Yield Curve Inversion Is a Signal, Not An Event
The name Arturo Estrella rings only specific bells whenever his name is brought up. Credited with “discovering” the predictive power of the yield curve, what Dr. Estrella really had done was write several influential papers during the 1990’s while working as an economist at the Federal Reserve Bank of New York. Their topic was how anyone might use the Treasury market to predict recessions.
There are today, as you may know, dozens of such papers littering the academic journals building further upon those works. In one place on FRBNY’s website, the Fed branch lists 114 of them which have entered the orthodox canon. It has become one of those things, a presumed truth everyone cites because, in this case, it is literally cited over and over again.
One of the 114 was a paper written in July 2001 by Andrew Ang of Columbia Business School and Monika Piazzesi of UCLA. The title is the usual dense jargon of statistics that I typically hold up in the way Ronald Coase once chastised the discipline for caring more about formulas than what’s going on in reality.
In this case, however, A No-Arbitrage Vector Autoregression of Term Structure Dynamics With Macroeconomic And Latent Variables is actually one of the few which does try to find a way to bridge that divide. We can see the yield curve works in the mathematical properties written up over the decades, but what does that actually mean is going on in the real economy?
The authors were refreshingly direct:
“The terms ‘short rate’ and ‘inflation’ are just convenient names for the unobserved factors. Another example is Knez, Litterman and Scheinkman (1994), who call their factors ‘level,’ ‘slope’ and ‘curvature’. Similarly, Dai and Singleton (2000) use the words ‘level,’ ‘slope’ and ‘butterfly’ to describe their factors. These labels stand for the effect the factors have on the yield curve rather than describing the economic sources of the shocks.”
In other words, when the bond market moves in such a way that it distorts the yield curve the public at least may have been conditioned to look at the yield curve as itself causing what may follow. It is instead an effect of something else going on in the economy. Inversion is a signal, not an event.
The rest of the paper is the typical econometrics, the fancy mathematics that are used to define a theory’s “evidence.” The results are, shall we say, interesting. The authors indicate that an inflation shock is the single best macro factor able to explain behavior on the yield curve.
Or, more accurately, parts of the yield curve.
“We find that macro factors explain a significant portion (up to 85%) of movements in the short and middle parts of the yield curve, but explain only around 40% of movements at the long end of the yield curve.”
Yield curve inversion, which is all anyone is talking about in this context, requires at least those two pieces. Typically, it means the difference between some short-term reference and a longer-term benchmark. In most academic settings, that has been the 10-year US Treasury yield on the one side and the 3-month T-bill equivalent yield on the other (3m10s). In others, and what has become conventional wisdom, same 10-year but now the 2-year note as the stand-in for short-term rates (2s10s).
We’ll come back to the long end in a moment. First, short-term rates are placed exclusively within the context of monetary policy. The Fed sets its fed funds target and that’s all there is to it. This is what we are all taught from the very beginning.
And that isn’t what Arturo Estrella found in his seminal paper co-written with Gikas A. Hardouvelis and published in The Journal of Finance in June of 1991 (Vol. 46, #2). In it, they lay out and test all the evidence you may be familiar with as it relates to the predictive power of the yield curve. But, they ask in Section III, what is the source of that power?
Most of what has come from this paper is taken from Section IV, the one following in which the authors document all the instances inversion and recession are clearly related. It doesn’t seem like there’s been much attention paid to the previous one which asks all the more pertinent questions, the answers to which today are so taken for granted they don’t seem to fit the actual findings.
Still on the coattails of Paul Volcker, very much in the shadow of 1979-80 monetary policies and more so the mainstream interpretations of them (the Fed, when it puts its mind to something, can accomplish anything and the markets all agree and obey), it was immediately assumed short-term rates and therefore the overall slope of the yield curve must be related to the main central bank setting more so than anything else.
Rising short-term rates as a matter of monetary policy is believed to be tantamount to monetary “tightening” and therefore an increase not only of nominal rates at that end but also real rates (“in the presence of price rigidities.”) Those are a signal in the economy of low opportunity and therefore lead to less investment and overall output.
Therefore, if rates at the long end are indifferent, the yield curve inverts and we have an explanation for the signal (inversion) and the consequences (recession).
When testing (using econometric methods) rates associated with monetary policy and proxies for them, such as the 3-month T-bill, the authors find, however, that the predictive power of the yield slope lingers onward far, far longer than you would expect believing monetary policy the source of all movement here.
“These results indicate that the information in the slope of the yield curve is mostly about variables other than current monetary policy.”
Ouch. To work their way around these well-grounded findings it has become convention to claim the entire Treasury curve must be under the influence of the FOMC – no matter how many repeated conundrums that may lead to. Maybe not 100% short to long, but enough all up and down the curve such that monetary policy must be the single biggest factor according to this way of thinking (recall Alan Greenspan’s analogy of the 10-year Treasury yield being like a series of one-year forwards all tracing back to that first one which he declared unequivocally under his thumb).
The implications run in both causal directions. First, the policy error. It is largely assumed that if the short rate rises above the long rate that’s because the Fed has made a mistake in taking its “tightening” too far. As the 1991 paper sets out, the chain of events is like a monetary contraction which leads to an eventual recession.
The long end in these mainstream scenarios is treated as completely detached, just kind of sitting there over in the corner not doing anything while all this other highly pertinent stuff is taking place.
That’s not exactly what Estrella and Hardouvelis had found, particularly when taking things a step further and thinking about them from the perspective of future monetary policy – what we talk about today as the expected future path of short-term interest rates. In short, sure, the Fed can be important but not as important as you might think – at least in this narrow setting of yield curves relating to recessions (and I argue that’s just the beginning)
What about the long end’s role in all this? Remember Ang and Piazzesi found that they couldn’t find out. Bonds are nearly a complete mystery even when Economists do attempt to demystify them.
But a monetary contraction that leads to an eventual recession, or any economic downturn, doesn’t have to relate to monetary policy. That’s what these Economists never once consider. Nor do they factor how there are the same considerations at the long end of the yield curve much as there are on the short side.
Effective monetary conditions are treated exclusively as monetary policy alone. The idea that markets let alone the monetary system might act independently is anathema to conventional thinking (despite, you know, 2008). Even when made to consider the behavior of long rates they are more often just dismissed as errors (remember also 2018’s “strong worldwide demand for safe assets.”)
No coincidence, then, that the name Arturo Estrella started popping up in the news late in August 2019 rather than in August 2018. Suddenly, the world was paying attention to the yield curve. It would invert as that particular month drew to a close, amplifying recession fears just then becoming serious for the first time.
CNBC brought Estrella onto the network where on August 22, the 2s10s right at zero, the yield curve guy said the dreaded “r” word.
“It’s been 50 years and 7 recessions with a perfect record. It’s impossible to be 100% sure about the future but I’d say the chances of a recession in the second half next year are pretty high.”
While it was technically true that the 2s10s were about to invert, the entire yield curve had been experiencing various levels of inversion for quite a long time before August, including the 3m10s. The 10-year yield had dropped below the 3-month bill all the way back in May 2019.
Long before either of them, there had been a “wrinkle” in the yield curve, a little blip in one crucial section of it dating as far back as December 2018. That had been the final month of Jay Powell’s “rate hike” cycle so naturally it was assumed the Fed was potentially facing a policy error.
And if that was the problem, quite naturally rate cuts would be if not a solution then tremendously helpful.
What was truly interesting about those first Treasury curve distortions was where they were happening on it; in the 6m, 1-year, 2-year areas. The short middle. These were precisely where the eurodollar futures curve has been inverted for more than six months before Treasury’s.
Eurodollar futures are the anticipated future path of short-term interest rates and this is where cause and effect really get screwed up in conventional thinking. Those investors (read: banks) buying up eurodollar futures at those particularly maturities aren’t buying them on the premise Jay Powell is in charge.
Yes, 3-month LIBOR which is what is used to settle eurodollar futures contracts is closely aligned with the Federal Reserve’s monetary policy. But future monetary policy may not be what present-day policymakers envision. Indeed, that is the whole point here.
Jay Powell spent all of 2018 telling everyone who would listen that the Fed would be raising its rate corridor gently if not more aggressively into the future. And the eurodollar futures market had spent at least half of 2018 saying he was wrong; that he would end up reducing rates whether in 2018 he agreed or not.
The importance of that initial little wrinkle of a distortion in the Treasury yield curve was in the one confirming the interpretation of the other. Back then it was high unthinkable. Today, we stand in the reality of it actually happening this way.
The difference wasn’t about who sets the short-term interest rate, the Fed does, rather it was about why it gets set the way it does at various times. Unlike what you’ve been taught and what gets reinforced every day in every way, the Fed can get it - does often get it - wrong even about where its very own policies might be concerned not that far into the future.
The predictive power of the yield curve lies in that direction at least in the reality we actually live in. And the yield curve itself pulls all those pieces together; central banks not understanding what’s going on (not a policy error), monetary contraction, and then the consequences of it. The long end, contrary to popular assessment, is very much in line of both monetary and economic conditions if not exactly in the same way as the short end.
In fact, we can actually observe this relationship in the one place the public was being reminded of at around the same time Arturo Estrella was most busy with media engagements. The yield curve inverted for the public anyway (2s10s) in late August. Mere weeks later the repo rumble.
As any of the 114 academic papers can tell you, the yield curve didn’t cause the repo mess last September. Those two were, however, related. The same thing going on in the one was responsible in large part for how the yield curve was being twisted and distorted, and long before last August.
According to the Treasury Department’s TIC data, there has been another large spike in repo transactions going on between US banks and unknown offshore foreign entities. In this most recent case, the biggest outbreak to date, dates back to...the last two months of 2018. These are classified as resales in the figures (because they are reported from the perspective of those US banks who are lending cash and accepting collateral, therefore “buying” the securities and “reselling” them back to their owner later) and over the last decade they correlate very strongly to the price of US Treasury bonds particularly in the middle as well as the long end of the curve.
The “strong worldwide demand for safe assets” can be matched up repeatedly with potential problems with offshore repo markets, which shows up in TIC data at these inopportune moments (always on the edge of globally synchronized downturns, in some places recessions) and at the very least correspond to sharp increases in the price of Treasury instruments.
Since UST’s are the most pristine form of US$ repo collateral, it doesn’t take a huge intuitive leap (but it does require a little bit of work, as I’ve done in other places) to connect them.
Thus, to put the last year or so into these terms: the rate hikes were not responsible for the yield curves “partial” inversions, it was more so the work of the repo market which kept long-term yields lower than they might otherwise have been. The eurodollar futures market was betting that such liquidity irregularities would become problems globally that would be severe enough to cause economic problems eventually leading to a reversal in monetary policy the Fed had neither expected nor was it prepared for.
And then the globally synchronized downturn which followed right on cue.
The question on everyone’s mind, though, continues to be recession. US recession specifically. When the part of the yield curve the public pays attention to inverted in August, it was incorrectly assumed that was an all-or-nothing, brace-for-impact warning shot. I mean, Arturo Estrella.
But then the yield curve steepened back out and “the” inversion went away. The 2s10s pulled up from -4 bps to as much as +34 bps. In large part because of this, sentiment has swung entirely optimistic. All predicated on this idea of a recession scare that the US economy easily weathered with the skillful aid of Jay Powell’s rate cuts and repo operations.
Perhaps. Then again, the yield curve “wrinkle” is back already. In those same places on the yield curve, the longer rates are below (some of) those in front of them. The difference between the 6-month bill and the 1-year, which looked like it was going to shoot positive early in November 2019, has gone back and remained underwater (inverted) ever since.
More and more it has been joined by the 1-year to 2-year space, mimicking the behavior of the curve in late 2018.
Behind that backdrop, TIC says US bank resales offshore continue to rise (through the latest figures for November 2019) and remain extremely high. The strong worldwide demand for safe assets unabated, which in all likelihood is why the middle to longer end of the yield curve really hasn’t budged all that much from its recent lows – set in August.
From the commentary surrounding the bond market over the final four months of last year, it sounded as if the thing had had a complete change of heart. It had earlier warned of a recession, instead changed its mind and sounded the all-clear to everyone’s great relief. The NYSE and Jay Powell most of all.
That’s not really what happened. The mainstream came in, as usual, late to the story and from its conventional view grounded in academic scholarship exposed to huge gaps in its framework misread the situation; a situation, it must be pointed out, that today isn’t all that much different. As January 2020 drags on, it seems less different by the session. Despite trade deals and, yes, three rate cuts. The curve continues to be distorted and twisted even if it isn’t “the” inversion anymore.