Yield Curve Inversion = Recession

It is common knowledge that the spread between ten year treasuries and 90 day bills possesses some correlation to recession prediction. But never before had I poured through weekly data to examine the correlation and relate the depth and length of inversion back to incidence of recession.



The last 46 years have seen yield curve inversions successfully signal an impending recession 6 out of 7 times within 14 months of their occurrence. Close scrutiny reveals a near perfect correlation between changes in the status of yield curve inversion and rate policy agenda changes before they occur. With exception to the 1966 yield curve inversion (induced by a round of government spending reductions that did result in an economic slowdown), all other inversions were shortly followed (by an average time of 10.3 months) by recessions. This is a worthy indicator to follow.



In evaluating a set of weekly closing yields on the 90 day T-Bill versus the 10 year note, I formed a criteria for evaluating yield curve inversions. I sought all periods of time during my sample set (of 1962 to 2008) where the result of the 90 day T-bill yield subtracted from the 10 year note yield was below zero as periods of yield curve inversion. Additionally, the length of time in weeks that each yield curve inversion condition persisted is also noted. The average depth of the yield curve inversion, in basis points (bp) is also noted. -50 bp equal 50 basis points of inversion (ie, 4% T-bill, 3.5% 10 Year Note). A positively notated number of +11 in the 6th inversion denotes a depth that is not inverted. It is included to show flat curve conditions are also correlated to recession.

Last, in parenthesis is a percentage number denoted as magnitude that is used to show the context of the number of basis points. It is simply the size of the respective mean or maximum inversion in basis points divided by the entire yield of the 10 year note for the respective sample week. 250 basis points divided against a 10% yield would derive a 25% magnitude. 25 basis points against a 2% yield would derive a 12.5% magnitude.



The above chart depicts two statistics. The red line shows the yield spread divided against the 10 year note yield, while the green line articulates the same yield spread number divided against the T Bill yield. The historically outsized result of the last several years is a result of long term debt offering many times the yield returns of short term debt, such as 90 day bills.

It may be worth noting that this gigantic disparity may have been a prime contributor to the credit excesses that are currently unwinding. A search for return with no fear of peril was a perfect setup for the many billions in losses we are today seeing. Most of this relative price spike was due to short term debt tracking yields on fed funds rates. In retrospect, it is easy to conclude there was already too much outstanding cash seeking a home, and that perhaps the loose monetary policy that followed the dot-com boom was almost entirely unnecessary to the degree it was employed. It is difficult not to wonder if the world economy already had more than enough cash in the system, since long term maturity yields gapped far away from short term debt. One could postulate that the price of long term debt securities would have followed short term securities up if there was no fear of inflation. That signals a system with more investable reserves than we perhaps realized at the time.

It should be noted that (usually correct) market anticipation of FOMC changes of policy guide the inversion status. With every inversion dip or return back above normal curve status, there is an accompanied FOMC direction change within 2 months.

The final rate hike in a string of 17 quarter point moves occurred on June 29th, 2006. With that, we’re left with an unresolved yield curve inversion that commenced not too long after on July 31st, 2006 and ended in the following May – the next FOMC move confirmed the validity of the inversion, acknowledging the risks of a slowing economy. It’s of interest to note that this 41 week yield curve inversion is the third longest out of the 9 total yield curve inversions in the 1962-2008 data set. As of January 2008, we sit 14 months away from the commencement of the inversion, approximately equal to the early 90s recession lag.

Another inversion is possible, and this double-dip effect has occurred with precedent, with two smaller inversions preceding the recession in 1968-1970. The jury is still out on the predictive value of the latest inversion. That makes a market.