Sunday, July 12, 2009

When Historical Market Studies Fail to Identify a Pattern

As readers are aware, I use historical market patterns to generate hypotheses to guide my thinking about the future movement of markets. Note that the emphasis is on the generation of hypotheses, not the formulation of conclusions. I am using historical patterns as a kind of heads up, not as a source of mechanical trading signals.

Sometimes a historical investigation will lead to no hypothesis whatsoever. That, too, is information.

For example, we have recently seen four consecutive weeks of decline in the S&P 500 Index (SPY). It is reasonable to expect some bounce from such oversold conditions. Going back to 1994, however, we find that, when SPY has been down for four consecutive weeks, the next four weeks average a gain of .51% (16 up, 15 down). That does not meaningfully differ from the average four-week change for the remainder of the sample (.45%; 459 up, 315 down).

Interestingly, a careful look at the data suggests that how the market behaves in the four weeks following four consecutive down weeks is a pretty good identifier of longer-term bull and bear markets. In bull markets, four down weeks tend to be followed by strength; in bear markets, the down weeks lead to further weakness. In this case, the pattern appears to be not so much predictive as indicative. (Note: the most recent instance occurred in early March of this year and led to four weeks of strength).