I am going to leave the deep thinking on the term premium to my friend, Brian Romanchuk. I share — and actually defer to — his view that the term premium is difficult to measure and that we need to take the various efforts with a grain of salt. But I am also going to work with what we have.
I realize that investors in the bond and (by extension) equity markets are interested in bond yields per se and not necessarily in the academic exercise of separating them into components representing the expected path of short rates and presumed compensation for duration risk.
Still, my interest, as often, is in assessing the importance of the portfolio balance channel of QE (and its going away), which is meant to operate on the term premium. My thesis, which seems to be playing out, is that QE operated mostly through rates signaling (and, in equities, behavioral channels) and that the portfolio balance channel was radically overstated by QE enthusiasts. And I am going to need a measure of the term premium to continue pushing that take.
One thing I think we do know is that the Kim-Wright term premium is pretty much useless in the QE era and probably recently as well. The problem with Kim-Wright is that it infers the term premium in large part from the gap between bond yields and the presumed risk-free yield that would occur if market participants shared the view of professional forecasters responding to surveys asking them to project the path of short rates.
The Kim-Wright term premium is currently available only to the end of 2016. Note the tight correlation and unit beta between it and just the 5-year note yield during the entire QE era, which overlapped with explicit rates guidance.
Very obviously, those survey responses became anchored, probably to Fed rates guidance, whose prominence was contemporaneous with QE, after 2008. You can see this in the totally bizarre relationship between just the 5-year note yield and the term-premium estimated for the 10-year maturity.
Not only is the correlation very tight, but the two concepts have the same range expressed in basis points. This is what one might expect if the survey respondents were anchoring AND markets were ignoring those professional forecasters. So I am going to just dispense with Kim-Wright. I am not confident I know exactly why that relationship is so tight, but it does seem disqualifying.
The failure of Kim-Wright may explain why Fed types have (seemingly recently, not sure) begun to emphasize the ACM measure of the term premium. For example, see former Fed Chairman Bernanke, here.) The ACM measure of term premium allows no role for surveys of short rate forecasts and instead infers the term premium by relating historical principal components of the term structure to historical excess returns.
The ACM term premium is shown at a monthly frequency with the last observation being April 17. QE dates are also shown at monthly frequency, which makes for some minor slippage. I define the QE era as starting with QE1, when the Fed broadened it purchases to include Treasuries and thus popularized the term “monetization”, although misleadingly. The Fed’s initial purchase of mortgages and Agencies in late 2008 is not generally seen as having been the beginning of QE, which is why I call it QE0. This is ironic because what I call QE0 hit a dysfunctional and particularly segmented bond market and seems to have “worked” as intended, even to my skeptical eyes. In Bernanke’s ” The Courage to Act”, the former Fed Chairman follows what I describe as convention in claiming that QE really began with QE 1. See page 420.
Leaving aside for now the question of whether such an approach is valid, the history of the ACM term premium is interesting. At first glance of a long history of the ACM premium, it looks to have fallen during the QE era. See left panel of chart below for that first glance.
But if you zoom in on the QE era itself, you see that the term premium on balance rose between 2006 and the end of the entire QE era and that it tended – on balance – to rise while applications of QE were ongoing. Moreover, it fell to a record low after QE ended and has barely responded to recent news of the Fed planning to sell assets more quickly than previously expected. The spike in late 2016 looks to have been about Trump, not QE-related sales. QE enthusiasts will say yes, but you are not doing the counterfactual. Whatever, bros. That can’t be your answer to all of these repeated failures of your basic thesis.
One bit of evidence that appears still to support the QE story still told by most Fed officials is that the ACM term premium (like the 10-year yield itself) rose steeply during the taper tantrum. Bernanke, along with the ACM authors themselves, attributes that to a rise of interest rate volatility.
Interestingly (to me), Bernanke is very clear that that had little to do with the portfolio balance channel or thus really to the taper itself. Bernnake (now?) shares my take that the market was responding to a more general turn towards hawkishness, including on rates, during that period. I find that interesting because it fits much more closely with a view I have been pushing and much less closely with Bernanke’s take on QE while in the Chairman’s seat at the Fed.
In contrast, the ACM authors themselves attribute the rise of volatility to QE and do not draw the distinction Bernanke now does. I may have more on that later, focusing in particular on the relationship between QE and vol. (Basically, don’t confuse being contemporaneous with causal.) In this post, I just wanted to show some pictures.
One final caveat. In my view, these pictures are very hard on the official QE story often told by Fed officials. I am aware that there are alternative interpretations of how QE worked. For example, Nick Rowe claims that QE worked through the liability side of the balance sheet and (I assume) signaling of future inflation intentions associated with that. I am not in that camp, but in fairness the pictures I show here are neither here nor there as regards it. I am leaning — as often — into the official and until recently consensus take on QE.