Oct 06, 2003

Last week, I presented a series of charts outlining the assets versus liabilities of customers of NASD-designated firms. The charts showed a tremendous jump in margin debt (i.e. liabilities) in customer accounts in June and July, rivaling what was seen at the top of the equity bubble. However, free credit balances (i.e. assets) had also risen dramatically, and the difference between the two, what we could call “available cash” or “net worth” was actually a positive $6.6 Billion.

For the figures just released, which covers the August period, margin debit balances fell $8.4 Billion, which is the largest one-month drop in at least six years on an absolute basis. On a percentage basis, this 33% one-month decrease in margin debt is the third-largest decline, second to a 38% decline in December 2000 and a 37% decline in October 2001.

Free credit balances, on the other hand, dipped by $4.5 Billion, or 14%, from what was seen in July. This pushed the “available cash” to a positive $10.5 Billion, up nearly $4 Billion from the previous month. This is one of the most positive “net worth” readings seen over the past six years, and is on a par with many of the other readings seen since October 2002, as the following chart shows:

This most certainly is not a timing mechanism, as the data is two months old, and the NDX can easily decline even though these customers’ brokerage balance sheets may be in good shape. But it should give some perspective to the hype given the increase in margin balances a couple of months ago. I’ve been asked to give the same information for the NYSE, so I have included it below.

We can see here that the “net worth” of these customers has been negative all along. Even though it has improved considerably from the levels seen in 2000, this “available cash” went from -$31 Billion in September of last year to -$61 Billion as of August of this year. This is due to a 15% increase in margin debt and an 11% decrease in free cash during this time.

It should be noted that according to the NYSE, these figures are for long accounts only. They try to separate out short sales and for the most part they are not included in the margin figures. Even though short sales are done in a margin account, technically they are kept separate (e.g. in a “type 3” account). The NYSE realizes this and makes efforts to keep those short sales out of their margin balance reporting.

Looking back on the S&P 500 since 1950, I can find only four other times when the index began October with a string of five straight positive closes, as it did this year. The table below shows how the index performed the given number of days after the fifth positive close:

DATE	3 DAYS LATER	5 DAYS LATER	10 DAYS LATER	15 DAYS LATER	20 DAYS LATER	30 DAYS LATER	60 DAYS LATER
10/07/70	-3.1%	-3.1%	-3.7%	-4.0%	-2.9%	-4.7%	4.9%
10/10/83	-1.6%	-1.3%	-3.9%	-5.3%	-6.2%	-3.8%	-2.2%
10/06/89	-0.5%	-7.0%	-3.2%	-6.6%	-5.9%	-4.8%	0.0%
10/07/97	-1.6%	-1.3%	-1.1%	-6.2%	-4.3%	-4.6%	-0.8%

AVERAGE	-1.7%	-3.2%	-3.0%	-5.5%	-4.8%	-4.5%	0.5%

The instance in 1983 should not technically be included, since the first day in October that year was actually down 0.2%, but after that it went on a string of five straight positive closes, so I included it.

The data was compelling enough that I went back even further. The table below shows the results from the same study, only this time using the Dow Jones Industrial Average from 1897-1950:

DATE	3 DAYS LATER	5 DAYS LATER	10 DAYS LATER	15 DAYS LATER	20 DAYS LATER	30 DAYS LATER	60 DAYS LATER
10/06/1920	-1.4%	-1.4%	-1.2%	0.2%	-0.8%	-9.4%	-18.8%
10/06/1922	0.7%	1.8%	0.4%	-1.2%	-4.7%	-5.4%	-2.6%
10/06/1936	0.4%	1.1%	1.7%	0.9%	1.6%	4.5%	2.7%

AVERAGE	-0.1%	0.5%	0.3%	-0.1%	-1.3%	-3.4%	-6.2%

It would be a stretch (to say the least) to attempt to draw any meaningful conclusions from a sample of seven occurrences, but it is hard to deny that that this is compelling data, since it coincides with what is supposed to be a relatively weak time of year, and most of the occurrences showed negative performance almost immediately. After 30 days, the S&P was lower by at least 4% in each case except one (well, two, if you count 1983 when the S&P lost 3.8%). Again, I wouldn’t hang my hat (or my dollars) on this kind of information, but it is something to keep in mind.

I continue to struggle to find any kind of consistent reading among the measures that we follow. The majority of them are neutral, and the others are mostly split between mild bullish and mild bearish readings. For every reading showing excessive speculation or optimism (e.g. the Rydex RSI Spread and Beta Chase indicators, and the sentiment surveys), there are others showing that traders are relatively subdued, especially so near new 52-week highs (e.g. most variations of the put/call ratios and the Commitments of Traders data). We are very mildly overbought now according to some of our breadth measurements, such as the Up Issues Ratio, the daily NYSE cumulative TICK and the Down Pressure gauges, though it is not enough to be concerned about on its own. As I said on Monday, I have a slight preference for the short side here, mainly because of the confluence of failures we saw after the last time we made new highs, but due to the numerous conflicting signals, would not have the desire to remain short should we make new highs going forward.

This disclosure is not intended as trading advice in any form. It is meant as a note to subscribers that the author may have a position directly affected by the market outlook reflected in the commentary. Although the author takes great pains to remain objective in any commentaries, it is only fair that readers should know that the author may have taken positions in accordance with his market outlook. Positions can and do change at any time, without notice to the reader.

I received a few inquiries after my last comment regarding the lopsided breadth reading last Wednesday (when advancing issues on the NYSE outpaced declining issues by a ratio of 4.7 to 1, a ratio not seen in nearly 10 years). All of the inquiries questioned the validity of the study, due to the two caveats I listed in the comment – the preponderance of non-operating companies in the NYSE statistics, and the effects of decimalization. These concerns are valid, and I think they should be addressed more fully.

Tony Dwyer, of FTN Midwest Research, addressed the issue of operating companies in a post on Minyanville.com, by quoting some figures by Lowry’s which showed that a ratio of operating companies only resulted in a breadth ratio of 9.4 to 1, which is even greater than the 4.7 to 1 ratio used in the study. This is confirmed by my own figure of a ratio of 12 to 1 using the combined issues in the S&P 500 and Nasdaq 100. Granted, a base of 600 issues makes it significantly easier to attain outsized breadth readings, but I believe it’s instructive to know that using these figures for operating companies only, the breadth figure not only holds up, but becomes even more extreme. However, I do not know how the 9.4 to 1 figure for operating companies only compares to historical extremes, say from the 1970’s.

The issue of decimalization, I think, has more potential impact, and I took a detailed look at it over the past few days. Each time the NYSE has tightened the minimum change for the stocks traded on their exchange, it has had an impact on the way the breadth figures are reported. This makes perfect sense – after all, it takes less to move a stock 1/16 than it does to move it 1/8. When the NYSE switched to “1/16’s” in June 1997, the average percentage of issues that closed unchanged on the day went from an average of 25% to an average of 15%, and it happened immediately. Similarly, when they switched to decimals in January 2001, the average number of unchanged issues immediately went from 15% to 6%. So, obviously, if fewer issues are being considered “unchanged”, then that means that more of them are being considered “changed” – meaning they either rose or declined on the day.

The most important question from this is…”do they rise and decline in proportion?”. If they do, then we don’t really have to worry about decimalization. After all, even though more issues are now considered “changed”, if they are “changed” in an amount that is proportional to before decimalization was introduced, then decimalization doesn’t really make a difference. Unfortunately, that doesn’t seem to be the case.

To get a grip on the potential impact of decimalization, I looked at the NYSE breadth figures for the two years prior to when decimalization was initiated and the two years afterward. It made sense to me to compare the largest up days and the largest down days both pre- and post-decimalization to see how the breadth figures stacked up. So, I looked at the 5% of days with the largest gains (in the S&P 500) and the 5% of days with the largest losses, both pre- and post-decimalization. For the “pre” period, it turned out to be days with gains of 2.2% or more, and losses of 2.1% or more. For the “post” period, I looked at days with gains of 2.4% or more and losses of 2.4% or more. These were the days with the most extreme moves relative to the others in their period. The point of this was to see if there was any difference in the breadth readings between the largest moves in the “pre” period compared to those in the “post” period. It turns out there was, as the following charts show.

First, we’ll look at the days with the largest gains in the two years prior to decimalization (aka “pre”) and the two years after decimalization (aka “post”). I wanted to see how the breadth differed in those two time periods, so I looked at the number of advancing issues as a percentage of total changed issues (aka “ADV”), and the number of decliners as a percentage of total changed issues (aka “DECL”). As we can see in the chart below, there was quite a difference in the average percentage advancing issues, but not much difference in the percentage of declining issues.

Prior to decimalization, advancing issues accounted for an average of 55% of total changed issues on the days where the S&P 500 rallied the most. After decimalization, advancing issues accounted for 67% of total issues.

To see if this is consistent, let’s look at the other end of the spectrum – the days where the S&P 500 recorded its largest losses:

Here, we see nearly an exact flip from the chart above. Now, advancing issues were relatively unchanged both “pre” and “post”, but declining issues went from an average of 59% to an average of 69% of total issues.

What this shows is that when the market made an outsized move, in either direction, the breadth figures were skewed in that direction. So, on large up days, there are more advancing issues than there used to be, and on large down days, there are more declining issues than there used to be. Let’s look at this one other way, and that’s by looking at the standard deviations of the breadth figures, which will help show us if we truly need to adjust these figures post-decimalization.

The chart below shows the lower and upper range of the “up issues ratio”, which is simply defined as: ADV / (ADV + DECL). From that figure, I added and subtracted one standard deviation. This will give us some kind of idea as to the likelihood that the higher average “post” readings were due simply to chance.

This chart shows us that prior to decimalization, on days the S&P 500 rallied the most, the majority of the days should have seen advancing issues fall in a range between 55% and 73%. After decimalization, on those same types of days, the percentage of advancing issues should have fallen somewhere between 65% and 77% - a significantly higher range than before.

The next chart shows similar information, except it is for the range of declining issues on the largest down days.

Prior to decimalization, on days the S&P 500 declined the most, the majority of the days should have seen declining issues fall in a range between 58% and 78%. After decimalization, on those same types of days, the percentage of declining issues should have fallen somewhere between 66% and 80%, which again is a higher range than before.

So, what’s the bottom line? The bottom line is that we cannot accurately compare breadth post-decimalization to breadth pre-decimalization. Does that invalidate the 4.7 to 1 breadth reading from last week? Not entirely. While it does appear that it is now more likely than before that we will see very skewed breadth readings, in the nearly three years since decimalization was initiated, the most skewed breadth reading we saw was 4.4 to 1, despite the S&P rallying much more than it did last Wednesday. In fact, last Wednesday’s performance in the S&P wouldn’t have even cracked the top 5% of returns.

To put last Wednesday’s reading into an historical context, it needs to be adjusted. I think that reducing the 4.7 to 1 figure by 20% is fair, as the percentage of advancing issues in the post-decimalization environment appears to run about 20% higher than it did previously when the market rallies significantly. That would give us an adjusted breadth ratio of 3.76 to 1. How does that compare historically? It’s still in the top 1% of all readings since 1965, though Wednesday’s reading would fall from 50^th place all-time down to 105^th place. If we look at how the S&P 500 performed 90 days after seeing a breadth reading of 3.76 or greater, it was higher 73% of the time, with an average return of 7.1% - not appreciably different from what I reported last week.

A few of you have wondered about the intermediate-term “score” of our indicators. The confusing part is that there are more “red” check marks than “green” (meaning more indicators are more negative for the market than positive), but yet the score itself is in positive (for the market) territory. The reason for this is that the score is shown on a de-trended basis, meaning it is given in relation to other recent scores, thus removing any longer-term secular trends in the data. This way we can be more comfortable when we are seeing extremes, as they are more comparable to other extremes.

Below, I’ve included a chart of the raw score, which is simply an addition of all the red and green check marks that you see on the indicator page. This paints a drastically different picture, as here the score is currently at 46% compared to the 65% we saw in March, even though on a de-trended basis they are about equal.

Over time, I would say that I am more comfortable with the de-trended score as opposed to the raw score as shown above. The reason is because of readings like December 2002 – in the raw score above, we never reached what I would consider an extreme. On a de-trended basis, however, as you can see on the site, we did reach the lower extreme and it tipped us off that we were seeing excessive optimism in relation to other readings.

There weren’t any major developments in any of the weekly statistics that came out this weekend. The sentiment surveys continue to show an excess of bullish opinion, the NYSE specialist vs. public shorting data remains market-friendly, and there was essentially no change in the Commitments of Traders data for the large S&P 500 futures contract. In the e-mini S&P contract, commercials added a significant number of longs while the small specs reduced their long position by a large degree, but as I said last week, I am assigning only minimal value to all of this COT data at the moment.

Right here, I’m torn. Although I am not surprised that we are now challenging the recent highs – in fact, in last Tuesday’s comment that’s what I thought was the most likely course of action off the undue pessimism that existed at the time – there does not appear to be a whole lot of “buy-in” on this leg up over the past few days. For example, the Rydex traders have not been overly aggressive, and small-trader put/call ratios have been muted. While we got the three distinct failures from a previous high that I had been harping on, suggesting the momentum had shifted to the downside, this reaction back up has been heavily doubted. I know that it is most helpful when I can express a clear opinion, but I’m afraid I can’t do that at this point. I can see compelling arguments for both higher and lower prices in all time frames, but I don’t see the odds heavily favoring either. My “preference” if you will, is that we will see more weakness in the weeks ahead, but it is not strong enough to be willing to be short if the broader market can take out the September highs.