In Lintner's model of dividend behavior of firms change in dividends is a function of current earnings and lagged dividends. We show that under a rational expectations hypothesis of management behavior change-individends equation should include lagged earnings as an additional explanatory variable, and that expected sign of coefficient of lagged earnings variable is positive. Fama and Babiak predicted opposite sign for a lagged earnings variable in such an equation. Estimation and simulation results based on panel data for U.S. and Japanese firms provide modest econometric support for our Rational model. A good descriptive model of firms' dividend policies is useful, for example, for portfolio managers and for studying aspects of firm behavior such as interactions between investment and financing decisions and the management's transmissions of signals concerning changes in expected future earnings.' The econometric specifications of dividend behavior favored in literature are Lintner model (Lintner (1956)) and its FamaBabiak (FB) variant (Fama and Babiak (1968)). In Lintner model change in dividends is regressed on current earnings and lagged dividends. Fama and Babiak (1968) note that forecasting ability of Lintner's model is increased by adding lagged earnings as a regressor. We show that under a rational expectations hypothesis dividend behavior of firms may be described by an extension of Lintner's model, that we call Rational model, that includes lagged earnings as an additional explanatory variable. An important empirical difference between our Rational model and FB model is that expected sign of coefficient of lagged earnings variable is negative in Rational model while in FB model it is implied to be positive (see Fama and Babiak (1968, equation 10)). Our results based on panel data for U.S. and Japanese firms provide modest support for Rational model.2 I. The Rational Model of Dividend Behavior Our point of departure is partial adjustment model of dividend behavior of a firm (Lintner (1956)) given by AD, = aO + c(D,* D,-) + u,; t= 1,2,...,T (1) where A Dt = Dt -Dtdenotes change in dividends, Dt is dividends paid out in time period (year) t, Dt* is unobserved target dividend payout, c is speed of adjustment to difference between target dividend payout and last year's payout, ao is a constant and ut is an error term often assumed to be independently and normally distributed with zero mean over time periods. In Lintner model target dividend payout is replaced by Dt* = ryt which means that desired dividend payout is a fraction r of current earnings (yt). Thus Lintner's model is A Dt= a0 + cryt cDt+ Ut. (2) This model fits U.S. data (at both aggregate and disaggregate levels) quite well. Suppose instead that management determines target dividend payout by

SOME economists have ignored corporate financing decisions on the assumption that they do not affect the investment and production decisions of firms. Yet even if this extreme position is accepted, firms must still make financing decisions which in turn have impacts on other sectors of the economy. Since interest payments on corporate debt are tax deductible, the reliance of firms on debt as opposed to equity financing directly affects the revenue the federal government collects through taxation of corporate profits. The corporate presence in bond markets and corporate borrowing from banks can be expected to affect both interest rates and the demand for money. Thus any large macro model must somehow account for the borrowing behavior of firms. Yet in assessing this portion of the 1965 Brookings Quarterly Econometric Model of the United States, de Leeuw (1965, p. 506) writes that the regressions for business borrowing are the least successful of the model. In this paper we derive an equation for the long-term debt ratio (capital structure) of a firm which can be estimated using available data. Unlike some recent qualitative studies' on the aggregate corporate debt ratio as related to inflation and taxation over time, our primary aim is to explain differences in the debt behavior among individual firms. Tobit estimation results, explicitly allowing for the fact that some firms have no long-term debt, are presented for both U.S. and Japanese firms. Our theoretical model implies that the long-term debt ratio which maximizes the present value of the existing stockholder's equity depends positively on the cost of equity and negatively on the cost of debt, capital productivity, and retained earnings. Our estimation results are generally in agreement with these expectations. In particular, we find that capital productivity, which has not been included as an explanatory variable in most previous studies,2 and the cost of equity capital are both important determinants of the firm's capital structure. Our empirical results also support the view put forward by Komiya3 that debt ratios for Japanese firms are higher than those for U.S. firms in part because the cost of equity has been historically higher in Japan than in the United States in relation to the cost of debt. Other important attempts to explain the debt ratio as a behavioral function of the price of debt and other variables include the studies of de Leeuw (1965) and Goldfeld (1969).4 Certain conceptual problems mar these studies, however. Short-term and long-term debt are not distinguished, despite the fact that long-term debt is usually used to finance capital spending while short-term debt is used to finance inventory and