In contrast to conventional setup, a type 2 Tobit model is proposed to characterize the dividend behaviour. In the model, the selection regression determines whether a company would pay dividends whereas the output regression decides how much dividend a company will pay given that the company has decided to pay dividends. This modelling allows for the possibility that these two decisions might be affected by different variables and a given variable might influence each of the two decisions differently. Estimation is carried out via the Gibbs sampler with data augmentation algorithm which has been shown to be conceptually easy as well as computationally feasible and provides exact small sample properties.
