Tax-Loss Selling: Evidence from December Stock Returns and Share Shifts

Abstract

This article develops a variety of new evidence on tax-loss selling and the January effect. Individual stocks are divided into categories depending on a tax-loss measure. By November, a given year can be categorized as a high tax loss or a low tax loss year. I find that December returns are significantly lower in high tax loss years than in low tax loss years. The subsequent January returns are higher in high tax loss than in low tax loss years. These effects hold across stocks of all sizes but grow in amount as stocks become smaller. The January increases are roughly the size of the December decreases. Tax-loss selling explains about one-half the January effect. Even following low tax loss years, the January effect is present. The size and tax loss effects across individual stocks depend on whether the year is a high tax loss or a low tax loss year. In low tax loss years, there is no size effect and no tax loss selling effect. I hypothesize that these effects occur because the public shifts shares to dealers in December, the tax-loss selling season. Data on Exchange member and nonmember transactions support this hypothesis. Over 60 percent of the temporal variation in the small firm January returns is explained by variation in end-of-year share shifts.