The risk of earnings information uncertainty and the January effect in Korean stock markets

Abstract

Numerous empirical studies have documented seasonal regularities in stock returns. The most prominent regularity finds that returns on common stocks, especially small stocks, in January are significantly and consistently larger than those in any other calendar month. This phenomenon, so-called "January effect," is found in Korean stock markets as in other major international stock markets. This January seasonality is one of the strong empirical inconsistencies with market efficiency since, if the markets are efficient, investors should eliminate any predictable abnormal returns in January by readjusting their portfolios. Several explanations for this phenomenon are offered. Among those, the most extensively investigated explanation is the tax-loss selling hypothesis. This hypothesis suggests that tax-motivated investors sell off previously declined shares in price toward the end of the calendar year in order to realize capital losses and take advantage of tax benefits, which create downward pressure on stock prices. After the turn of the year, prices bounce up as the selling pressure is relieved. However, the strong January effect is found in Korean stock markets even though there are no capital gain taxes in Korea. This evidence is inconsistent with the tax-loss selling hypothesis. To date, no proposed explanations account satisfactorily for the observed small firms' January phenomenon. The fact that this January phenomenon is robust over such a long period of time suggests that the January effect, in fact, may be a phenomenon consistent with equilibrium pricing. One of the reasons that previous attempts to explain the January effect have failed is that the misspecified models might have been used in calculating abnormal returns. The purpose of this paper, therefore, is to identify and develop a common risk factor that provides a risk-based rational explanation for the January effect. The risk factor to be suggested is related to earnings information uncertainty caused by earnings volatility, earnings quality or speculation. We use the standard deviation of the forecast errors (FESTD) as a proxy for earnings information uncertainty. The forecast errors are computed as the difference between the actual earnings and the forecasted earnings. Because of the scarcity of earnings data point, we regard the earnings of the previous year's half-year period as the forecast earnings of the current year's same half-year period. The common risk factor for information uncertainty is constructed in a similar way to Fama and French's (1993). That is, the common risk factor, FESTD, is calculated as the return on firms with greater FESTD values minus the return on firms with smaller FESTD values. When the FESTD factor is combined with the market risk factor, there is a significant improvement in explaining the January effect in Korea. With the adjustment of raw returns for risk by using the two-factor model with the market risk factor and FESTD, the systematic pattern in the residual returns (and the abnormal returns) across firm size disappears. The arbitrage residual return in January, which is the difference in the average residual returns between the smallest and largest size portfolios, also is not significant. The competing models considered in this paper do not perform as well as this parsimonious two-factor model.