Rethinking Lintner: An Alternative Dynamic Model of Dividends

Abstract

Empirical modeling of dividends has been dominated by Lintner (1956). However, Lintner's model suffers from the logical paradox that if companies have target payout ratios then in the steady state the companies will have reached those target payout ratios. Moreover as demon-strated by Bond and Mougoue (1991) Lintner's model is also poorly specified when earnings are serially correlated. This twin shortfall of Lintner's model motivated us to explore the possibility of an alternative dynamic empirical model of dividends. We test our model by cross sectional Tobit regression as well as by time series fitting. We find that the results of the Tobit regression are consistent with the predictions of our model. In time series testing, we find that one of our models fits the empirical reality at least 75% of the time. For firms with longer data series of 35 years or more, our model describes the empirical data succinctly in 96% of the cases.