Why you should follow market correlations
In the past year, I have been privileged to work with a variety of unusually
consistent, profitable traders. These highly
successful market professionals, I’ve found, operate very differently from
how successful traders are usually portrayed in the popular media. Indeed,
I cannot find a single one who utilizes the strategies that dominate most
trading books and periodicals: chart patterns, oscillator readings, Fibonacci sequences,
etc.
If I had to describe what they do in a very simple way, I’d say that they
examine multiple markets and the normal relationships among these. When
one or more markets deviates from these normal relationships, the traders
develop ideas to exploit this deviation. Sometimes the ideas look for the
deviations to amplify (one market trending vs. another); other times they look
for the deviations to return to normal (markets coming back into line).
Less successful traders trade a single market, and they look for directional
moves. A surprising proportion of the trades of the highly successful
traders occur across sectors or asset classes and attempt to exploit mispricings
of these.
Let’s model a simple example of this promising mode of trading. As I
recently noted on my research blog, I used ETFs to divide the equity market into
seven sectors: energy, finance, technology, consumer, health care, utilities,
and raw materials. I then computed 20-day correlations between each of the
sectors on a moving basis and, for each trading day, calculated an average
correlation among all the sectors. This overall average correlation since March,
2003 (N = 807 trading days) has been .48. It represents the degree to
which the seven sectors are moving in unison or independently.
The key question is: Does it matter if sectors are moving together?
Since March, 2003, we have had 97 days in which SPY
(
SPY |
Quote |
Chart |
News |
PowerRating) has risen 2% or more on a
five-day basis. Five days later, the average gain in SPY has been .25% (58 up,
39 down). This is not different from the average gain in SPY for the entire
sample (.28%; 465 up, 342 down).
When, however, SPY has risen 2% or more *and* we have a high intercorrelation
among the sectors (N = 48), the next five days in SPY averages a strong gain of
.42% (34 up, 14 down). When SPY has risen 2% or more but the intercorrelation is
low (N = 49), the next five days in SPY averages a gain of only .08% (24 up, 25
down).
This suggests that strong rises are most likely to continue when the sectors are
moving in concert. Returns are subnormal after strong rises when the sectors
diverge relative to one another.
How about relatively flat markets? Since March, 2003, we’ve had 84 days
in which the five-day change in SPY has been between +.20% and -.20%. When
SPY has been relatively flat on a five-day basis *and* the intercorrelation of
sectors has been high (N = 42), the next five days in SPY have averaged a solid
gain of .42% (28 up, 14 down). When we’ve had a flat five-day SPY and the
intercorrelation among sectors has been low (N = 42), the next five days in SPY
have averaged a loss of -.10% (21 up, 21 down).
Here again, we see bullish near-term results following from sectors moving in
concert and subnormal returns when sectors move on their own.
When it comes to declines, however, we see a fascinating picture. Since March,
2003, we’ve had 105 days in which SPY has dropped 1.5% or more in a five-day
period. Five days later, SPY has averaged a gain of .99% (72 up, 33 down)–much
stronger than the average gain for the entire sample as noted above.
When SPY has dropped 1.5% or more in a five-day period *and* the
intercorrelation among sectors is high (N = 53), the next five days in SPY
average a very strong gain of 1.44% (40 up, 13 down). When SPY has dropped
sharply and the intercorrelation among sectors is low (N = 52), the next five
days in SPY average a more modest gain of .54% (32 up, 20 down).
The bottom line is that when sectors are moving in unison, near-term results are
bullish. Large gains tend to follow through with further strength (momentum
effect), but large declines tend to reverse and produce large gains (reversal
effect).
Now extend this reasoning–and these patterns–to entire asset classes.
This allows us to look at intercorrelations among instruments along the yield
curve, among energy markets, among currencies, and among international equity
markets. Extend the reasoning even further and you’ll see shifting
correlations *among* these asset classes, as well as within them.
Within and across multiple markets, the largest trading edges occur when
markets deviate from their normal relationships with one another.
When you realize this, it’s very difficult to go back to reading tea leaves.
Brett N. Steenbarger, Ph.D. is Associate Clinical
Professor of Psychiatry and Behavioral Sciences at SUNY Upstate Medical
University in Syracuse, NY and author of The
Psychology of Trading (Wiley, 2003). As Director of Trader Development
for Kingstree Trading, LLC in Chicago, he has mentored numerous professional
traders and coordinated a training program for traders. An active trader of the
stock indexes, Brett utilizes statistically-based pattern recognition for
intraday trading. Brett does not offer commercial services to traders, but
maintains an archive of articles and a trading blog at www.brettsteenbarger.com
and a blog of market analytics at www.traderfeed.blogspot.com.
His book, Enhancing Trader Development, is due for publication this fall
(Wiley).