Tests of Dividend Signalling Using the Marsh-Merton Model: A Generalized Friction Approach

• Main Authors: WU, Chunchi, Kao, C.
• Format: text
• Language: English
• Published: Institutional Knowledge at Singapore Management University 1994
• Subjects: Business
• Online Access: https://ink.library.smu.edu.sg/lkcsb_research/806
• Institution: Singapore Management University

In this article, we propose a generalized friction model to test the information effect of dividends. The proposed model resolves the problem of dividend stickiness in the time-series data and provides a more reliable estimate for the relation between unexpected changes in dividends and earnings. Unlike previous studies, we specifically consider the dynamic partial adjustment of dividends in estimating the expected dividend component. Our results contain several important findings. First, the results support the contention that dividend changes reflect both expected and unexpected permanent earnings changes. Since Miller and Modigliani (1961) proposed the information hypothesis of dividends, financial researchers have sought to answer questions concerning whether dividends convey information and, if they do, what types of information are contained in dividends. The results in this study show that dividends are strongly related to a firm's permanent earnings estimates. Our results also show that dividend changes signal changes in management's views of the firm's future earnings prospects. The abnormal (unexpected) dividend, which is the difference between the current dividend and the conditional expectation of the current dividend, has on average a significant positive relation with the unexpected changes in the firm's permanent earnings. Second, there is evidence that the effectiveness of dividend signaling depends on firm-specific characteristics. Consistent with the existing dividend signaling theory, the strength of dividend signaling is negatively related to the firm's systematic risk, external equity financing, and size and positively related to net investments and the degree of dividend smoothing. Third, our results suggest that the generalized friction model is a more suitable framework for examining the information content of dividends. Watts (1973) took an approach similar to this study to examine the information effect of dividends.