Tax-loss Selling and the Turn-of-the-Year Effect: New Evidence from Norway 1

Transcription

1 Tax-loss Selling and the Turn-of-the-Year Effet: New Evidene from Norway 1 Qinglei Dai Universidade Nova de Lisboa July Aknowledgement: I would like to thank Kristian Rydqvist at Binghamton University, Bruno Gerard, Bernt Arne Odegaard and Tor Kildal at Norwegian Shool of Management, Tommy Stamland at the Norwegian Shool of Eonomis and Business Administration, attendants at the EFMA 2003 meeting in Helsinki, the AsianFA 2004 in Taipei, the AFBC 2004 in Sydney, faulty members at the Tilburg University and Fauldade de Eonomia Universidade Nova de Lisboa for valuable omments and suggestions. I am also grateful to Erik Fjaeli and Stein Ove ettersen at Statistis Central Bureau of Norway for providing tax information. Mailing address: Fauldade de Eonomia, Universidade Nova de Lisboa, Rua Marques de Fronteira 20, Lisboa, ortugal qinglei.dai@fe.unl.pt

2 Tax-loss Selling and the Turn-of-the-Year Effet: New Evidene from Norway July 2007 JEL No. H24, H26, G12, G18 ABSTRACT This paper re-examines the tax-loss selling explanation of the turn-of-the-year effet on the Norwegian equity returns over the 1984 to 1999 period. This sample is partiularly interesting as the Norwegian rules governing the taxation of dividend and apital gains had hanged substantially several times over the sample period. The market-adjusted stok return over the turn-of-the-year is found to be signifiantly influened by the limits on the apital loss write-offs and the length of the holding period during whih the apital loss is tax-dedutible. The evidene is strongly supportive of the tax-related explanation for the turn-of-the-year effet. 2

3 1. Introdution During the past few deades, a large number of studies have been devoted to the so-alled turnof-the-year effet, the phenomenon of abnormal exessive stok returns over the first few trading days just after the turn-of-the-year. Among the various hypotheses that were provided to explain for this phenomenon, the two most plausible ones are the tax-loss selling hypothesis 2 and the institutional investor window-dressing hypothesis 3. Using a unique dataset, this paper provides new evidene that the turn-of-the-year effet is at least partially explained by tax-loss selling of individual investors. The tax-loss selling hypothesis posits that, sine apital gains are only taxable upon realization, rational investors are motivated to systematially trade on the stok market and time the realization of apital gains and losses in order to minimize their tax liabilities. rior to the end of a fisal year (whih in many ountries oinides with the end of the alendar year), the motivation for the investors to realize a apital loss is espeially strong beause the assoiated tax shield an be reeived immediately after the year turns and is therefore least disounted. Towards the end of the year, strong selling pressure in loser stoks fores down the stok pries; right after the turn-of-the-year, the selling pressure disappears, and the stok pries rebound, resulting in the turn-of-the-year effet. On the other hand, institutional window-dressing hypothesis suggests that the strong selling pressure in loser stoks at the end of the year is mainly aused by the trading of portfolio managers of finanial institutions. These managers are about their portfolio ompositions at the end of the year, and are motivated to sell off stoks that have performed poorly in the past year so as to polish up their portfolio reports. Sine both of the two theories predit a linkage between previous stok prie depreiation and high exessive stok return over the turn-of-the-year, one problem that exists in earlier empirial studies on is that thu7ey are unable to distinguish the effets between the two explanations. For example, Dyl (1977) finds that, in Deember, the trading volume for loser stoks is abnormally large, while the trading volume for winner stoks is abnormally low. Givoly and Ovadia (1983) report that January returns are higher for stoks with 12-month-lows in Deember. These findings are onsistent with both of the two explanations. 2 See, for example, Constantinides (1984), Roll (1983), oterba and Weisbenner (2001). 3

4 Some more reent studies try to disentangle the preditions of these two theories by emphasizing their respetive underlying mehanisms: while institutional window-dressing is strongly influened by the size of institutional ownership in the stoks, the motivation for tax-loss selling is governed by tax provisions. For example, Sias and Starks (1997) examine the relationship between institutional ownership and the turn-of-the-year effet. They argue that if institutional window-dressing dominates individual loss selling, stoks with large institutional holdings should generate higher exessive returns over the turn-of-the-year. Instead, they find that stoks dominated by individual holdings exhibit a stronger January effet, whih is more supportive of the tax-loss selling hypothesis. Chen and Singal (2004) study stok returns around semi-annual losing when tax loss selling is not involved yet institutional window dressing is expeted, and find no evidene for window dressing hypothesis. Starks, Yong and Zheng (2006) doument a January effet for U.S. muniipal bond losed-end funds, whih are held mostly by tax-sensitive individual investors, providing evidene that this effet is largely explained by tax-loss selling. A strand of literature investigates the influene of tax law hanges on the turn-of-the-year effet. Shultz (1985) and Jones, Lee and Apenbrink (1991) report that the turn-of-the-year effet only ame into existene in the U.S. after the introdution of the personal inome tax in Bhabra, Dhillon and Ramirez (1999) doument a November Effet of the stok returns after the fisal year-end for the U.S. mutual funds had been moved to the end of Otober by the U.S. Tax Reform At of oterba and Weisbenner (W, 2001) examine hanges in tax provisions in the U.S. during 1960s and 1980s. They find that the differene between long-term and short-term apital loss write-offs is influential on the exessive stok return over the turn-of-the-year. This paper examines the relationship between hanges in apital gains tax odes in Norway and the stok returns over the turn-of-the-year between 1983 and It is very similar to W, sine both of the two papers fous on hanges in tax laws and investigate the linkage between tax motivations and the stok return patterns over the turn-of-the-year. What makes this study speial is that our dataset inludes muh larger variations in multiple tax odes among the tax provisions. If tax-loss selling an at least partially explain for the turn-of-the-year effet, we expet that the tax motivation related variables should have muh stronger explanation power in our study than in W. 4 Indeed, when the main finding by W is that the legal distintion between long-term and 3 Lakonishok and Smidt (1988) and Lakonishok, Shleifer, Thaler and Vishny (1991). 4 In W, the short-term-investment holding period varies between 6 months and 12 months; the short-term loss limit hanges between $1000 and $3000; and the tax value of a long-term loss as a perentage of a short-term loss varies between 50% and 100%. In our dataset, the short-term holding period runs from 2 years to infinity (where there is no differene between short-term and long-term investments); the short- 4

5 short-term losses is important in the determining how turn-of-the-year returns are assoiated with past returns, in this study we not only find the same result as W did but also disover that legal restritions on loss write-offs also play an important role in the return anomaly. An additional ontribution of this study is that we model numerial relations between tax provisions and the motivations for tax-loss selling, and derive testable preditions about the variations of the turn of the year effet under the tax-loss selling hypothesis. In the following part of the paper, Setion 2 examines the turn-of-the-year effet on the Oslo Stok Exhange and briefly disusses the window-dressing effet on the market. Setion 3 introdues the tax and institutional environment in Norway during 1983 and We examine various tax lienteles and onlude that we should fous our analysis on the Norwegian domesti individual investors. Setion 4 develops a model to present the investor s motivation for tax-loss selling. Setion 5 desribes the data, and Setion 6 reports our empirial results. Setion 7 onludes the paper. 2. Turn-of-the-year effet in Norway 2.1 The Oslo Stok Exhange The Oslo Stok Exhange (OSE) is the only stok exhange in Norway. It was first established in During the first 60 years of its history, it worked mainly as a ommodities market and a foreign exhange market. The first seurities were listed on the OSE on Marh 1, A shipping setor was introdued in As of today, the OSE is the world s leading market in the shipping setor. OSE is a member of NOREX, the Nordi stok exhange alliane whih also inludes stok exhanges in Stokholm, Copenhagen, Helsinki, Tallinn, Riga, Vilnius and Ieland. As of the end of 2005, 219 firms were listed on the OSE, of whih 28 were foreign ompanies. The industries in whih the listed ompanies operate in inlude tehnology, IT, finane, manufaturing, shipping and teleommuniations. The total market apitalisation of the listed term loss limit is raised from apital gains to ordinary inome; and the tax value of a long-term loss as a perentage of a short-term loss varies between zero (when long-term gains were tax-exempt) and 100% (when the differene between long-term and short-term investments is abandoned and all apital losses are entitled to the same tax benefit). 5

6 Norwegian firms was NOK 1,475 billion, with a total turnover of NOK 1,512 billion in By the end of 2005, foreign investors own 37 perent of the OSE Turn-of-the-year effet on the OSE rie and trading volume data for the OSE stoks is available sine January 1, However, trading was extremely thin during the first several years, whih are regarded by some as the dead horse years of the Norwegian stok market. Hene our dataset starts on January 1, 1983 and runs up to the end of The sample inludes altogether sixteen turn-of-the-years (exluding the turn-of-the-year of 1983). Using a value-weighted index of OSE stoks, for eah year during 1984 and 1999, we alulate the index return on the first trading day of the year ( turn-of-the-year return hereafter in this subsetion), the average daily return (inluding the first trading day) of the year, the monthly return for January and the average monthly return (inluding the January return) of the year. Table 1 reports the numbers. At the bottom of Table 1, mean and median figures are reported for eah of the four ategories. The average turn-of-the-year return for the value-weighted index is 1.19%, muh larger than the average daily return on the same index, 0.10%. The average monthly return during January is 6.43%, signifiantly higher than the average monthly return over the whole year, 2.08%. aired samples t-tests indiate that these differenes in means are statistially signifiant. Therefore it s safe to laim that a turn-of-the-year effet exists in the Norwegian stok market: there are abnormally high January returns on the OSE, and the stok returns are espeially high on the first trading day of the year. 2.3 Ownership struture and Window-dressing hypothesis As has been disussed earlier, both of tax-loss selling and institutional window-dressing predit the linkage between prior stok prie depreiation and exess stok return over the turn-of-theyear. However, this linkage is formed under different mehanisms in these two onjetures: given a ertain amount of apital loss, while the motivation for tax-loss selling is determined by tax provisions, the need to window-dress is strongly influened by the size of institutional holdings: the larger the institutional holding, the larger selling pressure on the stok by portfolio managers prior to the alendar turn. 5 Data soure: Fats & Figures Oslo Bors Stok Market

7 Although we hope that by fousing on the underlying mehanisms we ould disentangle the two hypotheses, spurious orrelations between tax laws and institutional holdings would severely undermine our apability of distinguishing their influenes apart. Therefore it is neessary to have a look at the orrelation between institutional ownership on the OSE and the tax provisions. Table 2 reports the fration of OSE market apitalization held by different investor groups during 1989 and Five basi investor groups are reognized: the Norwegian State (inluding muniipalities, ommunities, the state pension fund, the Churh, labour unions et, all of whih are tax-exempt institutions in Norway), domesti institutions, foreign investors, domesti individuals and domesti orporations. Among these investor lasses, we identify the first three as institutional investors (Foreign shareholders are regarded as institutions beause they are mostly international investment funds). During the eight years overed by the dataset, the State, the domesti and the foreign institutions own on average 18%, 17% and 30% of the OSE respetively. In another word, around two thirds of the Norwegian stok market is owned by institutional investors. Domesti individual and orporations on average own 10.2% and 24.8% of the market respetively. Table 2 shows that the ownership struture on the OSE has been fairly stable throughout the data period with very little variation. If there is window-dressing effet by institutional investors on the OSE, we expet that suh effet has been stable throughout the period, the size of apital loss ontrolled for. In drasti ontrast, multiple hanges in various tax provisions took plae from 1983 through 1999, as will be disussed in detail in Setion 3. If January effet is only aused by institutional investors window dressing and not related to tax-loss selling, there should be no relation between variations in the stok returns over the turn-of-the-year and hanges in tax provisions, whih we will try to verify in the rest of this study. 3. Institutional setting Before we link the stok return anomaly to the speifi tax ode hanges, it is neessary to introdue the institutional setting of the Norwegian stok market. Various lasses of tax lienteles oexist in the stok market, e.g. professional traders, tax-exempt investors, tax payable shareholders, and foreigners et, not all of them interested in year-end taxloss selling. For example, tax-exempt investors do not gain tax benefits from the loss selling. 6 Soure of data: Bøhren and Ødegaard (2000) Table 4. 7

8 Also, investors who an utilise the tax benefit immediately are not onerned with the timing of loss realization. The first part of Setion 3 disusses the loss selling motivation of eah tax lientele, and onludes that it is the domesti tax-payable orporations and individuals on whom this study should fous. Next, we turn to examine losely the tax regulations for the domesti tax payable shareholders during 1983 and We find that while for Norwegian orporations variations in the tax law have ambiguous effets on the firms loss-selling motivation, hanges in the personal investors tax odes are lear-ut and their impat on the tax-loss selling motivation should be fairly straightforward. The analysis for the rest of the study will therefore onentrate on the domesti tax payable personal investors. The third part of Setion 3 disusses other events on the stok market that also take plae around the turn-of-the-year whih might affet our results. One is the strong buying pressure from trading in tax-dedutible AMS shares; the other is the distribution of a tax redit whih amounts to the retained earnings after orporate tax, a unique approah to eliminate double taxation on the orporate inome. And finally, Subsetion 3.4 reports the disount rate during the sample period, sine it is an important fator whih determines the value of the tax benefit. 3.1 Tax lientele The tax-loss selling explanation for the turn-of-the-year effet is based on the assumptions that there are tax liabilities and benefits assoiated with apital gains and losses, and that the tax benefit of a apital loss is not reeived upon the realization of the loss but after the end of the tax year. Under these assumptions, the loser to the end of the fisal year a apital loss is realized (i.e. the shorter the time span between selling the stok and reeiving the assoiated tax shield), the less disounted the value of the assoiated tax benefit. The benefit to realize losses peaks at the end of the fisal year, resulting in large selling pressure in loser stoks. However, if tax liabilities and redits are omputed and paid ontinuously, or if apital inome have no tax onsequenes at all, over the years there would be no partiular patterns of realising losses prior to the turn-of-theyear. Different from the theoretial illustrations, the tax treatment of apital gains in the real world is muh more omplex and varies widely aross investors. 8

9 The following subsetion introdues different tax lienteles in the Norwegian stok market. By examining the tax-loss selling motivations of domesti tax payable, domesti tax-exempt, foreign investors and professional traders on the stok exhange, we attempt to find out whih investors prone to provide selling pressure in loser stoks towards the end of the fisal year Domesti taxable investor There are altogether three lasses of domesti taxable investors: individual investors, unlimited private ompanies, and limited ompanies. These taxable investors do not always pay tax after the inome year, i.e. under ertain irumstanes tax benefits of apital losses an be utilized within the inome year. 1) The tax liability of a domesti individual investor is usually due after the inome year. However, when an unexpeted apital loss ours with an amount that does not exeed the investor s expeted apital inome, the investor is allowed to use the losses to offset ertain taxable inome during instead of after the inome year 7. 2) An unlimited private ompany pays estimated tax to the government during the inome year 8. When a apital loss ours, the ompany an apply to the tax authority for approval of offsetting its inome with the loss. The approval is usually granted. 3) A limited ompany always pays tax after the inome year 9. To onlude, the Norwegian taxpayers who have a tax inentive to sell loser stoks towards the end of a fisal year are individual investors who use apital losses to offset unexpeted apital inome and labour inome 10, and limited ompanies that pay tax after the inome year. Unlimited ompanies and individual investors with large expeted apital inome an utilize the tax shield from apital loss almost as soon as the loss ours (or, at least four times a year for the unlimited 7 Before the start of a fisal year, the taxpayer estimates her inome of the oming year, inluding labour inome and apital inome et., and reeives a personal inome tax rate from the tax authority. Throughout the year, she pays tax aording to the estimated tax rate. When an unexpeted apital loss takes plae during the year, the investor an apply for using the apital loss to offset her expeted apital inome (for example interests from bank deposits). When the expeted apital inome is not large enough to be offset by the loss, the investor annot laim the tax shield from the loss until April of the next year when the tax statement for her previous inome year has been finalized. 8 ayments are made on Marh 15 th, May 15 th, September 15 th and November 15 th of the inome year. 9 Around one third of the inome tax is paid on February 15 th, another third on April 15 th. The rest is usually paid in late November after the tax authority finalizes the ompany s inome tax for the previous inome year. 10 As will be explained in detail in subsetion 3.2, before 1992, the investor ould only write off the apital loss against apital inome. However, the Tax Reform of 1992 stipulates that apital losses an be used to offset ordinary inome. 9

10 ompanies). Therefore we don t expet these taxpayers to be major partiipants in the year-end selling ativity. In fat, if these investors deide to realize ertain losses, given the persistent pattern of high stok returns in January, they would almost surely be better off selling the stoks at the end of January Domesti tax exempt The largest Norwegian tax-exempt investors inlude ommunities, muniipalities, the national pension fund of Norway, the Churh and labour unions. Sine they are not subjet to any tax in Norway, these institutions have no tax motivation for preferring realising losses on any partiular day of the year. It is possible that portfolio managers of tax-exempt institutions sell loser stoks at the end of the year to window-dress their performane reports. However, as disussed in Setion 2.4, beause of the stable state ownership in the stoks listed on the OSE, we expet the window dressing effet by the domesti tax-exempt investors to be onstant throughout the period. Furthermore, there is a speifi onern with the domesti tax-exempt investor s inentive problem: we expet that managers of state-owned funds are not as performane driven as their peers at ommerial institutions, and therefore not the likely andidates for window-dressing Foreign investors As shown in Table 2, foreign investors, mostly large institutional investors from the U.S. and other European ountries, own around 30 perent of the Norwegian stok market. The foreign investors apital gains and losses are not subjet to taxation in Norway. On the other hand, the Norwegian government imposes a withholding tax on dividend payments to these investors. The foreign investors influene on the year-end selling pressure in loser stoks in Norway is limited ompared to the domesti investors for the following reasons: First, some of these institutions are tax exempt for their apital gains/losses both in Norway and their home ountries, e.g. U.S. pension funds, who have no inentives for tax-loss selling. Seond, some tax payable institutions may not be interested in the loss selling towards the end of Deember. For example, throughout most of our sample period ( ), the aounting year for the U.S mutual funds ends in Otober, whih means that, if these funds wish to realise losses for tax benefits, they would rather sell at the end of Otober than Deember 11. Third, foreign taxable investors apital gains/losses are subjet to tax in their home ountries but not in Norway. During 1983 and 10

11 1999, foreign ountries rules governing the tax onsequenes of these investors trading have hanged in various and possibly ounterating diretions, leading to an ambiguous total effet on the year-end tax-loss selling by the foreign investors. And finally, foreign institutional investors might sell loser stoks to window-dress their performane. However, as argued in Setion 2.3, due to the fat that foreign ownership on the OSE had been rather stable, we expet the windowdressing effet to be onstant throughout the sample period. For these reasons, this study fouses on Norwegian domesti rather than foreign investors rofessional traders Taxation of professional traders on the OSE is fundamentally different from that of the professional traders in the U.S.: while Amerian traders are taxed on the hanges in the value of their inventory, their Norwegian peers apital gains are taxed upon realization 12, in a similar fashion as all the other limited ompanies in Norway are. Tax is paid after the inome year; therefore the professional traders on the OSE also have inentives to realize losses before the turn-of-the-year. However, as will be shown in Setion 4.2, when taking up round-trip trades around the turn-of-the-year, the arbitrageurs indifferene ondition is irrelevant of their tax status. Hene the professional traders an still be interested in arbitraging in the event, although they might also be motivated to realize tax losses themselves. 3.2 Capital gains tax during 1983 and 1999 As disussed above, the domesti tax payable individuals and orporate investors (inluding professional traders) may be the major partiipants in the year-end loss selling. In this subsetion, we will have a lose look at the tax odes for these investors ersonal apital gains ersonal apital gains tax was first introdued in Norway in Table 3 summarizes key provisions of the apital gains tax in Norway from 1983 to Before 1992 there was a distintion between long-term and short-term investments. The maximum holding period for a short-term investment was initially 2 years, lengthened to 3 years in While proeeds from long-term investments had always been tax-exempt, the tax rate on short-term apital gains started at 30% in 1983, rose to 35% in 1986, and further up to 40% in The Tax Reform of 11 Bhabra, Dhillon and Ramirez (1999) doument a November Effet for U.S. mutual funds. 11

12 1992 abandoned the distintion between short- and long-term investments. All apital gains are taxed at a flat rate of 28% disregarding the length of the holding period. In addition, several restritions on loss write-offs have been relaxed: investors are entitled to write-off apital losses against their general inome instead of apital inome; and the arry-forward limit for a apital loss is inreased from 4 years to 10 years. The influene of the tax reform on investor s loss selling motivation is several folded: First, the lower tax rate redues the size of the tax shield brought by apital losses. For example, if an investor had a apital loss of NOK 100 in 1991, she would have been entitled to a tax saving of NOK100*0.40=NOK40. The same amount of apital loss would only bring her a tax saving of NOK28 after the Tax Reform of The deline in the tax rate redues investor s tax benefit from realising apital losses, and the investor s inentives for loss selling as well. Seond, the removal of the distintion between long-term and short-term investments weakens the motivation for tax-loss selling. When short-term apital gains are taxable while long-term proeeds are tax-exempt, investors with short-term losses benefit from realizing losses and apture the assoiated tax benefits before the investment goes long-term. When the distintion between short-term and long-term investments is removed by the Tax Reform of 1992, there is no more selling pressure brought by the hange of the tax status of an investment. Therefore investors tax-loss selling motivation may be weakened. Third, as shown in Column 5 and 6 of Table 3, the Tax Reform of 1992 relaxes several restritions on apital-loss write-offs: First, before 1992, apital losses ould only offset apital gains, inluding dividends 13 and interest inome; after the tax reform takes plae, apital losses an be used to offset general inome 14. This hange in poliy enlarges the benefit assoiated with loss selling: both small investors without soures of large apital gains and those who have 12 The professional traders on the OSE have to arefully keep reord of every transation in order to measure gains and losses. 13 Dividends were subjet to apital gains tax before The Tax Reform of 1992 stipulates that dividends beome tax-exempt so as to ahieve a symmetri tax system, avoiding double taxation on orporate inome. Setion 3.3 will explain more in detail. 14 The Tax Reform of 1992 brought to Norway a dual inome tax system, whih divides inome into two ategories: general inome and personal inome. A flat tax rate of 28% is imposed on the general inome for all taxpayers, inluding both individuals and firms. General inome overs all taxable inome from work, business and apital assets. Tax allowanes and relieves are dedutible in the omputation of general inome, inluding apital losses and interest payments on debt. ersonal inome is the inome earned by labor input, on whih basis soial seurity taxes and surtax are alulated. A progressive tax shedule is applied to the personal inome. Currently, the top tax rate on personal inome is 55.3% (64.7% inlusive of employer s soial seurities ontributions). For a disussion on the soial effiieny of the dual inome tax system in Norway and other Nordi ountries, see Sørensen (1998). 12

13 exeedingly large apital losses are now interested in realizing losses before the turn-of-the-year for their tax benefits, whih would be infeasible before Seond, before 1992 apital losses ould only arry forward for four years. If within the four years the investor did not manage to aumulate enough taxable apital gains, the exessive apital losses ould not be utilized for tax advantage. The Tax Reform allows losses to be arried forward for as far as ten years ahead, whih inreases the tax benefits of apital losses. To onlude, we predit that the tax reform of 1992 has mixed effets on the investors loss selling motivation: on one hand, it inreases the tax benefit of apital losses by relaxing loss limit and inreasing the length of loss arry-forward; on the other hand, it negatively influene loss selling by reduing the tax rate and removing the differene between short-term and long-term losses Corporate inome tax Before the Tax Reform of 1992, the Norwegian taxation on business inome was haraterised by high orporate inome tax rates and large dedutions from the tax base. On one hand, the statutory inome tax rate was as high as 50.8%; on the other hand, the government generously allowed ompanies to use a number of speial provisions to redue their taxable inome. The tax odes for dedutions varied widely aross industries and asset types 15. In order to eliminate tax disrimination, the Tax Reform of 1992 introdued a flat 28% tax on business inome, and abandoned all speial allowanes in the previous tax system. How these hanges would influene a firm s motivation for year-end loss selling depends on not only the firm s industry type but also the firm s asset struture. Firms that were previously entitled to large dedutions, e.g. savings banks that were paying very little taxes thanks to speial tax treatments, now need to pay more tax after the reform. For some other firms the effetive tax rate ould remain lose to the previous level, e.g. eletriity ompanies that were taxed at the rate of 27.8% before the reform. And for firms that were not able to utilize the various tax allowanes, their tax rate is indeed redued by the tax reform. In fat, without knowing the exat identity of the orporate investor, it would be very diffiult if not impossible to tell exatly how the tax reform influenes the effetive tax rate for the speifi firm. Therefore, in this paper, we only fous on the individual investors loss selling behaviour, and ignore the effet of the tax paying ompanies. 15 Before the Tax Reform of 1992, the Norwegian orporate tax law inluded liberal depreiation, allowane for inventory reserves, and a yearly tax dedution for transfers to a onsolidated reserve whih 13

14 3.3 Other events around the turn-of-the-year Many events take plae around the turn of the year. Apart from loss selling, whih is the fous of this paper, there are other fators that might ause abnormal market movements, inluding the trading of tax-dedutible AMS shares (Aksjesparing Med Skattefradrag, meaning tax-dedutible stok investment), and the adjustment of stok ost basis with RISK (Regulering av Inngangsveidien med Skattlagt Kapital, whih means retained earnings of after-tax orporate inome) AMS shares Between 1983 and 2000, a speial tax redit alled AMS was distributed to Norwegian investors in order to enourage investment in the stok market. Every year during this period, eah domesti investor is entitled to a tax redit equal to a ertain proportion 16 of her investment in stoks or equity funds (at the purhasing prie) up to a ertain limit 17, whih she an use to write off taxable inome. Typially, a lot of the investments in the AMS shares are made in late Deember. The assoiated buying pressure an result in higher stok returns towards the end of the year, ounterating the effets auses by tax-loss selling. Consequently we expet our results to be weakened beause of the trading of AMS shares RISK: a apital gains tax redit In order to ahieve a symmetri tax system and to avoid double taxation on investment inome, the Tax Reform of 1992 stipulates that dividends are tax-exempt, and retained earnings of orporate inome are distributed to investors as a apital gains tax redit alled RISK. The tax redit of year t-1 is distributed on the first day of year t: the shareholder steps up her ost basis for the stok by an amount equal to the undistributed after-tax orporate inome per share. 18 The effet of the distribution of RISK is very similar to that of a dividend payout: it negatively influenes the stok pries. Dai and Rydqvist (2003) show that on average RISK redues the ould be as high as 23% of the profit. In addition, apital gains tax on fixed assets was allowed to be systematially deferred through speial provisions % between 1983 and 1992, 15% sine Maximum tax redit was NOK3000 before 1992, NOK5000 for Class 1 investor and NOK10000 for Class 2 between 1993 and The holding length of the investment should be no less than four years, otherwise the tax authority will relaim the tax redit. 18 Dai and Rydqvist (2003) gives detailed information on the Norwegian imputation tax system and this speial tax redit. 14

15 stok return over the turn-of-the-year by Taking this into aount, we should expet weaker turn-of-the-year effet after 1992, other things being equal. It is reasonable to expet that the size of the tax redit distributed at the end of the inome year and stok prie movements during the inome year are orrelated. For example, a firm whih has been operating poorly an suffer from market value depreiation, and have very low net inome to be distributed as the tax redit as well. A good firm on the other hand is likely to experiene stok prie appreiation during the inome year and bring a large RISK to its shareholders at the end of the year. If the above assumptions are true, for loser stoks, the stok return of over the turn-of-the-year will be influened positively by tax-loss selling; while for winner stoks, it is negatively affeted by the RISK distribution. Hene loser stoks will have even higher stok returns than the winners over the turn-of-the-year. In another word, the effet of tax-loss selling will be magnified by the RISK distribution. We examine the relationship between the size of the tax redit RISK of year t and the average daily stok return during the same year, and find that the two-tailed earson orrelation is only Therefore we onlude that the effet of tax-loss selling on stok return over the turn-of-the-year is unorrelated with that of RISK distribution. 3.4 Interest rate One of the basi assumptions of the tax-loss selling hypothesis is that there is time value of money. When the investor estimates the benefits from realising the apital loss towards the end of the year, an important onern for her is that if she sells after the turn-of-the-year, she will have to wait for an additional year to laim the tax shield. The higher the disount rate is, the more disounted the present value of the tax shield beomes, the more willing the investor is to sell before the end of the year, and the larger the turn-of-the-year effet gets. Table 4 reports the NIBOR 360 day interest rate on Jan 1 st of eah year between 1983 and During 1983 and 1993, the interest rate was floating between 10% and 15%. However, in the mid 90 s the level of interest rates dereased substantially and hovered between 4% and 7% for the rest of the deade. Aording to the tax-loss selling hypothesis, the lower interest rate after 1994 should negatively influene loss selling pressure. 4. Tax-loss selling hypothesis In this setion we set up a numerial model whih illustrates the relationships between various fators that influene the magnitude of the tax benefit and the investor s motivation for loss 15

16 selling. The following analysis assumes that (1) there is no prie unertainty; (2) time between the last and the first trade day around the turn-of-the-year elapses; (3) tax is due immediately after the turn-of-the-year. 19 Wash sales are not allowed in Norway, so the investor who just sold her shares annot buy the same stoks bak immediately to re-establish the ost basis. As established in Setion 3, this study fouses on loss selling by Norwegian taxable personal investors. In this model, we only onsider two tax-lienteles: the domesti tax-payable individuals who may onsider realising their apital losses at the end of the year, and the arbitrageurs who ompete until zero profit. 4.1 Sellers Constantinides (1983) posits that, in the presene of transations osts, investors will postpone selling until the ost of not selling outweighs the transation osts. The ost of not selling depends on not only the tax status of the apital loss whether the loss an bring the same amount of tax shield before and after the turn-of-the-year but also the amount up to whih the loss an offset inome, and the disount rate at whih the tax benefit will be deduted if the selling is postponed Seller Type 1 Seller Type 1 s apital loss will reeive equal tax treatment around the turn-of-the-year. The investor is not onerned with the tax status of the loss. However, she does need to onsider the disount fator if she deides to sell after the alendar turn. If the investor sells before the turn-of-the-year, disregarding the ash flow for the initial purhase, her ash flow is b ατ ( 2) b where is the prie at whih the stok is purhased, the stok prie on the last day before the turn-of-the-year, the fixed transation ost per trade, α a number between 0 and 1 representing the fration of the apital loss that is tax dedutible, and τ the apital gains tax rate. 19 As stated in Setion 3.2, for the domesti taxable individuals, the tax liability is due four months after the turn-of-the-year. But for analytial simpliity, we assume that the tax is paid one the alendar year turns. 16

17 If she deides to sell after the turn-of-the-year, by postponing trading for one day, the tax shield is delayed by one year, and the net present value of the tax shield has to be disounted for one year aordingly. e b e ατ ( 2) (1 + r) *1 e where is the prie on the first day in the new year, and r is the rate at whih the tax shield is disounted. The investor will prefer to sell the stok before the turn-of-the-year if e b b e ατ ( 2) ατ ( 2) > 1+ r For the seller to be indifferent to the timing of the sale, the following equation must hold, e r = * ( r aτ b b ) * ατ (2 + ) *ατ is the size of the tax shield from realising the loss before the turn-of-theyear. 1+ r r is a multiplier term. Assume that r =10%, α =1, and τ =30%, we get aτ r =0.125, i.e. the seller would be indifferent to the timing of realising the loss if the 1+ r aτ prie drop over the turn-of-the-year is about one eighth of the size of the tax shield. The indifferene ondition an also be written as e b rατ 2 = ( 1+ ) 1+ r aτ (1) Let us onsider a numeri example. Assume that an investor bought a stok at NOK150 per share, and by the end of the year the stok prie has redued to NOK100, providing a apital loss of NOK50 per share. The disount rate is 10%. The investor s apital gains are taxed at 30%. The transation ost is NOK0.5 per share. We also assume that the investor an fully utilize her apital loss to shield taxable inome both before and after the turn-of-the-year, i.e. α takes the value of one. Then aording to (1), a turn-of-the-year stok return of 1.95% will make the investor indifferent to when to realize her apital loss. If the stok return is less than 1.95%, she prefers to sell before the end of the year. 17

18 Seller Type 2 To seller Type 2, after the turn-of-the-year, her investment will immediately hange tax status and beome a long-term investment, wasting the tax benefit of the apital loss. When a stok is treated as a short-term investment before the turn-of-the-year but a long-term investment afterward, the investor will find it benefiial to sell the stok before the turn of the year as long as b e ατ ( 2) > To the left of the inequality is her ash flow if she sells before the end of the year (exlusive of the initial purhase); to the right if sells after the turn of the year. After rearranging, the indifferene ondition for the timing of the sale beomes e b 2 = ατ ( 1+ ) (2) Notie that both of the slope and the interept of (1) are 1 r + r times of those of (2). Sine ατ r both of α and τ take a value between 0 and 1, is less than 1, whih means that (2) 1+ r ατ has a steeper slope than (1), and that the turn-of-the-year effet is a lot stronger in (2) than in (1). Let us return to the above-mentioned numeri example. Everything remains the same exept that now the investor will lose the short-term investment status immediately after the-turn-of-the-year. Aording to expression (2), the turn-of-the-year return will have to be as high as 15.6% in order to make the seller indifferent. Any stok return lower than that will indue the investor to sell before the year-end The Marginal Investor All sellers in the market belong to either Type 1 or Type 2, i.e. the tax benefit of the apital loss will either remain or disappear immediately after the turn-of-the-year. The indifferene urve of the marginal seller, whih determines the stok return over the turn-of-the-year, an be expressed as: e b b * rατ 2 2 = D 1 ( 1+ ) + D * ( 1 ) 1 r 2 ατ + + ατ D 1 = 1 if the marginal investor is Type 1, 0 otherwise. 18

19 D 2 = 1 if the marginal investor is Type 2, 0 otherwise. After rearranging, the above expression beomes e b = ατ 2 r ( 1+ ) * D1 * + D r ατ (3) After 1992 when there is no distintion between short- and long-term investments, all marginal investors are Type 1 investors. However, when short-term apital losses are treated differently from long-term losses, it is possible that the marginal is Type 2. We hypothesise that the shorter the holding period for the short-term investment, the higher the hane that the marginal seller belongs to Type 2. As we have shown above, when the marginal seller is Type 2, the turn-of-the-year stok return is higher than when she is Type 1, other things ontrolled for. Hene the shorter the holding period for the short-term investment, the stronger is the turn-of-the-year effet. 4.2 Arbitrageurs The turn-of-the-year effet gives rise to arbitrage opportunities. As argued by Roll (1983), when investors realize that suh trading pattern is persistent, they ould easily make profit from buying stoks before the year end and selling immediately after the turn of the year. Consider an arbitrageur who follows this strategy and onduts a round-trip trade. Her ash flow is e e τ ( 2) + 1+ r The arbitrageur will only partiipate in suh trading if the above ash flow is positive. And she will ontinue trading until her profit is ompeted away, and the non-arbitrage ondition in the market is e 2 = (4) whih implies a horizontal line with the interept equal to the transation osts of the round-trip trade. Notie that the arbitrageur s indifferene ondition is independent of the rate at whih arbitrage profits are taxed. 19

20 4.3 Testable hypotheses Figure 1 depits the relationships between past performane, b, and stok return over the turn- b of-the-year. In the middle of the figure is a dashed vertial line mn where equals 1. All stoks ended up in the area to the right of mn are loser stoks with purhasing pries higher than the last trade prie of the year. In this area, the further away from line mn, the stronger the motivation for the shareholder to realized the apital loss. Those to the left of mn are winner stoks for whih the optimal strategy is to keep the stok and delay the realization of the apital gain. Type 1 seller s indifferene urve, desribed by expression (1), is line ab. In the area below the indifferene urve, she prefers to sell before the turn-of-the-year. Similarly, Type 2 seller s b 2 indifferene urve is line d. The two lines, ab and d, interset at point e, where = 1. 2 The non-arbitrage ondition, defined by expression (4), lies along fg, with an interept of on the y-axis. When transation osts are high, i.e. fg has a large interept on the y-axis, the stok prie will be mainly influened by the indifferene urve of the marginal seller, i.e. line ab or d, depending on the seller type. But when transation osts are low, arbitrageurs ompete to zero profit and the equilibrium pries lie along fg. The small interept of the indifferene urve for arbitrageurs on the y-axis determines that the stok return will also be limited. The above analysis provides us with a few testable hypotheses: When the transation osts are high, the stok return over the turn-of-the-year is expeted to be b - ositively related to, the size of the marginal investor s apital loss; - ositively related to α and τ, the fration of the apital loss that is tax-dedutible and the tax rate respetively; - ositively related to r, the interest rate; 20

21 - ositively related to, the transation ost. - Negatively related to T, the length of the holding period for the short-term investment. 5. Data and regression models We start with a dataset that inludes all ordinary shares listed on the Oslo Stok Exhange during 1983 and 1999, inluding 16 turn-of-the-years and altogether 2316 stok/year observations. In the following setion, we losely examine the stok returns over the turn-of-the-year, define explanatory variables based on the above-mentioned testable hypotheses, ondut desriptive analysis and develop regression models. 5.1 Stok return over the turn-of-the-year The turn-of-the-year return (TR hereafter) of stok i in year t is alulated as e it is the first losing prie of stok i in year t, and e it it, where it it its last losing prie in year t-1. Market-adjusted turn-of-the-year return (AR) is the differene between the stok return over the turn-of-the-year and the value-weighted index return on the Oslo Stok Exhange. The index return during the two days is subtrated in order to ontrol for the market movement during the period. Table 5 summarizes the TR, AR, and the number of alendar days between the last trade of year t- 1 and the first trade of year t. The average TR for the whole sample is 2.13%, while the median is 0.83%. The measurement of mean is influened by a few extreme outliers whih are either small firms with very low trading pries, or rarely traded stoks. For example, the maximum stok return over the turn-of-the-year is %. It is from a firm whih was not traded for 283 alendar days around the turn-of-the-year. The seond row in Table 5 shows that both of the mean and the median AR are 0.40%. Again there are a few outliers for the same reason as in the ase of TR. The third row of the table reports the number of days for the turn-of-the-year return, with a median of four alendar days and a mean of fifteen days. As mentioned above, inluded in the whole sample are some rarely traded stoks. The maximum reord is from one stok that was not traded for 501 days straight. 21

22 Table 5 also reports the size of the listed firms. The firms listed on the OSE are relatively small, with a median market apitalisation of NOK304 million, whih is around US$40 million. In order to keep outliers from distorting the results of OLS regressions, we exlude from the sample 22 observations with absolute values of AR larger than 0.5. Of the 22 ases, 8 are from year 1993, the year in whih the Norwegian stok market started to reover from a severe finanial risis originated in the late 80s. The 22 observations are all small stoks, with an average market apitalisation of NOK80 million. Another harateristi of these stoks is that they are rarely traded. On average, they are not traded for 128 days around the turn-of-the-year. Hene our sample size drops from 2316 to Our seond filter is the number of alendar days between the last trade before the turn-of-the-year and the first trade after. If a stok is not traded for more than 5 alendar days, the observation is taken away from the sample, beause too many fators other than tax-loss selling an influene the stok prie during a prolonged period of time. 663 more firm years are exluded from the sample. Eventually the filtered dataset onsists of 1631 observations. 5.2 Capital loss and transation ost We first set out testing the hypotheses that the motivation for loss sell is positively related to both of the size of the apital loss and the transation ost. Based on expression (1) and (2), we develop the following regression speifiation: ARit = a0 + a1lossit 1 + a2 * GAINit 1 + a3 * LN( CA) + εit (5) Loser vs. winner stoks b The size of the apital loss,, is different aross investors who purhase the stok with b various. We use two proxies for the marginal investor s apital loss: The first measurement is the aumulated stok return of stok i during year t, R n = (1 + it r i j= 1, t, j ) 1 where n is the number of days that stok i is traded in year t, and r i t, j, the return of stok i on the th j day that it is traded during the year. The seond measurement is a trading-day weighted average daily return for stok i during year t: 22

23 R it 1 = n n r k i, t, j j= 1 i, t, j where k i t, j, is the number of trading days between the previous trading day and the th j day. R it aords a higher weight to the period when the stok is more frequently traded. It is similar to a volume-weighted approah but is more appropriate than the latter for the Norwegian stok market, whih is haraterised by small and illiquid stoks that are sarely traded 20. The twotailed earson orrelation between these two measurements is In order to ondut the regression speified by expression (5), we reate two pairs of proxies for loser stoks and winner stoks. The regression variable LOSS1 ( LOSS2 ) takes the value of 0 it it when R it ( R it ) is positive, and the value of the proxy measurement when it is negative. Inversely, GAIN1 ( GAIN2 ) equals 0 when it it R it ( R it ) is less than 0, and the value of the proxy when it is positive. Sine LOSS1 it and LOSS2 it only arry non-positive values, and the loss selling hypothesis predits the loser stoks to have positive abnormal returns over the turnof-the-year, we expet the two proxies for loser stoks to have negative regression oeffiients. On the ontrary, we expet that the oeffiients for signifiantly different from zero. GAIN1 it and GAIN2 it will not be Aording to the major hanges in the tax provisions reported in Table 3, we divide the whole sample period into four tax regimes, and in eah of the tax regimes we ompare the mean and median TR and AR of loser and winner stoks. We use both of the two above-mentioned approahes to distinguish losers from winners. Table 6 presents the statistis. anel A inludes all stoks that are traded within 5 alendar days around the turn-of-the-year. anel B and C broaden the observation window to 10 and 20 alendar days. Our major findings are: First, aross all tax regimes and observation windows, TR and AR are persistently larger for loser stoks than for the winners. The differenes between losers and winners are statistially signifiant for most of the time, whih is onsistent with the loss selling hypothesis. 20 Consider a stok that is traded only twie during a given period. The first trade takes plae after k1days, and the assoiated stok return is r1. The seond trade after k2 days with a return of r2. k1<k2. We onsider it appropriate to alloate a higher weight to r1 than to r2. 23