This paper exploits yearly accounting data from 1977 to 1994 to test the relative signaling power of dividends and net stock repurchases. The specification controls for potential agency cost and asset dissipation effects. Specifically, we regress changes in future income before extraordinary items on changes in dividends, changes in net stock repurchases, and a host of control variables. We also split the sample at 1981 to measure the impact of changes in the relative taxation of distribution methods. For the full twenty-year sample, only dividend changes are correlated with changes in future income. Moreover, the dividend coefficient and the repurchases coefficient differ statistically different in every future income equation. Splitting the sample reveals that the pre-1981 subsample drives the full-sample results. Put another way, the empirical link between changes in dividends and changes in future income vanishes just as a revision of the tax law reduced the tax disadvantage of dividend distributions. This evidence supports the notion that, at least for a period in time, firms deliberately exposed shareholders to punitive taxation to signal favorable prospects.