Discrete Dividend Policy with Permanent Earnings

Abstract

We develop an empirically tractable dynamic model of discrete dividend policy based on an inter-temporal coarse signaling framework in which dividend adjustments signal only substantial variations in the permanent earnings of the firm. Our theoretical framework relates the extent of dividend smoothing to the information content of dividends and also generates refutable predictions on the determinants of high or low smoothing by firms. Using an empirical methodology developed to test our predictions, we show that dividend smoothing is positively associated with factors that adversely impact the investor demand for the firm's shares. These factors include risk factors such as earnings variance, low liquidity, and high probability ofbankruptcy, as well as the expected return on capital investment by the firm. Firm dividend policy, especially understanding how managers set and change dividends, is a central issue in financial economics. This paper presents a dynamic signaling model of firm dividend policy. In the model, dividends are discrete in equilibrium (i.e., they are typically unchanged for a number of periods and are adjusted by a non-infinitesimal magnitude), even though permanent earnings are continuous. Thus, the equilibrium dividend policy reflects the primary characteristics of observed firm dividend behavior. The paper uses this theoretical framework to develop and empirically implement a dynamic discrete dividend model. The empirical model is one of the first time-series firm dividend models with a direct relation between dividend adjustments and changes in the permanent earnings of the firm. The validity of the empirical model is illustrated by showing that the model outperforms the Lintner (1956, 1963) partial-adjustment model on both long- and short-term samples, using both in-sample fit and out-of-sample forecasting. The theoretical framework is also used to empirically examine the determinants of dividend smoothing by firms. The model makes an important connection between dividend smoothing and the information content of dividends. This link derives predictions on the relation between firm-specific risk-return factors and high or low dividend smoothing in equilibrium. Although researchers have studied the presence of dividend smoothing by managers, there is relatively little work that explores the determinants of dividend smoothing. As in other signaling models, the empirical implementation of the model is challenging because the theory does not make transparent the empirical determination of the equilibrium dividend levels. Implementing the coarse signaling equilibrium presents a further challenge because the permanent earnings change thresholds that trigger the (non-infinitesimal) dividend adjustments have to be empirically determined. In this paper, these issues are addressed in an internally consistent fashion. Therefore, the approach presents one set of resolutions for several empirical