The nature of the dividend signal of future earnings.

This research investigates the empirical content of dividend signalling theory. Revisions in expectations of future earnings are examined to determine whether unexpected dividend changes signal expected future earnings. The dividend signal is modeled as both a direct signal of future earnings and as an indirect signal of future earnings through a signal of the persistence of current period earnings. Revisions in Value Line forecasts of future earnings measure exante revisions in expectations of future earnings and the market reaction to subsequent earnings announcements measure ex-post revisions in expectations. The results of this research are consistent with dividends signalling information about earnings for up to four future quarters. A positive association between unexpected dividend changes and revisions in analysts' forecasts of future earnings, after controlling for current period earnings, provides evidence that unexpected dividend changes signal future earnings. Revisions in analysts' forecasts are not consistent with unexpected dividend changes signalling the persistence of current period earnings, but are consistent with an interaction between unexpected dividend changes and current period unexpected earnings. Analysts discount the impact of current earnings on future earnings when unexpected dividend changes and unexpected earnings have the same sign, but place more emphasis on current period earnings when the signs differ. The market reaction to earnings announced subsequent to an unexpected dividend change confirms the effect of the dividend signal. The market reaction to subsequent earnings announcements is smaller than when no dividend change occurs, consistent with the information in the dividend announcement preempting the information in the subsequent earnings announcement. This research raises additional questions about whether unexpected dividend changes convey new information to the market or reflect information available from other more timely sources. This research confirms that unexpected dividend changes reflect information but provides little evidence that they convey new information.PhDFinanceAccountingUniversity of Michiganhttp://deepblue.lib.umich.edu/bitstream/2027.42/162022/1/8907007.pd