ALEX ROSENBERG: It's really in recessions that market structure matters most. And that's something that you'll see in this piece where Hari Krishnan talks about the dangers of ETFs. He says that ETFs, like the HYG provide a false sense of security because there's a liquidity in the ETF that's not there in the underlying asset. So when things really start to go south, they can go south in a hurry because people are struggling to sell, institutions are struggling to get out of these products, and the underlying basket of assets that they actually want to sell is very hard to come by.
I think that this is a very useful piece as we continue to think about this recession question, as we continue to probe what in the market, where the states of vulnerability are, and also to think about how to hedge ourselves and protect against these worse outcomes. Because in addition to his market structure analysis, Hari gives a very interesting philosophical thesis about how you might want to hedge. So with that, please enjoy the full video from Hari Krishnan.
HARI KRISHNAN: My name is Hari Krishnan. I'm a PM Doherty Advisors in New York. And I focus on volatility strategies with an abiding interest in global macro.
Where are your biggest market volatility concerns?
I focus on two areas. And one of them was reflected in my first book. And the idea there was even though hedging is cheap in quiet markets, nobody seems to want to hedge when volatility is low. And investors are complacent. So what can you do progressively as market conditions get worse? To hedge in such a way that you don't overpay for insurance. That was step one.
Step two is more related to try to identify market tremors or market weaknesses, structural problems in the system. And that's the focus of the second book, which will be called appropriately enough market tremors. Really what I'm focused on are situations where volatility is low, but the embedded risks in the system are high. I've been trying to target things from an agent based perspective. Now, that sounds fancy. But really what I'm thinking of is trying to identify situations where markets aren't gyrating very much, but certain players in the market are becoming too big and too risky.
And we're in a situation where a small move in prices could trigger a large sequence of events that could be very damaging in the markets. I did a postmortem like many other people did, on the Vol Magadan in February of 2018. And there, it was fairly clear because you had ETNs dominating the system, and trading according to fairly structured rules as per their prospectus. And you could almost say exactly how much the VIX would explode given an initial impulse or impetus.
Here, it's a bit more authority, because I'm delving now into the corporate bond markets. And I think the agents are more diverse, but the situation is pretty ugly. And I'll go into that in some detail. You now have a situation where the high yield markets in the US are about, let's say, $1.25 trillion. The ETFs, the high yield ETFs, which are predominantly the JNK and the HYG and a few others, they account for about $50 billion.
Now, you might say, well, that's small, that's about 4% of the whole market. But if you look through the structure, you look through the system, you'll see that things could get dicey very quickly. Now, why is that? ETFs are chameleons, especially bond ETFs. The reference basket is a set of bonds, but they trade like shares. So people can easily get confused.
And I'm no- I've been a victim to this too, looking at the realized vol of something like the HYG, which has historically been about 3%, at least since the recent bull market and saying, oh, this is a stock effectively, that spits out a high dividend. And that can be traded in and out, intraday with a refresh quote every 15 seconds or so and so all sorts of strategies that you wouldn't think would be possible in the bond market are now suddenly possible, trading the HYG. This is a mistake. And the reason it's a mistake, and I want to pound home this point, that realized vol is not a true barometer of risk.
By saying that this is a very illiquid market, what underpins many of these ETFs, especially in the fixed income space, is stuff that doesn't move. It's either closely held by pensions, insurance companies and so on, or it's held by mutual funds who are very vulnerable to outflows. If you operate under the illusion that the ETF is trading like a share, you could be in for a rude awakening. Now, the question could arise, what could happen to the ETF? Why would the ETF trade in line with the reference index? And so on.
How can we better understand the problem with ETFs?
There is a middleman in the ETF space. And that's the AP with a set of APs, authorized participants. So basically, when you buy an ETF, you're buying something that looks like a share, but there's an Index Provider. And then the provider which could be a BlackRock or a Barclays or someone like that. And basically, what they do is they appoint authorized participants to try and ensure that the ETF stays in line with the reference index. The authorized participants tend to be dealers in the space, which the index is based on.
Now, let's say that the ETF takes a hit. It's trading far below the reference index value. In theory, the authorized participants should step in, buy the shares from the market at a discount presumably, and then deliver them to the provider in exchange for a basket of bonds. So in theory, if the ETF drops 1% below the reference index, they'll go in, buy the bonds- or get the bonds delivered, and then sell them into the market for an arbitrage profit. But that's on the presumption that the bonds themselves are liquid.
Now, these bonds are not very liquid. And what's happened in the marketplace is that the dealer community is no longer taking principal risk the way they used to. Now, you can look at this in the data. The Federal Reserve Bank of New York publishes primary dealers statistics. And what you'll find is that across the fixed income space, particularly in the higher yielding corporate bonds space, dealers will not warehouse positions. Whereas they may have taken a 10 million slug of bonds previously on a sale, now, they'll only take one or 2 million, and then try and find a buyer for the rest. This slows down the entire process.
Now, the problem with this setup- and again, I want to bang this point home as hard as I can is that you have high frequency traders and other algos involved in the ETF, imagining that those ETFs are trading like shares. And at the same time you have an illiquid reference basket. Now, you might say, well, how many flash crashes have there been? What could drive the ETF price down? One might think that we haven't seen anything since May of 2010. This is not true.
We've looked through the data and what we found is that if you look at limit down and limit up statistics, and you even take a semi-nasty month, like December of 2018, was a horrific month, the S&P was down 9% or so. But what you'll find is that there were 500 instances of limit down circuit breaking events in December alone, which is a huge number. And many of these occurred in the bundled products, the ETFs and ETNs. So, it's not hard to imagine that something like the HYG or the JNK could shank 5% at a circuit breaker would be triggered.
Now, if that were the case, what would happen? Would the AP step in? Would the dealers who've been appointed by the provider step in, buy shares at a 5% discount and then take delivery on the bonds? I would argue no. I don't think they would because a 5% drop on a $50 billion ETN complex or even on a smaller component, let's just take the HYG, would be at least a $1 to $2 billion sale of high yield debt would require that for the arbitrage mechanism to work.
Now, you may then go in and say, well, bid ask spread seem ties. The high yield bond market seems healthy, the corporate bond market at large seems healthy, everyone is issuing corporate bonds. We've had this massive transformation where many corporations in their treasuries are acting aggressively. They're issuing corporate debt, they're either using it as a mechanism for the repo markets, or simply as a way to buy back their shares or engage in other activities that are constricting the supply of stock and ballooning the supply of debt. In theory, this should mean that debt is attractive, it's abundant, and it's liquid.
And again, if you go into the databases, you'll see tight bid on spreads. But again, one should not be confused, as I've mentioned, ETFs give the illusion of a share price for a stock price world in an environment where the biggest growth area is bond ETFs. And when quotes are made in various databases for bonds, these are indicative only. There's no guarantee that you can get filled anywhere close to where the quotes are given.
As well known that for illiquid assets, you'll have say, five trading desks, and the first desk will be called, someone will want to sell some bonds or some inventory and they'll call the first desk and the first desk won't know what the price is so they'll call the second desk, second desk won't know so they'll call third desk, and you'll go around the circle. And finally, the fifth desk will call the first desk and no one will be the wiser. And so no one really knows where the prices should be. No one knows how much depth is behind these quotes. And they're indicative only, they're not binding.
One can easily imagine a situation where a $2 billion sale of high yield debt, if it were necessary, would trigger a 5-point down move in the reference basket of bonds. This would be disastrous. Investors are looking at things like the HYG and seeing a 3% volatility number around it, which is very low. Again, this is an illusion of safety in an environment where everybody thinks they can get out in time. The whole basis of this explosion in algorithmic approaches to risk is based on the idea that, oh, yes, now we have all this data. We can sell and pull risk at higher and higher frequencies, so we can get out in time.
So there's this notion that simply by having a quicker system, risk can be managed in real time. And again, this is the illusion of safety. Because if everyone's doing it while I can imagine very, very sharp jobs that occur more quickly than any risk management system can reasonably get out in. And so what you've got now is a situation where in markets like the high yield market, there's a real lack of true liquidity in the system.
What are your concerns regarding contagion?
One of the important agents that I didn't mention is the mutual fund space. Now, if you think of the US high yield bond space, as I mentioned, it's about $1.25 trillion in size US. About 250 billion is held by mutual funds. And the interesting thing about the bond market that is not so true in the stock market is the way that mutual funds operates, and also the relationship between flows and performance. That's a key point.
Now, for equity mutual funds, at least in the glory days of active investing- and we can go into that too. There were enough Graham and Dodd-type value investors out there, that at least at some level, they would provide a floor for the market. So if equities dropped enough, price to earnings would become attractive or price to book or anything that looks like price divided by one would drop enough value whereas investors would go in and buy. And they weren't so affected by outflows that they were handicapped in this way.
For example, the flow performance relationship for equity funds historically has been concave. That's a fancy phrase. What does that mean? It means that for every percent of outperformance, the inflows are greater than for every percent of underperformance. So you get an asymmetry, good performance is rewarded more than bad performance is penalized. In the bond markets, it's completely the opposite. Bonds are illiquid.
So there is a first mover advantage to selling before everybody else does. And the flow performance relation historically is concave. What that means is that bond investors, even the people investing in mutual funds over the longer term, let's say, are quick to get out if a fund is underperforming. And if you think of the high yield market as a whole, as a segment of the bond market, a 5-point drop triggered by say a single ETF could trigger some significant outflows. And the back of the envelope estimates that we've made suggests that a 5-point drop would probably lead to about 1% to 2% level of redemptions from high yield bond funds, which would trigger another potentially 5-or 10-point drop.
So you can have a sequence of events going from the ETF that has a random flash crash. Again, I can't predict when or if this will happen. All I can say is that it has happened a lot recently, more than most people choose- care to remember, even in the last year. And if it were to hit the high yield complex, it would quickly funnel into the mutual fund space.
Now, you might say, well, even taking that into account, we're only at about 300 billion total. Only is a strong word, I would say. But who holds the remainder? You've got these hold to maturity characters out there. I probably shouldn't call them characters, but they're pension funds, insurance companies and other people who are trying to manage their duration risk.
Now, we've had a few bullets which have been dodged in the past four or five years, especially in the public pension fund space in the US where people have or institutions have quite a bit of latitude in terms of how they apply discount rates. Now, buying and holding some maturity corporate debt, especially high yield debt is a nice way to apply a good fat discount rates to your liabilities, creating a better balance sheet situation. At least assuming no capital goals.
These investors are unlikely to step into the market and act quickly. Now, there are buyers of last resorts. Again, I won't go into the names, they might be hedge funds, they might be in the distressed space, and so on, but they're going to require a significant can discount to come in play. And if you think of the way that the corporate bond markets have been extremely compressed globally, but also in the US over the past many years, I could easily argue in terms of a major liquidation in the event of a 10-point down move. Even if it's a flash move.
Bond investors are not used to this. It's almost a new generation of players who haven't seen anything like this in the past 10 years. I would argue that this is a market where there are some significant market tremors, there are no indications of volatility in terms of realized movements. But there is significant downside risk.
Now, the way I think about this is I'm trying to do responsible scaremongering. That's my brand at this point. And responsible scaremongering to me means this, it doesn't mean that anything could happen at any time. Of course, that's true. What it means to me, though, is that if there were an initial move that were a random shock, maybe even within a normal distribution with 3% to 5% volatility for the HYG, what can you see in terms of follow-through? What will the ramifications be? Who would be flushed out of the market?
I can give you two examples. Imagine a situation where you have a single asset, there's just one asset in the world, it's trading at 100. But there are two scenarios, there are two parallel universes or planets, and the asset trades in both. In the first case, the volatility of the asset is say 5%. But there's no overhang. There's nobody who's levered and will be forced out if that asset goes to 95.
In the second case, there is a whole series, there's a litany of stock losses, five points below or three points below. They're all these people positioned where they have to get out if there's a down move, which is riskier. In both scenarios, the vol is 5%. But in the second scenario, you can imagine a normal size random shock, leading to a cascade of selling. That's what I'm looking at. That's what I'm thinking of in terms of agent based risk.
I'm not trying to solve the whole package here. I'm not saying who are all of the agents, and how do they interact. I'm not coming up with a massively big data representation of the market. I know there are groups that do this. I believe they're getting some interesting theoretical results. But I'm more interested in what physicists I call a mean field problem, where you've got all these agents and in classical economics, you think of all of the agents as being averageable.
Each of them have slightly different constraints, preferences, and so forth. But you can still build an efficient frontier or whatever by just averaging them all together. And then the world of neoclassical economics emerges. I don't think that's a good place to be. It highly understates risk. It's a misrepresentation of changing structural dynamics in the system.
But by the same token, I don't think the full-blooded agent-based representation is possible at this point. We don't know who all the agents are. We can't track all the transactions in real time and make sense of them in a holistic way. Whereas what we can do and what I'm trying to do in a lot of the research I'm focused on is to say, well, can we ever find situations where one or two agents become big enough, that we can map them out, we can say what they're likely to do, and then average everybody else.
So you have two forces at play. One of them is a distribution that's generated from all of those neoclassical assumptions. You got a distribution, it generates normal looking shocks, these are random. But then you have a few big agents that are emerging in the background, whether it's high frequency traders in the ETF space for high yield bonds, or it's somebody else. And you say, well, if there is a random shock, what's the follow-through likely to be?
That's what I'm focused on. And I think the high yield space is ripe for this, not because there's something imminent, again, I'm focused on doing things responsibly. But where I can say something about initial random shocks