Previous research suggests that stock prices are inefficient because they only partially impound the implications of current earnings on expected future earnings. In other words, current earnings have implications for expected future earnings, but when subsequent future earnings are announced, market prices react as if the predict- able effect on expected future earnings had not been fully impounded in market prices. Prices seem to underestimate the implications of quarterly earnings changes for subsequent earnings by 50 percent, on average, across the four subsequent quarters.

The authors explore the limits of these findings by focusing on a component of earnings—special items. Special items can be consid- ered to be the nonrecurring items identified by Compustat from the income statement and the accompanying footnotes—for example, current-year results of discontinued operations, natural disaster losses, and nonrecurring profits or losses on the sale of assets, investments, and securities. Whereas aggregate earnings are persistent, special items are commonly viewed as transitory.

Previous research done on market response to the release of special item information and managing earnings through the use of special items includes that of Elliott and Shaw (Journal of Accounting Research, 1988), who concluded that material asset write-offs resulted in negative one- and two-day stock returns when announced, as well as Francis, Hanna, and Vincent (Journal of Accounting Research, 1996), who found negative market reactions to inventory write-offs but positive reactions to restructuring charges (meaning the nature of special items made a difference to market reaction). Elliot and Hanna (Journal of Accounting Research Supplement, 1996) found that the market attached less weight to unexpected earnings before special items once large special items had already been recognized and that investors believed write-offs to be more transitory than other earnings components. Kinney and Trezevant (Journal of Financial Statement Analysis, 1997) found that negative special items tended to be shown on the income statement, whereas positive special items were described in notes, and that when firms had large positive (smoothing) or negative (big bath) special items, they chose to recognize the negative special items if they had large earnings changes (whether positive or negative).

The authors ask whether the market fully impounds the implications of special items on future earnings and find significant differences between the effects of positive and negative special items on future earnings. Positive special items are less than completely transitory in the sense that they are followed, on average, by a small but nonzero amount of earnings of the same sign in subsequent quarters. Negative special items, in contrast, are followed by earnings of the opposite sign in subsequent quarters. This finding is consistent with the conjecture that negative special items sometimes represent a shift of expenses from future periods into the current period (e.g., through restructuring charges), which reduces current income but increases future income.

The authors also find that prices reflect a larger proportion of the implications of special items than do the remaining components of earnings. This result is consistent with the proposition that because the implications of (largely) transitory special items on future earnings are relatively clear, market prices reflect relatively more of the implications of special items.

Focusing on the announcement of future earnings four quarters subsequent to the original announcement (when the implications for future earnings of special items differ most from the implications of the remaining components of earnings), the authors find that the effect of special items on future earnings (four quarters hence) is underestimated by about 27 percent. In contrast, the effect of an innovation in current aggregate earnings on future earnings is underestimated by about 75 percent.

The authors also find that market expectations reflect the differences in the implications for future earnings of positive versus negative special items. Although the implications differ, prices impound approximately the same proportion of information for positive versus negative special items.

To provide further evidence of a potential inefficiency, the authors examine the return to a trading strategy in which a portfolio is formed by tracking a long position in companies reporting negative special items four quarters earlier and a short position in companies reporting positive special items. This portfolio earns a small but statistically significant three-day return around the earnings announcement four quarters subsequent to a special item.