Markov discrete choice process for dividend policy

This thesis aims to understand the dynamics underlying payout policies for companies listed on the public stock exchange. The dividend policy affects the liquid assets of companies both directly, given the size of the dividend paid, and indirectly, affecting the ability of companies to attract sources of financing in the immediate future. The thesis proposes an optimal dividend payout model that describes managers' behavior. Every year, managers have to choose whether to change their payout policy (by increasing or decreasing dividends) or to maintain the level of dividends paid in the previous period. We assume that the dividend policy is based on observable state variables (earnings at the beginning of the period and payout policy during the last period) and unobservable state variables (conflicts between managers and shareholders/bondholders, idiosyncratic risks and growth opportunities). We derive the optimal dividend policy from the solution of the stochastic discrete choice dynamic programming problem. The model depends on unknown primitive parameters that regulate the expectations of managers on future values of state variables. The maximization of the utility function provides the optimal strategy for the manager. Using annual balance-sheet data for companies operating in the Euro area, we estimate the structural parameters of the model using a nested fixed-point algorithm, and we test whether managers choices are consistent concerning the model predictions.