A New Dividend Forecasting Procedure that Rejects Bubbles in Asset Prices: The Case of 1929’s Stock Crash

Abstract

We develop a new procedure to forecast future cash flows from a financial asset and then use the present value of our cash flow forecasts to calculate the asset’s fundamental price. As an example, we construct a nonlinear ARMA-ARCH-Artificial Neural Network model to obtain out-of-sample dividend forecasts for 1920 and beyond, using only in-sample dividend data. The present value of our forecasted dividends yield fundamental prices that reproduce the magnitude, timing, and time-series behavior of the boom and crash in 1929 stock prices. We therefore reject the popular claim that the 1920s stock market contained a bubble.