The principles of behavioral psychology can explain how crashes occur. In particular, the concept of “stimulus generalization” tells us that organisms tend to respond in the same way to similar stimuli. In a crash, or pre-crash, context, several stimuli—including rising prices, above-average equity allocations and prices perceived as excessive in relation to fundamentals—are aversive because they have been associated with prior market crashes. They induce escape behavior on the part of some investors; that is, these investors sell stock. The sale of equities and consequent lower stock prices are particularly aversive to other investors, whose escape behavior produces more selling and further price declines. Before long, the crash has arrived. Prices, however, eventually reach a point where aversiveness is no longer a factor. Investors’ escape behavior ceases; selling ceases; the crash ends.

Economists are trying to approximate such behavior with a new economic model that distinguishes between value-based investors and portfolio insurers. In this model, a market dominated by value-based investors is characterized by the traditional economic model: The demand curve slopes downward to the right, with positive information and preference changes increasing equilibrium price and negative information and preference changes decreasing price. Here there is a strong tendency for price to return to equilibrium, and the market is continuous and stable. In a market dominated by portfolio insurers, however, the demand curve is Z-shaped; prices can increase or decrease discontinuously with positive or negative information or preference change. The market is discontinuous and unstable.

Of course, the market was not dominated by portfolio insurers before the ‘87 crash (or earlier crashes). But the behavioral model suggests that ordinary, value-based investors can behave, in the aggregate, like portfolio insurers during a crash. In ordinary times, then, investors’ behavior may be characterized by a “mostly value-based investors” model, but as a crash develops, a “mostly portfolio insurers” model becomes a more accurate approximation of their behavior.