One of the more interesting and useful market indicators that I watch is the relationship between stock prices (S&P 500) and the 10-Year Treasury Note Yield ($TNX). Under normal circumstances stock prices and bond yields will move together in unison as money sloshes back and forth between the two markets. For example, stock prices drop and investors flee to the relative safety of bonds, thus pushing the yields down. On the charts you'll see both stocks and bond yields moving down together. The inverse effect (stocks rising and bond yields rising) is also pretty much the norm.
This positive correlation is sometimes called the "flight to quality" and you hear about it a lot on big market down days.
Unfortunately, this intermarket relationship on a daily basis is very chaotic and you can't really use it as a timing system: it's just too noisy and prone to "gimbal lock" on the shorter time frames. However, I've found that it can be useful as confirmation in conjunction with other technical systems for spotting major market trend changes.
Here is a screenshot of a graph I keep of the 20 and 50 day correlations between the S&P 500 index and the 10-Year Treasury Note Yield (click for larger non-blurry version):
The black line is the S&P 500 index starting back in December of 2006. The red dashed line is the 20-day correlation with the 10-Year Yield and the green dashed line is the 50-day correlation. Generally speaking the average correlation over the 400-day period I'm tracking is quite high at around 0.7.
I won't go into much detail on how correlation works (that's what Wikipedia is for), but if you remember your Statistics classes, the Correlation is simply a measure of how closely two sequences are moving together. A reading of 1.0 means the two sequences are moving up together very closely. On the other extreme, a reading of -1.0 means the two are moving down very closely. A reading of zero means there is no correlation and things are just sort of moving randomly.
Anyway, in the above chart you can see that the 50-day longer term correlation tends to spend quite a lot of time wiggling about around that average 0.7 level. The shorter term 20-day correlation swings wildly around, sometimes reaching negative readings or going up close to 1.0 but also always returns to the 0.7 level. It often overshoots on the upside and the downside.
It is during these "overshoots" into extreme readings far from the statistically average value (0.7) that makes this relationship worth keeping an eye on. During all the major peaks and troughs during the past 400 trading days the 20-day correlation was showing extreme readings. In some circumstances the 20-day correlation seemed to forecast market turns and periods of flat returns (again, click for larger non-blurry version):
In the above chart I've drawn black arrows where the 20-day correlation appears to mark those times when the markets became extremely oversold or overbought, triggering a major inflection point. The yellow areas, where extreme readings in 20-day correlation were made, seems to forecast periods of flat trading or marks false bottoms.
The longer-term 50-day correlation seems to be more useful as a measuring stick for calibrating the 20-day correlation. The farther the 50-day is away from that 0.7 average level the more accurate the signals generated by the 20-day are.
As with most technical indicators, and especially custom indicators, this one has a fair amount of room for interpretation. I've tried to develop swing trading systems based around this indicator and discovered that the daily noise makes short-term entry timing suboptimal. It's been pretty accurate as of late but in backtests it can break down badly - badly enough to blow out your account if you let it. So you simply can't look at the correlation readings (OMG!!!! It's at +/- 1.0!!!) and go enter market orders for the next trading day. However, if you keep it in context and use it as just one part of a broader, more stable system then it can definitely be useful.
Comments