THE PRESENT VALUE MODEL WITH STOCHASTIC DISCOUNT RATE AND AN ANN PROCESS FOR BROAD DIVIDENDS

This paper uses an artificial neural network (ANN) model to forecast broad dividends, and computes fundamental stock prices with a stochastic discount factor (SDF). Broad dividends are used because they measure payouts to shareholders more accurately. Since nonlinearity is found in broad dividends, an ANN process is fit to these. Empirical results show that the consumption-based broad dividends model with ANN forecasting procedure predicts fundamental prices better, compared with models using linear dividends process, narrow dividends, or a constant discount factor. Nonetheless, actual stock prices remain largely detached from fundamental prices. Deviations between actual and fundamental prices, positive or negative, are found to coincide with business cycles, a result not consistent with alternative models considered in the paper.