forecast the positions of small bundles of molecules in a uid, treating
them like Newtonian objects leading to very precise models of turbulent ow. All such models rely on the assumption that the most important variables can be precisely specied and that the premises underlying the model do not change in any fundamental way. Unfortunately, in most important realms of forecasting, neither of these assumptions is likely to hold. Variables such as culture, values, beliefs, and behavior have proven to be very dicult to rigorously and precisely specify. They are not ob- jective like the ination rate. Human consciousness enters into all of these variables and, as a result, makes them self-referential and vulnerable to surprising shifts. Culture shapes values and beliefs, which shape what we selectively pay attention to, and reinforces culture and values. The assumptions underlying models are also vulnerable to surpris- ing and sudden shifts. The oil shocks of the 1970s threw most economic models out the window. The Phillips curve, linking unemployment and ination, predicted that when ination rose, unemployment would fall. Ination results from growth in excess of the real potential of the econ- omy. So high growth leads to tight labor markets, high wage growth, and high ination. But a reallocation of global energy income toward OPEC led to ination, low growth, and rising unemployment. Stag- ation entered the economic vocabulary in the 1970s. In the 1990s the economists were fooled again. They could not imagine that productivity gains from new technology could be so large as to restore a high growth, low ination environment. Fortunately, the chairman of the Federal Reserve, Alan Greenspan, was not too much a prisoner of the models. He kept searching for hints that the rules might be changing and man- aged the U.S. monetary policy accordingly. As I noted earlier, the intellectual orientation of the forecaster also inuences the choice of the model. There seem to be two very dierent kinds of intellectual mindsets. One sees the world as mechanistic and reductionist. A forecast should lead to a correct prediction. The intel- lectual problem is simply one of better models and better data. These forecasters will invest all their energy and eort in rening and elabo- rating their models and the pursuit of more data. This is the route to the right answer. The other intellectual orientation sees the world as in- trinsically complex, holistic, and indeterminate. For those who follow this orientation, learning and better decisions are the goals not better