The informational content of dividend announcements and earnings forecasts in Canada
Foltz, David A.
Corporate dividends have been the subject of a great deal of research. Most of the past work on dividends has focused on the effect of the dividend policies in American markets with Canadian markets receiving little attention. Regardless of the outcome of past research, one thing that is certain is that dividends are considered costly to both firms and shareholders. Firms may have to forego positive net present value projects in order to pay dividends. At the same time, shareholders are required to pay taxes upon receiving them. Thus, several theories have been developed to explain why dividends are paid. In addition, empirical studies have been conducted to determine the effects that dividends have on share prices around dividend announcements as well as their link to taxes around the ex-dividend date. One theory that explains why dividends are paid and is the subject of much interest is dividend signaling. Signaling is defined as an attempt by management to reveal valuable information to the markets in order to reduce the prevailing information asymmetry. The postulation is that dividend signals assist the market in resolving some of the uncertainty regarding the future profitability of the firm. Early dividend signaling theory focused on dividends as a single event that signals future investment opportunities to market participants. In recent years, dividend signaling theory has progressed to include at least two stages of inter-related events. In other words, dividend signals are interpreted as part of a series of other relevant events. In this thesis this signaling scenario is tested using dividend announcements that are preceded by the release of analysts' earnings forecasts. This process sheds more light on the informational content of dividend signals. This study has two main goals. The first goal is to investigate the role that dividend announcements play in Canadian markets given the prior release of analysts' earnings forecasts. The second goal is to test the reliability of dividend signals in predicting future changes in earnings by examining the link between the dividend signal and subsequent earnings performance. The first goal is achieved by integrating a methodology developed by Best and Zang (1993) to determine the level of uncertainty prior to the release of dividend signals and a procedure designed by Elfakhani (1993) to define the role of these signals. The sampled dividend announcement events are classified first into three portfolios by the sign of the quarterly dividend change (i.e., increase, decrease, or no change). Each portfolio is categorized further according to the sign of the change in the earnings forecasts made by specialized financial analysts prior to dividend announcements (i.e., an upgrade, a downgrade or flat). These two steps divide the sample of dividend announcement events into nine portfolios (i.e., a matrix of three dividend changes by three earnings forecast changes). Each of these nine portfolios is then classified according to the level of information uncertainty prevailing in the market (i.e., high or low). Information uncertainty is proxied by the historical level of accuracy in the analysts' earnings forecasts (known as the prediction error). In the end, the sample is sub-divided into eighteen portfolios based on the sign of the dividend change, the direction of the analysts' earnings forecast change and the level of market uncertainty.1 Following Elfakhani (1993 and 1995), the dividend announcement event in each of the eighteen portfolios is assigned one of three possible dividend signaling roles (confirmatory, clarificatory or unclear) and one of three possible news directions (good, bad or flat). Both the dividend role and the news direction take into consideration the sign of dividend change, the sign of change in earnings forecasts, and the level of uncertainty in the market. A confirmatory dividend signal validates the information direction that has already been conveyed by the previously released analysts' forecast (i.e., potentially good, potentially bad or flat). The confirmation is needed because of the uncertainty that exists regarding the future of the firm particularly in a world of asymmetric information. A clarificatory dividend signal resolves any ambiguity resulting from the analysts' forecast. An unclear dividend signal leaves the market in a state of confusion regarding the future of the firm. Of course, it is expected that the dividend signaling role has an important part in determining the value of the signal to the market. The contention is that a clarificatory signal will be of greatest value and will elicit the strongest market reaction. Fallowing clarificatory signals in order of proposed value are confirmatory and then unclear signals. The reaction of the market to dividend announcements is examined across these 18 separate portfolios (only 15 contain data). The results show that the reaction of the market to dividend announcement events is consistent with the proposed role of the dividend signal and the news direction. The tests confirm that, of the three signals, the strongest is the clarificatory signal. This is consistent with both the signaling theory I proposed in this thesis and the results of recent American research by Elfakhani (1993 and 1995). The results also indicate that the market appears to assimilate the information revealed through the clarificatory signals more efficiently than with the other two roles. Further, the results from these tests indicate that, consistent with recent American evidence (Elfakhani, 1993 and 1995), dividend increases are not always associated with positive market reactions. Therefore, it can be concluded that the market does rely on the role that the dividend announcement plays. Tests were also performed on the data classified only by either the sign of the dividend change, the direction of the analysts' earnings forecast and the level of uncertainty. Based on these results, it appears that dividend signals categorized only by the sign of the dividend change do provide signaling power. Statistically significant positive (negative) market responses are generated at the time of the announced dividend increase (decrease). These particular tests also indicate that, unlike the evidence presented by Elfakhani (1993 and 1995), dividend decreases are more informative (i.e., they draw a stronger market reaction) than dividend increases. In this sense, Canadian evidence regarding dividend decreases appears to differ from American evidence. The second goal is to determine the relationship that exists between dividend announcements and subsequent earnings. Following Manakyan and Carroll (1990), the tests here involve a comparison of the earnings surprise metric to the level of earnings that is implied in the dividend signal. The results indicate that future earnings performance is generally consistent with the intention of the dividend signal (e.g., if the dividend signal predicts strong performance then future earnings rise to confirm the prediction). These results are consistent with recent American based studies on the subject such as Manakyan and Carroll (1990), Aharony and Dotan (1994) and Carroll (1995). This thesis makes some significant contributions to the field of research into dividend behavior. Evidence is presented on the reaction of the market to dividend announcements in Canada in a manner never before reported. For the first time in Canadian dividend research, the focus is on the dividend announcement date and its effect on the markets. To the knowledge of this author, the prior Canadian based studies have focused only on the stock behavior around the ex-dividend date or the dividend announcement month. In this sense, this study fills the Canadian void in the existing body of literature. Further, the methodology utilized in this thesis is also unique in that it provides a framework for a two stage dividend signaling model that utilizes financial analyst's earnings forecast information following dividend events. This study also looks into the reliability of the dividend signal in a manner never before attempted. Previously when looking at the relationship between dividend signals and the level of subsequent earnings performance, researchers assumed that a dividend increase (decrease) was good (bad) news and the market should react accordingly. The current study looks beyond this good news I bad news generalization and accepts the possibility that dividend increases (decreases) could signal bad (good) news. The contributions of this study stretch beyond the uniqueness of the testing procedures. The use of the Canadian database presents its own significant contributions. The Canadian and American markets are very different. Individual tax situations differ as do the make-up of the equities markets (Canadian markets are thinly traded and are largely made up of small firms while American markets are the opposite). Furthermore, while Canada is a resource based economy, the United States economy is influenced more by industrial firms. Given this separation between the two markets, it is also possible to replicate this study's methodologies using American data as well. This study and the results generated herein provide practical, everyday value to a variety of individuals. Investors will be able to utilize the results in order to better understand how to interpret earnings forecasts and dividend changes. By the same token, firm management will be better equipped to influence the opinion of the market and understand how it is likely to react to a given dividend event. In addition to these groups, professional investors (e.g., financial analysts, portfolio managers and institutional investors) will be able to use the results of this thesis to obtain a better understanding of the motives of management regarding its dividend practices and formulate their portfolios and strategies accordingly.