Good morning. I'm Thomas Philippon, professor of finance here at NYU. So it's my pleasure to introduce the next panel, which is on theory on measurement of financial crisis. I think we've heard a lot in this conference about carelessness and maybe some complacency in the private sectors among regulators. I would argue that there was just as much carelessness and complacency in academic research definitely precrisis. And the good news, of course, is that even Academics learn at some point, albeit slowly if you want. And I honestly cannot think of a better list of speakers to discuss
and review the faults and the progress that have been made since the crisis to think about the theory and the measurement. So we're going to have four speakers, and they're all going to cover different topics related to theory and measurement. So we're going to start with Arvind Krishnamurthy from Stanford. And Arvind is going to give us also an overview of how the crisis has changed research in academia with a focus a bit On the theory both of corporate finance and asset pricing. We're going to have then Tobias Adrian from the IMF, who is going to
focus on leverage and financial conditions. We're going to have our own Rob Engle, who's going to talk about systemic risk. And so that's going to be much more narrowly focused on what is it that we can actually measure in real time and how do we do it and are these measurements that we build-- are they useful out of sample. And finally, we're going to have Andrew Metrick from Yale, who is going to conclude this panel with a wider overview and a bunch of open questions Both for regulators and for academics. So I think this is
about as good as it gets, and I'm very happy to hear them today. So Arvind, please. Thank you, Thomas. And thank you all for coming here. So as Thomas said, I'm going to talk about two things. One, I want to talk a little bit about academic progress, research progress inspired by the financial crisis. And then I want to talk a little about measurement. I think the other speakers will talk more about measurement, But I just wanted to say a little about that as well. So maybe I can start here. And if I go back to,
say, 2005, 2006, the dominant way in which people talked about asset pricing is through an equation and a model along those lines. Here P is a price, D is a dividend, some cash flow is coming the future. And M is the way an investor is valuing stochastic payoff streams, how he's thinking about how to price the risk of cash flows. The dominant models, probably the two most popular models, The two most dominant models of this kind in thinking about how investors price risk are the Campbell Cochrane consumption habits model, which emphasizes how investors think about
consumption risk over the business cycle-- so thinking about recession risks, for example-- and the long run risk model of bond selling your own, which thinks about how investors think about how a long term consumption growth might be risky and how that impacts asset prices. These are beautiful, elegant models. Everybody in graduate school is still taught these models. They are simple models, and they have a great virtue. The way I like to think about these models is they are like looking at the world through a telescope from Mars. You look through the world through a telescope
from Mars, the world looks very smooth. The fact that garbage is picked up on East 4th Street doesn't really matter. The world looks very smooth and lots of little things along the way drop out. So in the financial market context, what this type of model smooths over is the fact that investment requires delegation, You hire somebody to make investments for you, markets are sometimes segmented, there are different players in different asset markets. There's financing needs. There's liquidity needs. There's a whole collection of institutional things that make up the asset market structure that this type of
model [INAUDIBLE] neglects. And for many purposes, that's OK. But as we know, models are wrong. This is a wrong model. And where is this model particularly wrong? It is a model that's been particularly wrong for thinking about the crisis. So I'm graphing here the departure from covered interest parity, the euro dollar basis, which as probably many of you know, essentially holds precrisis. Covered interest parity works extremely well until about the crisis when it blows up, It's very hard to make sense of this through the lens of thinking about consumption risks or long risk. In a
way, the point is that what happens in a crisis is all of these things that are smoothed over Suddenly start to reassert themselves. The granularity of the world shows up as being important in how asset market prices are determined, segmentation, repo markets-- all of this stuff starts to come to the fore and plays a pretty critical role in determining asset prices. Covered interest parity is a particularly nice one to make this point, but there's many things that one can point to-- many collection of spreads, the CDS bond basis, a whole collection of places where you
see this type of phenomenon. And it's a phenomenon that shows up particularly in a crisis. So how should one think from a modeling standpoint about the transition from non crisis to crisis? So how does the world grow from a place in which things look smooth to a world in which the granularity of the world starts to show up and starts to affect prices? So what I'm showing you here is a graph from a paper of mine with a co-author [INAUDIBLE] from the AR from 2013. And what this graph shows on the x-axis is a measure--
this is a theoretical model. So in the x-axis, this is a model Based measure of the equity capitalization of the financial sector. So to the right, as you go further to the right, you're better capitalized. And then on the y-axis, I'm guessing graphing the risk premium that comes out of this model-- so the required return per unit of risk that comes out of this model. And you can see that the main feature of this model is-- if you are to the right, if you are in the part of the state space where equity capitalization is
high, risk premium are low. And probably more interestingly shocks to equity capitalization have very small impacts on the risk premium. In a way, the model looks like a non granular world. Shocks are absorbed. The system is self-correcting. Essentially, what happens is you have shocks. There's slack in the system that takes up the shocks. And the world functions as normal. I have a red line there. And if you look to the left of the red line, you see that the risk premium starts to take off. And what's happening there is the crisis effect Starts to show
up. So the outstanding feature of this model is the transition from a period in which markets look relatively well functioning, granularity is not an issue, to a part of the state space where that is an issue. And moreover, the further left you go, the higher the risk premium gets. So for example, if you're on the left side of this red line and there are losses to the financial sector, they'll push you further to the left. And the risk premium goes up even further. There's this spiraling fire sale effects that many people have talked about that
comes out of a model like this. What I've just talked about is asset pricing. But one can also think about how these types of effects will show up for macro. So again, what I'm graphing here is I'm graphing two things from this model. I'm putting in a shock into the model, a 1% shock to productivity, and tracing out what the model implies for risk premia, as well as investment, aggregate investment in, this model, Depending upon which state of the world you're in, whether you're on the right sided so the unconstrained well capitalized part of the
world or to the left side, where you are in the constrained granular part of the world. The dashed line gives the normal period, and the solid blue line here gives the crisis period, the granular period. And you can see what happens is, in the normal phase, a 1% shock just reduces investment by 1%. It's kind of what you'd expect. In the crisis period, the same shock Gets you a much bigger impact on investment. That slowly reverts. And why does this happen? It's basically through the risk premium channel. What's happening is, if you are in the
part of the state space where the world looks more granular, a shock increases risk premia. So it increases required returns, and that immediately affects lending and investment. And that creates an effect on investment and for macro. So what I've just taking you through is really a sampling of research in this area The crisis triggered a re-evaluation of standard models, and it has also led to a whole bunch of new models. And I've just giving you a taste of some of the models that have come out. But there's many models of this kind, and they all
have this type of feature. So the outstanding questions are, when does intermediation matter, when does it affect asset prices, how does it affect asset prices, how can we detect these effects in the data. There's a whole bunch of empirical work. I talked about covered interest parity. But there's a whole bunch more empirical work that then thinks about the data and thinks about how to match the data to models. There's a related set of issues outside asset pricing, thinking about how these same effects affect macro, how in crises do these asset pricing phenomenon show up in
macro. And likewise, there is a bunch of data work in thinking about crises and macro outcomes. So my read on the academic work in the last decade prompted by the crisis is it's actually been very healthy. There's been a whole bunch of work, lots of modeling. And we have learned a lot, and I'm expecting we will continue to learn a lot about these issues going forward. The second thing I want to talk about is measurement. I'm putting this picture up, which I put up before. And to remind you what this picture shows you-- is how
the risk premium depends upon the capitalization of the financial sector. So if you look at this picture and you're thinking about the world and trying to ask questions about, Is the world a stable place, the central question you're going to be asking yourself is, where is this red line. This model has this feature that there's a cliff. And if you're to the right of the cliff, everything goes on OK. If you're to the left of the cliff, things take off. So if you're looking at the world, you want to know where this cliff is. And
I don't mean this from a modeling perspective. You'd like to know this in the data. Are we near the cliff of are we farther from the cliff? So for example, in the world today-- and this is to pick up on a point that Doug also made. There's been lots of concern about say leveraged loans. Leveraged loans have grown dramatically. It's probably the case that corporate leverage is stretched in places. There's a bunch of places in which it looks like valuations are stretched and quantities are stretched in a way. And there's a good chance that when
things stretch they unstretch. The question for looking at the world And thinking whether this-- it means that there is a macro risk involved-- is a question of how close we are to the cliff. If we are way to the right of this point, we can hit shocks. But the system absorbs the shocks and lets it go. If we're to the left, then we have a problem. So the question of systemic risk is a question of, if I add a shock to the world, am I likely to end up on the left side of this red
line. Am I likely over the cliff? And centrally, what you'd need to know for that is you'd need to be able to figure out where w hat is. You'd need to figure out where the cliff is. And within the context of a model that I write down, which I have written down a model, w hat is very clearly defined. Everybody in the world in the model knows exactly where w hat is. But if I back out of my model and I start thinking about the world, the central question that comes up is, what is this.
What data do I need in order to figure out whether we are at the cliff? So the model has a very clear answer. The modeled answer, it's the capitalization of the financial sector loosely leveraged. And so if I take the model literally, I would say, go measure leverage and go ask yourself, is capitalization sufficient. Now the model that I've written down is a very, very simple model. It has really only one variable that drives financial risk. It's a capitalization. That's the only thing that matters. In the world there's much more. There are issues about segmentation.
There are issues about debt markets and how well debt markets are functioning in financing trade. And so if I step back and I ask the measurement question, the measurement question being, do we know from the data whether we are close to w hat or not, are we close to the cliff or not. I am not sure. I really don't know. And in fact, in many ways, we probably could have asked this question in 2007 as well. What data do I have to know whether we are close to the cliff? It's very clear, looking at the
crisis, going back through the crisis, we would like to notice. If we had known this in 2006, we would have had a different trajectory I'm guessing than we do today. And so one thing that worries me looking at models and thinking about data is that our measurement systems have not kept up. They have not responded to the need of the crisis, which is we really would like to know what w hat is. I don't feel in any significant way that we've gotten there. So I'm going to put the slide up. This is a quote from
a writer commenting on the Depression. And I put this up just because it tells you that measurement problems are problems that we've had throughout history. And measurement problems are solved sometimes by better measurement. This is a quote that comes from a article that Gary and Markus Brunnermeier and I wrote called "Risk Topography." And what we argued in that chapter is that a useful thing for the world-- this is a chapter that we wrote in 2011-- A useful thing for the world is to try to figure out measurement systems to understand where w hat is, which
led to a NBER book that invited suggestions from a whole bunch of authors, including some in this room, for how we would measure risk in a way that we could, in the end, form something like a risk topography of the US economy so that we can use that data to figure out where w hat is. I'm listing a bunch of chapters here. And I still think this is a good idea. And as I'm thinking about this question of leveraged loans, I don't know and I certainly don't Feel like we're anywhere close to where we should
be on this measurement challenge. So I'm going to stop there. Thank you. Good morning. So in 2007, I wrote a paper with [INAUDIBLE] for a BIS conference on the leverage cycle. And so this session is about thinking about theories of financial crisis. And so our theory of the financial crisis was one of a leverage cycle. And so what I'm going to talk about today is, in retrospect over the past 10 years, how would we look at the leverage cycle and what has been the impact of regular [INAUDIBLE] forms. On the leverage cycle. So the first
of the impact is visible in this chart. So this is showing leverage defined as total assets over equity for banks in advanced economies in red here and then a couple subregions, like the United States, Europe, UK, and Japan. And you can see very nicely that leverage ran up in the run up to the crisis. So for example, for major advanced economies, leverage was around 20 in 2002. It went up to around 25, and then peaked in something like '27, '28 before declining sharply in the crisis. And then of course, even though financial conditions are very
easy today-- so both Doug and [INAUDIBLE] pointed to the easiness of financial conditions in the current environment, even though financial conditions are very easy, we don't see a recurrence of the leverage cycle. So one might expect that, well, we're in another boom, why isn't leverage going up. And of course, there's an answer to that. And the answer is this. A lot has happened in the past decade, which is related to regulations. There's Basel III. There's the stress test. They're market risk amendments, tier 1 capital rule, the Dodd-Frank Act, leverage liquidity ratios, TLAC for resolution purposes,
capital conversation buffers, G sub surcharges, et cetera, et cetera. When you talk to strategists from banks, when they try to optimize the balance sheet of a bank, they write down something like 40 constraints, like 40 regulatory constraints. And then you have to optimize how to run your balance sheet in order to maximize profits for the firm. So the world has not only become more regulated. It also has become a lot more complicated because you have so many constraints to take into account. And so clearly that has done something. So leverage is lower even though financial
conditions are very easy. And we do see leveraged lending going through the roof and underwriting standards are deteriorating. In the market system, clearly in the banking system, leverage is not going anywhere. So point them one is, there is a leverage cycle and the leverage cycle has been tamed due to regulation. So what are some of the first order facts of more bank capital? So this is a scatter plot. So this is a scatter plot for a sample Of banks from the mid 1990s to the mid 2010s across the globe, 200 large banks. And basically, the
first order thing is that, when you have more capital, you are funding costs. So these are the non-equity funding costs are lower, because you're less risky. So this very clearly in the data. Secondly, banks with lower leverage raise funds and expand the lending at a faster pace. So when you only have less leverage, you react more to changes in financial conditions. And this is visible both in the asset and the liability Side of the balance sheet. So this is very good news. Now of course, the impact of less leverage on the cost of funding is
declining with the level of leverage. So the less leverage you have, the less the margin of impact on the cost of funding, which I think is both intuitive and consistent with theories we write down. Now having said that, of course, we do have less leverage. But at the same time, the mechanics of the leverage cycle are still valid. And so this is the first order finding that Were presented back in 2007. And this remains true. So you can verify that even in post crisis data. And so when you look at the balance sheet of banks
and other intermediaries for that matter, when you look at the correlation of total assets with liabilities, you basically get an R square of 99.99%. So banks adjust balance sheet size by adjusting non-equity liabilities, while the equity cushion is pretty much fixed. So the blue line, the blue dots-- this is the equity cushion. So as banks are expanding and contracting, their balance sheet size bank equity is pretty much flat. And so really what is moving around is leverage. And so this notion of leverage in the leverage cycle continues to be true, even though we are of
course operating at a much higher level of bank capital. And so the leverage cycle is very much related to what Arvind talked about earlier and what Doug talked about earlier and what Raghu talked about last night, which is that during times of easy financial conditions, Risk management systems are going to tell you that you can take on risk. And of course, you take on risk, balance sheet sizes expanding, leverage is expanding. And eventually bad shocks occur. The system is fragile and deleveraging occurs. So one place where that does play ours is in the repo market,
where of course haircuts are set relative to the riskiness of collateral. And this is in particular in the repo market that is funding the buy side. So this on the asset side of the dealer balance sheets. So their haircuts are calculated according to risk management metrics. So when you're in the boom, volatility is endogenously low, haircuts are low, and leverage is high. And then there's the reverse of the leverage cycle once bad shocks occur. Now when we look at the dealers in particular-- so dealers, on the left hand chart. You see the dealer balance sheet
normalized to 100 in 1990. And you see the dealer balance sheets used to grow exponentially just like household balance Sheets and non-financial corporate business balance sheet grow exponentially. And they used to grow at a faster rate because the market based system used to used to expand quite rapidly. And the share of market based finance used to expand, and dealers are at the center of the market based financial system, which as economists, we all appreciate as very important of course. Now what has happened is that, of course, in the crisis, there was a slowdown in this
growth. So total [INAUDIBLE] balance sheets came down from close To $1 trillion to something like $600 billion-- so a very, very sharp decline. And you also see such a decline in the total balance sheet size of households and non-financial corporates. But of course, the households and the non-financial corporates went back on the exponential growth path. So you had exponential growth. You go back to exponential growth. But this is not true for the dealers. So the dealers just completely stagnated. So dealer balance sheet sites stagnated roughly at the level That they first reached in 2003. And
that goes hand-in-hand with the deleveraging, which is shown on the right hand side. So of course now this is not causal because it's hard to get causality. We don't have an instrument, et cetera. But it is suggestive that the way in which things are regulated and the way that intermediation is done has shifted quite dramatically. So I wanted to zoom in a little bit more at the actual composition of the dealer balance sheet. So when you look at that, you can See that dealers had a very large corporate bond position. So this from The New
York Fed Data. So here what is called copper bonds does include a fair amount of securitized assets as well. So they basically ran up a long short position being long corporate bonds, including securitized assets, short treasuries. And then this long short position was unwound very, very rapidly. And basically since 2008 or so, they have a flat position long in both corporates and in treasuries-- so very, very dramatic change. And of course this kind of balance sheet behavior is related to expected returns. So when you look at the share of debt securities in the financial assets
of the dealers, this is very, very closely tied to what I proxy as an expected return for fixed income security. So here just taken a credit spread plus a term spread. So basically what you earn as a dealer is like a credit spread and the term spread. And this correlates very, very closely with the share of fixed income assets on the dealer balance sheet. So of course, I mean the way that the dealer balance sheets are managed is related to returns. Now another way to ask this puzzle is that normally, when volatility is low, valued
risk constraints are loose, which allows to take on leverage. And so this is what is shown on the left chart so these are the value address so you can actually pull valid risk numbers for the largest institutions. And this is very closely tied to market volatility. I use the MOVE index, which is treasury of volatility here as a proxy. And so despite the fact that volatility is very low, as I showed you earlier, leverage continues to be very low. So Only regulations have done something. And when you talk to market participants, the way you can
see that first order is when you look at things like basis. So basis are things like the CDS bond basis. So this is in principle-- If you think about textbook finance, something that should be arbitrageable, because you could put on positions In the bond and positions in the CDS and you can have a convergence trade that is going to pay off for sure in the long term. The problem is it, costs balance sheet. There is a balance sheet. Cost and with all these regulations, of course balance sheet cost has increased. And so these bond bases
are suddenly very large similarly for the 10 year interest rate swap basis. So it's just that, because it costs balance sheet space, you now have risk free arbitrage opportunities in the market, or what is in the textbook a risk free arbitrage Opportunity. Similarly, for effects swap basis-- these continue to be very wide. And there's very interesting work at the New York Fed going on, where they link-- bank by bank positions' data to the tightness of these balance sheet constraints. So yes, these are more constrained, and our dealer balance sheet capacity clearly has not returned. And
hasn't kept up with the growth of the economy. And that does have an impact in terms of intermediation. I mean, furthermore of course, securitizations activity also is massively smaller. So the way I think about the world of finance is that there's bank intimidation there's market based intimidation, and then they're securitization nation activity. And securitization activity also ran up a lot, and there were all the problems with the underwriting standards, that we're extremely severe. And those problems are solved because the activity hasn't returned to a large extent. So there's some securitization activity. CMBS people do worry
about underwriting standards, but just the magnitude of overall securitization activity is just nothing today compared to what it was 10 years ago. So these are very, very significant shifts. Now when you look at what the FSB called shadow banking-- so it was just renamed. It used to be called shadow banking, and now it's called non bank credited mediation So the FSB has started to collect all theis non bank credit intermediation data. And basically what has happened over the past decade Is that the risky type of shadow banking credit intermediation, which is related to securitization, has
declined quite dramatically while the other part, like the healthy part of credit ammunition, which is like asset management essentially, has increased. It's never clear to me why the FSB classified asset managers as shadow banking in the first place. I think there's some maturity transformation and some credit transformation that is happening in the asset managers. But basically asset management is the healthy kind of market based intermediation that we want to see as economists. And so of course, credit intermediation in the asset management sector has increased dramatically. And leverage in that sector is very, very low. I
mean, typically the asset manager has much lower leverage than any bank. It doesn't mean we don't have to worry at all. Now we do have to worry because there are some weak players out there. So for example, you remember Third Avenue was a high yield bond fund. That was taking massive amounts of risk. So you have the core, high yield funds out there That are well managed, but then there's some guys that are taking huge amount of risks. They were not even invested in rate of products. I mean, they were you know just taking a
lot of risk. Now when they blew up, there was a little bit of a spillover to the rest of the sector, but it was not like the kind of massive spillover that some people are worrying about. But of course that was an environment of easy financial conditions. So there's this question of, once we are in the next downturn, one there Is the next financial crisis, are they going to be more spillovers across asset management investors because these guys are giving the asset managers assets to manage. But if they pull out, what's going to happen. So
we don't quite know because of course, with more leveraged loans, more corporate bonds more high yield bonds in that segment, we don't quite know what that is going to do in the next crisis or so. So in 2009, there was the Pittsburgh G20 summit. And basically that laid out the strategy For the regulatory reforms, which was on four pillars, four pillars plus 1 here, which is the other stuff. And so number one-- building resilience, basically having more capital, higher quality capital, and more liquidity. So that has clearly been achieved. So the banks are much safer.
Ending too big to fail-- I didn't talk about that. So the resolution regime has been phased then pretty much in the jurisdictions that house [INAUDIBLE]. And the TLACs-- so the long term debt that can be bailed in has increased. It's not quite at the level where it should be, but it's heading in the right direction. Making derivatives safer-- there has been a massive shift to CCP-- so to central clearing. Now we have of course the new institutions for the CCPs, which have to be regulated appropriately. And there's quite a lot of work on that going
on at the FSB at the moment. So I would say there is movement in the right direction. But it hasn't quite finished yet. And then there's shadow banking. And there, clearly, I think that we have moved to a much better place where the off balance sheet vehicles that were sponsored by a credit lines that were capital arbitrages as for the banks-- those have been replaced by more like America based credit at intermediation, which is exactly the kind of development that makes the system safer. Which doesn't mean that there aren't risks, but certainly the risks are
much less Than what we had in the old shadow banking system. And then there are all these other areas-- macro [INAUDIBLE] frameworks is in particular something that the fund has done a lot of work on. And there are, of course, the emerging areas, like fintech, where there is a lot of developments and accounting standards. I mean, part of what has made the system safer is the change in accounting standards. So SPVs now to a large extent have to be consolidated on balance sheet. And so the conduits are basically No longer arbitrages-- not just for capital
reasons, but also for accounting reasons. So I want to finish this talk with the following picture. And so the title here is a little bit misleading. So it only applies to the left chart, but not to the right chart. So we have seen all this tighter regulations. So the dealers now clearly have stagnated. Leverage has been contained. There has been the substitution into the healthier market based system away from the old, very risky shadow bank maturity And liquidity transformation. But with that of course, market making has changed dramatically. So when you talk to market participants,
either on the buy side or the sell side, they will tell you that market making is no longer what it used to be. You can Google this great photo of the UBS trading floor. Out in Stamford, that was the biggest trading floor in the world. This was the size of something like two or three football fields. And this is now completely empty right. And it was unwound. And this is just a symbol of what has happened. And then you're like, is that costly or not. And one of the things-- of course, it's too early yet
to really evaluate the unintended consequences of regulations. But here is one initial look at that. So here we are using trace data, and we're calculating price impact for corporate bonds. So this is one very useful liquidity measure. And what you can see is that the price impact For the retail size trades on the left hand side basically is cheaper today than it was precrisis. So retail investors can trade very cheaply. And this is the important thing, but institutional investors are something like 1 and 1/2 to 2 times more expensive. And so, again, when you talk
to buy side or sell side in that space of market making, they will tell you that there has been a transition from the principal trading to agent based trading. And this is what I showed you earlier. So it used to be that the dealers had huge inventory of corporate bonds, And they were happy to take on directional trades. And now they are no longer. So if Doug wants to sell an illiquid security to Stan, it used to be that, if Bob was the dealer, Bob was like, yeah, I'm going to take Doug's risky stuff. And
now he's like, no, I'm not going to take that. I'm going to look around for somebody to buy this. And then Stan perhaps happens to say, yeah, I'm going to take it. But this might take a couple of weeks. And so this is why you see that the training costs have increased for the institutional-sized trades Because the dealers are no longer inventorying this kind of bonds. Now how costly is that? Does that have an impact on the inside borrower? It's hard to see because of course QE also happened at the same time. So liquidity costs
are higher. Trading costs are higher. But then, again, interest rates are very low. So how do you calculate that? So how do you find instruments to really calculate the costs? It's very hard. I've done some work on that. So you're certainly using different type of approaches. You certainly do find significant statistically and economically statistically significant effects. But then you also have to weigh that against the fact that, yeah, we do think that the system is somewhat safer today. So I think there's a lot of very exciting work to be done there over the next couple
of years. And so with that, I want to conclude right on time. I think the regulations have made the system safer. There are some consequences in terms of the cost of credit intermediation, But I think that the overall punchline for the moment is that probably the system is safer and the unintended consequences are acceptable in terms of magnitude. Next speaker is Rob Engle from NYU, and I just want to emphasize we already have one key takeaway from this conference, which you're not supposed to say shadow banking anymore. You have to say non bank credit intermediation
bank. Thanks, Thomas, and it's great to be here. I appreciate so much economic reviews, annual reviews, doing this whole volume, and the conference. And it's been a terrific two days. I'd like to talk about a slightly different aspect of systemic risk, which is illustrated in this picture. That is, the financial system continues somewhat like a volcano to be inactive, but possibly not. In other words, if you look closely at this picture, you might see that there's-- well, on my monitor I can't see it, but there's a little wisp of smoke coming out the top. And
the question is, is the financial system Going to erupt again or not. And can we build monitoring tools that give us an idea where the risks are and how close we are to them? I mean, this is exactly a response to Arvind's I think question. So I'm going to show you stuff that we have on V-Lab, which is the website. Which you could see on your phone if you choose to do so, which is updated for many things every day and for systemic risk once a week. If you Google it, you'll find out where it
is. But if you open it up, you'll see immediately a map of the world. And the map of the world is a volatility map, which tells you, where it's red the equity markets are highly volatile. And where it's green they're low. And this is a day or two ago. You see we are in a high volatility period, but it was only a couple of weeks ago it was low volatility. So volatility is changing very rapidly on a day by day basis. Further measures of risk are in the website, And we have sections on, as I
say, volatility, correlation, systemic risk, which is what I'm going to talk about today, long run value at risk, liquidity, fixed income, and climate risk. So let's see what kind of monitor we can put together on systemic risk. There's a lot of view maybe which comes initially out of Basel, but I think we might even have a consensus here that financial crises are really caused by excessive credit growth. I mean, the phrase is sufficiently vague That we almost all could agree with it. And the real issue is, how on Earth do you measure when credit growth
is excessive. And I'm going to propose a measure, which again would probably be different from what everybody else does. But it's designed to measure when credit growth is excessive. And let me motivate it by an example, which is from the mortgage market. In the end of a credit cycle, what you would imagine to see is that you would see banks and other financial institutions Lending money to underqualified borrowers and using overvalued collateral. What that means is that the market value of these mortgages will be less than the accounting value. So the difference between market and
accounting values immediately is important. This difference is accentuated if there's a downturn in the economy because then the borrowers become even weaker, the collateral becomes weaker, and the market value of the mortgage will go down. At this new level of market value, a lot of the bank's capital reserves Will be allocated to covering these losses in the market value. And if there is not a sufficient capital cushion, the bank will either face bankruptcy or look for a buyout or a bailout. And so with this kind of motivation, what we call excessive credit growth is when
the financial institutions do not have adequate capital to cover losses in a downturn. So this is actually very close in spirit to stress tests, but we're going to do it with market values. And we're going to do it without seeing the supervisory data. So this motivates the definition of S risk, which stands for systemic risk. Which many of you have seen many times over the years. How much capital would a financial institution need to raise in order to continue to function normally if we have another financial crisis? It's a bail out estimate. And many of
my colleagues have contributed to this. And Matt and Thomas and [INAUDIBLE] and [INAUDIBLE],, and there's a lot of references on this slide. So what is capital shortfall? We think that there ought to be some relationship Between the amount of equity that a firm holds and its assets. Let's call that capital ratio little k. And if the desired capital is greater than the amount of equity that they do hold, we say there's a capital shortfall. We're going to use a measure of assets, which is the accounting value of debt plus the market value of equity. So
it's a little bit of a mixture there. And so what we want to know is, what is the expected value or the median value of the capital Shortfall if we have a crisis. So this is a counterfactual. We need a model to estimate this. The model we're going to use is totally familiar. We're going to recognize that the equity change for a particular firm is it's beta times the growth of some aggregate equity market measure. We use a world equity index, an all world ETF AWCI. And we're going to ask, if the global capital market
falls a certain amount, what do we expect is going to happen to this firm. We calculate that. The key issue there is, what is the beta. And so we need to talk a little bit about the beta because this is not an ordinary beta. This is a time varying beta. And the beta changes actually every day. In fact, we know simply that the beta is going to be the correlation times the ratio of two volatilities. And I just showed you that volatilities are changing all the time. So it makes sense to think that the beta
is actually Changing over time. In fact, it also changes because correlations change. So we estimate this with something called dynamic conditional beta, which essentially takes volatility estimates, puts them in the beta formula, dynamically estimated correlations-- we use the dynamic conditional correlation model to put in there. And we have then a beta, which is an optimal forecast of the beta using information that was in the past. As a check that what we're doing makes sense, We run what is really a non-nested test of whether this beta works. That is, we put this beta times the market
return into the original regression to see what kind of a regression coefficient we get. And we put in the same regression a fixed beta. And so if this is really a good estimate of beta, we'll find no fixed beta. And we'll find that the theta in this model is close to 1. It often turns out that it gives a coefficient of theta, which is around 1. But it also turns out sometimes it's considerably less than 1, Which suggests that we shrink this estimate of time varying beta toward a constant value. So this is the way
we do it. One of the things that I've done in this paper is some analysis of how much risk there actually is in these kind S risk measures. And this includes sensitivity analysis to changes in the capital rate and changes in the stress. It includes econometric uncertainty because there are standard errors on these pages. And the standard errors are a little complicated to figure out because it's three different models. So I bootstrap them-- and because you've got some thousands of observations, the standard errors actually are pretty small. There's model uncertainty, and we've actually got several
different models on the website. And so I compare them in here. And then there is, of course, uncertainty in whether the data is well measured. And that's particularly important for the accounting data. We don't do any specific adjustment for it, but it's certainly worth paying attention to that. So here's what you got. Here's the beta for Citibank over the last 10 years, I guess. That's basically what we're doing here. And what you see is that on average is somewhere around 1-- not a surprising number. But during the financial crisis, it rose to something up to
4, which meant that every time the global market went down by 1%, Citi stock would go down by 4%. Why is that? Well, it's because Citi was very exposed to the same products that were driving the global capital Markets, which were basically these mortgage backed securities. So I think it makes sense and it shows how vulnerable Citi was at that point. If you look at Goldman, for example, you see it did not go up nearly as much during the financial crisis because, in fact, Goldman was pretty well hedged against these risks. But that Goldman's average
beta is a little bigger than 1 because it's riskier than Citi. BNP Paribas, for example, has very little effect during the financial crisis. But it did go up a lot during the European sovereign debt crisis. And if you look to the right, you'll see it actually went up fairly substantially where-- this narrow peak on the right, which is Brexit. So the beta for BNP went up during Brexit for a while and then came back down again. And Barclays had the misfortune of participating in each of these crises, and its beta went up for the financial
crisis, the sovereign debt crisis, and Brexit. So if we put these things together, We get estimates of S risk for something like 1,200 financial firms around the world, which we update every week. What do we see? So this is actually a picture of risk. And it's a personal picture, actually, because, if you can see that pink visor, well, that's my wife. And I'm next to her, and my kids are in front. And the guy with the oar in the back is the risk manager. And he has about a second a half left before he loses
all its credibility. So what do we see when we look at this? So here is a summary picture. This is what it would cost to bailout all the financial institutions in the world if we have another crisis like the last one. And so you see, the number for the financial crisis that we have been focusing on is something like $4 trillion. It was a little more than that during the European sovereign debt crisis. And then there's a third peak there, think which doesn't have a name. But I think it's due to the slowdown in the
Chinese economy That led to resource prices collapsing all over the world and putting pressure on financial institutions and all resource rich economies. And you can really see it in a lot of ways. But there's also one more. What's this last little one doing up there? I mean, this is the advantage of high frequency monitoring. I think this is a second China problem, which is associated with the debt in China. But it's also associated with the trade war that is going on. And by looking at this in a high frequency basis like this, we can actually
see these events as they're happening. So oops, five minutes. China is the biggest contributor to this. If you look at the US, you'll see S-risk has been going down. We've been talking about that today. In a lot of ways, things are a lot more stable in the US. If you look at China, you'll see a totally different picture. The debt is increasing very rapidly. The under capitalization of the banking sector is dramatic. And you see this last peak, which I think of as the trade war, peak being very substantial. A moment to look back-- On
the website that we have details. One of the details is you can look at what this looked like at any moment in time. And this is what it looked like August 29th, 2008. So this is two weeks before Lehman's bankruptcy. And if you look at the S-risk numbers, Citi is the highest. It looks like it would have needed $138 billion. Then we have JP Morgan, Bank of America, Morgan Stanley, Freddie Mac, Merrill Lynch, Fannie Mae, AIG, Goldman Sachs, Wachovia. We don't get to Lehman until we get to number 11. But if you think about it,
the first 10 Were all basically rescued. And so essentially, my view is that the decision, assuming it was all just done on economic grounds, was put in a very reasonable order. But there a question of where you draw the line. And there's Washington Mutual down there a couple later. And if you had drawn the line below Lehman, then you'd have had to deal with Washington Mutual. I don't know quite how you decide where the line ought to be. But it seems consistent with these numbers. There was advanced notice. This is a year and a half
before. This is three years before. So the question that I wanted to close with this how much S-risk is really too much. And this is a picture that's obviously too much risk. So there is a separate paper, which is on how much S-risk is too much, which is with [INAUDIBLE]. And basically what we try to do in that paper is calculate, given the S-risk that you see, what's the probability that this country is in a crisis. And then we also want to know, how much S-risk can this country really tolerate before the probability of a
crisis gets to 50% or more. So how do you do that? Well, we need an economic model. But the economic model is that, when firms have high S risk, they're vulnerable. And either the regulators or the risk managers are going to tell them to reduce the risk. The most common way to reduce risk is to sell assets to reduce their liabilities. So the question is, what's the probability You get a fire sale out of this kind of asset sale problem. And that depends on how much assets need to be sold relative to the total assets
that are out there. If the fraction of total assets that need to be sold is 80%, you've got no buyers, you're going to have enormous price impact. I think that's what we saw with the mortgage backed securities. There was no market for those during the financial crisis. So that's the variable we want included in this model. We want to try to predict Romer and Romer's crisis measure, which is something between 0 and 15. And we're going to predict it with a panel regression model that estimates using a [INAUDIBLE] because all these zeros mean that we
don't want a straight line. We want a hockey stick for our prediction method. And what do we see? Well, here are the regression results. But I'm going to show you some pictures from the green curve, which is the global model. Which says the risk depends on S-risk divided by total assets, but it also depends on what's happening in the rest of the world. It depends on whether other countries have high risk, and it depends on whether other countries' risk is growing or declining. So with that much justification for this model, which really deserves a lot
more than that-- but anyway, what do you see? Well, first of all, I want to say what this does is it focuses on two externalities that we all think are important. One is that the riskiness of having a financial crisis depends, not only on one firm. It depends on all the firms in the country. And so no firm has sufficient motivation to reduce its risk because of the externality. But that's also true across countries, that the risk in one country depends on what's going on in the rest of the world. And this is an argument
for why we need to have some kind of regulation or coordination of risk across countries and within countries. So here's the same picture I showed you, but with two new features on it. This is the S-risk for the US, but in the background is the capacity. So this is how much S-risk the US can tolerate without having the probability of a crisis go to above a half. And what you see is mostly the US was below its capacity, but not during the financial crisis. In fact, the capacity number drops behind the S-risk, meaning that the
capacity that the US would have had to cut its S-risk by a whole lot in order to avoid the crisis in 2008, 2009. But that's because the rest of the world was weak too. So the bottom curve gives the probabilities, and you can see it was about 100% That we were in a financial crisis. We're not too shocked at that. Today it's very low. Here's Spain, for example. You see that it really was devastated by the European sovereign debt crisis. Here's Greece. It was even worse. But it has a secondary peak as the IMF and
the triumvirate renegotiated the loan package. Australia is a much happier story. It never really got close to its capacity And didn't really have a financial crisis. And these things are updated every week. Take a look at them on V-Lab. And my final slide is two of my grandsons looking out over this lake. What's in their future? Do we have another financial crisis or not? And I think we've got some things to worry about, but I leave you with that. Thank you. That's a perfect transition for Andrew Metrick, who's Going to tell us about the open
issues in regulation and research. And I want to say, I argued earlier that there was too much complacency also in economic research. Now this is an all academic panel, and we are exactly on time. So one place where we can much [INAUDIBLE] crisis is enforcing time requirements. Good morning. What do I have to do to get my slides up? Are they up? OK, great. I should say this is based on the paper that is in the annual review, and this is joint work with my colleague [INAUDIBLE] from Yale. Raise your hand, June. June is here.
So the last three presentations that we heard, I think, really I would summarize as good news overall. They've shown a lot of progress that we've made since the financial crisis in research and particularly in these areas of a better Understanding and mapping of how financial intermediation and the macro economy are linked-- so I think of as some of what we learned from Arvind and Tobias' presentations-- and also measurement and a better ability to monitor what's going on, which we just saw from Rob. These are topics that have very, very big academic incentives. If you do
a really good job at these things, you can win a Nobel Prize. Rob can win another Nobel Prize. So these are good topics for people to do. What I'm going to talk about here Is some open questions in financial regulation, systemic risk regulation in particular. And here I think actually the pure academic incentives are relatively weak. Not because, if we could solve the whole problem, yeah, sure, that would be great. But a lot of the research here requires first that the researcher get very deep in the weeds of the institutions. That's a pretty big investment.
And then there's not a whole lot of data on some of the questions that I'm going to be talking about. So you wonder what kind of thing can I do and what will people see. So I think here we have-- this is much more depressing talk. It's depressing, I guess, about the state of knowledge. Maybe not depressing if you're a young person who wants to make that investment and you see that you won't have as much competition as you'll have in some of the other places. So let me talk about what my scope is here.
So I have the least exciting slides and not too many of them either. So unfortunately, Rob's slides were the most exciting. So the contrast is going to be really bad. The scope here is-- this is based on the paper-- to look at the regulation of systemic risk and how that has changed since the crisis. So it's called regulatory reform. There are a fair number of things that were done in regulatory reform that were not about systemic risk. I don't consider the Consumer Financial Protection Bureau of the United States to be a systemic risk thing. We
don't talk about that in the paper. What we're looking are what are some of the open questions in the regulation of systemic risk. And you can break them up into, I would say, three different categories. There's preventive powers, trying to make sure we don't have another crisis. There's what do you actually do in the emergency, and then there's what do you do to clean stuff up. That's the resolution and restructuring. Now the clean stuff up part can also be thought of as partly preventative and partly emergency, but it seems to be its own animal. And
I'm not going to cover absolutely everything here. So you see in total there are six things-- three under preventative, two under emergency powers, and one a resolution and restructuring that will make up, what we'll say, are six big open questions. Not on this list is bank capital. Bank capital has received a fair amount of attention and also has received its own treatment in the annual review. So we didn't have to cover that in our paper. Nor stress tests-- so stress tests is a very nice annual review paper. It's technical. It's a new thing since the
crisis. But we're not looking at the research on that. So we're going to be trying to cover some of the other things. So what I'm going to do is quickly just talk about what some of these big changes have been in the last 10 years to give just a little bit of institutional background-- maybe some of my other academic colleagues don't know all of it, although I'm sure a lot of people in this room do-- and then talk about what we think the big open questions In this area are, where there's a ton of research
that can be done. So on preventive powers, the first thing are new liquidity guidelines that came out of Basel III and have now been implemented-- or almost implemented-- across all of the major economies. And Tobias mentioned these in his talk. So the two big ones are the LCR, or the credit coverage ratio, which operates mostly on the asset side of the balance sheet forcing banks to have a certain amount of high quality liquid assets. And then the Net Stable Funding Ratio, and NSFR, Which is more focused on the liability side-- so making sure you have
enough stable funding on the liability side. So these are new things. Second is the optimal design and oversight of central counter parties. So these were created in the United States. We have a whole title under Dodd-Frank. That created a lot more central clearing for swaps. Europe has a similar type of directive. So they also put a lot of things into central clearing. We don't know all that much about how those things work, And so there's going to be a lot of open questions there. There is a migration of traditional banking activity to the-- what are
we calling it-- Non-bank credit intermediation. We'll fix that in the paper. So there's been a lot of migration from traditional to, let's call it, nontraditional banking. We saw that before the crisis. And there are some new powers that have been put into place that didn't exist before to try to capture some of that. So in the United States, you could think of some of those new powers as being designation authority that the Financial Stability Oversight Council has. And we have similar things in Europe where overall we can look at activities we weren't necessarily able to
look at before and bring them under the regulatory umbrella. The next group are, I would say, newly imposed restrictions on emergency powers. This is mostly in the United States, but they are significant things. For lender of last resort activities, the Federal Reserve's powers under section 13.3 have been somewhat curtailed, although how creative they can be to get around that curtailing is still an open question. But also globally, we've certainly lost a fair amount, I would say, of political room to be able to do things here. And it makes it even more important that we study
what worked and didn't work in the crisis. And then in the United States, we've largely lost the ability outside of a congressional process to broad based guarantees, as the FDIC did over Columbus Day Weekend in 2008. And that's a very significant restriction. That's a change based on Dodd-Frank, and one that is understudied. And then on the resolution restructuring side, really we've done a lot. We have Title II of Dodd-Frank with ordinary liquidation authority. That's a totally new thing and untested. In Europe, they have the Bank Recovery and resolution directive, which has been implemented across many
different jurisdictions, which Is a completely new approach and the single resolution mechanism that has been created, a new approach for dealing with the failure of large banking institutions in Europe. So those are just those are just the basic institutional changes that have happened. And what I can say is, economists had, as far as I could tell-- certainly in Basel III, I know there's a lot of economists who were involved in that, although I don't know to what extent they had the final word on the liquidity types of guidelines. But as far as the administration in
the United States-- I was working in the administration at the time that we were working on Dodd-Frank. I was there. David Sharfstein was also there. And we lost a lot of arguments. And we lost a lot of arguments because we did not have a literature that we could fall back on. People would want to do things. We would say, this is a terrible, awful idea. They would say, why. And we would say, it just is. It became very difficult to have an ar-- And the basic tension that was there frankly was that there was a
lot of lawyers. The lawyers were running the show, and lawyers are very focused. At least they were in my experience when we were trying to do things. Lawyers were very focused on making things ex post, nice and clean. So if there's a failure, we want to be able to rely on a whole bunch of laws so that we can go and clean things up nice and fast. And economists often here are really Focused on the ex ante incentives for all of the different players based on those rules. And so we would make a lot of
arguments that, if you write the rule this way, you know everyone's going to run a whole six months early. And they would say, well, how do you know that. And we would say, it's obvious. And they say, no, it's not. And we would lose the argument. So we need to be ready. We're not going to be able to change all these rules, change any of these laws prior to the next financial crisis, Although there's still some rules that can be adjusted on the margin. But we need to be ready that, after the next financial crisis,
we can hand a draft piece of legislation to Congress in the United States and the equivalent elsewhere so that we can do better. So I let me talk about what I think the six big open questions are. So there's one for each of these areas and where I think we can do some work. So the first is, what are the implications of Basel III's liquidity rules for overall For systemic risk for crises. So we have these new liquidity rules. And you have to understand here-- or at least I think we should understand-- I don't know
how much economists got to really be involved in figuring out what those liquidity rules could be. But they're strange. So remember, the main idea of those-- what motivates these liquidity rules is the idea that banks were doing too much of the lending long and borrowing short. They were doing too much maturity transformation. We have to keep them from doing that. That's not that's not a bug. That's a feature. That's what banks are. And so to the extent that we place restrictions on banks' ability to do maturity transformation, it's probably going to go somewhere else. And
I don't think we have any really great estimates of where it's going to go or what that's going to mean overall for systemic risk. And there are opportunities to do some work here because these LCR and NSFR things are being rolled out Kind of in an exogenous order across countries and different sizes of banks. I think there's stuff to do. But we need people doing those things, and to do them you've got to bury yourself a little bit in how these rules get built. It's not that much fun. It's not as much fun as a
DSGE model. But it's important. Second-- how do CCPs affect the probability and cost of a financial crisis? How should they be designed and regulated To maximize their net benefits? This was another area. It's very big. All these new rules-- we're going to put things into central clearing. Overall huge problem leading up to the crisis-- Doug talked about it in his speech about how really you can get a run that just comes from people coming and demanding more collateral. And in fact, in a world where everything is bilateral trade, you can have situations like Lehman Brothers
that had effectively a flat book and got totally crunched By collateral calls. And then furthermore, their flat book had an enormous amount of breakage after they went bankrupt that caused great damage to the estate. You have situations like AIG, where the collateral calls also killed them. But it's tricky. Lehman Brothers also got a ton of hurt from the place where they were doing things kind of centrally. In their third party repo, JP Morgan kept asking for more and more collateral from them. That's almost like JP Morgan was acting like a central clearing party. And 5
billion, and they handed it over and another 5 billion. And then they really couldn't. So we don't know which of these two methods-- what it does to make a crisis more likely going in or what's going to make it more difficult to clean up afterwards. And when the rules were being written, all David and I were able to do was continue to call Darrel Duffy and ask his opinion. Darrell had written some papers on this. Nobody else really had papers on it. And since the crisis, Darrell has written more. Some other folks have picked it
up. But there's a lot more, say, in the operations research literature, where people have tried to sort out exactly how these things might work and not so much about the economic incentives that happen when you have this in place versus a system where things are largely going through bilateral clearing. Third, how to post GFC regulations affect migration to the non-bank credit information, whatever we're calling it? And what tools should we use to monitor and regulate that sector? So we saw very clearly-- I've talked about this with regards to liquidity. With regards to bank capital, with
regard to any of the rules that we put on banks, if we think our main tension is, these are big institutions. They're scary. Let's look at them more carefully and supervise them, make them hold more capital. We will push stuff outside of them. And the evidence on this from before the crisis Is very strong. People in this room have written papers on it, the basic result being, if the convenience yield-- however you want to measure it-- widens, we get a whole lot of short term safe asset creation outside of traditional banking. And you see it
in asset backed commercial paper. You see it in exactly when people take stuff out of the warehouse to securitize. And it happens at very quick horizons. So we need a real way to monitor that and figure out, well, who's making all this collateral and where is it going And what are the rules doing for that. And there's work on that, but I think not really answers. Fourth, moving to the topic of what we do actually in a crisis, the emergency powers-- how should we design lender of last resort policies for modern panics? And in particular,
lender of last resort has a mean tension here between stigma and moral hazard-- long term moral hazard and short term moral hazard, in some sense. So we think that banks consider it to be a bad signal that they are borrowing From their central bank. And so there is a temptation-- however, we want them to feel a little bit bad because the idea is, well, if they don't feel bad at all, then they're going to take tons and tons of risk going forward. So there's a strong feeling that we have to stigmatize it. It should be
bad, and they must be punished. And that will help them take less risk going forward. On the other hand, if it's viewed that way, they don't want to borrow from the central bank. We saw that clearly in the United States in the early stages of the financial crisis. So the Federal Reserve had to design and put out the term auction facility in December of 2007, which had a variety of really interesting design features that destigmatized borrowing from it. And we see that, at the worst times of the crisis, effectively borrowers were willing to pay 100
basis points more to borrow from the TAF then they were willing to go for the regular discount window. Otherwise, the rules were all the same. The TAF money was harder to get-- same collateral, same haircuts, all that stuff, 100 basis points to get away from the stigma. TAF has been studied pretty carefully. And we've learned a lot, I think, from the academic literature on that. There were hundreds of emergency lending programs around the world where we can learn something about how to design these things. And especially in a world where we have a little less
flexibility about what we can do, I think that's important. Fifth here-- the extended guarantees of the GFC-- there are two very big ones. We did one in the United States. The best estimates that I've seen are that it was quite successful, that what the FDIC did. Ireland did a really big one that was very unsuccessful. We have a lot of things in between, and we believe that these guarantees, along with capital injections, are important parts of the crisis fighting arsenal to come out at the same time. But how exactly should you design it? Looking across
about 30 different programs, as we're doing in a project at Yale, you see a lot of variation. Did banks have to put up collateral? What exactly were the spreads that they were able to borrow? How much was it linked to the sovereign? Which banks were eligible? Was there a cap? All these things have subtle effects on sometimes the stigma of the program and its long term effectiveness. We don't know. And it seems like an important thing to know. Whoever is advising the Ben Bernanke's of the next crisis really should be able to give them answers
about what works And what doesn't work. And then finally, can the new regimes for resolution and restructuring of [INAUDIBLE] work? If not, how should they be altered? So we have Title II in the United States. It is a monster, really interesting, lots of moving parts. Zero research, I would say that really was able to inform us beforehand about why do design it one way versus another. In Europe, they've had a little bit of experience. And they've gotten extremely good At figuring out how to evade the regime that they've put into place because it would be
so politically painful to actually go through it in some of the cases. So we don't know. It's great to have this option that stands in between bailout and bankruptcy. We have something else. But we don't really know how well it's going to work, and we don't know how the individual decisions that were made about how to design it are going to affect that. So that's it. I think I'm just on time.
