I’ve been engaged in a very interesting discussion with Kash at Angry Bear on economic forecasting which will appear in the Wall St. Journal’s Econoblog later this week. Kash has a couple of posts ,  based on his contributions over at Angry Bear. Here are the remarks that I used to open the discussion:
People often complain about the inaccuracy of economic forecasts. But for many variables, the inability to forecast the variable can be an indication that our economic understanding of the market is right on target.
The classic example of this is the stock market. According to one theory of stock prices, in a properly functioning market it should be completely impossible to forecast the change in stock prices. The argument is that if somebody knows that the price of the stock will be higher at the end of the week than it is now, they should buy right now while the price is still cheap. But such purchases drive the price of the stock up immediately. This theory thus maintains that the current stock price should reflect all available information, so that no predictability remains in the change between the current price of the stock and that at some future date. This is sometimes described as the “random walk” theory of stock prices, or, more accurately, the view that the stock price should follow a martingale.
There is both theoretical and empirical evidence against this martingale hypothesis. When investors care about avoiding risk, for example, there could be higher expected returns in times when the stock is more risky to hold. But there is no question that the martingale description is a pretty good first approximation to the behavior of most stock prices.
Actually economic theory suggests that not just stock prices but a great number of other economic magnitudes of interest should exhibit near-martingale behavior. Predictable price changes in any commodity that can be stored, for example, can be arbitraged away by increasing or decreasing current inventories. Such inventory adjustment again would result in the current price jumping to the value that best reflects the expected future price. Although costs of storage, changing inventories, and other benefits from holding inventories can introduce the possibility of more complicated dynamics, a martingale proves to be a good approximation to exchange rates and many commodity prices.
Indeed, it turns out to be a standard feature of any dynamic optimization problem in the face of uncertainty that a certain function of the magnitudes you choose should prove to be impossible to predict ahead of time– if it were ex ante predictable, you’re not using all the currently available information in the optimal way.
I’m sympathetic to the popular perception that if you really understand something, you should be able to predict what’s going to happen next. But there are at least some examples in economics where our theory suggests that if we really understand the phenomenon, it should be impossible to predict the variable. When someone tells me they can predict with a great deal of accuracy what the stock market is going to do next, my first reaction is that the person may well be a charlatan. Whereas when someone tells me they don’t have a clue, I think, now there’s a real expert!