ans taxes. Fure 1 llustrates the effect of a captal ans tax chane on stock prce and the nteracton between the two opposn forces: captalzaton and lock-n. To facltate the dscusson n our emprcal analyss, we examne the effect on stock prce of a captal 0 ans tax cut. At frst, suppose that the ntal captal ans tax rate s τ. The demand and the supply for any partcular stock are depcted as D and S n the raph and the ntersecton determnes the equlbrum prce P 0 and float of shares Q 0. Now, we 0 1 ntroduce a captal ans tax cut from τ to τ and τ <. The demand curve shfts to 1 0 τ the rht from D to D due to ncrease n demand assocated wth the captalzaton effect. At the same tme the supply curve also shfts to the rht from S to S due to ncrease n supply assocated wth the lock-n effect. In equlbrum, the new demand and supply 7

9 curves ntersect wth each other at new equlbrum prce P 1 and new float of shares Q 1. It s obvous that the new prce could be hher or lower dependn upon whch effect domnates. However, the float of shares s clearly ncreased. In the event of a captal ans tax ncrease, the shft n demand and supply s reversed. onsequently, the float of shares s unambuously decreased. However, the chane n equlbrum prce remans ambuous dependn on whch effect domnates: the captalzaton or the lock-n. In the appendx, we formalze the demand and supply analyss and analytcally demonstrate the effect of captal ans taxes on stock prce to be ambuous dependn upon the relatve mantude of the captalzaton to that of the lock-n. Our analyss above has the follown emprcal mplcatons. Frst, when the captalzaton effect domnates the lock-n effect, a reducton n the captal ans tax wll cause an ncrease n the stock prce (hher current stock returns). Ths wll arse when buyers are more responsve to an mmnent captal ans tax cut than are current sellers. onversely, when the lock-n effect domnates the captalzaton effect, a reducton n the captal an tax rate wll cause a decrease n the stock prce (lower current stock returns). Ths wll happen f current sellers are more responsve to the captal ans tax cut than are buyers. Second, the float of shares s nversely related to the captal ans taxes. When the captal ans tax s reduced, both the captalzaton and the lock-n effects renforce each other to ncrease the number of shares actvely traded. The above mplcatons apply to all stocks wth embedded captal ans and thus represent market wde reactons to captal ans tax rate chane. Because of dfferent effects of captal ans taxes on stock buyers and sellers, stocks wth dfferent characterstcs wll also be affected dfferently n the event of a captal ans tax chane. Growth stocks are expected to offer larer future prce apprecaton than ncome stocks. A captal ans tax cut wll reduce the buyer s future tax lablty and attract more demand. These stocks wll experence a reater prce ncrease and hher returns than ncome stocks n the event of a captal ans tax cut. In eneral, dvdend-payn stocks are more lkely to be ncome stocks whle non-dvdend payn stocks are more lkely to be rowth stocks. Ths means that for the captalzaton effect the stock returns are lkely hher for non-dvdend payn frms than dvdend payn frms. From a stock seller s perspectve, tax senstve nvestors holdn stocks wth lare 8

10 lon-term prce apprecaton wll have lower current tax costs upon selln when the captal ans tax s reduced. Thus, tax senstve nvestors wll be more nclned to sell stocks wth lare embedded captal ans to rebalance ther portfolo. Ths mples that n the event of a captal ans tax cut, stocks wth lare embedded captal ans wll experence a larer prce declne than wll other stocks. These mplcatons pertan to ndvdual stock characterstcs. We thus call them cross-sectonal effects of a captal ans tax rate chane. In addton, nvestors antcpatn the captal ans tax cut may wthhold selln shares wth embedded ans before the tax cut becomes effectve (seller s strke). In ths case, the supply curve may reman unchaned or even move up from S to S. The demand and supply curve may ntersect at pont or D. Ths wll lead to a temporary ncrease n both the stock prce and float of shares. Ths may provde an alternatve explanaton to a domnatn captalzaton effect before the captal ans tax cut becomes effectve. In the next secton, we emprcally test both the market wde and crosssectonal effects of a captal ans tax chane by jontly consdern the captalzaton and the lock-n effects. III. Emprcal Methodoloy To emprcally test the above mplcatons, we use the Taxpayer Relef Act of 1997 (TRA97) captal ans tax reducton as our event. The TRA97 lowered the top tax rate on captal ans from 28 percent to 20 percent for assets held more than 18 months. TRA97 s partcularly attractve as an event because the tax cut was both lare and relatvely unexpected. Often tax leslaton follows a protracted process wth radual chanes n the probablty of a partcular bll becomn a law. In TRA97, however, onress provded researchers wth an attractve research settn by comn to rapd areement on a lare, unexpected reducton n captal ans tax rates. Havn a well-defned event s partcularly mportant n ths study because we need to defne separate event wndows for two opposn effects. The key to jontly dentfy the captalzaton effect and the lock-n effect s to understand that stock buyers 9

11 and sellers perceve the expected captal ans tax cut dfferently as we dscussed above. 1 They dffer not only from requred rate of return or valuaton, but also from when they react to the news/event. A buyer, n order to capture the expected tax cut beneft, wll react to the captal ans tax cut nformaton before the tax cut becomes effectve. On the other hand, a seller who s subject to captal ans taxes wll more lkely sell shares wth embedded captal ans to rebalance hs portfolo after the tax cut becomes effectve. As a result, the captalzaton effect s more lkely to domnate before the tax cut announcement and the lock-n effect s more lkely to domnate after the tax cut becomes effectve. For the Taxpayer Relef Act of 1997 (TRA97), lttle nformaton was released untl Wednesday Aprl 30, 1997, when the onressonal Budet Offce (BO) surprsnly announced that the estmate of 1997 defct had been reduced by $45 bllon. Two days later on May 2, the Presdent and onressonal leaders announced an areement to balance the budet by 2002 and, amon other thns, reduce the captal ans tax rate. These announcements reatly ncreased the probablty of a captal ans tax cut. 2 On Wednesday May 7, 1997, Senate Fnance harman Wllam Roth and House Ways and Means harman Wllam Archer jontly announced that the effectve date on any reducton n the captal ans tax rate would be May 7, Gven the above backround, we choose Wednesday, Aprl 30 to Tuesday, May 6, 1997 as the week that we expect the captalzaton effect ( WK ) to domnate as demand ncreases n reacton to the ncreased lkelhood of a captal ans taxes cut. The same event wndow s used n Bloun, Hal, and Yetman (2005). Lan and Shackelford (2000) use a smlar event wndow (Aprl 29 to May 5, 1997). We then choose Wednesday May 7 to Tuesday May 13, 1997 as the week that we expect the lock-n effect ( WK L ) to 1 If an nvestor has to rebalance hs portfolo by addn postons n some stocks and reducn postons n other stocks around the captal ans tax cut perod, he s more lkely to add the postons when the tax rate cut announcement s mmnent, and delay hs selln untl the tax rate cut s effectve f he has embedded captal ans on the postons. 2 There were some conflctn snals on the captal ans tax cut before Aprl 1997 as detaled n Sna and Gyourko (2004). Based on artcles n the New York Tmes and Wall Street Journal they fnd that Presdent lnton had already sad that he mht be wlln to cut captal ans taxes to reach a budet compromse n the frst week of February, thouh at that pont the Admnstraton s preference s for a captal ans tax ncrease. However, on February 13 and February 23 there were addtonal news reports suestn that the Whte House would accept a tax cut. We vew the surprsn announcement by the BO as the stronest ndcaton for an mmnent captal ans tax announcement. 10

12 domnate as current shareholders sell ther apprecated stocks to rebalance ther portfolos. Another mportant factor s that the captal ans rate reducton only appled to ncome that s reported on personal tax returns,.e., captal ans from the selln of shares held drectly by ndvduals or held ndrectly by ndvduals n flow-throuh enttes, such as mutual funds, partnershps, trusts, S corporatons, or lmted lablty corporatons that pass dvdend ncome to nvestors personal tax returns. aptal ans taxes are not leved on tax-deferred accounts (e.., qualfed retrement plans, ncludn pensons, IRAs and 401(k)), tax-exempt oranzatons, and foreners. orporatons pay captal ans taxes; however, the rate reducton n TRA97 dd not apply to corporatons. Thus, the ensun tests predct varaton n returns based on the amount of holdns by ndvdual nvestors and mutual funds. To capture the roup of nvestors who are the most senstve to the captal ans taxes, we construct proxes for the percentae of tax senstve nvestor ownershp of a stock (ndvdual nvestors and/or mutual funds) usn data on shares outstandn and shares owned by nsttutonal nvestors. The data on the nsttutonal nvestors ownershp are obtaned from ther quarterly flns wth the U.S. Securtes and Exchane ommsson (known as Form 13F). Let R be frm s stock return at tme t. To test the effect of a captal ans tax rate t cut on stock prces, we formulate the basc emprcal reresson equatons as follows: for frms wth postve embedded ans R t = α + β WK β WK + β WK * Gans + γ ontrols + ε, 2 t L ( + β WK 3 * TSO ( Dv + β Beta _ market 6 ( + β WK 4 L Gans * TSO + β Beta _ momentum 7 ( ( (1) for all frms R t = α + β WK + β WK 5 + β WK 8 1 * Gans Gans + γ ontrols + ε + β WK 2 t L ( ( + β WK * TSO * L 3 ( ( Dv * TSO ( ( + β WK + β Beta _ market 6 + β WK 9 4 L L Gans Gans ( ( * L *(1 L ( ( + β Beta _ momentum 7 * TSO ) * TSO ( ( (2) 11

13 where WK represents the dummy varable for the week when we expect the captalzaton effect to domnate (hereafter, we call WK the captalzaton week for brevty), WK L represents the dummy varable for the week when we expect the lock-n effect to domnate (hereafter we call WK L the lock-n week for brevty), Dv( s the dummy varable whch takes the value of 1 f there was no dvdend dstrbuton n the pror year and 0 otherwse, Gans( s the embedded captal ans pror to tme t and n our baselne case s measured as the past fve year stock prce apprecaton, TSO ( 1) t s the percentae of shares of stock owned by tax senstve nvestors at tme (t-1) and s measured ether as the sum of ndvdual nvestor and mutual fund ownershp n the most recent past quarter or ndvdual nvestor ownershp alone, L( s a dummy varable that takes the value of 1 f Gans 0) and 0 otherwse, Beta_market s stock s ( ( < market return beta, Beta_momentum ontrols refer to all other varables that may affect stock returns. s stock s beta on the momentum factor, and Our specfcatons above consder both the broad stock market reactons to the captal ans tax cut and the cross-sectonal dfferences n the tax effect for dfferent stocks wth dverse characterstcs. Intutvely, a captal an tax cut wll ncrease the demand for stock. Thus, n the event of a captal ans tax cut, the coeffcent of the captalzaton week dummy ( WK ) wll be postve ( β 0 ). Smlarly, for exstn tax senstve shareholders contemplatn selln shares wth embedded captal ans, a tax cut wll nduce them to sell the stock to rebalance. When the lock-n effect domnates the captalzaton effect, a tax rate cut wll lead to lower returns on stocks wth embedded captal ans resultn n a neatve sn for the lock-n week dummy varable ( β 2 < 0 ). These consttute the broad market effect on stock prces of a captal ans tax cut. Frms dffer n ther dvdend polcy and rowth potental (and consequently ther future captal ans tax lablty) and the sze of embedded captal ans and current tax costs upon selln. Thus, the mantude of reacton to the captal ans tax cut wll lkely vary wth the characterstcs of these frms. Our analyss n the prevous secton suests that n the event of a captal ans tax cut, the mpact from demand sde on stock return wll be larer for rowth stocks than for ncome stocks. Ths s captured by a postve 1 > 12

14 nteracton term (WK *Dv (t-1) ), ndcatn a larer prce ncrease for rowth stocks than ncome stocks durn the captalzaton week ( β 3 > 0 ). On the other hand, for a current shareholder who faces a captal ans tax lablty, a captal ans tax reducton offers a stron ncentve for hm to sell shares wth lare embedded captal ans to rebalance, consequently leadn to lare downward prce pressure assocated wth the lock-n effect on the stock returns. Therefore durn the week the lock-n effect domnates the captalzaton effect, stocks wth larer embedded captal ans and a hher percentae of tax senstve nvestor ownershp wll experence lower stock returns. In our specfcatons above, ths s captured by a neatve ( β 4 < 0 ) nteracton (WK L *Gans (t-1) *TSO (t-1) ) for frms wth postve ans and nteracton (WK L *Gans (t-1) *TSO (t-1) *(1-L (t-1) )) for the specfcaton on all frms. As dscussed n Secton II, an alternatve explanaton to a domnatn captalzaton effect n the week before the captal ans tax cut s that tax senstve nvestors may wthhold selln stocks wth lare embedded ans, leadn to a prce runup. Ths s referred to as the seller s strke. To emprcally test f a prce run-up n the week before the captal ans tax cut announcement s caused by a domnatn captalzaton effect or seller s strke, we nclude n our specfcatons the nteracton (WK *Gans (t-1) *TSO (t-1) ). If the seller s strke causes the prce run-up, we would expect a postve coeffcent for the nteracton. Otherwse, the prce ncrease s lkely to be the result of a domnatn captalzaton effect. In addton, we also nclude measures of systematc rsk of ndvdual stock returns to the market returns and the momentum factor n our reresson analyss. Ths s motvated by the standard captal asset prcn theory and exstn emprcal evdence suestn that systematc rsks are mportant determnants of ndvdual stock returns. Gven that a tax cut s a market-wde event, the returns to the market portfolo and the momentum themselves wll lkely be affected by the event as well. onsequently, n our baselne specfcatons we nclude betas of the market returns and the momentum factor as control varables. We wll dscuss the results of usn the beta adjusted stock returns as dependent varables n our robustness check later. Our specfcaton for all frms ncludes two addtonal nteracton terms: (WK *Gans (t-1) *L (t-1) *TSO (t-1) ) and (WK L *Gans (t-1) *L (t-1) *TSO (t-1) ). These nteracton 13

15 terms focus on the responses of loss frms to the captal ans tax cut n the captalzaton week and the lock-n week. They allow us to further dentfy f tax senstve nvestors wll enae n tax effcent tradn n the event of a captal ans tax chane. Intutvely, tax senstve nvestors should more aressvely sell stocks wth lare embedded captal losses before the captal ans tax s reduced to beneft from hher tax deducton. Ths would create some downward pressure on these stocks durn the captalzaton week. onsequently, the coeffcent for (WK *Gans (t-1) *L (t-1) *TSO (t-1) ) wll be postve ndcatn that the prces of stocks wth embedded losses are lkely to decrease n the sze of losses durn the captalzaton week because Gans 0). On the other hand, ( ( < snce there s no lock-n on stocks wth embedded losses and the captalzaton effect remans durn the lock-n week, the coeffcent for (WK *Gans (t-1) *L (t-1) *TSO (t-1) ) wll lkely be neatve ndcatn ncreased demand assocated wth the captalzaton effect. To test the predcton on the effect of the captal ans tax cut on the float of shares, we need to frst provde a measure for the float of stock. Unlke the shares outstandn, the float of shares measures the number of shares actvely traded and s usually less than shares outstandn. For example, shares owned by nsders sometmes are subject to certan restrctons and cannot be quckly sold n the market thus not ncluded n the float; some lon-term buy-and-hold nvestors are also less nclned to churn ther portfolo for short-term prce fluctuaton. Ther holdns are not ncluded n the float durn normal tme. However, for a major event, such as a captal ans tax cut, an nvestor may fnd t optmal to buy addtonal stocks and/or to sell some stocks wth a lare prce apprecaton to rebalance hs portfolo. Tradn from these nvestors s lkely to temporarly ncrease tradn volume. In partcular, the ncrease n tradn volume caused by the captal ans tax cut s lkely to be concentrated n the few weeks when the tax cut s announced. We use tradn volume as a proxy for the float of shares of stocks. Let vt be stock s loarthmc weekly tradn volume. Follown Mchaely and Vla (1996) and Dhalwal and L (2006), we frst compute the excess tradn volume as the dfference between the weekly tradn volume at tme t and the averae weekly tradn volume n the most recent past three months relatve to the three month weekly averae tradn volume,.e., 14

16 as follows: t 1 1 vt vj 13 j= t 13 v t =. (3) t 1 1 vj 13 j= t 13 We then formulate our reresson equatons on the tax effect on the float of shares for frms wth postve embedded ans v t = α + β WK β WK for all frms v t + β WK + β WK L = α + β WK β WK 2 Gans Gans Gans ( ( ( + β WK 2 L L * L + β WK 3 * TSO + β WK 3 * TSO ( ( ( * TSO Dv ( + β WK + γ ontrols + ε, Dv + β WK 6 ( ( 4 Gans t + β WK 4 L ( + γ ontrols + ε L Gans Gans * L t ( ( ( * TSO * TSO * (1 L ( ( ( (4) ) * TSO where the varables are defned the same as above and the controls are dscussed n next secton. Our predcton for stock float suests that the coeffcents for WK and WK L wll be postve reflectn the market wde response n tradn volume to a captal ans tax cut. The nteracton term, ( WK Dv ( ), wll also be postve because more tax senstve nvestors should buy shares of stocks wth rowth potentals durn the week the captalzaton effect domnates. Furthermore, the nteracton term, ( WK L Gans( TSO( ) wll be postve because more tax senstve nvestors should sell ther holdns wth lare embedded captal ans to rebalance ther portfolos durn the week the lock-n effect domnates. In our dscussons of the emprcal analyss, we use tradn volume to refer to the excess tradn volume for brevty. ( (5) IV. Emprcal analyss A. Sample and Summary Statstcs We use stocks ncluded n the RSP dataset between January 1, 1990 and December 31, Follown Lan and Shackelford (2000) we focus on weekly returns. Explanatory varables nclude dvdend dummy, embedded captal ans, the percentae 15

17 of ndvdual and/or mutual fund ownershp of a stock, week dummes defned to dentfy the event perod, measure of ndvdual stock s exposure to the market return (Beta_market ), measure of ndvdual stock s exposure to the momentum factor (Beta_momentum ), and varous nteracton terms to dentfy the tax effect. where We calculate the weekly return as follows R =lo( d +1) (5) t r t d r t s the daly return and t runs from Wednesday to the follown Tuesday to be consstent wth the event wndows. The loarthmc weekly volume s smlarly calculated as where d v =lo( ) (6) t Vol t d Volt s the daly tradn volume of stock on day t and the summaton runs from Wednesday to the follown Tuesday. We use both volumes n shares traded and n dollar amount for our emprcal tests. We obtan daly stock returns and tradn volume from the daly RSP data set. Dvdend, stock prce, and shares outstandn are extracted from the monthly RSP data set. To obtan the percentae of shares of each stock owned by ndvdual nvestors, we extract nsttutonal nvestors ownershp as of March 31, 1997 from Form 13F submtted to the SE by nvestment manaement companes. 3 We then compute two measures of the tax senstve ownershp on stock at tme t (TSO t ) as follows the percentae of ndvdual nvestor ownershp (IND t ) IND t = 1 Percentae of shares owned by nsttutonal nvestors at tme t the percentae of ndvdual nvestor and mutual fund ownershp (IND&MF t ) IND&MF t = IND t + Percentae of shares owned by mutual funds at tme t. We exclude non-common shares such as preferred stocks from our analyss. Frms wth mssn observatons are also removed. For our baselne case, we follow Klen (1999, 2001) and defne the embedded captal an as the prce apprecaton n the last fve years. Specfcally, the fve year embedded captal an s calculated as the prce 3 We thank Rabh Moussaw for provdn the nsttutonal stock ownershp data. Ayers, Lefanowcz and Robnson (2003), Dhalwal and L (2006), and Dhalwal, Krull, L and Moser (2005), amon many others, also use ths measure to capture the extent to whch ndvduals hold shares n the frm. 16

18 apprecaton n the past fve years up to the most recent month pror to tme t for each stock. For nstance, the fve embedded an for 3/31/1997 s calculated as a stock s prce apprecaton from February 1992 to February For robustness check, we also use embedded captal ans measured at two years and eht years and fnd that nferences larely hold. To obtan measures of exposure of ndvdual stock returns to the market return and the momentum factor, for each stock, we estmate a multple reresson of the frm s weekly excess return on the weekly market excess return and weekly momentum factor usn data on these varables between 1/8/1992 to 4/16/ The reresson slope for the market return and the momentum factor s used for Beta_market and Beta_momentum, respectvely. For the emprcal tests, we use weekly returns n the last three years (1995, 1996 and 1997). Our control varables for the weekly return reressons nclude the dvdend dstrbuton dummy, the percentae of tax senstve nvestor ownershp, the embedded captal ans, the nteracton terms WK L *Gans (t-1), WK L *TSO (t-1), WK *Gans (t-1), WK *TSO (t-1), Gans (t-1) *TSO (t-1), TSO (t-1) *Dv (t-1), the sze of the frm measured by the loarthm of frms market captalzaton at t-1, and the calendar effect represented by month and annual dummes. For the volume reressons, we use the dvdend dstrbuton dummy, the percentae of shares owned by ndvdual nvestors and mutual funds, frm sze, the nteracton terms Gans (t-1) *TSO (t-1), Gans (t-1) *Dv (t-1), and the calendar effect as control varables. Table 1 presents the basc summary statstcs for varables used n our reresson analyss for both the subsample of frms wth postve embedded captal ans and the full sample ncludn frms wth embedded captal losses. Each varable s defned at the bottom of the table. The subsample conssts of nformaton on 2,565 frms wth postve embedded ans for the past fve years and a total of 266,252 observatons. The averae weekly return for frms wth postve embedded captal ans s 0.31 percent wth a standard devaton of 6.3 percent. In the meantme, the full sample ncludn frms wth 4 To construct the weekly momentum factor, we use the daly data on the momentum factor (Up mnus Down or UMD) obtaned from Kenneth R. French s webste. Sx value-wehted portfolos formed on sze and pror (2-12) returns are used to construct UMD. The portfolos formed monthly are the ntersectons of 2 portfolos formed on sze and 3 portfolos formed on pror (2-12) return. UMD s the averae return on the two hh pror return portfolos mnus the averae return on the two low pror return portfolos. 17

19 embedded captal losses conssts of 412,730 frm-week observatons and has a lower averae weekly return of 0.29 percent and a slhtly hher standard devaton of 6.83 percent. On averae, about 48 percent of the frms n the ans subsample and 49 percent of all frms n the full sample dd not pay dvdends (recall that Dv takes a value of one f the frm does not pay dvdends). The averae fve-year prce apprecaton for the ans subsample s 210 percent wth a standard devaton of 941 percent. The fve year prce apprecaton s much lower at 123 percent wth a standard devaton of 765 percent for the full sample. The fve-year embedded captal an s hhly skewed wth half of the frms ann less than 79 percent for the ans subsample and about 27 percent for the full sample as ndcated by the medan. The averae percentae of shares owned by ndvdual nvestors s 68 percent wth a standard devaton of 24 percent for the ans subsample. Indvdual and mutual funds toether own 78 percent of stocks on averae wth a standard devaton of slhtly less than 17 percent for the ans subsample. The mean and standard devaton for the ndvdual and/or mutual fund ownershp are of smlar mantude to those of the full sample. The averae ndvdual stock beta on the market return s 0.6 wth a standard devaton of 0.53 for and the averae beta on the momentum factor s much smaller at 0.02 wth a relatvely lare standard devaton of 0.40 for the postve ans subsample. Smlar mean and standard devaton are found for the full sample wth all frms. The tradn volume for the ans subsample has a mean of 11.7 and a standard devaton of 2.2 when measured n loarthmc shares. The statstcs are slhtly hher at 14.4 for the mean and 2.7 for the standard devaton when measured n loarthmc dollar volume. The summary statstcs for the tradn volume for the full sample ncludn both the ans and loss frms are of smlar mantude to those of the ans subsample. The excess tradn volume has an averae of wth a standard devaton of 0.72 for postve ans frms when measured n loarthmc shares and an averae of wth a standard devaton of 0.75 when measured n loarthmc dollar volume. We also report the summary statstcs for the adjusted weekly return after removn the systematc components assocated wth the market return and the momentum factor. The mean adjusted weekly return for ndvdual stocks ranes from 0.05 to 0.07 percent wth a 18

20 standard devaton rann between 6.1 to 6.7 percent for the two samples, respectvely. The adjusted weekly return s much lower on averae than the unadjusted return. We use eneralzed least squares to estmate our reresson model n order to account for correlated resduals n reressons. Specfcally, we use clustered standard error estmates, whch are shown to be unbased n reresson analyss usn panel data sets by Peterson (2005). 5 B. Return Tests for Jont Tax aptalzaton and Lock-n Effects Our frst set of reresson results s based on equaton (1) and reported n Table 2. The frst two columns of estmates and p-values are for the percentae of tax senstve ownershp (TSO) measured by ndvdual nvestors and mutual funds combned, whle the last two columns measure tax senstve ownershp usn ndvdual nvestors only. The coeffcents for key varables are qualtatve very smlar. onsstent wth our predctons, the coeffcent estmate assocated wth WK s postve and statstcally snfcant at 1 percent. Ths ndcates that the market reacted postvely to the possblty of a captal ans tax cut. The weekly return for the captalzaton week s 8.18 percent hher than the averae weekly return when the tax senstve ownershp s measured by ndvdual nvestors and mutual funds combned (IND&MF), and s 6.04 percent hher when the tax senstve ownershp s measured by ndvdual nvestors only (IND). The estmates are lower than the averae return reported n Lan and Shackelford (2000) for the week between 4/29/1997 to 5/5/1997, but reasonable ven that more controls are used n our reresson. The estmate for WK L s neatve and also statstcally snfcant at 1 percent. The estmated coeffcent suests that the averae weekly return n the lock-n week s about 2.0 percent (1.0 percent) lower than the averae weekly return usn IND&MF (IND). Ths provdes emprcal support for a market wde domnatn lock-n effect durn the week mmedately after the captal ans tax cut became effectve. onsstent wth our predctons on the cross-sectonal behavor of stock returns, the nteracton term (WK *Dv) s postve and hhly statstcally snfcant wth a p- value less than 1 percent. The estmated coeffcent shows that non-dvdend payn 5 We use SAS PRO MIXED procedure to estmate our models treatn frm as our subject so that each frm s one cluster. The oodness of ft for ths procedure s ven by the -2 resdual lo lkelhood. 19

21 stocks yeld 1.7 percent hher returns on averae durn the captalzaton week than do dvdend-payn stocks for the same perod for both measures of tax senstve ownershp. Lan and Shackelford (2000) report that non-dvdend-payn frms experence 4.25 percent hher weekly return on averae. The coeffcent estmate assocated wth the nteracton term (WK L *Gans*TSO) s neatve at percent based on IND&MF and neatve at percent based on IND. Both are hhly statstcally snfcant wth p- values less than 1 percent. Ths suests that stocks wth lare embedded captal ans and hh ndvdual nvestor ownershp have lower returns durn the lock-n week. The coeffcent mples that for frms wth the averae percentae of ndvdual and mutual fund ownershp, a one standard devaton ncrease n fve year embedded captal ans wll yeld 2.5 percent ( 0.34% 941% 77.8%) lower weekly returns durn the lock-n week. For frms wth an averae fve year embedded captal ans, a one standard devaton ncrease n the percentae of ndvdual and mutual fund ownershp leads to 12 bass ponts ( 0.34% 210% 16.8%) lower weekly returns durn the lock-n week. When the tax senstve ownershp s measured by ndvdual nvestors only, the correspondn reducton n the averae weekly returns durn the lock-n week s lower at 1.4 percent and 11 bass ponts, respectvely. The above fndns are consstent wth both nvestors buyn stocks as the probablty of a captal ans tax cut ncreases before the announcement and tax senstve nvestors selln ther shares after the effectve date of captal ans tax cut. Investors respond by ncreasn ther demand for stocks and drvn up prces durn the captalzaton week. Ths s partcularly evdent for non-dvdend payn rowth stocks, whose returns are more lkely to face captal ans taxaton. After the lower tax rate became effectve, ndvdual nvestors senstve to captal ans tax labltes were more nclned to sell postons wth lare embedded captal ans to rebalance ther portfolos. Ths leads to a lower prce for stocks wth lare embedded captal ans and a hher percentae of ndvdual stock ownershp durn the lock-n week. Our fndns suest that a domnant lock-n effect s stroner n a subset of stocks that have both lare embedded captal ans and hh tax senstve ownershp n the week after the effectve date of the tax cut. Gven that TRA97 drectly affects 20

22 ndvdual nvestors and mutual funds wth lare embedded captal ans, ths fndn s consstent wth our predcton on the relaton between stock prce and captal ans taxes. The coeffcent for the nteracton term (WK *Gans*TSO) s postve but statstcally nsnfcant for both measures of the tax senstve ownershp. The p-value for the coeffcent estmate s above 90 percent for both the IND&MF and IND. Ths suests that there s no emprcal support for the seller s strke, an alternatve explanaton of the prce run-up durn the captalzaton week. The reason s, f nvestors were wthdrawn from the market after the announcement of a rate cut but before the effectve date, we would expect share prces to ncrease durn the captalzaton week for the apprecated shares that were on to be sold after the effectve date. Snce the coeffcent on nteracton term (WK *Gans*TSO) s nsnfcant, we nfer that ncreased demand larely accounted for the ncrease n prces durn the captalzaton week, rather than a seller s strke. Our reresson analyss also provdes the follown fndns. onsstent wth the captal asset prcn theory, a frm s exposure to systematc market rsk has a postve effect on the frm s return. The estmated coeffcent for Beta_market ranes from 0.11 to 0.12 percent and statstcally snfcant at less than 1 percent level across the two reressons. Frms exposure to the momentum factor also has a postve effect on returns. The coeffcent for Beta_momentum ranes from 0.07 to 0.08 and snfcant at around 1.5 percent level for across the two reressons. Stocks wth hher tax-senstve nvestor ownershp experence about 0.2 percent hher weekly returns on averae. The nteracton between WK L and the embedded captal ans s snfcantly postve. Ths may be caused by demand from nsttutonal nvestors that dd not face a captal ans tax cut. Stocks wth hher tax senstve nvestor ownershp experence a lower averae weekly return durn the captalzaton week and a hher averae weekly return durn the lock-n week. Non-dvdend payn stocks wth hher tax senstve nvestor ownershp also yeld a lower weekly return on averae. Frm sze has a neatve and snfcant effect on stock returns. Althouh not reported n the table, we fnd that the annual dummy s hhly statstcally snfcant for year 1995 but not for year The monthly dummes on the 21

23 other hand are all statstcally snfcant ndcatn the exstence of monthly return varaton. Gven our panel data, frm characterstc varables (ans, dvdend, sze and ndvdual and/or mutual fund ownershp) also act as controls for the fxed effect n the test.. Return Tests wth all Frms For frms wthout embedded captal ans, the lock-n wll not lkely have a snfcant effect on stock returns. However, tax-senstve nvestors wth embedded captal losses may sell ther holdns before the tax cut becomes effectve so that they can beneft from the hher tax rebate under hher tax rate. 6 If so, frms wth embedded captal losses and hh tax senstve nvestor ownershp wll experence some downward prce pressure durn the week before the tax cut announcement. For these stocks, the prce run-up caused by the captalzaton effect, whch ncreases the demand from nvestors who have no exstn postons n them and nvestors who are not subject to taxaton, wll lkely be attenuated. On the other hand, because there s no lock-n effect on stocks wth embedded captal losses after the tax cut announcement, the selln pressure wll lkely be small whle the captalzaton effect remans for these stocks. To check the robustness of our fndns of jont captalzaton and lock-n effects and to test the hypotheses on frms wth embedded captal losses, we estmate equaton (2) whch ncorporates two addtonal terms: (WK *Gans*L*TSO) and (WK L *Gans*L*TSO) usn observatons on all frms. Table 3 reports the reresson results wth the two addtonal nteracton terms. Both the sns and mantude of the effects of key varables are smlar to those for frms wth postve embedded captal ans only. Specfcally, the coeffcent for the captalzaton week s 8.61 percent (6.42 percent) wth a p-value less than 1 percent when the tax senstve ownershp s measured by percentae of shares owned by ndvdual nvestors and mutual funds (ndvdual nvestors only). The coeffcent for the lock-n week s percent and percent for the two measures of tax senstve 6 omplex nettn provsons, whch are beyond the scope of ths paper, overn the taxaton of captal ans and losses (see Shackelford, 2000). That sad, our predctons about the ncentves to sell stocks wth embedded captal losses before the rate fell assume that the nvestors can utlze the captal losses. More specfcally, we assume a net captal an poston,.e., total captal ans exceed total captal losses. 22