Tax-loss selling and seasonal effects in the UK

Abstract

We examine monthly seasonal returns for the UK during the period 1955 to 2003. We identify four distinct tax regimes during which both the incentive and ability to tax-loss sell varies. In support of the tax-loss selling hypothesis, we find that the relationship between past losses and both January and April returns is strongest during tax regimes in which the incentives to off-set tax is high and weakest during regimes in which the incentive is low. Most intriguingly, our evidence suggests that tax reforms introduced in 1998 that had the aim of reducing short-term trading have been successful in limiting tax-loss selling by the company sector but not for the personal sector. Finally, neither the January nor April effect appears to be driven by the size effect.

Notes

¹ The issue is further complicated since the January effect is linked to other anomalies such as the winner's curse, momentum and the size effect.

² Of the twelve sub-periods, there are four significant April effects, three of which are prior to 1965. And there are four significant January effects, all after 1965.

³ We also find that the final month of the other two quarters are popular year-ends, with 11 and 9% of companies having a June and September year-end, respectively. The corresponding figures for 1994 were 7% (June) and 8% (September).

⁴ There are a number of causes for this latter increase. The Office for National Statistics cite the following: ‘The increase since 1994 partly reflects international mergers where the new company is listed in the UK, floatation of UK subsidiaries of foreign companies on which the parent has retained a significant stake, companies moving their domicile to the UK, improvements to the classification of holdings and the 1997 reclassification of £6.1 billion of shareholdings in the UK offshore islands to rest of the world (ONS, 2003).’

⁵ Although as Reinganum and Shapiro (1987) note, individuals who are classified as traders may have the gains and losses from their trading taxed as income.

⁶ And particularly from 1989, from when shares acquired before 1982 could be rebased and added to the ‘new pool’.

⁷ The 30 day rule is intended to counter bed and breakfasting. As from April 1998, the chargeable gain or loss for a share that is bought within 30 days of selling is based on the difference between the selling price less the repurchase price.

⁸ For individuals this order was: (1) Against acquisitions on the same day; (2) Against acquisitions within the 30 days following the disposal; (3) Against acquisitions preceding the disposal, but after 5 April 1998, LIFO basis; (4) Against shares in a Section 104 holding, but without identifying any particular shares in that holding; (5) Against shares in a 1982 holding, but without identifying with any particular shares in that holding; (6) Against shares held on 6 April 1965, again on a LIFO basis; (7) Finally against acquisitions following the disposal (and not already identified under stage 2 above), taking the earliest acquisition first. For companies and other concerns within the charge to Corporation Tax, disposals of shares of the same class in the same company held in the same capacity must be identified in the following order: (1) Against acquisitions on the same day; (2) Against acquisitions in the prescribed period which is either one month or six months; (3) Against acquisitions in the previous nine days; (4) Against shares in the ‘Section 104 holding'; (5) Against shares in the ‘1982 holding'; (6) Against shares held on 6 April 1965 on a LIFO basis; (7) Finally against subsequent acquisitions of shares, taking the earliest acquisition first.

⁹ The introduction of this taper was gradual. In 1998/98 92.5% of the gain on business assets held for one year is chargeable, falling to 50% for assets held for ten years. By 2002/03 the corresponding figures were 50 and 25%, respectively.

¹⁰ For Regime 1, our January premium is slightly higher to that estimated by Reinganum and Shapiro (1987) for the same period, while our April premium is a little higher. This difference is explained by our different filtering system as we only include stocks with at least 6 months of prior observations. We also include stocks for which we do not have size information.

¹¹ We also tried defining LOSS over the previous 12 months. Results were qualitatively similar, though less pronounced. On balance we believe our 6-month definition is more appropriate because tax-loss traders are more likely to be frequent traders and so the 6-month LOSS measure is likely to be the more relevant of the two.

¹² One explanation for this finding is that, although the incentives and ability to tax-loss sell may have declined, individuals are now more aware of personal money management issues and as a result are more likely to trade actively to minimize the tax burden. This is consistent with the relatively low April LOSS coefficient for Regime 2.