Wednesday, March 22, 2017

A look at yield curve compression

 - by New Deal democrat

Since the mid-1950s, an inverted yield curve has been perhaps the single most deadly harbinger of a recession in the next 1-2 years.  The most typical measurement has been the spread between 10 year and 2 year treasuries:

Typically these inversions have happened because the Fed raised interest rates in an effort to tame inflation deemed to high. Thus, because we are in a very low inflation and interest rate environment, I suspect this version of the yield curve is one of the most likely long leading indicators to fail to signal before the next economic downturn.  Most notably, the yield curve between short term and long term bonds never inverted at all between 1931 and 1954, as indicated by calculating the spread of the archival "long term government securities" data with 3 month treasuries:

My suspicion is -- and here unfortunately I do not have any old data to compare with -- that a compression at closer points along the yield curve is likely to be a more accurate signal in this environment.  To show you why, let's take a look at the spread between 30 year and 10 year treasuries (blue in the graphs below), 10 year and 5 year treasuries (red), and 5 year and 2 year treasuries (green).

Here is 1977 to 1997:

and here is 1997 to 2017:

The first thing I want to point out is that in advance of all recessions in the last 40 years, yield curve inversions happened across the board.  All three measures inverted.

Secondly, a compression of all three measures on the order of +0.5% or less was associated with stock market corrections (in the case of 1987, a crash!), but not an outright recession in the near future.  

Finally, the most likely measure to invert, including a number of "false positives" for recessions, was the 30 year minus 10 year measure.

With that in mind, let's focus on the last 5 years:

In this era of very low rates, the shortest term measure (5 year minus 2 year) has crossed the +0.5% threshold to the downside several times. The measure next out in the range (10 year minus 5 year) crossed once -- the middle of last year.  The longest term measure (30 year minus 10 year) has never crossed the threshold.

We can put this information together by calculating the average spread among the three measures, by taking each value, adding them together, and dividing by 3.  When we do so, here's what we get:

No matter how you look at it, at least compared with the last 40 years, no aspect of the yield curve is signaling any danger now.  

But if we suspect that it is not necessary for the yield curve to outright invert before the next recession (noting how little of an inversion there was in 2006 before the 2008 recession), then at least we can raise the caution flag in the event that either one or both of the following two things happens: (1) all three term measures declined below +0.5%; and/or (2) the average of the three term measures declines to +0.3% or less.  

It is certainly not a perfect work-around.  Although the data series are different, no compression at all between long term and 3 month securities occurred before the 1945 demobilization recession, nor before the severe 1938 recession (which appears to have been caused fiscally rather than by Fed action). 

But even so, if we think that this low interest rate and inflation environment will function more like that of the 1920s-early 1950s, then measuring yield curve compression across maturities will at least keep us on our toes.