Study: stock market acts oddly before a crash. Feb. 15, 2006 Special to World Science Physicists have found that shortly before and after stock market crashes, stock prices start to follow distinctive patterns, somewhat like those found in heartbeats and earthquakes. The findings might lead to a system to predict stock-market crashes, they added-but not necessarily to stop them. In the study, University of Tokyo researchers analyzed the Standard & Poor's 500 index, a closely watched measure of U.S. stock performance. They found that over time frames of more than a day or so, short-term changes in stock prices usually obey a class of statistical patterns in which the most likely changes are relatively small ones. But for two-month periods surrounding crashes-such as the "Black Monday" plunge of October 19, 1987-fluctuations of all sizes are equally likely, they found. One result of this behavior, they noted, was that a graph of the fluctuations looks statistically similar if plotted over different time scales, between four minutes and two weeks. The study appears in the Feb. 17 issue of the research journal Physical Review Letters. This type of change in behavior is known as "criticality," the researchers wrote. According to an article in the Feb. 14 issue of Physics News Update, a newsletter of the American Institute of Physics, they drew an analogy with a class of metals that can become magnets when placed in a magnetic field. These "ferromagnetic" materials are each associated with a characteristic temperature known as the critical temperature. Below it, the metal arranges itself into distinct regions, in each of which the spins of the atoms become aligned the same way among all atoms. These regions, which contribute to the magnetism, look similar at different magnifications. This self-similarity is also found in graphs of the time intervals between heartbeats, or between earthquakes, the researchers added, according to the publication. But there is also a key difference: in those cases, the self-similarity stems from yet a different mathematical pattern, called a power law. Things that obey power laws become less likely with increasing size according to a characteristic formula. The university's Zbigniew Struzik told Physics News Update that it's unknown whether the new findings could lead to an early-warning system for crashes. One problem is that the warning could actually produce the crash, by inducing panic, he told the publication: thus, the tipoff could either "compensate for or neutralize the crashes, or make them worse." He added that his team will be researching how individual trading decisions lead to "criticality" in the stock market. * * * jose '-)