There have been no empirical attempts to estimate and verify the dividend-earnings relation reflecting both the signaling and dividend smoothing hypotheses. This study proposed a cointegration model to test both hypotheses in an integrated framework in order to provide better insight into the dividend and earning relation. The purpose of this paper is two fold. First, we test the informational content of dividend. For this purpose, we construct a model to show the predictive relations among dividends and future earnings at the firm level. We are particularly interested in the issue of whether the model can detect a presence of inter-temporal relations between dividends and earnings. Specifically, this study pursued two main objectives: (1) what is the characterization of the optimal dividend payout patterns on the individual firm basis; (2) what is the effect of information on the firm's dividend payout policy. The implications of the signaling model and smoothing model of dividends were empirically tested using the recent 26 annual series data of dividends and earnings up to year 2006 for 222 firms listed on the Korea Stock Exchange. The firms with long-term equilibrium relationship between earnings and dividends are identified through a cointegration test and cross-sectional properties of these firms are investigated. The results of t-test and logistic regression show that the presence of cointegration is positively related to the degree of information asymmetry, a result consistent with the dividend signaling hypothesis. In addition, dividend smoothing is identified as an underlying force to make dividends and earnings cointegrated. To ensure the robustness of the empirical results, several firm characteristic variables are added in the logistic regression analysis. The results are virtually unchanged.