Dividend initiations and earnings surprises

Financial Management (Financial Management Association), Autumn, 1998 by Marc L. Lipson, Carlos P. Maquieira, William Megginson

In an economy that levies taxes on investment income, dividends are clearly a disadvantageous means of transferring wealth to shareholders. To justify dividend costs, two explanations are usually given: dividends are used to solve agency problems within the firm, or dividends are used to communicate information to the market.(1) These explanations are particularly appealing because they are consistent with the well-documented fact that announcements of dividend increases are accompanied, on average, by positive abnormal stock price movements.(2)

Recently, DeAngelo, DeAngelo, and Skinner (1996) raised two objections to these explanations. First, they find little evidence linking dividend changes to subsequent earnings surprises. Second, they suggest that actual cash commitments associated with dividend payments are small relative to the operations of a firm, and therefore have little value as either incentive or signaling mechanisms.

The results in DeAngelo et al. (1996) are somewhat surprising since Healy and Palepu (1988) document earnings increases, both in raw and industry-adjusted levels, subsequent to dividend announcements. Clearly, the methodology employed by DeAngelo et al., which examines earnings surprises (realized earnings in excess of growth-adjusted expectations) should better capture any signal embedded in dividend decisions. However, DeAngelo et al. examine firms that reduce dividends, and Healy and Palepu (1988) examine dividend initiations, and it is possible that dividends function as a signal in one context and not the other.

This paper uses the methodology of DeAngelo, DeAngelo, and Skinner (1996) on a sample of newly public firms that initiate dividends (henceforth "initiating" or "dividend-initiating" firms). Restricting our attention to firms that have recently gone public allows us to identify a matched sample of firms, which are comparable in terms of life cycle and future growth opportunities, but have not initiated dividends. Specifically, we identify firms that went public at the same time and in the same industry as our dividend-initiating firms, but ones that did not initiate dividends (henceforth "noninitiating" firms).

This matched sample is interesting because a firm should engage in signaling activities specifically to differentiate itself from firms that investors might perceive as having similar future prospects. Our matched sample comprises a set of firms likely to have similar future prospects. We also examine a matched sample of firms in the same industry that are approximately the same size as the initiating firm, but are already paying dividends (henceforth "size-matched" firms).

Consistent with Healy and Palepu (1988), we find that both raw and industry-adjusted earnings increase for our initiating firms in the first year after a dividend initiation, but not in the second year. However, in contrast to DeAngelo, DeAngelo, and Skinner (1996), we find that earnings surprises for our dividend-initiating firms are more favorable than for the matched sample of noninitiating firms in each of the two years following the dividend initiation. Tempering these results is the fact that the earnings surprises of our initiating firms are indistinguishable from either the size-matched sample or industry averages.

These results suggest that if dividend initiations signal future earnings prospects, it must distinguish one newly public firm from other newly public firms, not from established firms in the industry. We explore this possibility further by considering the magnitude of the dividend commitment associated with dividend initiations. DeAngelo et al. (1996) correctly point out that changes in dividend levels cannot be a valid signal of future prospects unless they represent a significant commitment of cash.(3) The essential argument is that firms with weaker future prospects will duplicate any actions by firms with better prospects unless the costs to the lower quality firm are significant. In this paper, we focus on the potential cost of dividends to firms that do not signal. Specifically, we examine the resource commitment required of noninitiating firms if they initiate similar dividends.

We find that dividend commitments represent 5% of earnings for our initiating firms. More importantly, had noninitiating firms matched the dividend yield, dividend-to-sales ratio, or dividend-to-assets ratio of the initiating firms, dividend commitments would have been about 8.5% of earnings. This 8.5% marginal commitment by noninitiating firms is substantially larger than the 3.5% documented by DeAngelo et al. (1996) for the dividend-reducing firms in their sample.(4) Furthermore, the difference in dividend commitments between initiating and noninitiating firms is significant. The large marginal dividend commitment required of noninitiating firms suggests that dividend commitments may be sufficiently large to support signaling equilibria in the context of dividend initiations.

Our results are related to a number of dividend studies. Consistent with DeAngelo, DeAngelo, and Skinner (1996), Lang and Litzenberger (1989) find no significant changes in analyst estimates of future earnings accompanying dividend announcements, and Benartzi, Michaely, and Thaler (1997) find no association between dividend changes and subsequent earnings changes. On the other hand, Venkatesh (1989) shows that greater dividend initiation announcement effects are associated with decreased announcement effects for subsequent earnings announcements. Asquith and Mullins (1983, 1986), Healy and Palepu (1988), and Venkatesh (1989) all document average positive price reactions to the announcement of dividend initiations, a result confirmed in our sample. Barber and Castanias (1992) find that dividend-initiating firms have significantly higher levels of cash flow from operations and less volatile common stock returns than firms that do not initiate dividends. Similarly, we find that dividend-initiating firms are generally larger and more profitable than the noninitiating firms that went public about the same time.