An alternative event study methodology for detecting dividend signals in the context of joint dividend and earnings announcements

Abstract

This paper has two purposes. In the first instance, it is an event study which sets out to determine whether or not a dividend signal can be detected, given that earnings and dividends are announced jointly by publicly listed New Zealand companies. The second purpose is much more important. In place of the usual simple OLS procedure known as the Market Model to estimate risk-adusted expected returns (and from these, abnormal returns), this paper contributes to event study literature by employing a friction model in which the values of relevant parameters are determined by a maximum likelihood estimation procedure. This friction modelling approach is more robust than the Market Model when an event study is run on company returns data in the context of a thinly-traded market as it is not immediately reliant on the assumptions underlying the employability of OLS regression. In addition, given that traditionally a piggy-back procedure involving restricted least squares regresssion has been used by past researchers to sort out the confounding effects of simultaneous dividends and earnings disclosures, this paper further contributes to event study literature and dividend signalling literature by positing several friction models which effectively achieves this same disentanglement. The models in this study are not necessarily restricted to investigations of dividends and earnings, but are employable (with appropriate specifications of variables) to event studies in general.