5/24/01
The Speculator
An investing tool you can do without
The Arms Index pinpointed the market bottom on March 22, but does it
have any
practical use as an investing tool? If it does, we couldn't find it.
By Victor Niederhoffer and Laurel Kenner
The NYSE Short Term Trade Index ($TRIN), or the "Arms Index," flashes
past the eyes
of CNBC viewers every 20 minutes and is one of the most widely used
indicators in
market technicians' toolkits. An examination of it seems like a nice
supplement to
our recent workouts on the value of the stock-bond ratio and Volatility
Index
($VIX.X).
The Arms Index represents the ratio of New York Stock Exchange up
volume scaled down
by number of rises, divided by the ratio of down volume, scaled down by
the number
of declines. The index has had some favorable press recently. Don Hays,
in an
interview with Barron's, used it to call the exact recent bottom of the
Dow ($INDU)
on March 22. Richard Arms, the inventor of the index, did likewise in
his
$8,000-a-year advisory for institutional clients. Some traders use a
five-day
average for short-term forecasts.
So TRIN works, right?
Not so fast.
TRIN may have worked on March 22. But that doesn't begin to make up for
all the
times it didn't work.
We studied the index from dozens of vantage points. We looked at the
correlation of
one-day, two-day, five-day and 10-day TRIN with S&P 500 Index ($INX)
moves over the
next one, five and 10 days.
Briefly, we found the results to be completely consistent with
randomness, with
correlations close to or right at zero over the last three and a half
years. In
other words, TRIN doesn't predict the future.
We're not the only ones to find difficulties with TRIN. In "The
Encyclopedia of
Technical Market Indicators," Robert Colby and Thomas Meyers tested the
TRIN from
January 1928 to March 1987. They performed many different tests using
daily data to
come up with predictions over the next one to 12 months. Their
findings: "We can
only conclude that it has relatively limited forecasting value for
stock prices."
Some may feel that conclusion is premature, as it is built on only 59
years of data.
But in "The New Technical Trader," Tushar Chande and Stanley Kroll
noted three
difficulties with using TRIN.
The ratios used in TRIN often obscure market action. This is
particularly true when
we have "mixed" market action.
The very process of smoothing TRIN with moving averages distorts the
picture of
relative volume flows.
The TRIN has a bounded scale for upside activity, but an unbounded
scale for
downside activity.
We talked to Richard Arms about how he uses TRIN (by the way, he
prefers to call it
the Arms Index), and about our findings. In his words:
Laurel: Do you trade much?
Arms: I trade some money for myself, but my primary work is talking
with my clients.
I put out an advisory for dozens of institutions. I don't trade much
any more,
because I'm occupied with other things.
Laurel: Millionaire? Centimillionaire?
Arms: Somewhere in the middle.
Laurel: From trading?
Arms: From trading, and from the fees I charge my clients.
Laurel: How would you advise people to use the index?
Arms: If you're a trader, watch the five-day moving average. When it's
in the 0.75
sort of area you become a seller and when it's over 1.25 you become a
buyer. It's a
judgment thing....The 10-day moving average allowed us to call the
bottom perfectly
on March 22.
Laurel: Can you elaborate on how to use judgment?
Arms: After working with it for so many years it becomes a little …
instinctive. I'm
not sure I can explain it really. I explain it to my...institutional
clients. I look
at a lot of things. I'm not just looking blindly at this index and
saying buy or
sell. It's not just a simplistic sort of thing.
Laurel: How many times have you used the Arms index as a primary signal
in the past
five years?
Arms: It's given only two of those big major signals in the last five
years. This
one and in October 1997. Then we have to go back to 1987 to get a
similar signal.
If there's proof, we can't find it
The problem is, the "buy" levels for the 10-day average don't happen
enough to
scientifically evaluate TRIN. You can't estimate uncertainty based on
something that
happens twice every five years.
Also, the signals happen to occur at times when the market is most
down, so
basically what you're taking advantage of is the old adage to buy in
panics.
As for the five-day indicator, not even Arms relies on it alone.
We told Arms that we had found no correlation between TRIN's moving
averages and
subsequent S&P 500 moves, and he said it bothered him to hear us say
that. "All of
Wall Street uses the index," he said, "because it works."
We can't prove TRIN doesn't work, any more than we can prove that there
are no black
swans. It's like a will o' the wisp, or, as the song goes, like holding
a moonbeam
in your hands.
Statistically, it's very difficult to evaluate a ratio of two ratios.
In technical
terms, the distribution of the ratio of two normal variables has an
infinite mean
and an infinite variance. In other words, larger samples don't help you
reduce your
uncertainty about the mean or variability of the indicator.
Arithmetically, the ratio is also intractable. Its variations are
dominated by
changes in volume. The volume moves are highly correlated with price
changes. The
ratio itself varies from 0 to infinity. A much more well-behaved
measure would be
based on the differences between the volumes and advances vs. declines,
perhaps with
a logarithmic operator thrown in for stability.
Whatever we say about TRIN, we have no doubt that someone will write in
to say we
are misinterpreting it, and that we should have looked at it this or
the other way.
We wish users of this ratio of ratios good fortune, and hope there are
experts out
there who have nonlinear ways of using it or selected levels at which
they have
confidence it will magically give a significant augury.
David Simon and Roy Wiggins, the Bentley professors we cited in our
volatility
article, did in fact conclude that TRIN has a small incremental value
as a daily
forecasting tool, when combined with VIX and the put-call ratio.
But as for us, the indicator brings to mind the story of the Hydra, the
nine-headed
monster allowed to live by Zeus as a challenge to heroes. The fumes
from any of its
heads were enough to kill any hero. And when one head was destroyed,
two grew in its
place.
Agree? Disagree?
In preparing this article, we, together with Dr. Brett Steenbarger, ran
hundreds of
regressions with many lags and leads for many different back periods.
We welcome
reader comment, and will be happy to share our results, some of which
are quite
technical, with interested readers. Mail us here. |
