OK, this is weird. The number of NYSE stocks that currently have a negative beta (computed on last 60 monthly returns) appears to be a predictor of S&P 500 performance over the next 12 weeks:

Negative-beta companies | <100 | 100-200 | >200 |
---|

More specifically, if the number of companies with negative beta is less than 100, you have a 73% chance of negative market returns over the next 12 weeks; if that number is greater than 200, you have a 76% chance of positive return over the next 12 weeks.

Perhaps, this has something to do with the fact that correlations between returns or different assets are higher in down markets.

Data mining is also a possibility. These results have been obtained using a sample from Zacks database; the sample included 64 overlapping 12-week windows beginning every 4 weeks starting September 1, 2000. To rule out data mining, the study should be repeated on a larger sample spanning a longer period of time…

While the statistical significance can (and should) be questioned, the economic significance is clearly there. $10,000 invested in S&P 500 on September 1, 2000, would shrink to $8,283 by July 29, 2005 (including reinvested dividends). If, however, the investor chose to be out of the stock market when 200 NYSE companies or less show a negative beta (September 1, 2000 through January 17, 2003), her $10,000 would instead grow to $13,582, assuming cash was held in a non-interest-bearing account. Assuming that cash earned an interest rate of 0.3% every four weeks (about 4% annually), the investor’s portfolio would be worth $14,903 on July 29, 2005.