Banner for 2025 Financial Markets Conference

May 19, 2025

Research Session 2: Asset Purchase Rules: How QE Transformed the Bond Market

UCLA Anderson School of Management professor of finance Tyler Muir will present his paper "Asset Purchase Rules: How QE Transformed the Bond Market" for the second research spotlight of our annual Financial Markets Conference. The session, moderated by Symmetry Investments chief US economist Troy Davig, will examine quantitative easing and tightening policies and how each constitutes a dynamic state-contingent plan.

Transcript

Troy Davig: Okay, we're ready to start the next session; it's going to be with Tyler Muir. Tyler is the Donnalisa and Bill Barnum endowed term chair in management at the University of California, Los Angeles (UCLA). He's got a distinguished research record on the intersection of finance and macroeconomics, so I'm very pleased to have him here. And on that note, I'd also like to thank President Bostic and the team at the Federal Reserve Bank of Atlanta for putting together a great conference.

And so, with that, I'll turn it over to Tyler to give us an overview of the effects quantitative easing (QE) has had on the bond market since the global financial crisis. So, take it away! Thanks, Tyler.

Tyler Muir: Thanks a lot, Troy. And thanks so much for inviting me here. It's a real pleasure to get to present this work here. I should note, this is joint work with Valentin Haddad, who's at UCLA, and Alan Moreira, who's at Rochester.

So, let me start. I'll start with a picture of central bank balance sheets, across a few different countries over the last 30–40 years. No surprise; we've probably all seen some pictures like this. Central bank balance sheets have grown very substantially. They've primarily grown through asset purchases—things like QE.

And these big increases are not random. They occur specifically in bad economic times. For the US, you should be thinking about 2008 or 2020. For the euro area, you have the sovereign debt crisis increasing those, as well as 2020, too. We want to think about, in this paper, the impact of having these asset purchases be dynamic and state-contingent and used as a repeated tool.

So, the main thrust of the paper is, we're going to think about both QE and quantitative tightening (QT) as a dynamic and state-contingent policy. So, market participants are not going to view QE or QT as one-off, independent decisions. They're going to have expectations on the path of purchases and the path of the balance sheet, depending on the state of the economy. That's going to have a really significant impact on the level and dynamics of long-term rates, even during periods where you're not actually purchasing or significantly unwinding.

I think it's, in a lot of ways, analogous to conventional monetary policy, where we know that there's a lot of power in having monetary policy be cyclical, and market participants understanding a bit the reaction function there. But there'll be some different economic mechanisms in play here, and I'll try to highlight those.

So, let me think about this. I want to think about moving to a dynamic, state-contingent view of the world, but let me think about one-off or static effects of purchases, that I think are often a lot of the ways that people have thought about this. Well, if you absorb a lot of bond supply in a given period, you're going to push long-term bond prices up, and push yields down. So, you're effectively reducing the term premium, and you'll have lower yields after the intervention.

Now, if we think about a dynamic policy that's going to systematically do purchases in large amounts during bad economic times, well, you're going to generate two additional effects here. You're going to generate an insurance channel. The fact that you're raising prices higher than they otherwise would be, or pressing yields more than they otherwise would be, in really bad economic states, is going to be valuable to investors. That's going to generate an insurance effect.

It's also going to generate this crowding in from investors, in that if you're helping support the safety of long-term bonds, that's going to make investors more willing to invest today. They know they can sell those at a higher price during these really bad economic states of the world. It's going to be really potent, because those effects are going to be priced in, even ex ante. That's going to lead to permanently lower, and more stable, long-term yields.

So, I'm going to go through some evidence that's consistent with that. I'm going to have three things in my talk: I'm going to look at low frequency, the slope of the yield curve, and what long-term yields have done since we've begun QE so, effectively since 2008. I'll show you the slope of the yield curve looks lower, by about 100–150 basis points from what we might expect, and I'll talk about that and lower volatility and sensitivity to shocks in the Treasury market as well.

Then I'm going to zoom in, because there could be lots of things that are going on at low frequency that are affecting the yield curve. I'm going to zoom in on specifically QE announcements, so we can really understand the mechanism. The announcement responses that you see, if you accumulate them all up, they're exactly equal to this amount that the slope is lower in the low-frequency evidence. There's also strong evidence that with the announcement effects, you get sharp decreases in implied volatilities, even across horizons for long-term rates.

The last piece of the paper that we have sizes this up through a quantitative model, where we have a quantity-based term structure model, introduce a purchase rule, and see what the effects are. We estimate about two-thirds of the total effect of QE policies is through this dynamic and state-contingent piece, rather than just what has been bought at a particular point in time.

There's lots of related literature. I'll skip that, but I just want to say, we're obviously not the first to work on or think about this. Let me give you a couple of slides on the framework that we're using to think about this, and then I'll go straight to looking at the data.

So, we have a model: if you know this Vayanos and Vila model—this is a standard model to use for interpreting QE evidence—or the Greenwood Vayanos model. We're going to be along quite similar lines here. We're going to have two types of investors in the model: inelastic investors, that either move slow or just hold Treasuries or other long-term assets very passively, and then active investors that are going to be the ones absorbing demand and supply imbalances at any given point in time.

The term premium in this model is going to be made up of three things: the quantity of bonds or risks that these active investors are holding, how risky the bonds are, and then the price of risks that they're assigning to that. So, in equilibrium, active investors pricing bonds—they're going to require a larger term premium when they have to hold more of that risk.

So, a one-off announcement, if I just think about what immediate purchases do to the term premium, well, they're pushing prices up and yields down. Based on the slope of that investor demand curve, they're affecting this quantity of bonds, quantity of duration risk, that these active investors are exposed to. So, it's just going to be the expected reduction in net bond supply, times the expected risk of bonds.

When we move to—we're going to call it a "purchase rule," but really, I just mean purchases that are dynamic and stay contingent and occur during bad economic times. All three of these channels are going to move, so you're still going to get that static effect that I highlighted, but you're also going to get this insurance channel, if you're doing purchases at times when either risk is high, or people are particularly worried about risk so the price is high—you can get extra bang for the buck that way.

But by doing these purchases dynamically, you also can affect the risk of the bonds themselves, because you're raising their prices or lowering their yields during particularly bad economic times, and you can feed through to affect the price of risk that investors are going to require for holding that. So, more stable long-term bonds increase investors' willingness to hold long-term bonds. This is going to have a much bigger effect on the term premium from all of these three channels, and not just the first static effect.

So, let's look at the data. Let me look at the yield curve after QE. Here's what I'm going to do: I'm just plotting the slope of the yield curve—that's plotted here in blue—and then I'm going to fit the slope of the yield curve, using a regression based on three variables. I'm going to use debt-to-Gross Domestic Product (GDP) (technically, I'm going to use maturity-weighted debt-to-GDP, because we care about the total supply of duration that's out there), the Treasury bill rate, and the unemployment rate—but you can use another macro variable if you want. I'm going to fit that up until 2007 (so, before QE happened). I'm going to take those coefficients, and then I'm going to project forward in the red line what the predicted slope would have been.

And you can see the actual slope being substantially lower than what would have been predicted here. The gap there on average is somewhere around 150 basis points. So, that's evidence that the slope looks too low, relative to what these other variables would predict. Now, there are lots of other things that could have happened after 2008 that could have generated a change in the behavior of the yield curve. I want to highlight QE as a pretty plausible mechanism for doing that.

The green dashed line is what happens if I take out changes in long-term yields or changes in the slope, just on QE announcement days, and then replot what the slope would have looked like, taking out just those changes—and there's only about 10 or 15 of those days in our sample. That's the green dashed line; that lines up much better with this predicted slope based on this ex ante criteria.

So, that's pretty suggestive evidence that something about QE looks like it's responsible for this significantly lower yield curve. This is basically just doing the same thing in regression form, so I'll highlight maybe column three here, where we just have this dummy variable for the post-2008 episode; and you can see that negative lower slope during that period.

Now, why has that occurred? Partly that gap has occurred because in the pre-QE period, increases in debt supply—increases in this maturity-weighted debt-to-GDP—pushed up the yield curve, and that hasn't been true in the last 15 years. We've had a big expansion in supply without an increase in the yield curve.

Long-term Treasuries have also been less volatile in the post-QE era, so this is just saying implied volatility of a 10-year Treasury is down around 17 percent since QE. That's what's supposed to happen, working through this risk channel that we've highlighted, of dynamic QE.

That's just focusing on Treasury yields. What about transmission to broader assets, and to the broader economy? What we see is, there are similarly strong effects for mortgage rates or mortgage-backed securities (MBS) yields; we find a gap there of about 130 basis points, so almost complete pass-through there. Maybe that's no surprise. Part of what they were buying were mortgage-backed securities as well.

What about pass-through to corporate yields? We see a positive but imperfect pass-through there, so corporate yields look like they're lower by around 100 basis points, versus that 150; that means part of that lower slope is going into benefits that accrue to Treasury specialists, rather than just to long rates more broadly (although there's a substantial amount of pass-through).

We talk a lot in the paper about the zero lower bound, and whether that's the driving factor in what we see in this slope; we have evidence against that, but I'll refer you to the paper for that. And then, all these other trends that could be going on, post-crisis: differences in regulation, there was discussion about secular stagnation at the time, all of these things.

One thing that's going to help with that is to look at the international evidence and the international experience. So, the US is not the only one that's adopted this policy. What we're going to exploit here is that there's been a staggered introduction of this throughout the world. So, the US and the UK adopted these in 2008, 2009, Japan being earlier, the euro area later in 2015, with purchases targeted at specific countries during the sovereign debt crisis earlier, Canada and Australia, these are experiences that they've had later on.

So, we're going to exploit that here. This is showing that as a regression table. I'd prefer to just show it as a picture. So, this is now the slope of the yield curve, using all of that international panel data, where time, is when they're adopting these policies—so, think of it as an event study. And what you see is permanently lower, or at least—the x-axis here is in years—many years out, much lower slope of the yield curve by somewhere around 100 basis points. We find the same result, that one of the main reasons that that is lower is that supply has increased in those areas, but the slope has not, term premia look like they have not.

There's a bunch more evidence in the paper looking at higher frequency evidence of the Treasury market being less sensitive to supply shocks. We looked at these refunding announcements, as well as Treasury auctions; it looks like, again, the sensitivity to that news is dulled in the post-QE era.

So, now I've showed this broad evidence. I want to zoom in a little bit now on the actual mechanism by looking at the QE announcements. Here are cumulative event study responses across all the QE announcements that are in our sample; again, this is about 10 or 15 announcements, roughly speaking, and I'm plotting the cumulative change on yields across the term structure, for different maturities.

Focusing on the 10-year, cumulative response has been a lowering of about 140 basis points—so again, that's consistent with this low-frequency evidence that I showed you before. It's hard to rationalize that by just what the Fed had bought over this period, even though they've bought quite a bit of supply. If I take the static view of just what they've bought, and some estimate of what the slope of that demand curve is through these regressions that I showed you earlier, that would only imply a drop of somewhere around 30 basis points.

So, we attribute that missing piece to the value of this dynamic and state-contingent policy tool, that investors know that you will come in during these bad times and do even larger purchases. This view also helps explain why initial announcements—so, that was pooling all the announcements together—but initial announcements look like they're very powerful, and later announcements look like they're much weaker. Well, the initial announcements are, in part, going to be what you've announced that you purchased and the effect of those purchases. But they're also, in part, going to be that the market now understands that you have this new tool that you can use, and there's going to be value to them knowing that you're going to be able to use it again in the future.

The later announcements being much weaker does not mean that the policy itself is weaker. It's partly just, a lot of those expectations get built in—so, the next time you announce, the market's a lot less surprised by that, if they think that you're following this dynamic and state-contingent policy. So, it's a different way to view and make sense of what looks like a weakening announcement effect.

Here's one of my favorite pieces of evidence, actually, looking at option price responses—in particular, implied volatility responses—to these QE announcements. And so, what we're looking at is implied volatility. I'm doing, again, just these QE announcement dates, and I'm doing the cumulative percentage effect on implied volatilities. We've got 10-year Treasury futures—that's probably the most direct thing in the first three columns, but the option maturities there don't go out really that far.

Still, you can see implied volatilities cumulatively across those announcements, falling by somewhere between 30 percent and close to 40 percent—so, really big reductions in implied volatilities. Now, we could also move to looking at swaptions, where we're still looking at a 10-year rate, but now we can make the option maturity a lot longer. And so, you're finding similar evidence there.

What that means is, the market is now saying implied volatility of long-term rates—even quite a ways out into the future—looks substantially lowered by these policies. It's hard to think about how that happens, without investors thinking about this in a dynamic and state-contingent way, that you'll use this tool in a state-contingent way in the future.

Now, even though there's strong learning about these tools at initial announcements, there's also, of course, going to be updates to that. The market's going to learn about that reaction function over time. So, two examples of that: the taper tantrum in 2013, a fear of a rollback of QE policies. We saw yields increase quite a bit, and an increase in downside risk there. The second one is expanding the scope of asset purchases we saw in March 2020, an expansion into corporate bond interventions; and there we saw a huge effect on the corporate bond market as well.

So, I'll end that with this picture again. What we're saying is, by having this as a dynamic and state-contingent policy that the market understands you will use in crisis states, it's leading to a really big unconditional effect on the yield curve.

I want to spend just a couple of minutes walking through the model that we used to quantify this, and how we went about doing it. So, I outlined the model before. We have these inelastic or more passive investors that are either slow or not responding that much, and then this smaller set of active investors that are absorbing all the supply and demand imbalances. We're going to have two state variables in the model, which are going to be captured by this capital X. We're going to have the short rate—which is, of course, going to matter for yields—and the amount of supply that's out there.

We're going to estimate those up until the pre-QE period. We're going to calibrate the rest of the model to match some dynamics of the yield curve, and then we're going to see what happens in the model when we introduce this rule. So, the rule that we're going to start with, you can, again, play around with this. Obviously, purchases can depend on both the short rate and the term premium. Here, I'm going to have them effectively just depend on the term premium—that's the supply variable that's going to be driving all term premia variation in the model. We're going to calibrate that to the US experience, where they're buying essentially about a third of supply deviation.

So, on the bottom here I have changes in the Fed's balance sheet in blue, and the change in government debt in red, and you can see clear, positive correlation between those two. So, we're going to put that into the model and just see what happens, how much can it quantify.

The model is in blue before the rule is introduced. The red, dashed line is what the model produces after you put this rule in. The Xs are the data, so when we run these same regression exercises in the data of Treasury yields on the supply variable, you get this effect that in the post-QE period, yields look a lot less sensitive to supply than they did before.

What are the unconditional effects on the yield curve? Well, again, in the model, that's the blue curve; switching to the rule, it's the red dashed curve; and then the Xs here are the data, where I'm using those QE event study announcements to quantify how much it reduced yields across the term structure. In the model, we get about a 115-basis point reduction from introducing this rule.

And we can decompose that in the model, about how much is coming from the changes in the future risk, and all of these other things that come from having a state-contingent rule versus how much is coming from the actual amount that you're purchasing. We're getting about 75 basis points in the model coming from this rule, and only about 40 basis points coming from these realized purchases.

So, the model is useful in helping quantify that. It matches this reduction in volatility of bond returns, as well. So again, blue is before and red is after in the context of the model. It implies that the volatility of the 10-year yield should fall by about 10 percent. In the data, that looks like it's a little bit more, maybe. But we're qualitatively getting this thing that you reduce the volatility of long-term bond yields by doing this dynamically.

I don't want to take too much more time, and maybe we'll get to some of this in the Q&A as well, but we have this simplified view of putting this into a model, but we don't really know exactly what the states are that would trigger purchases. So, which are the sources of fluctuations in term premia or long-term rates that market participants expect the Fed to offset? Financial sector imbalances in market functioning, the rest of the world moving away from Treasuries, supply shocks—is this only something that should happen just at the zero lower bound?

There's a broader question here of, what are the sources of fluctuations that the Fed should offset—so, thinking more about optimal policy. And those are questions we're working toward, and getting to, in the research.

Let me conclude there, and we can move to a broader discussion. We're thinking of QE as a policy rule. It's now been a relatively routine tool used in a countercyclical way. Buying bonds in bad times creates extra safety, and crowding into investors looks like there's a huge value assigned to that and has a big effect on both the level of long-term yields and their dynamics. And then we have this simple model with the rule, that can help quantify this and shed some light on the channels.

Thanks a lot, and I look forward to the discussion.

Davig: Good! Very good. Thanks, Tyler—very clear and intuitive results. I think, for many of us, it does resonate, and it's consistent with, really, the way the Federal Open Market Committee has communicated this. If we look back at the statement on longer run goals and strategy, the second paragraph is very explicit. It says when rates are lower now on average, therefore more likely to be constrained, policy is going to be more likely to be constrained by the effective lower bound, and the Committee is therefore prepared to use its full range of tools to achieve its goals.

That statement has literally laid out what you have proposed and documented so carefully—and, I think, pretty convincingly. I think the question really is, is this unconditional, and we're always going to be seeing this effect? Or is it conditional on the effective lower bound (ELB)?

I have a few quick slides—if I can get those up—because I did a little exercise, just looking at post-COVID-19 dynamics of what we see in short-term markets versus term premium. So, the green, thin line here—what I do is, I construct...the probability secured overnight financing rate (SOFR) is going to settle below 1.25 percent in 9-15 months ahead. So, that's the range when you're in the zone, say, of the effective lower bound. It's not a perfect measure, but hopefully it conveys the intuition of what we're talking about here.

So, when Silicon Valley Bank failed, we see the probability of being in that zone—below 1.25 percent in a year—rise. It rose to close to 12 percent. I think it's pretty interesting, too, because we all know the Fed was basically hiking through this episode, and despite that increase in movement away from the ELB, the probability of hitting it was rising.

And I think it's interesting; this really backs up the points Tyler was making. We see the term premium; here's the 10-year, Kim-Wright-term premium in black. It falls during the episode when that probability of hitting the ELB is rising; and then, to borrow the great term from the last session, as the toupée was working in addressing the SVB situation, the probability of very low rates started coming down and term premiums started rising. So, it does feel like, as the proximity of policy moves away from the ELB, you should have higher term premium, it seems.

Another quick case study: so, in August 2024, the Sahm rule was triggered. Lots of discussion about recession risk, the probability of, again, getting into this "low interest rate" zone in a year was increasing, according to market pricing. The Fed cut 50 basis points in September, and during that sequence of events we see term premium rally and fall, and then we get some good data and other developments, and term premium starts rising.

When I look at this and think about this, in my mind, it's very much a function of the ELB. So, a question to Tyler is: Do you think about it as predominantly driven by these ELB episodes and risks of returning to the ELB, or is there also a sense in the market that the Fed will step in for—call it "market functioning QE," a little bit less well-defined, but that's always a discussion market participants have when there's any hiccup in Treasury markets.

So, if you had to separate out ELB effects from market functioning effects, away from the ELB, how would you divide those up?

Muir: I think that's a great point, and a great question. I love these slides and pictures. I think it's hard to quantify exactly how much is probably driven—I mean, if you want to think about very persistent, long-run variation in the balance sheet, purchases that are going to be bought and held for quite a long period of time, then I think you do start thinking about probably more the ELB. That's most of the experience that I would say we've had.

These market functioning purchases, though, I think, still can matter quite a lot. We were talking before about the UK experience in 2020, and the gilt market meltdown—where you're in a tightening phase and then switching around and actually coming in and purchasing to make sure that market doesn't completely melt down or dysfunction. Those types of things, I think, have a lot of value, even though they're relatively short-term.

How much of that is what's baked in that the market is really thinking about these market functioning shocks, versus return to this ELB? I don't think we cleanly know or can cleanly separate those—both in terms of their probabilities, but also how much value of this is coming from both of those pieces. Seems to me quite a bit from both, though.

Davig: That makes sense. Any time there is a hiccup in Treasury markets, it's an immediate discussion among market commentators about, is the Fed going to do something? So, it makes sense that it's affecting term premium.

Just one other question. I don't know how convinced I am of this myself, but I wanted to ask you: This is five-year, five-year, breakeven inflation; so, a major change did occur in 2012, when the Fed adopted its inflation target. Now, it codified what I think the market already understood about how the Fed conducted monetary policy; but on an unconditional basis, we do see the average five-year, five-year expected inflation drop, post the 2012 adoption.

And if you just think about yield-curve slope, it would make sense that if you have a credible inflation target, you're going to have a flatter yield curve than if you live in a world where inflation's high and volatile and the central bank lacks credibility. So, you know it's a little bit of an identification problem. I'm sympathetic to your results, but I wanted to ask you, to see if you thought this was also a major factor in the yield curve flattening, post-GFC.

Muir: For sure, there are other factors that are going on and driving the yield curve—both over that whole period, and at any given point in time. We're certainly not saying that there's nothing else going on. What helps me be comfortable with saying that QE was one force that seemed like it mattered a lot in the term premium and flattening out the yield curve, is, again, when you look just at these times when the news is really QE announcements; when we zoom in on those, we see these huge effects that all accumulate to this missing slope part that we had in the low frequency.

So, those are the parts that make me more comfortable that QE is certainly playing at least a big role there—certainly not to say that other factors aren't involved.

Davig: Yes, that makes a lot of sense; okay. So, if we think about, let's say, the outlook for term premium. There's a good chance—a fair chance—we're moving to a world with larger, more persistent supply chain shocks (supply shocks). So, in a supply shock world you have higher inflation, and lower growth. So, equity is lower, bond price is lower. In some ways, the hedging value of US Treasuries isn't there to the extent it was in a world where it was mostly demand shocks driving macroeconomic activity.

So, this is taking us a little bit afield from your exact paper, but I'd be curious: If we are in a world with a different set of shocks impinging on markets, what is the role of the expectations of Fed QE? How does that interact with this potentially new macroeconomic backdrop?

Muir: That's a great question. I think something that you said earlier, too, which is if we're farther away from the ELB, well, then there's going to be less value (if this is going to be used much more substantially when you're close to that). I would say I wouldn't expect any direct interventions to those shocks, per se; but again, if those shocks are causing some major form of market dysfunction, and some of the other tools aren't alleviating that—again, like the UK example, I wouldn't be that surprised if there were temporary purchases or measures taken to at least get market functioning back. And so, there's still—again, there's quite a bit of value from that. How much is hard to exactly say.

Davig: Yes, okay. Continuing with this theme, as we think, again, looking forward—a lot of fiscal policy changes are afoot. Lots of debt issuance, lots of questions about how the Treasury is going to manage this debt issuance. When you think and look forward, there could be episodes in the future where market functioning QE potentially may be appropriate, maybe not appropriate, where we're a long way from the ELB. To what extent do you think the Fed needs to think about this in an in-depth way in its framework review?

So, to what extent should the Fed—they have, in the current statement, codified: "We're going to act at the ELB with our additional tools." To what extent should they also address scope, when (if appropriate), on using these tools away from the ELB? Is that an important consideration, or is that something—

Muir: It seems to me like it is an important consideration, because market participants are forming their own beliefs about this and about when you're going to intervene and in what states, and what you really don't want is for them to think you're going to do something and then not do it (or really, vice versa). And it's hard in this case, unlike standard monetary policy, I would say—it's hard in this case to read off-market expectations from anything.

So, in the case of "are we going to raise rates, during the next meeting?" or whatever—we have the fed funds futures market, we have the one-year yield, the two-year yield. You can read off what the market expects the path of short-term rates to be. We have all kinds of surveys. It's a bit harder to know what the market expects about some of these interventions.

That does not mean that they don't have expectations about them, so at least trying to communicate that seems valuable to me. Also, trying to measure that through surveys—I know there's the primary dealer survey, where they do ask some questions about this—but through other means, where you can gauge what the market is expecting. So you know if you're going to surprise them or not, it seems to me to be important.

Davig: Yes; makes sense. How do you think about QT in the context of your results? So, the QT run-off the last couple of years, I think by almost any metric, has been very successful. I don't know if you had a chance to do any checking. Did the effects of the QT balance sheet run-off—did it align with your results, that used the period of QE to measure those effects?

Muir: Yes, it aligns pretty well. And so, in the model that I was quantifying there, we have both—so, it's the whole portfolio is increasing during certain states of the world, and then coming back down in other states of the world. So, we do have some way to think about QT in those episodes.

Now again, what our framework says is what's really important is, it doesn't matter so much whether or not you're doing QT. That will have an effect, but as for how yields are going to react, it's if you're doing it in a way that's really going to surprise the market. And so, this time around, that seems like it's gone quite smoothly. If we go back to the taper tantrum episode, that's something where you surprise people in how fast you're going to tighten, and that can certainly generate effects.

So, again, it's understanding what those expectations are that are built in, for how fast people think that you're going to tighten; and knowing whether or not you're going to—that's going to lead to surprises, or not.

Davig: Yes; makes sense. What about—I think we can gather the answer for this, but in terms of thinking about outright purchases versus twist operations, because your framework does incorporate a measure of duration, but did you pick up any difference between doing outright QE operations and doing twist-type of operations?

Muir: In the framework that we have, those things are basically the same. What matters is what you said: it's the total supply of duration that's out there, and so any way that you're affecting that, you're going to end up affecting the term premium. In the empirical evidence, I don't know if it looks like there's any really big differences between those two or not—for long-term yields, at least.

Davig: So, this raises, I think, a really interesting, I think, potential separation principle; so, Treasury is now doing buyback operations, and these are very narrow, limited—basically issuing bills to buy, say, off-the-run securities that you could argue are mispriced. But the tool kit is there now, and there have been comments made to that effect.

So, if there is a market functioning episode, it seems like if the Treasury has the ability to issue bills with a high capacity and step in and take duration out of the market. There's that toolkit—which is basically the same, then, for the Fed doing just outright QE. So, I don't know. Does it matter who does that? I think it matters immensely from a broader policy perspective.

Muir: Yes. In terms of the effects that you might see, I don't know if it matters so much; from a broader policy perspective, I definitely agree. But no, in the framework and the way that we're thinking about it, the Treasury could effectively do that and it's sometimes referred to as stealth QE, or something like that—changing the maturity structure of what it's issuing, or doing some buybacks where they're...and again, shorten or change the total supply of duration that's out there.

I'm not sure, practically speaking, whether or not that could move at the speed and scale for dealing with these market disruption shocks, as the Fed has done in the past. We remember the 2020 episode; the Fed bought $1 trillion within three weeks, of long-term Treasuries. That type of scale, for dealing with the market-functioning shocks, might be something that's still best handled by the Fed. I'm not sure.

Davig: Yes, that makes sense. It's also interesting, you've addressed MBS pricing, corporate bond pricing. But presumably, it affects all asset prices, equities, you can make a case for it affecting commodities. I'm just curious if you looked at any other asset prices and could see other spillovers from QE—because a big initial argument for QE was the portfolio balance channel (you're pushing investors down the risk spectrum, and so on). So, I'm just curious if we have persistently higher equity prices than we otherwise would have had.

Muir: I think there's two things going on: one, there is a lot of pass-through. We see that really strongly in MBS—partly, they're also buying MBS. But if you look at corporate bond yields, for example, you can go down the credit risk spectrum there. The largest effects that we're seeing in terms of pass-through is AAA, and then you see a little bit less pass-through, and then you see a little bit less pass-through. Now, it's also—it becomes a bit noisier when you go to things that are farther away, too, to cleanly measure that in the data.

But it looks like there is some role for this as actually adding specialness or safety to Treasuries themselves. Roughly speaking, the number that I had there was—that's about a third of the effect that you're seeing—is special to Treasuries. But about two-thirds of that effect is passing through directly to corporate yields as well—but there is a spectrum within there, too, as far as credit risk. It seems to be more passed through on the investment grade side and less passed through when you get down to high yield.

But yes, I think this is something that is going to affect long-term rates more broadly, affect equities as well, broadly affect the cost of capital for firms, have an effect on house prices—all of that should feed through, through those long-term rates.

Davig: Yes. It makes sense. Excellent. Okay. That is time. Thank you, Tyler, for a great paper—and I'm supposed to remind us that it's time for lunch, and we'll reconvene at 1:00 p.m.—so, thank you.

Muir: Thank you.