25 years ago Meese and Rogoff showed that exchange rate changes were largely unforecastable. This result continues to hold. Now Flood and Rose say, Don't worry, be happy because we can't forecast aggregative stock index changes either (I am not making this up).
Their paper is called "Why so glum? The Meese-Rogoff methodology meets the stock market"
Here is a link. Here is the abstract:
This paper applies the Meese-Rogoff (1983a) methodology to the stock market. We compare the out-of-sample forecasting accuracy of various time-series and fundamentals-based models of aggregate stock prices. We stick as close as possible to the original Meese-Rogoff sample and methodology. Just as Meese and Rogoff found for the case of exchange rates, we find that a random walk model of stock prices performs as well as any estimated model at one to twelve
month horizons, even though we base forecasts on actual future fundamentals of dividends and earnings. Using this metric and for this sample period, aggregate stock prices seem to be as difficult to model empirically as exchange rates.
Note that saying a random walk model works best means that the best predictor of tomorrow's price is today's price which means price changes are not forecastable.
I think more disciplines should adopt this trend. Math guys could write papers saying hey, don't worry that we can't prove conjecture X, we can't prove conjecture Y either!