Breadth Inspection

Bottom Line: New 52-week lows on the NYSE have now accounted for 20% of all issues. Over the past 20 years, this has been an unequivocal “buy” signal.

The incessant selling the past few days has given us a few records, and a few “almosts”. On Friday, we recorded 3125 issues as having closed the day lower than the day before, which is a new all-time record of declining issues. Of course, such references must be put in the context of total issues traded, since issues listed on the NYSE have nearly tripled in the past 40 years. So as a percentage of total issues, Friday’s declining issues came in at 90% - the third-highest in the past 4 decades. The two readings above Friday’s were recorded on 10/19/87 (the day of the crash, when 95% of all issues declined), and 10/09/79 (which hit 91% declining issues). Not only was the number of declining issues extraordinary, so was the number of new 52-week lows on the exchange. At 710, we saw the most number of lows since the market low on July 24, 2002. Again, we must look at these in terms of issues traded, but we still saw a very unique event.

The chart below shows the last 20 years of NYSE new highs and new lows, both expressed as a percentage of total issues traded. New highs are in green, while new lows are in red. The dotted horizontal line in the bottom of the chart highlights the 20% level – the percentage of new lows we saw on Friday. The green arrows show the other times in the past two decades we have breached that 20% level. Below the chart is a table outlining each instance, along with the return in the S&P 500 the given number of days later.

DATE	NEW LOWS	% OF ISSUES	5	10	30	90	120	250
10/19/87	1,068	51%	1.3%	13.7%	3.2%	16.7%	20.2%	22.4%
10/20/87	1,174	57%	-1.5%	5.9%	-1.4%	13.1%	14.6%	16.3%
10/26/87	467	23%	12.3%	6.8%	3.2%	17.4%	13.9%	24.3%
10/28/87	477	24%	6.7%	3.7%	0.1%	15.5%	9.8%	21.0%
08/21/90	426	21%	-0.2%	0.8%	-3.2%	2.1%	14.5%	19.8%
08/23/90	711	35%	3.8%	5.3%	1.4%	6.3%	20.2%	23.6%
09/24/90	444	22%	3.4%	2.9%	3.3%	12.9%	22.7%	27.2%
09/27/90	407	20%	3.9%	-1.8%	2.2%	16.7%	22.2%	28.8%
09/28/90	439	22%	1.8%	-2.0%	2.5%	17.0%	19.8%	26.4%
04/04/94	637	23%	2.5%	0.8%	2.4%	4.5%	4.7%	14.4%
08/27/98	991	28%	-5.8%	-3.2%	-5.6%	22.0%	18.9%	32.5%
08/28/98	813	23%	-5.2%	0.3%	-2.9%	23.6%	23.9%	32.6%
08/31/98	1,183	33%	6.9%	8.4%	3.9%	33.2%	32.8%	40.8%
09/01/98	901	25%	1.2%	5.2%	1.1%	27.1%	26.1%	33.2%
10/08/98	930	26%	9.2%	12.4%	20.1%	29.0%	34.8%	38.1%
09/21/01	784	23%	7.8%	10.9%	12.6%	17.0%	20.7%	-12.5%
07/23/02	708	21%	13.2%	7.8%	12.0%	17.7%	16.1%	22.7%
07/24/02	943	27%	8.1%	4.0%	4.2%	11.0%	10.5%	17.2%

Average Return			3.9%	4.5%	3.3%	16.8%	19.2%	23.8%
Percent Positive			78%	83%	78%	100%	100%	94%

From the table, we can see that 10 days after at least 20% of all issues hit a new 52-week low, the S&P 500 was higher 83% of the time (15 out of 18 occurrences), with an impressive average gain of 4.5%. Even more impressive is that the average gain was more than twice the average loss. Going out to 90 days, the S&P was higher every single time, with an astounding average gain of nearly 17%.

It’s important to note that I’m being selective with my choice of time frame here. Instead of using only the past 20 years, why not use the past 40 years? Part of the reason is because 20 years is a nice round number that encapsulates “recent” market dynamics, another part is because prior to that we had fewer issues traded and another part is honestly because the numbers are very, very impressive. On Friday, I show that there were around 3500 issues traded on the NYSE, compared to around 2000 in 1987 and 1300 in 1965. The fewer issues traded, the easier it is to push these new high / new low figures into extreme territory. In the late 1960’s and early 1970’s, it was a not uncommon occurrence to see new lows as a percentage of total issues reach 30%, 40%, even 50%. The argument could be made that NYSE decimalization in 2001 has made it easier to see similar extremes, and while I think that is true in a minor way, I don’t think it has had an outsized effect on these numbers.

When we go back to 1965, new lows crossing the 20% level didn’t lead to anything spectacular in aggregate. While future market returns were generally still better than average, it preceded some awful declines (e.g. June 1969, April 1973 and July 1974). Perhaps it would be more accurate to use the whole data set, the use of which would suggest that Friday’s extreme in new lows tells us nothing consistent about what is most likely to happen going forward. But I always prefer to give recent market conditions more weight than historical ones, and market performance after recent extremes in new lows has been nothing short of extraordinary. One thing that gives me pause is that this new low figure is coming so soon after the broad market averages topped out. We’ve only declined a little over 5% in the S&P 500, which is the smallest decline to trigger this type of extreme in new lows in the past 20 years. Even the occurrence in 1994 saw an 8% decline before new lows reached such a level. Whether this is due to decimalization, or the inclusion of closed-end bond funds or whatever, it strikes me as curious and is a hint that this time may not be as indicative of a low than the others.

Bottom Line: Declining issues on the NYSE swamped advancing issues by 12 times. This has been a very rare occurrence, usually leading to outperformance.

Above, I mentioned the very high number of declining issues we saw on Friday. As a percentage of total issues, it was the third-highest in 40 years, but when we look at the number of decliners as a ratio of advancers, it is not quite so extreme. There were 12 declining issues for every advancing one on Friday, the first time we’ve seen such a thing since the mini-crash on October 27th, 1997. When looking at this historical data, however, we run into a similar problem as that with new lows above. Since there were significantly fewer issues, this time going back to 1940, we saw many more instances of “extreme” advance/decline figures. The most skewed figures I have on record occurred during another crash in May 1940. On May 13^th, 1940, there were 66 times more decliners than advancers, on May 14^th there were 26 times more and on May 21^st there were 64 times more. In all, there were 30 occurrences of at least 12-to-1 down days just in the 1940’s alone.

If we look at how the Dow Jones Industrials Average performed after 12-to-1 down days just during that decade, the results are nothing to write home about. In fact, they are just about right in line with average performance during that decade. The extreme down days preceded some nasty crash-like sell-offs but they also preceded some very nice rallies as well. In all, they pretty much averaged out.

Beginning in 1950, when these types of extremes began to get less frequent, future market performance improved astronomically. Out of those 50+ years, there was really only one long-term failure, which occurred on October 16^th, 1987 – the day before the 23% “dip” on Black Monday. There were a couple of others that required some short-term pain, such as in June 1950 and October 1979, but buying even then would have resulted in a nice gain after one year had passed. The chart below shows each occurrence in the past 50 years of these 12-to-1 down days (the dotted vertical lines).

Since I receive many requests to include the actual data points when possible, below is a table showing each of the instances.

NYSE One-Day Negative Breadth Extremes Declining Issues at least 12 times Advancing Issues 1950 - 2004
DATE	ADV	DECL	RATIO	5	10	30	90	120	250
06/26/50	59	1124	19	-2.6%	-4.4%	0.7%	6.8%	7.0%	17.9%
06/29/50	73	1015	14	0.9%	-3.3%	4.0%	9.2%	10.5%	21.2%
10/26/50	80	1035	13	-0.4%	-1.1%	-1.9%	12.5%	8.8%	21.4%
11/06/50	56	1046	19	3.2%	4.0%	1.5%	13.3%	13.1%	21.9%
11/28/50	69	1054	15	-2.7%	-0.6%	5.4%	8.5%	14.0%	18.0%
12/04/50	60	1039	17	2.2%	1.1%	10.1%	12.1%	17.5%	18.8%
03/13/51	81	996	12	0.9%	1.3%	4.1%	1.8%	8.3%	10.5%
10/26/51	59	812	14	1.3%	0.5%