A THEORETICAL AND EMPIRICAL STUDY OF EARNINGS UNCERTAINTY AS A DETERMINANT OF THE DIVIDEND PAYOUT DECISION

KENNETH MURRAY EADES, Purdue University

Abstract

This work is a theoretical and empirical extension of Modigliani and Miller's (MM) (1961) information content of dividends hypothesis. Even though this hypothesis has received considerable empirical scrutiny, it lacks a rigorous theoretical base from which testable implications can be drawn. In fact, the information hypothesis implies only that dividend changes should be correlated with stock price changes. This theoretical limitation is eliminated in this paper by using a dividend signalling model patterned after Ross (1977) and Bhattacharya (1979). Several testable implications are drawn from this model and subsequently tested using both non-capital and capital market determined data. The non-capital market data (annual dividends and earnings per share) is used to test the theoretical implication that dividend payout (DP) is negatively correlated with earnings uncertainty (U). The capital market data (common stock monthly returns) is used to test the relative signalling strength implication of the model. Since high (low) U firms will have low (high) DP's in equilibrium, U will be a determinant of dividend signalling strength. In particular, for equal changes in DP's, the lower U firms will be giving the stronger signal to the market. The corollary of this is that for firms with equal U's, the firm making the higher change in DP will be giving the stronger signal to the market. Much of the methodology used for the non-capital market tests parallels that used by Kalay (1978) who did not find the hypothesized negative relationship between DP and U. However, the results of the tests of the DP-U relationship in this research clearly indicate the existence of the predicted negative correlation. In addition, it is shown that change in payout is more consistent with the data as a proxy for signalling strength than is the percentage change in dividends as used by Pettit (1972). The methodology used for the capital market data tests is an adaptation of the Black and Scholes (1973) paired sample design or, as termed in this research, the arbitrage portfolio method. The results of these tests clearly support the relative signalling strength prediction of the dividend signalling theory. Because of this unanimous empirical support using two different sets of data, we can conclude that U is a determinant of the corporate dividend decision. This is a conclusion which could not be reached solely on the basis of the information hypothesis of MM.

Degree

Ph.D.

Subject Area

Management

COinS