Bear Market Threshold

Does a 20% Drop Really Matter?

"Dow falls 166.75 to Bear-Market Level"
--WSJ July 3, 2008

 Back in 2008, the July 2 market drop nearly brought the DJIA decline below its 20% “bear market” threshold.  The S&P was off 19.40%, within less than a single day’s normal fluctuation of also reaching a 20% loss.

No economic logic underlies the widespread belief that a bear market begins only after a 20% market drop.  It is arbitrary and much too “neat” to be analytically credible.  It is also so retrospective as to let nearly half a trend escape before detection.  I have disdained the 20% mantra for years.(*) 

But the fact is, we find empirical evidence that crossing the 20% threshold may actually have some forward-looking import.  There have been five occasions of 20% decline over the past 40 years.  In every case the market continued downward.  The ensuing decline (after 20% loss) was greater than would be expected by chance.  On four of the five occasions, ensuing decline was far greater than could be expected by chance. 

S&P 500 Crossing the Bar

Bear Market

Total Pct Decline(**)

Decline up to 20% Bear Threshold

Still-to-go after -20%












-  8.9%





                       (**) Total decline is less than sum of parts due to downward compounding; e.g. top line (1 - .568)/(1 - .20) = (1 - .460).

The findings in the “still-to-go” column are astonishingly improbable, but real.  Consider this; of all declines from any S&P high, 87% are smaller than -5%, and 94% are smaller than 9% (which is the smallest follow-through above, in 1980-82).  That implies 15-to-1 odds against a -9% drop from a new high  But after passing the -20% mark, the average additional loss was -28.7%.

On the five bear occasions of the past 40 years, once the –20% threshold was crossed, the S&P produced another –28.7% on average.  (Earlier data from the ‘40s ‘50s ‘60s suggest lesser but still significant follow-through effect.  Email for data.)  The media’s favorite Bear threshold may be arbitrary or even irrational, but it’s actually dangerous.
Click Profile for a tabular profile of 249 S&P corrections and follow-through. 
(*)  For statistically defensiible thresholds, see our U-Turn model.  U-Turn synthesizes magnitude, duration, and variance into objective probability metrics.  Email for details.
Copyright© 2008. All rights reserved.