Monday, February 22, 2010

Assessing Sentiment With Intermarket Correlations

When markets go through periods of fear, there is a tendency to jump out of riskier assets and into ones that are perceived to be safer. If that's the case, during periods of fearful sentiment, we should see higher intercorrelations among asset classes than during periods of relative complacency.

Above I looked at the absolute values of moving 10-day correlations for stocks (SPY), oil (USO), 10-year Treasury rates ($TNX), and the U.S. dollar (UUP). I computed how each of the above was correlated with every other one over moving 10-day windows, took the absolute values of those correlations, and then averaged all of those correlations into a single average correlation across the asset classes.

As you can see from the chart above, we tend to get high correlations during and immediately after periods of market correction, as traders and investors bail out of risk assets and then back into them. At relative price peaks in markets, especially when volatility has been lower, the average intermarket correlations have been lower.

This is a rough look at an important phenomenon, as it captures portfolio-related decision making across a broad range of assets. Only large institutions can affect these correlations, which makes the average correlation a useful tool in gauging the sentiment of active money managers.



ivanhoff said...

The average correlation is not only a useful tool to measure institutions' sentiment, but it also seems to move in a range: .20 to .70 and extreme values could be insightful of what is likely to come.

JimRI said...

Very interesting. I looked at the method of Jesse Livermore in his book, "How to Trade in Stock" and while I did not really figure it out, it seems that he was in fact making decisions based on correlations.