SocGen has become the latest in a long and illustrious line of (so far wrong) forecasters, to predict that the 30-year-old bond rally, unleashed by Alan Greenspan's "great moderation" and having gone through QE, thousands of rate cuts, and NIRP, is finally over.
As Bloomberg notes, Societe Generale SA is the most recent firm to go back to the drawing board, and is using a "different" approach in an attempt to come up with the right formula for why bonds are now massively overvalued. "Its bond team recently tweaked a long-standing macro model to incorporate two decades of bond prices from Europe, Japan and the U.K."
Based on that new model and statistical norms, there’s less than a 1 percent chance U.S. 10-year yields fall below 1.1% especially as the Federal Reserve moves to raise interest rates.
What SocGen is forgetting is that it was precisely the Fed's rate hike that sent the long end plunging, and yield curve going horizontal, on fears that the US - and global - economy is not ready for tighter US financial conditions. What it is also forgetting is that reliance on any historical models, and thus precedent, is laughable at a time when ever central bank is unleashing never before tried financial experimentation, and helicopter money may be next.
That said, here is why SocGen is convinced that the 10Y will not hit 1.1%: “We had to revert to a model-based approach to figure out how low yields can go after we broke below 1.4%,” a scenario that the firm didn’t think would happen unless the Fed did an about-face, said Subadra Rajappa, SocGen’s head of U.S. rates strategy. In our view, “it still doesn’t make any sense for the Fed to change its policy stance from tightening to even on-hold or easing.”
Cited by Bloomberg, Bruno Braizinha, the architect of SocGen’s models, says the new one implies a “fair value” for 10-year yields of 1.95 percent. That suggests Treasuries are still extremely overvalued after last week’s selloff. Yields were at 1.59 percent today, up from a record low of 1.318 percent on July 6. Alternatively, what the model may be implying is that global growth has been massively overestimated and misrepresented as political powers across the globe fabricate data to restore confidence among the population. But surely such a proposal would be considered nothing more than "conspiracy theory."
And then there is the model's own abysmal performance. As Bloomberg puts it very nicely, without hurting SocGen's feelings, "the fact that SocGen’s original model implied a fair value of 2.85 percent -- a level last seen in early 2014 -- reflects just how bewildering the bond market has been." Put in trader terms anyone who shorted the 10Y when it was at 2.85%, has lost about 15%, and that excludes the cost of carry.
It’s not hard to see why bond shops are looking for new methods: after all they have all been dead wrong. But before you mock SocGen, consider that at the start of 2016, strategists surveyed by Bloomberg projected that yields on 10-year Treasuries would end the year at 2.75%. That has proven to be... wrong. Today, the median forecast is more than a percentage point lower.
This is what SocGen's model says explicitly:
How rich are Treasuries - a guide for the perplexed
Our macro model for treasuries reveals a significant decoupling between Treasury yields and US fundamentals since the start of the year, yielding a fair value for the 10yT of 2.85%.
- The drivers for this decoupling are well understood: QE, a sustained bid for the back end of the US curve in a search for yield in a global low-yield environment, the recent risk-off environment on Brexit, the perception of the risks that a China hard-landing scenario may pose to the global outlook, and signs of a deterioration of domestic data in H1.
- The strength of these exogenous factors can be incorporated into our macro model by the inclusion of a global yields variable. The analysis shows a roughly 90bp impact of these exogenous factors on the 10y sector of the US curve, roughly 65% of the gap between our macro model fair value and the market, yielding an adjusted fair value for the 10yT of 1.95%.
- The maximum historical deviation from the fair value in this adjusted macro model was three standard deviations. This currently would correspond to 1.10% for the 10yT, where we expect significant resistance.
It is hardly perplexing to say that treasuries look rich – yields at the back end of the curve are around record lows after all – but quantifying by how much, and relative to what, and in what context is slightly more complicated. For a simple analysis versus fundamentals, we can use our macro model (a PCA framework that includes inflation, growth and employment macroeconomic variables, along with short-end rates and a vector for QE purchases) or a simpler approach where we regress the 10y on long-term CPI and GDP averages to gauge the level of richness. The macro model (see Graphs 1a-c) shows a significant decoupling between Treasuries and fundamentals since the start of the year, with the current level for the 10yT roughly 3 sigma rich relative to fair value (roughly 2.85% – see Graph 2).
The regression versus long-term CPI and GDP averages yields a fair value of 2.35% for the 10yT implying that the 10yT yield is currently 1.5 standard deviations rich to fair value (see Graph 3). This is less impressive than the signal from the macro model but still rather significant, with this level of richness not seen since mid-2012 in the context of QE. Macro approaches therefore yield significant rich signals for the Treasury market and confirm the perception of a decoupling between the Treasury yields and macroeconomic fundamentals. It is also important to note that the maximum deviation from the mean in this model was seen in 2008 when yields were two standard deviations rich to fair value. How low can yields go? In today’s market, and in the context of this model, a two standard deviation from fair value would correspond to 1.15% for the 10yT yield, which is where we expect to see significant resistance.
Of course, the core fallacy in this model is the assumption that CPI and GDP are accurate. Which is also why we stopped reading there.
Incidentally, a far more compelling case was made not by some iteration of a SocGen model, but by its global asset allocation strategist , Alain Bokozba, who also tries to make the case that the "bottom of 30Y rates is in view", however using a far simpler metric: trader positioning. This is what he says.
Peak in long 30y T-bond positions suggests that the bottom of 30y rates may be in view
Currently, 30y Treasuries are yielding a dismal 2.2%. Meanwhile, higher Treasury prices only seem to increase demand for them. Hence, hedge funds remain positioned for even lower 30y yields, and increasingly so. In September 1998 and December 2004, net long positions on 30y Treasuries also peaked (chart below). A subsequent bottom in long-term yields was reached in the two weeks and six months thereafter, during which yields fell some 50bp to 4.7% and 4.2%, respectively. Those levels marked the bottom for 30y rates during the four and 2.7 years thereafter.
Ok great, however just one question: how many trillions in 30Y TSYs had central banks bought in 1998 and 2004, and how many trillions more were they expected to buy in the future? And yes, Alain may be right in the short run but, absent the launch of helicopter money, or direct central bank monetization of deficits and welfare which inevitably leads to runaway inflation, we are confident that when rereading this post one year from now, we will all have a hearty chuckle at someone calling a 2.2% yield on the 30Y "dismal"...