I remember my academic colleague at The Ohio State University (now at Notre Dame), Paul Schultz saying “Why do you find fixed-income and the yield curve interesting?” I have always found the yield curve to be interesting … at least until The Federal Reserve hammered down the short-end with it zero-interest rate policy (ZIRP) and tried manipulating the 10-year Treasury Note yield through Quantitative Easing (QE) meaning The Fed’s purchase of Treasuries and Agency Mortgage-backed Securities (MBS). No, I still think the manipulated yield curve is interesting.
Here is today’s Treasury actives curve (green) versus the yield curve at the peak of the previous housing bubble in 2005 yellow). That is a 300 basis point shift as the short-end. And a 243 basis point shift for the 10-year Treasury Note.
(Bloomberg) — The yield curve is one of the most-powerful forces in the observable financial universe. While much of the price action that we see on a day-to-day basis may be driven by some sort of dark energy, the curve provides a highly visible lodestone indicating the state of policy settings and the likely trajectory of the economy. That being said, the curve is often misunderstood — a bear flattening often produces plenty of hand-wringing, when it’s the bull steepening that you should really worry about. In fact, referring to “the curve” itself is something of a misnomer — while different iterations of the yield curve often travel in tandem, sometimes their paths diverge. That has been the case recently, though perhaps not for much longer. The recent rise in two-year yields looks more than justified, as various fixed income models demonstrate in a roundabout way.
For the past year and a half or so, most of the focus on the yield curve in this column has been on the 5s-30s iteration. The rationale for this has been relatively straightforward: With the Fed funds rate locked in near zero for the foreseeable future, the two-year note has been moribund. As such, 2s-10s has really just been another articulation of the 10-year yield. And much like recent price action vis-a-vis my 10-year model, the curve briefly traded where it “ought” to in March before once again becoming too flat in recent months.
At least 5s-30s has had the benefit of containing a useful forward-looking component on both legs of the spread. Yet even as I type that, it is interesting to note that 2s-10s and 5s-30s exhibited virtually identical price action at virtually identical levels earlier this year. While they remain positively correlated, of course, a clear wedge has emerged between the two curves as five-year yields have broken decisively through 1%, pricing greater conviction that a monetary tightening cycle will fully emerge over the next half-decade.
Yet I am left to wonder about the two-year note. The eurodollar strip is pricing that the bulk of monetary tightening will come by the end of 2023, a period that’s now largely captured by the shortest-maturity coupon security. To be sure, the appropriate level for 2s is a function not only of the ultimate magnitude of monetary tightening, but when it begins. After all, a 150 bp hike in Q4 of 2023 carries very different implications for the current two-year note than a 25 bp rate rise every three months from Q3 of next year onwards.
It occurred to me that I could back out a model for two-year yields by simply subtracting the output of my yield curve model from that of the 10-year model. I had no real idea of what to expect from this exercise, but even with the proviso that short-end yields rarely stray too far from the policy rate, I was pleasantly surprised at how close the fit is from this “derivative” model for the two-year.
The question then arose, naturally, of what actually went into the calculation of this “model.” After all, knowing the formulae of the two constituent models — for the 10-year and the yield curve– should allow for the distillation of a separate equation for the two-year note. Because that sort of thing is more fun than unpacking more boxes, that’s how I spent a few minutes on Wednesday night. The outcome isn’t necessarily an optimal model for the two-year, but more of an accidental one.
A bit of high school algebra
For what it’s worth, the resultant formula is 2y = 1.24 * FDTR + 1.3 * (ED2 – ED6) -0.015 PCE CYOY + 0.08 * USURTOT – 0.25 * (10y average of FDTR) + 0.12 * (10y average of USURTOT) – 1.27. I am pretty sure that one could get similar results with a simpler framework; the notion that a 2% rise in core inflation is worth just 3 bps on the two-year yield, all else being equal, leaves me simultaneously amused and bemused.
What does seem evident, however, is that henceforth there is going to be considerably more signal generated from two-year yields than has been the case in recent quarters. As such, 2s-10s are going to be worth following again, just as much if not more than 5s-30s. Both nominal yields and the curves are clearly constrained by the notion that all of this inflation kerfuffle really is transitory at its heart, and that, with r* remaining in the gutter, the long-run lid on nominal policy rates is going to be extraordinarily low.
That’s probably as good a null hypothesis as any, and possibly better than most. That being said, if we’re still having a lot of the same inflation conversations a year from now, we’re gonna need a long hard think about whether some of the post-GFC lessons need to be unlearned. In the meantime, at least fixed income is interesting again. I wonder where the yield curve and the model will eventually meet up to shake hands again… -Cameron Crise
The yield curve will become more interesting if Powell and The Gang take their foot off the monetary accelerator pedal.