Keynote at the NBER Long-Term Asset Management meeting, Spring 2026.
Tyler Muir, “Market Macrostructure: Institutions and Asset Prices” (joint work with Valentin Haddad).
This is a lightly edited transcript of the talk and Q&A. Verbatim filler words and false starts have been cleaned up; substance is preserved. Section headers correspond to the chapter timestamps above.
All right, thank you so much. Thanks a lot for that introduction, and thanks for having us on. We're really happy and excited to talk about this work. This is joint work with Valentin, who unfortunately couldn't be here. It's based on an Annual Review paper we wrote called “Market Macrostructure: Institutions and Asset Prices.” We want to lay out a research agenda that a lot of people in this room — and in the profession — have been contributing to, including ourselves, and think through it a bit.
I'll start with three large shifts we see in financial markets and in macrostructure (I'll define what I mean by “macrostructure” in a minute):
(1) Over the last 30–40 years, there's been a huge rise in passive investing. It basically didn't exist 40 years ago, or was very small. Recent estimates differ, but some now suggest passive has overtaken active, more than 50% of the market.
(2) Central banks and quantitative easing. Twenty years ago, central bank balance sheets were quite small. Since then they've been key active players in bond markets — true in the US and in a number of other developed economies. They've grown to roughly 40% of GDP and very large fractions of bond markets.
(3) CIP deviations — arbitrage spreads or mispricings — were very low before the financial crisis. The crisis hit, financial institutions got into trouble, those spreads blew out, and due to subsequent regulation on banks, those mispricings have persisted for a long time. (This picture is from Du, Tepper, and Verdelhan.)
Those are three big changes happening in financial markets, and we want a framework to help understand them.
Now, we can try to understand some of these things through the lens of neoclassical finance — perfectly competitive markets, rational agents, no frictions, maybe even a representative agent. In that view these things don't really matter. Anything the Fed does with QE isn't changing overall risk in the economy; it's reshuffling, so it doesn't really affect bond yields. A large fraction of investors shifting to passive and being price-insensitive is no big deal because other investors will pick up the slack. And if banks are devastated, that's irrelevant for prices because other investors step in. There's always an offsetting effect.
That's not how we think the world works. The market-macrostructure view, supported by a lot of evidence, is that these big changes show up in prices and price behavior. With QE, yields drop substantially when announcements are made. With the rise of passive, Valentin has a paper with co-authors arguing it makes the market more inelastic, because you've taken out a big chunk of investors that are no longer price-sensitive. And there's a lot of evidence that banks being devastated has pretty large effects on risk premia. These are macro shifts, and they have relatively big effects.
Let me define market macrostructure more formally. We think about it as the broad organization of financial markets into key players and institutional features, and we want to understand how this affects the level and dynamics of asset prices. Those three examples are my working examples through the talk, but you can imagine many others. So: who are the key players in any given asset market? Then we want to know something about how they trade — by which I mean their objective function (what they're trying to do), their constraints, and the shocks they face. Finally, how does this shape prices? A lot of this literature has brought in lots of information about how these investors trade and what they care about — some of it actual quantity data, some just things we know about how they behave.
I'll give you a simple framework with two conditions for macrostructure to matter, so we're on the same page. Then I'll go through the three examples — intermediaries and crises, the rise of passive investing, and central banks in asset markets — emphasizing results we and others have found, but also a relatively common research approach across them. Then I'll take stock and talk about how to apply this to other interesting questions.
Our approach: we like to say we'd rather be approximately right than rigorously wrong. We're not going to write down a single model of every single thing in the world — that would be the rigorous extreme. We understand the world is more complex, with a bunch of forces going on. At the same time, we don't want to abandon all rigor and structure, ending up with case studies and no common thread. The goal is to relax enough to include complications, but still have common structure.
Households and institutions both invest in asset markets. Households invest directly, but they also own and so indirectly invest through institutions. Sometimes that's explicit, sometimes implicit (e.g., a central bank). I embed the idea that household demand can depend on institutional holdings — that lets us check the “it's just reshuffling” case. Market clearing means the two demands have to add up to supply, which pins down price. (This follows the framework in our JF paper; lots of others have written similar models.)
What do I need for macrostructure to matter? Two conditions.
Condition 2 (let me start here): “stuff has to happen and not be completely undone.” Stuff = institutional shifts — a financial crisis where banks bear less risk; QE/asset purchases by the Fed; regulatory changes affecting institutional demand; passive growth.
Condition 1: that stuff actually ends up mattering — it's not all undone in financial markets. I'll have less to say about this today. Reasons: trading costs, expertise, plain old inertia or inattention, behavioral or other frictions that prevent perfect offset.
These aren't completely mutually exclusive, but roughly:
(a) Natural experiments — exploiting shocks for causal effects.
(b) Micro foundations — writing down models with specific frictions and rich dynamics. Think He–Krishnamurthy intermediary asset pricing, or Vayanos–Vila / Greenwood–Vayanos for QE and bond supply.
(c) Quantities and demand — flexible, data-driven approaches using portfolio holdings across investors. Work by Ralph Koijen and Motohiro Yogo, and by Xavier Gabaix, fits here.
Some papers fit multiple buckets, but that's roughly how I'll think about it. Now let me go through the three examples and show what I mean by “incorporating information” and a “consistent research approach.”
I'll start here partly because I know it, partly because it's where some of this literature got going. This picture is from my own QJE paper looking at asset prices around financial crises — large failures of financial institutions, banking panics — using historical data across many countries going back quite far in time (you need a lot of episodes). The argument is that financial crises are special in what they do to risk premia. The change in dividend yields and credit spreads spikes in financial crises, when the financial sector is in trouble. That's a sharp place to look for the effects these models predict.
Of course, financial crises are also bad macroeconomic times, so you want to compare to other bad macro states — deep recessions, or wars (the kind used in the disasters literature). When the financial sector specifically gets in trouble, risk premia rise dramatically.
That doesn't mean these models are only relevant for crises. In a paper with Matt Baron, we found this isn't only a crisis effect — it applies at lower-frequency, normal times. We went across countries with detailed data on bank balance sheets and securities firms. When intermediaries substantially expand their balance sheets, future risk premia are low; when they contract, premia are high.
We can also use information about where these effects should especially show up. Intermediaries matter more for some asset classes than others. In a paper with Valentin, we find effects are strongest for asset classes where intermediaries are more the key player — using holdings data from Flow of Funds and BIS, plus 10-Ks where intermediaries report risk exposures.
Last one I'll mention here: not just 2008 but what's gone on since. CIP deviations were roughly zero before the crisis, blew up in the crisis, and have persisted. The information being brought in is that post-crisis regulation made bank balance sheet space scarce, so engaging in these arbitrages got harder — which is when these arb opportunities open up. There's an even nicer version: regulation that binds especially at quarter-ends, and you see the spreads blow up right around quarter-end.
Pattern in this example: we start guided by theory that tells us how intermediaries should matter for asset prices versus standard mechanisms. That gives us a place to look. Then we use the fact that we know a lot about intermediaries — the episodes when they're in trouble (financial crises, LTCM), which assets they matter for (more for CDS, MBS, derivatives than the stock market), what regulations they face — and combine these to develop a quantitative understanding of how institutional frictions affect asset prices. The same approach applies to the other macrostructure shifts.
A side note: post-crisis, banks have pulled back — partly due to regulation — even in things like C&I lending, so new players step in. Mutual funds have stepped in particularly in the corporate bond market, and we saw fragility from that during COVID (a nice paper on this). Private credit is another — we'll see a paper later today. We're not totally sure yet what fragility might be there.
Does it matter that 50% of the stock market is passive? Here's that plot — this one is from a recent paper by Brightman and Harvey. I don't want to say it's the only estimate; if you have one, I'm happy to cite you. We all know the rough number doesn't matter — passive has grown substantially.
There's the intuition: if everyone is passive and price-insensitive, no one's around to make prices efficient. The question: as passive grows, do remaining active investors compensate, or does it actually change financial markets?
Valentin's work with Paul Huebner and Erik Loualiche addresses this. (He's not here; he told me if I misrepresent something, that's okay, he'll give me a break.) What they find is: there's some response, but not perfect offset. The equilibrium result is markets become roughly 20% more inelastic. They use direct information from how institutions trade and respond to each other to estimate this. So the rise of passive has made markets measurably more inelastic.
They're not the only ones. You can focus on trading data (Koijen–Richmond–Yogo have a paper). You can use natural experiments like index inclusion. You can go directly to information content in prices. The similarity is using information about trading and what we know institutionally to quantify the effect.
How has QE affected bond markets? Here's the central-bank-balance-sheet-to-GDP picture for several developed economies (the US in blue). They've peaked around 40% of GDP and become a new dominant player. True in levels of share of public debt held, and staggering in flows at times — most salient, in March 2020 the Fed bought over a trillion dollars in long-term Treasuries within a few weeks. A very big new player.
Increases in purchases aren't random — they tend to happen in crisis episodes (2008, 2020 for the US; the sovereign debt crisis for the euro area).
Event studies — yields around QE1 announcements, directly based on Annette Vissing-Jorgensen and Arvind Krishnamurthy's work — show pretty large effects. The 10-year Treasury moves about 110 basis points (yields drop). Only five announcements, with a couple big ones doing the moving. Effects on MBS too. There's a term-structure pattern: long-dated maturities respond most. (For some reason I made the maturities decreasing on the graph, which looks weird — but it's unconventional policy, so I went with an unconventional graph.) These effects are big — 110 bps in a couple of days is a big deal — and that's not just relative to magnitude; QE1 wasn't a huge share of the bond market. So we have to think through why it can have such a large effect. Very different from the complete-offset story.
In work with Valentin and Alan Moreira, we argue this isn't just an announcement effect — it's a persistent shift in long-term yields once the new player is introduced. Let me explain the graph: blue is the slope of the yield curve; red is a predicted value from running the slope before the QE era on debt-to-GDP and macro variables (unemployment, the short-term rate). Good fit pre-QE. Then post-QE you see a big divergence — the actual slope and long yields look low relative to that prediction. Most of that comes from the supply of government debt increasing while the slope hasn't kept up. The big persistent level shift matches the announcement effects exactly: the green line shows what the slope would have looked like if you omit changes in long-term yields just around QE announcement dates — and you get back to the red line.
How do we think about it? Through a model with a new large dynamic player. The logic, in reverse: we know the Fed does QE in bad economic times (2008, 2020). If those purchases push bond prices up, bonds appreciate more in bad times. Roll back to ex ante: if investors know the Fed will step in in those states, bonds become ex ante safer, demand rises, yields fall — permanently. You know that if you get into trouble, you can sell to the Fed at a high price. So the trading rule matters, not just the individual purchases. Investors anticipate the strategy, and it has effects even when there's no actual buying happening.
In another paper we tried to bring in more information about this state-contingency by looking at option prices — specifically, the announcement of corporate bond purchases in March 2020, using options on corporate bond ETFs. The effect of that announcement looks like adding a put option to the market — that's what the slope/shape looks like. So what investors found valuable was exactly the idea that the Fed would step in and buy more if things go south. We use this to infer the market's state-contingent expectations of central bank purchases.
What more information can we bring in? We can look at what central banks and investors say, and check whether they're thinking about this the way we think they are. This is new work, not totally ready yet. A quote from Bernanke around the taper tantrum: “It's critically important in our communication that we emphasize state contingency.” Exactly the thing — the market really values state contingency, when QE/QT will happen in bad/good times. From the investor side, we started looking at mutual fund investor letters, and you see them saying things like “we tactically extend duration as we expect the Fed to continue purchasing mortgages.” They think the Fed is going to do something and they change their portfolios now.
The broader goal of that paper: the Fed does a lot besides changing the short rate — QE and balance-sheet policy; liquidity and lending facilities; discretionary regulatory choices on banks. We try to take all those documents and categorize across these sources. The big finding is that the Fed has been a lot more involved in markets generally since the financial crisis — not just bonds — through these various facilities, swap arrangements, regulatory changes.
There's consistency across these — the same playbook a few times. For financial crises: we know when intermediaries are in trouble and where to look; we know which asset classes they matter more for; we know the regulatory changes affecting them. For passive investing: we know who's passive, which helps isolate price-insensitive demand. For central banks: we observe what they've done; they tell us what they will do and like signaling about the future; we can use all that, plus equilibrium models, to trace effects on yields and term premia.
The argument is to exploit institutional information while still using theory as a guideline — not abandoning structure. The workflow: theory tells us what institutions should matter and why; holdings and prices tell us who knows what and how they trade; trading strategy/objective functions/constraints exploit outside information about their condition, the rules they face, mandates, communications; then quantify and validate using counterfactuals and natural experiments.
I focused on three questions, partly because I've worked on them. There's a lot of other areas applying similar methods, with lots of open questions. When you look across countries, there are big differences in macrostructure — the key players and institutional features differ a lot, and there's certainly work doing this. Annette Vissing-Jorgensen and Robin Greenwood have a nice paper that fits both the pension-demand bullet and cross-country differences: countries with a really large pension and insurance sector have a special demand for very long-term assets, and yields at the very long end are lower when that sector is big.
Other areas: private credit (a lot of interest given growth — outstanding questions about fragility); pension demand and portfolio changes; sovereign wealth funds (work by Jinyuan Zhang, my UCLA colleague, with a UCLA PhD student) — they're trying to think about whether SWFs' objective functions are just risk and return or more geo-economic; international finance and macro (the “...” isn't because the area is wide open and untouched — it's because there were too many papers to cite, and we saw some yesterday in the asset pricing meeting taking this approach).
To close: we want to think about market macrostructure as the broad organization of markets and how it affects the level and dynamics of asset prices. In our view these are first-order forces, and we have good tools to make progress. The slides are online — on the conference website and on my own website. Looking forward to the Q&A. Thank you.
Question: A very inspiring presentation, thank you. I have a question about going forward and about our students. What I'm finding is that students are afraid to go into this area. Why? Because empirical work is difficult — it's just hard to identify anything. You need a mechanism, instruments. So what's your view? How can we make it more inviting for PhD students? Because, as in your opening quote — that quote was excellent, “approximately right rather than rigorously wrong” — exactly. So what's your view?
Tyler Muir: Should I collect them or just answer? Yeah. My view: it is hard. Macro finance in general is hard. But I wouldn't characterize most of what I've done as that hard — I'm maybe not smart enough to do the really hard versions. There are simpler approaches where you can still make a lot of progress. Take the example: financial crises are times when intermediaries should really be in trouble, and so if intermediaries affect asset prices, it should show up there. That's not relying on instruments and complicated identification — it's a relatively simple approach. That was true for a couple of the things in the talk. It will remain hard, but part of our goal is to inspire PhD students to do this kind of work, and then maybe they can also make progress on the parts that are still really hard.
Question: The text on your slide takes the institutional features and broad organization as exogenous — as the starting point — and goes from there to the level and dynamics of asset prices. In some cases that makes a lot of sense — bank regulation after the financial crisis. Yes. But particularly with the rise of passive, I find that unsatisfying. Why have we had the rise of passive? Partly because of all the evidence that active mutual funds charge fees that don't leave anything for retail investors. So that's perhaps endogenous, right? What do you think?
Tyler Muir: Completely. I agree. That goes back to the comment that it's hard. There is a tension. My own view is that just because something is hard doesn't mean we should ignore it either. We want some balance. I agree — taking those things as exogenous is a bit of a stretch and depends on context. There are still features we can use — event studies, natural experiments — where the effects we're talking about show up and we get cleaner identification. But yes, in general, trying to think through why these things change would be a useful step. We haven't done all that much of it, I'd say.
Question: Super interesting as always. I want to ask about a concept many people have been working on — elasticity. When you talk about levels and dynamics of asset prices, very naturally we ask how elastic markets are, whether at the micro or macro level. A fundamental difficulty I learned from your work: it's not about a one-time shock. The standard concept of elasticity — change price by a dollar, how much does quantity change; or change quantity by one unit, how much does price move — but QE isn't a one-time shock. It's a trend, a commitment, central banks “love” it, people love it. Same for passive investing and other slow-moving trends. I'd love your thoughts on how to revise our understanding of elasticity and how to incorporate slow-moving trends into that framework.
Tyler Muir: Really good question. I thought the paper this morning was nice on this — you want to think of QE as a dynamic version: it's not just that the Fed buys today, it's that you believe they'll continue to buy, particularly in some states of the world. That maps well to that paper. There are contexts where going to a really flexible, purely demand-driven approach can be useful — most useful in settings where you know less about what's driving demand and you have a huge cross-section of positions and portfolios to estimate flexibly while incorporating dynamics. As that paper highlighted, we didn't quite go that route with the Fed and QE. In a sense we did — we wrote down a Greenwood–Vayanos-style model with a dynamic strategy by the Fed and their positions — so we are solving for dynamics. But there's only one series for the Fed, so flexibly estimating their demand and how it moves over time is hard. That's where bringing in information is particularly useful: we know stuff. We know why they did what they did and they tell us when conditions are likely for them to do it again. That's a setting where you can directly write down a process and then estimate it through a quantitative model to see the effects. That's my take.
Question: Really fascinating presentation. I was wondering whether market macrostructure can tell us anything about how markets should be organized. For example, in the Fed context, maybe their large effect on out-of-the-money options speaks to a greater ability to stabilize, and maybe that's a good thing. Conversely, the fact that intermediaries have so much influence in MBS and other OTC markets — does that suggest a regulator should push for more standardization or more exchanges? How do we think about welfare effects?
Tyler Muir: Really good question. If the first part was hard, this part is even harder. We haven't quite gotten there yet. We've still been on the agenda — broadly, not just us — of showing these things look like they matter quite a bit and trying to understand why. To take the QE analogy: what we did not try to do is write down “what's the optimal QE policy by the Fed?” — that's the welfare analysis. Obviously this literature should ultimately go there, but my own view is I'm not quite comfortable enough to do that yet.
Question: What's always hard for me about understanding the Fed is that you need two skill sets that pull in very different directions. One is institutional detail — the mechanisms of triparty repo, the plumbing. The other is the long-run question: looking at how Fed actions have evolved over time. What the Fed does now is very different from 1990. Used to be: only set the short rate, bond vigilantes set the long rate. Then it turns out the Fed can set the long rate just fine in 2008, and the credit spread in 2020. So is there a trend that more and more things are set as government decisions, by fiat, rather than market forces? If you extrapolate, you ask things like: at the next intervention, will they just buy equities directly? I don't know how to think about the bigger-picture aim they're moving toward.
Tyler Muir: That's a hard one. The work I highlighted that we have in progress agrees with that — the Fed's footprint in financial markets has really increased over time, not just from QE but all these different facilities. Silicon Valley Bank goes down and they set up the BTFP (Bank Term Funding Program) — exchanging bonds with banks at par, not recognizing losses. There do seem to be a lot of these interventions every time something happens. I'm more reluctant to say whether that's necessarily bad or good — I see both sides. Would it have been good for them to let the corporate bond market completely collapse in COVID? No — probably wouldn't have been good for the economy, would have had severe effects. But our work suggests there is a trade-off: you build in expectations that they'll come in any time something really bad happens. The scope of those interventions has also increased — that's a fact. In 2008 it was Treasuries and similar; “we don't want to get involved in corporate bonds or municipal bonds” for obvious reasons. In 2020 they did do those. I don't know if I addressed your question exactly, other than: I agree, and I want some way of measuring it. To measure it, you have to try to measure what the market thinks they'll do in what states of the world, and how that affects markets.
Question: Tyler, this is a more general question — feel free to speculate. In the spirit of what you said, we don't want a collection of episodes where we just see what institutional frictions led to something. From that collection, what features of institutions are we learning matter more? Think about a situation where we might fall into a recession. What institutional features of the microstructure of institutions will make a difference between a recession and a financial crisis? Are you learning from these episodes that you can start constructing features of how institutions are organized that have an amplification effect on pure economic shocks?
Tyler Muir: Great question. A couple of things. As I highlighted, we know banks have stepped out, relatively speaking, of a lot of these markets, replaced in some cases by private credit, in some by mutual funds (e.g., corporate bonds). So those are places where there's likely continued fragility next time around. But also the thing we were just discussing: how the Fed will intervene in the next crisis and where that stops. There's a lot of discussion now with a new Fed chair and possibly a new policy regime — about shrinking the balance sheet, focusing more on short-term rates, less on intervening in various ways. I don't know if that's what will end up happening, but it would probably be a big deal — markets have gotten used to the Fed being there to help when things go bad, not just in financial crises but in any Silicon Valley Bank type event. If that shift does happen, it would have pretty significant effects.
Question: I wanted to ask whether you think your second and third facts are related. After the financial crisis there was a lot more regulation of the banking sector, which basically caused a lot of things to be very inelastic, with effects on yield spreads, and that then caused the need for the Fed to solve all the problems — because it had been overregulated… or, let me not make that normative — regulated. Going forward, as we think about optimal policy, we can't think of the second and third facts in isolation. To give an example: if you change liquidity regulation, banks' reserve demand falls and the Fed's balance sheet falls. So it's all linked.
Tyler Muir: I completely agree. They're very linked and connected. One of the actions the Fed takes in 2020 — while regulations had restricted what banks were willing or able to do — is to relax those in the middle of a crisis to make the market more elastic. Think of relaxing the supplemental leverage ratio in 2020, a big one. Setting up the Primary Dealer Credit Facility and similar things — it's like, “We don't want you to take too much risk, but that limits how elastic you can be in absorbing shocks; in the middle of a crisis, we'll relax some of that to keep markets from freezing up.” That seems to be the approach taken. I agree those things are very linked.
Question: Very quick. At what frequency does market macrostructure operate? There's the “younger cousin” — market microstructure — that says institutional features are very important at the nanosecond frequency for bid-asks. This is saying actually, even at longer frequencies — monthly, quarterly — these institutional frictions are really important. Is there a very long term where households undo these frictions? At which frequency does market macrostructure play the biggest role, and when do the effects dissipate?
Tyler Muir: Good question. If you go really into the long term, we don't know. It will be too hard to answer — number one, you won't have sharp enough empirical work to distinguish. It also relates to John's point: at lower frequency, these things adjust, and they adjust for reasons — people went into passive investing because the other thing wasn't really working. Over longer horizons, I agree. But exactly how long — it depends on the context. I don't think we have any easy way, in anything we've done so far, to distinguish.