ebook essays pieces of the puzzle
Tuesday, April 2nd, 2013
fun with correlations

13 0402 fun with correlations

You hear a lot about correlations between markets, so let’s take a look at them for the S&P 500 versus three key indicators.  All use the same scale and have a zero line for easy comparison.  These are 120-day trailing correlations.

The top panel shows the correlation with the U.S. dollar; it has been mired in negative territory since the financial crisis.  In fact, the correlation has been more negative since that time than it was for all of the previous sixteen years.  Yesterday, Sober Look examined recent changes in this relationship and wondered whether it is a sign that the risk-on, risk-off pattern is changing.

Another abrupt departure from history took place in 2008 for the stocks/commodities correlation, shown in the middle panel.  Prior to then, you can see why commodities got their reputation as uncorrelated with stocks.  The number chopped around zero in a narrow band until it didn’t.  Oops.  Yesterday’s Wall Street Journal had a headline proclaiming “Stocks, Commodities Break Up the Band.”  The article included a correlation chart as well as one showing the divergent returns on the two asset classes of late.

The bottom panel is a surprise to those who think of Treasury bond prices and stock prices as inversely correlated.  Much of the nineties (and the eighties too) featured positive correlations between the two, as the decline in very high interest rates pushed stocks and bonds together.  Flipping the situation over, could a rise from very low interest rates do the same thing?

That’s an important question, but only one of many that you could ask the next time someone uses a correlation to help you frame the future.  For example, the mean-variance optimizations that underlie financial planning and portfolio construction use correlations — as do the algorithms that executed during the last blink of your eyes.  After looking at the chart, what do you think they ought to be?

In addition to quantitative calculations, qualitative judgments about how markets work are formed by these relationships.  It’s good to look at history to see just how temporary our constructs turn out to be.

And this is only a bit more than twenty years of that history, likely not a representative sample of years to come.  The next time someone makes sweeping generalizations about correlations and recommendations about how to behave as a result, you might show them this chart.  (Chart:  Bloomberg terminal.)