Gavekal Capital: Are Correlations Made To Be Broken?

Tuesday, June 10, 2014

Are Correlations Made To Be Broken?

For the last three years, the 10-year US Treasury bond has moved pretty much in lock step with Fed asset purchases.  Int he chart below, we show the 10-year UST (with a 4 month lag) and the three month accumulation in Fed assets. As the three month sum of purchases went from basically zero is mid-2012 to over $300 billion by mid-2013, 10-year rates went from 1.6% to 3%.  As the Fed has slowed the rate of increase in its balance sheet by about half (going from $300 every three months to $159 billion over the last three months), 10-year rates have fallen back over 40bps.  There has been an 85% correlation.


One of the big tests this year is whether this dynamic changes and rates can sustain higher levels after QE is over.  So far, we don't see much of change in the correlation between bonds and QE.  In the chart above we lag 10-year rates by 17 weeks (the calculated optimal correlation), so this suggest four more months of rates under pressure.  A break above 2.8% in rate would surely signal the possibility that rates have detached from central bank activity.

Next we extend the blue line in the chart above to model the completion of the taper.  By March of 2015, the three month increase in assets by the FRB will fall to zero.  We bring the 10-year bond back in here, showing daily rates.  Here again, we don't see much of a change in the relationship.  Unless these correlations break down, this suggests sustained pressure lower on rates all year.


We understand correlations are made to be broken, but recent data suggests the correlation is rising, not falling.  Here we look at just the last two years of data.  The correlation rises to 92%.