Today I am joined by Hari Krishnan, Head of Volatility Strategies at SCT Capital and author of the books Second Leg Down and Market Tremors.

We begin with a discussion of Hari’s newest book, Market Tremors, and the main theoretical idea: Mean Field Theory.  Hari lays out both the philosophical underpinnings of the concept as well as how one might interpret it in practice.  This leads into a natural discussion of dominant agents, including examples of who they are, how we might go about identifying them, and why they are so important to consider.

In the back half of the conversation, we tackle some more practical considerations of tail risk hedging.  This includes key differences between equity and rates markets, how we might structure hedges in today’s market environment, how to navigate path dependency, and why it’s all just a “bag of tricks.”

Please enjoy my conversation with Hari Krishnan.

Transcript

Corey Hoffstein  00:00

Okay, Hari, well, are you ready? Cool. All right 321 Let’s jam. Hello and welcome everyone. I’m Corey Hoffstein. And this is flirting with models, the podcast that pulls back the curtain to discover the human factor behind the quantitative strategy.

Narrator  00:22

Corey Hoffstein Is the co founder and chief investment officer of new found research due to industry regulations. He will not discuss any of newfound researches funds on this podcast all opinions expressed by podcast participants are solely their own opinion and do not reflect the opinion of newfound research. This podcast is for informational purposes only and should not be relied upon as a basis for investment decisions. Clients of newfound research may maintain positions in securities discussed in this podcast for more information is it think newfound.com.

Corey Hoffstein  00:53

If you enjoy this podcast, we’d greatly appreciate it. If you could leave us a rating or review on your favorite podcast platform and check out our sponsor this season. It’s well it’s me. People ask me all the time Cory, what do you actually do? Well, back in 2008, I co founded newfound research. We’re a quantitative investment and research firm dedicated to helping investors proactively navigate the risks of investing through more holistic diversification. Whether through the funds we manage the Exchange Traded products we power, or the total portfolio solutions we construct like the structural Alpha model portfolio series, we offer a variety of solutions to financial advisors and institutions. Check us out at www dot Tink newfound.com. And now on with the show. Today I’m joined by Hari Krishnan, head of volatility strategies at SCT capital and author of the books second leg down and market tremors. We begin with the discussion of Harvey’s newest book market tremors and the main theoretical idea of mean field theory. Hari lays out both the philosophical underpinnings of the concept as well as how one might interpret it in practice. This leads into a natural discussion of dominant agents, including examples of who they are, how one might go about identifying them, and why are they so important to consider? In the back half of the conversation, we tackle some more practical considerations of tail risk hedging. This includes key differences between equity and rates, markets, how we might structure hedges in today’s market environment, how to navigate path dependency, and why it’s all just a quote bag of tricks. Please enjoy my conversation with Hari Krishna. Hari Krishna, welcome to the podcast excited to get you on. Right before we started recording, I said never really want to talk about what date or time we’re recording this because I like these to be evergreen. And yet, I think I’m gonna break my own rule right out of the gate. This is June 16 2022. And it’s been an interesting week, maybe it’ll come up in the episode. But it seems like a very timely day to be talking tail risk management, which is what we’re going to be talking about on this episode. So thank you for joining

Hari Krishnan  03:16

me. It’s a pleasure, Cory. I’ve followed up with interest for a long time. So I’m happy to be here.

Corey Hoffstein  03:22

Well, before we dive into all things technical, let’s maybe start at the beginning with your background, because I do know your path into tail risk management wasn’t necessarily direct. And I’m a big believer that those formative years ultimately end up being very telling about how we think about the world. So maybe you can step back and tell us how you really got into the industry.

Hari Krishnan  03:43

Well, I occasionally get this new idea, this buzz in my head that I have a new idea. And I’ll walk around and think about it. And I’ll realize that I actually did have that idea 15 years ago, or 20 years ago. And it just was kind of circulating in my brain. And now it’s back. And so yes, the path that you take to whatever it might be that you do, whether it’s tail risk, or something else has a big impact on your thinking. For me, it’s pretty simple. I was probably groomed to be a quant even though I didn’t realize it in graduate school. I had gotten interest in chaos theory as everyone else in my generation did by reading James likes book, chaos. So I pored through that as recreational reading and I thought I wanted to do this. And I wanted to do it because it’s kind of cool. You know, people like the sound of it. And there were all these pretty pictures. When I used to go to a running club, the coach always said, gladly do chaos, because your life is just a mess, you know, so it suits you well. So I cleaned up since then, but I did that and then I once I got into finance, I thought this isn’t going to work. No one’s interested in chaos theory. And the reason no one was interested is you couldn’t predict anything. You could say things like well, if the world were a bit different if the inputs were a bit different, you’d get you could get wildly different outputs you need to control for that instability. People would say, Well, how do I make money off that, and there was no clear path to be found, at least for me. So I started doing traditional things, I jumped around a bit. My first real job was helping a an options market making firm, build off floor hedging models, they would trade individual names and sectors and the options markets. And historically, the idea was you had to just get out of your positions or delta hedge everything at the end of the day, on a position by position basis. But you know, Blair Hall and others had this idea that maybe you could aggregate some of the volatility risk that was embedded in the portfolios or in the books of various market makers. And so we tried to develop Vega hedging models to do that. So we looked at things like I don’t know relative value in implied volatility, dispersion, implied correlation matrices how best to estimate them. In the absence of full information, all that jazz, we did that. Subsequently, I took a big jump, there was a fork in the road where I thought of joining a CTA, they recruited me to systematize their discretionary trend following system. But instead, I went to Morgan Stanley, and took a proper looking job building asset allocation models for clients. I didn’t actually trade, I just had various longer term risks slash predictive models that I built, that were used to set templates that were then presented to institutions and ultra high net worth. So I did that for a while I learned about the business, I didn’t really learn that much else. And then I had a yen to start going off and doing it on my own. So I ran a CTA for about two years, which did quite well. It was used as an ATM in 2008. The other business I was involved in which was kind of a fund to fund a hedge fund to funds business didn’t do as well, because of the liquidity mismatch between the funds I had invested in, and the hedge. So that was a bit of a challenge. And so from that point, I thought, I’m going to try it myself, I’m going to invest myself, I don’t want to outsource things that I don’t fully understand had since then, I’ve been doing macro, slash tail risk, which I didn’t I didn’t even know the term until fairly recently, long vol and currencies as well. So I do overlay management predominantly now. But I have a significant background and macro having run an FX fund for a while, and a volatility fund as well. So that’s kind of my background. UK, US kind of a hybrid. So there you have it.

Corey Hoffstein  07:34

The one thing you didn’t mention, at least explicitly is early on you were actually working in Weather Derivatives, if I’m not mistaken. Oh, I forgot. I forgot that. That. Yeah, I find Weather Derivatives to be fascinating. And I know you haven’t necessarily been working with them recently. But I’m curious, how would you think about the similarities and differences between markets like Weather Derivatives versus more traditional financial derivatives?

Hari Krishnan  08:04

This is a pretty complex question. But one main difference is that you cannot hedge on the underlying, I haven’t seen anyone buy and sell a sunny day or the level of rainfall in a region over time. So if you’re going to trade options on the weather, you basically making a directional bets. That’s point number one. So maybe Weather Derivatives are a little bit like the VIX. But they’re not the VIX, because they cannot be assembled from tradable financial instruments, however, roughly, they’re also bit like fixed income, in a way because at least the historical models have a degree of mean reversion priced in. Maybe that’s all broken nowadays. But historically they did. The classical model, I think, which was due to Vasa check had an explicit mean reversion and subsequent iterations also did. And the weather is a bit like that. It doesn’t stray that far, probably it’s even more well behaved than rates are. But by the same token, it’s not tradable in the same way that rates are because rates can be decomposed into forward rates, and it can be spread across country cross credit, blah, blah, blah. And with the weather, you’re kind of stuck. I mean, yes, you can trade Cincinnati versus Columbus, fine. But it’s much harder to assemble the payouts of the contract using other contracts. And so it’s, I never traded the stuff I was involved in a pricing system for it, which was built by it. I think it was a PhD from the University of Chicago physics. He built the thing. I analyzed it, tested how much variation the model produced relative to the real variation in the weather and so on. So it’s pretty different. It’s totally indirect, but there was a reason for it, which is that I was working at the Earth Institute at Columbia at the time, and I was looking for a job and I was looking at the weather. So I put two and two together. Maybe I should mix the weather and It’s a real job. And so I did.

Corey Hoffstein  10:03

So let’s now fast forward almost through the rest of your career and land in September 2021. And you, at that time just published your second book market tremors, which in many ways, as I read, it seemed like a spiritual successor to your first book, second leg down. But perhaps a little less practical than second leg down was and a little bit more academic in nature. I’m curious as to what sort of the intended endgame of the book was when you were writing it.

Hari Krishnan  10:35

One motivation for me through my career for work has been that I was dissatisfied with something that I did before. And so the first book, although I was happy with it, because I felt it was authentic, I knew that I’d written everything that I believed, maybe I left a few things out. But it was an accurate representation of how I felt about hedging for chapter eight, which is the final chapter in the second leg down was sketchy. I didn’t really understand central banks as well, how they work. I didn’t really understand the various crisis prediction models, what what could legitimately be done. So I didn’t really have a foundation for saying, How can we modify standard risk estimates using things that traders flow to people who are aware of flows, people who are aware of leverage, hopefully incorporate into their assessment of sizing. So I wanted to tell people basically, don’t do too much. If there are these latent risks in the market, you want to trade that aren’t being reflected in the price action of the assets that you’re trying to manage. So I felt that that was a necessary book to write because I wanted to triangulate between the more the very good but more fanciful book by Rick Bookstaber, the end of theory, which I thought was excellent, but it didn’t, you couldn’t really take anything too solid away from it, I wanted to sort of bridge that gap and come up with something that looks a little bit like a standard risk model, but made adjustments in the presence of credit and positioning risk. And so that was the goal of the book. You’re right, it’s less practical, it’s more open ended, it requires more DIY type thinking where every situation is different, you might miss the relevant situation that could make the big fortune or safer, safer skin. But I felt that such a book had to be written. And when I, at least I wanted to write it. When I heard about mean field theory and so forth mean field games, I thought that was exactly the right framework to use, because it gave at least some hope for coming up with concrete risk estimates. In the presence of whales,

Corey Hoffstein  12:39

you lead right into where I want it to go. Because I might be mistaken. But I believe you said that mean, field theory was, in your book, the main theoretical concept that you were trying to get across, I think, I think that’s the quote directly your book, the main theoretical idea of the book. So maybe you can explain what is mean field theory? And what are sort of the theoretical holes you’re trying to plug in using it?

Hari Krishnan  13:02

Sure, there is a whole branch of physics called statistical physics. And it basically looks at averages, averages of large numbers of particles. And often it’s easier to measure the, to measure average quantities than it is to do everything at the molecular level. So temperature is a good case in points. I’ve given this example before, but I’ll give it again, how does a room get Haas? Well, there are a bunch of particles buzzing around with high kinetic energy. And that energy is released as he’s, that’s hard to model because there are lots of tons of particles in any given room at any time. There lots of interactions, potentially, and so on. So what’s much easier is to think of volumes in space, instead of thinking of particles, so fixed volumes in space, and to model the variation of temperature in the volume. So we know that heat spreads out over time, it diffuses and it becomes more even, that’s in the heat equation that’s almost in Black Scholes in some sense, which is kind of the heat equation. And it applies to networks. And the crucial condition that the mean Field Game Theorists talk about is indistinguishability. Which means that so long as no two agents behave, or agents behave approximately in the same way across the network, the average configure configuration is good enough, the average analysis is good enough. And what that equivalently means that is that if you took out any agent, or any handful of agents, the system would look about the same. So if you believe that, and you believe classical financial economics, at least some loss, you can sort of replace the network dynamics with a distribution say pick an asset, you know, the 10 year or the s&p or something. And you look at the distribution of returns historically, and you say, well, as long as things have to change too much, that’s probably a good way to think about risk. Okay, it doesn’t cover the tails. It doesn’t account for changing dynamics over time. And so On, but it’s okay. But what happens when there are major players who are distorting the network, mega players whales, call them what you want, then you have to use a modified, you have to use the network, at least at some level. So the core idea of the book was to say, let’s assume the distribution is okay. But the risks can emerge when players get too large too levered or too active in a given direction, at the margins. In that case, you need to model the interaction between the network ie the historical distribution and the actions of the mega agent. So if you believe that the mega agent will have to sell if a given market goes down by 5%, and sell it sighs, then if you can estimate the impact of selling in that size, you have to adjust your risk estimate if Corey or or I or whales, and we were going to sell 10%, or even 2% of the open interest in selling futures contracts, it could have a negative impact on price action, at least in the short term. So the goal of the model was to take an initial random draw from the the original distribution and see if it would force the whale into action. If so, estimate the follow through and come up with an adjusted number. That’s the goal.

Corey Hoffstein  16:15

Again, you led me right to where I want it to go. Because I feel I love this concept right to me, I sort of interpreted and maybe this is incorrect, but I interpreted as saying, there is a distribution that becomes conditional upon certain actors when those actors have an outsized influence in the market. And if they are forced to act, it can have cascading effects, and therefore, meaningfully change the potential return distribution, going from theory to practice isn’t always easy. So I was hoping maybe you could talk a little bit about practically how mean field theory ends up getting implemented?

Hari Krishnan  16:50

Yeah, well, roughly speaking, you need to know who the players are. Who can be a well, one of the whales is increasingly known, at least in the FinTech community, and so on, which is dealers, market makers, in options. And the basic thesis is that they that institutions like to do certain things they like to buy puts on equity indices, maybe sell calls, and so on. And so dealers take the other side, they typically don’t have big balance sheets like the institutions do. So they have to hedge actively. And if they’re short options, and the market starts moving in the direction of the shorts, towards the strikes, they have to start hedging in a way that exaggerates moves. So that’s been talked about by squeeze metrics, and Jim Carson and various other people. But it’s a good case where you can at least heuristically if you know roughly where dealers are short in size, you can make certain assertions about how much movement you expect in those regions. That’s been done to death recently, much better than I do it. So I’ll start with that. But there are various statistical techniques applied in the book to explain why there was such wild gyrations in March 2020, over and above changes in the news flow, and why markets get pinned in certain conditions. And so that’s an obvious one and other now the big one was some of the Etn and ETF providers. I had talked about this for quite some years. One of them was the VIX, of course, others have talked about the volume again. But I think what ash and I did ask Paddington what I did was we came up with the first fully quantified specification of how much follow through could be expected, given a garden variety initial spike in the VIX futures from month vix futures in January and February of 2080. So we were able to do that, I needed some help, because my math skills have eroded every year that goes by. So there was a there is a mathematician Stefan stone at Worcester Polytechnic Institute, he helped me with some of the models. And so we were able to actually to glue together a follow through model that was pretty close to what happened on February the fifth or sixth as the fifth in 2018. That was another one, the ETFs are in the book, they just talk about liquidity mismatches. But the final one, the elephant in the room is the central bank. Central Bank is obviously a whale. And they’re increasingly acting as well as given the size of their balance sheets and the impact they have, not only on markets, but expectations of market movements. You know, what I’ve been thinking about a lot in that case is coming up with connections between balance sheet size and equity market moves. This has been done by others as well but have some interesting models that do that. And also thinking about inflation, and the cycles of inflation and deflation risks that can occur based on changes in central bank policy over time. And I’d love to talk about that too, but I’ll leave it to your cue.

Corey Hoffstein  19:48

Well, when I put your book down at the at the risk of sort of miss characterizing your own work, I sort of interpreted the main thrust being that historical distributions of different asset classes. As can be applicable, if you sort of have this small network of players that are all equally influential, but when you get a dominant agent present, there will inevitably be a latent risk that is just never fully captured by historical risk measures, which makes the identification of that dominant agent, really critically important. And you just laid out many that you discuss in the books, whether it’s levered ETFs, or central banks, or option dealers. But I’m curious how you think about identifying those dominant agents on a go forward basis?

Hari Krishnan  20:37

Well, it’s largely asking route, I mean, you know, you can go to banks and say, what are the big flows that are going through what sort of structured products is selling? Well, who’s buying them? Why are they buying them. And then you can go into if these are hedged using listed options, you can go in and see where the open interest is clogging up and try and make decent guesses about how institutions are positioned. So if I knew nothing about this, let me start with the most basic thing, I might look for distortions in the relative prices of options over time. And across strikes. Even naively, if I thought that the put skew in the s&p 500 was too steep, given what had happened. Historically, there are two things I could say. One is that historically, there haven’t been returns as big as the true return distribution in my head would suggest. The other is that there’s some structural pressure that’s pushing up the skew, people are over bidding for it because they have constraints or needs in terms of what they need to deliver to their clients. So even looking at distortions in the skew, and trying to disentangle the ones that are based on non structural reasons from ones that are is a useful exercise, I’ll give you an example. Often there’s a dip in implied volatility across markets in December, the Santa effect, it isn’t just the Santa rally anymore, it’s the Santa squashing volatility I know. I think it was in 2018. December was a bad month. But historically, December has been pretty stable. And those sorts of things aren’t really based on flows, so much as on structural biases so much as the ways people think about markets. But other things are caused by structural distortions. And that’s the sort of thing you want to look at. And you’re right, that the book is less pragmatic than the first one. And the reason for that is that it does require domain knowledge. That’s something I try and emphasize in the book. I mean, offline, I could talk to you and say, What do you see. And I like to know someone in rate someone in equity land, someone in physicals, and hopefully someone that affects as well given that I don’t do that anymore, and just comparing notes, not to copy their trades, but to see what they’re seeing going through is vital. And that’s one of the takeaways. The other one, which I think is easy to implement, is if volatility seems artificially low, in some asset, don’t do full size. There are hedge funds who have financed positions using pegged currency pairs that didn’t go well, in many cases. There are people who have rigidly sized positions in a very dynamic way as one over volatility that often doesn’t work. It probably works okay for trend followers, because they’re intrinsically long gamma. But if you’re not, I think that’s a very dangerous play, meddling with your sizing too much even and CTA land, in my humble opinion, is a way to sort of trim the convexity, the natural convexity in your portfolio. So maybe that’s not ideal either. So sizing, I think should not be one over variance, or one over sigma, whatever volatility needs to be, especially at the extremes, it needs to be more less dogmatic than that, in my view,

Corey Hoffstein  23:53

you started to get into it a little. And I want to push maybe a little bit further. Sure. Because after reading this book, and starting to think about dominant agents, I think you start to see dominant agents everywhere in markets and maybe misclassifying them, but they are almost certainly in the water in which we swim. And if we want to be invested, I don’t think there’s any other choice but to recognize that these dominant agents are going to influence our returns. And so my question becomes, how do we think about playing defense in the presence of these dominant agents?

Hari Krishnan  24:26

I think I did write about that. Let me wrack my brain here. sizing is one important thing. Avoiding overcrowded strategies is another important thing. And now the easiest way to avoid an overcrowded strategy is to look at the stuff that’s performed the best not at the asset level. I’m not speaking of oil here, or metals or something I’m speaking of value or certain types of short volatility trades. You don’t want to be the final person to get to the party for those things. So avoid crowds in things that have bounded out upside and unlimited downside, that’s a big thing. I mean, it’s hard to do. I remember reading a lot of the stuff that DDA saw net did, and others where they were looking for parabolic price moves in anticipation of a crash. That’s one way to do it. But you miss out so many of the short volatility trades that are actually grinding up until they collapse. That should be a big takeaway in the book, if you see something with an unbelievable Sharpe ratio that doesn’t have Renaissance attached to it, or it’s not some form of high frequency trading, don’t do too much, whether it’s made off the grinding up trade, which is making money every month, or the guaranteed the asset back lender who guarantees you 1% per month in a distressed market forgot about fraud, the risk is not apparent in the return stream, and generally should not be should be sized appropriately. I think that’s essential. If something doesn’t move, that doesn’t mean it’s safe. In fact, it may just be highly illiquid, or artificially, it’s basically performing some alchemy, where there’s a natural volatility, and it’s being pushed through this funnel, but at the cost of an explosion on the other side,

Corey Hoffstein  26:12

I think that brings up a really natural comparison, right, because I think a lot of people would say five years ago in a market environment like 2017, a lot of the incredibly low realized volatility that we saw was due to that short vol dominant agent player that was in the market, whether it was dealers large funds that were short vol or some of the short poly teepees that had gotten very popular, you can compare and contrast that to today’s market, we have much higher, persistently higher levels of implied volatility. And I would argue it’s more economically event driven today than it was back in 2017. I was wondering if you could maybe do a little bit of compare and contrast how you think about structuring tail hedges in these different types of environments. 2017 versus 2022,

Hari Krishnan  27:02

I’d like to go to 2006 2017 and 2020s. If you don’t mind, floors, yours, Halifax will appreciate that 2006, the banks were still mega players. And 2006 was a big volatility suppression here to the banks, from my experience always had the viewpoint that they would ride the horse until the horse fell over. So if the horse was making money for them, or if something was working, they would just keep doing it, they knew that their downside, individually was bounded, and their upside was big. So you’d get a lot of long, and the two things that would eventually unraveled 2017 was somewhat similar, but probably less bank driven, there was artificial suppression of volatility. But I think by that point, there was the realization, it takes a long time for us to realize these things that interest rates probably weren’t coming back up soon. Of course, the realization was earlier than that, but it was clear at that points. And so people were desperately looking for proxies for carry, then has been ahead of the curve. And all of this, certainly, and you know, the carry trade is bigger in Asia in many ways in Korea and Japan than it is in the US. But various people wanted to substitute carry for alternative forms of carry for yield, classical Treasury based yields. And that cause volatility to be trading at incredibly low levels. Now, it’s hard to manage. On the one hand, it’s easy to manage such a book, you just go out into the warehouse, cheap of all, and you put it in the cupboard, and you don’t look at it and you do other things. You don’t micromanage your portfolio, you just put it in there. And whatever you do for the rest of your investing, you do that, that you don’t try and nickel and dime at the margins, because you’ve got all this cheap convexity embedded in your book, which allows you to take risk everywhere else. And of course, eventually, that blew up people made a huge amount of money in February 2018. I was actually not one of them, because I was moving from one gig to another. So sad to say I wasn’t the beneficiary of the thing that I’m somewhat known to have talked about, but so be it. But that blew up. And so there were lots of opportunities that market tremors was trying to indicate where there was a zombification of markets. And I was pushing hard to get the book out because I knew that the cycle could turn, but I should have pushed harder. I didn’t think it would turn when it did. And now that brings us to 2022 I think a good lead up to 2022 was December of 2021, where the VIX was pretty high. And most of the points, s&p options across various tenders were quite expensive. Implied volatility was high. In fact, until today, I don’t think it’s much higher now than it was for a fixed term than it was at the close of yesterday. It might be about the same. That caused a huge problem for the tail risk community. Now, I have never defined tail risk because I don’t know exactly what it is. I don’t know what a benchmark is. But I do define hedging as being a number of things and a number of markets. But if I had to pick one market, let’s say the s&p Pay. It’s a combination for me of three things gamma hedging at the short end spread trading for middle term maturities, let’s say three to six months where other people like to buy insurance and Vega plays, where interest rates slash Vega plays going out at a year or more. So I break the book up into those pieces. And as you point out, 2022 has been quite different because the standard tail risk idea, which is to lock and load, let’s say 20%, or 25%, out of the money put hasn’t works. I haven’t checked my screen recently, but not concerned about that, specifically, why hasn’t it worked, because volatility didn’t go up enough from where it was. And he also faced a headwind, where if all goes from 10 to 20, you make a heck of a lot more than if it goes from 30 to 45. Because you can just buy that many more units of protection, you can buy that many more contracts, when fall is low. So if you have a fixed budget, you can really go to town, if volatility is low, you have to be much more conservative, conservative if it’s higher. So that’s been one problem. The other one is that the market has gone down, at least up to now in a trend where the savvy slow way, which points to the idea that just this idea of value, buying volatility, and waiting for the big one doesn’t always work, the evangelism of tail risk, which I have been somewhat involved in, but nowhere nearly as much as Nassim Taleb and various others, Spitz Nagel and others probably has changed the nature of the game a bit. The tails are the true tails are pretty thin, they’re not actively traded. And you could argue that they’re underpriced. We could debate this ever appear one day, but I’ve always said that they’re on price. And via the tails is just a way to keep it in the game. You’re almost pricing different up to a point as to what the cost of a one month 30% out of the money put might be. Of course, if the VIX is at 70, maybe you care then but otherwise, generally not. And you try and do all the stuff in between. and So classical tail risk hedging, which again, I haven’t defined, because I don’t know exactly what it is static buying puts from a value perspective, or even mechanical buying puts hasn’t worked. And a lot of people who bought long vol had been confused about this. But as you and I know, that’s one good reason to mix trend. And while they are nice partners, they’re nice partners in crime because trend doesn’t require a repricing of risk. It just requires moves of big enough magnitude Vall doesn’t require necessarily big moves. But if something comes out of the blue, that changes the perception of risk, it can be hugely, hugely beneficial. So they’re good companions.

Corey Hoffstein  32:45

somewhat frustrating thing about this year from talking to different practitioners has been that increased levels of vol of all these sort of swings between say, that just means the VIX, vix at 20 versus vix at 35. Increases the sequence risk of instruments that have convex payoffs, you know, when you choose to roll or how when precisely you monetize becomes a lot more important, the more violent those swings become, how do you think about trying to manage during periods of heightened path dependency? In the

Hari Krishnan  33:17

futures market, some people use explicit stops most of us well, I don’t do a lot of this now. But most people do have exit points for futures. And so they ride winners, and then they cut losers at some point, or they take profits at some point. But only after a reversal, at least in the classic trend following setup in options in the options world, it’s much harder to do that having stops for one month to maturity put is a pretty tough thing. Because you can have made 10 times premium paid one day go out for some food or something the next day that you’ve given everything back. And that goes back to your point, which is that you’re not only have directional risk, you also have vol risk. So if the market bounces and volatility gets crushed the next day, you may have given up everything. So that begs the question, how do you trade these things? And the one thing that I would say is what I pointed to earlier, which is mix up different tenors don’t just load into one trade. You have discussed this as well don’t just buy one strike by a range of strikes, take some profits along the way, but be increasingly patient. Once you’ve monetized something. In other words, I there’s nothing better for me than buying an option at 10 having a go to 40 Taking off 25% And then just saying I’m letting it run. For longer dated options. You do have the luxury occasionally of working stops. They don’t reprice that quickly. Now, there’s a negative side of that too. They don’t reprice in your favor that quickly either. But you can use more traditional techniques with them. And I don’t know if you know, I think Jerry Hayworth has talked about this and various other people that if you buy stuff that say has over a year or two to maturity, you can work downside stops. So you might Buy at 100 goes to 500. And you say I won’t sell until it goes down to 250, something like that. And it’s a legitimate thing to do, because it’s not going to be whipping around as quickly as your one day lottery ticket that maybe up 100 times. And that gives it all back two minutes before the close.

Corey Hoffstein  35:18

We’ve been talking mostly about options in the equity market, because I think that’s sort of the the dominant example, when people talk about tail risk hedging for most people that the big risk that they have is the equity side of the equation. Not true for all institutions, though, where primary risks might be on the right side. And I think even most individuals, and certainly professional investors are becoming more aware of the tail risks in the rates market in 2022. And I know you have started offering tail hedging mandates in rates and equity and rates, markets have their own unique economic drivers, as well as dominant agents, I was hoping you could spend a little time talking about maybe some of the structural differences that affect how you think about constructing tail hedges in those markets.

Hari Krishnan  36:06

That’s great stuff. Well, that’s why the clock back a couple of years, I have some friends who used to trade SKU and rates, let’s say in the 2000s. And then they just quit doing it because the SKU went away. So let’s pick us the most the easiest thing to conceptualize, let’s say options, two month options on the US 10 year note futures, the SKU became a little bit of a chameleon over time, where if, in those brief periods where Bond notes were in a downtrend, a bit of a Putski would form. And whenever there was a melt up in notes, or bonds, meeting a meltdown and rates, a call skew would form as people went scrambling to protect against a crash. So the SKU became a little bit of a chameleon, and rates were low. And I would argue that in the period from 2010, to 2019, the SKU was fairly priced. But volatility across the board was a bit depressed, it was quite depressed. So it was hard to make money by trading relative value along the skew. And you could sit there being long vol for years and make nothing. So that’s one big thing that has changed. Now. I mean, even going into this year, rates, fall was pretty attractively priced. And that of course is all changed rates are now seem to be the source of real fall. And I think rates fall has been a more attractive place to make money in 2022, than equity vol. In many, many respects. That’s where the real risk is emerging from. Now, if you ask me how I think about all this in terms of hedging, I’ll give you a few ideas, I’ll just throw them out there. If you go back to the beginning of the year, already, there were rate hikes beginning to be priced into the US curve. And the hiking cycle was predicted to be earlier than that in Europe. So the ECB had no view looks at your eyeball, you didn’t see many if any hikes priced in for quite some time, negative rates were assumed going out well over a year. And that is the sort of thing that the gives opportunity. So even if you restricted yourself to trading the European versus us curve, there will be leads and lacks. And historically, the ECB has lagged the Fed. And so if you if the Fed is doing something, and the ECB hasn’t done it, yes, you can assume that they may do it in the future, especially. I mean, of course, the inflationary environments different in Europe, you do have to factor various things in but even mechanically, you can do trades like this, the more interesting thing that I’ve been thinking about, please stop me if it’s confusing, because I haven’t written about it. Yes, is regarding central banks in general. So I’d like to take an off piste moment and just chat with you about it. I hope you don’t mind. So I’m friendly acquaintance of the craft sia reichlin, who used to be the headquarters of the ECB. And so I used to go to her office now together, and I was friends with the various people there. And she went along with another Italian Domenico giannone, developed a model for estimating GDP now using a bunch of other releases. So it was a big statistical model. And the Atlanta Fed now does it, which people know about, which basically compresses all the data releases that occur at different times into a single number that, hopefully is tracking the growth rate of the economy today. So predictive is trying to specify that that’s a big model. It’s been a quite a bit of interest to people. And over time, people started to other people started to ask, Well, why don’t you have an inflation model does the same thing. Now MIT has the billion prices project and the various other people who have things but finally Lucrezia and her team came up with a model that may not be true in Maybe entirely false, but accurately reflects based on numerous releases from Fed governors. And now Janet Yellen that basically characterizes the Feds view of inflation. And I’d like to take one moment to discuss that. And please cut me off when you want. There are basically three factors to inflation according to their model, which as I said, may or may not be good, but is accurately reflecting what the Fed thinks. First model is trend inflation trends, that is anchored by expectations. And given the admitted genius that the Fed, even if inflation trend is running a little hoarse. corporate executives know that the Fed will get it right, eventually, the inflation expectations going forward will be relatively tethered to their target level, let’s say 2%, two and a half percent, that this has been true in the previous decade. The second factor is kind of a Phillips Curve factor, which relates outputs where the output gap to inflation. So in other words, if the factories are running costs, relative to where they should be, relative to an optimal state, then inflation will pick up and vice versa. So you get this kind of medium term cycle that can be controlled with monetary policy, according to the Fed. So as they raise rates, growth will slow and inflation will also be temporary. And the final factor is things like oil prices, anyone out in the real world worries about oil price spikes, at least to the extent that they have occurred. But in the Fed model, if you believe this characterization, it strongly mean reverting so not to be worried about so if it doesn’t, if inflation doesn’t rear its ugly head expectations or in output gaps, then fed doesn’t have to do anything. Okay, fine. Now, what does this mean in practice? Well, the Fed is also worried about a deep recession. So there’s a loss function. If they decide to raise rates to adjust the Phillips Curve PCE, they’re gonna hit growth. And given that we’re not growing that quickly, if they hike too much, they’ll hit their loss function, and they’ll have to stop liking. Now the problem, as I see it, is the difference in delays. If the Fed raises interest rates by a points, it takes longer for that, on average, to affect inflation, especially sticky inflation than it does to hit growth. So the inflation dampening that they’re hoping for may never occur. It won’t be observable before we hit the skids. And they have to stop. So what I’m seeing here now is a sequence of events where huge amounts of inflation are priced in, and then there’s sudden episodes of sharp deflation, and vice versa. And so I think the trade going forward is weighing the risks along the US and European yield curves, among others between steep hiking, unusually steep, hiking, unusually steep yield expansion at the longer end, for other reasons, and a recessionary shock. And the market has sort of sort of calmed down to that. I mean, at least until recently, you saw a slight dip up to June 23. In the Euro dollar curve, which was an implication the implication was, they cannot keep hiking, as far Some people felt they couldn’t keep hiking as far as had originally been thought or as you might originally extrapolate. But I think the full ramifications of that can be exploited in a hedging program. And that’s my sales speech for the for the day.

Corey Hoffstein  43:34

I think this is what you were alluding to, in our pre call when you said that he thought you could demonstrably show that it was theoretically impossible for the Fed to even control inflation at all.

Hari Krishnan  43:46

Yeah, without pushing the economy into a horrible recession. Yes, into an unpalatable recession.

Corey Hoffstein  43:53

And I think that’s counter to a lot of the narrative, which is about can the Fed ultimately break the back of inflation and engineer a smooth landing? And I mean, obviously, it seems like there’s profound implications of your belief that perhaps that’s not possible.

Hari Krishnan  44:09

Well, I can only accuse the Fed of hubris, not of doing the wrong thing all the time nowadays, maybe they did the wrong thing for the previous decade, or specially in 2021, when they could have been a little bit more hawkish. But if the problem really is is I have said it out, which is that the control problem has no solutions. Given this current state, again, putting my chaos theory has on then it’s not that the Fed is messing it up now, or that it’s doing a great job. It’s more than it’s a bit out of their control. It was much easier the other way you could if inflation was structurally low, it wasn’t that hard to kickstart the economy by pushing up the leading indicators of the economy, stock prices, asset prices, improving the shape of the yield curve and so on. Going the other way. If the delays are as I say they are, of course, the delays are hard to specify, then it’s much harder because you don’t Add the desired effect to inflation, pushing it down until you smash growth. So I think they had an easy position where if the delays are as I states, it was much easier to stimulate than it is to withdraw stimulation.

Corey Hoffstein  45:14

Is it an overgeneralization to interpret that view is maybe simply delays are ultimately a source of instability? I mean, certainly that’s, that’s something we’ve we see in it has been along documented in supply chains would certainly seem relevant from a policy perspective. So is that sort of the evidence you’re seeing delays are a source of instability? And that ultimately affects how meaningful the impactful policy can be? Yeah, well, I

Hari Krishnan  45:41

think so. I mean, I think that’s a very good way to say it. I think a lot of economists are econometricians live in a world where they fix the delays. So they have a series of lags, they have some auto regressive model or something. And so they’re a bunch of fix legs, and then estimate the coefficients in front of the legs. And they come up with a result, but they never think of the delays the way you would in physics, or in dynamical systems, where you think of the delay maybe as being fluid, or was being a variable of some kind, you don’t rigidly fix it. And what’s known in dynamical systems is that solutions aren’t smooth with respect to delays. If you move delays a little bit, under certain conditions, the solutions become highly unstable. I think that’s missing. I think it was Milton Friedman, who talked about long and variable delays and this and that in monetary policy transmission, but the variability is problematic. And I don’t think that’s fully understood.

Corey Hoffstein  46:36

We’re sort of talking about this idea of delays as a source of instability at the macro economic and policy level, do you see it play out within financial markets as well?

Hari Krishnan  46:48

Yes, I’m a bit stumped to give you a good example. But I’m speaking a little bit loosely here. So but I do think there is a behavioral aspects, I hate to talk about it too concretely, because I’m not an expert here. But it takes a while for people to realize that a regime has changed. And maybe that’s the way it has to be if you’re constantly move bobbing and weaving with the market, you’ll just get tops and sailed all day long. But there are delays in the way people think. I mean, crypto had a long delay for, for the belief systems to develop and for the arguments to be formed that would justify the moves and outs had to delay in there took a while for people to change their view, at least temporarily, maybe to question the whole thing. So psychology obviously has delays, you know, if we hung out together 10 times, and you’re nice to me, the first nine, and then the 10th time something goes off, I’m still probably going to think you’re a decent guy. Right? But you know, three or four times, I shouldn’t use you example, a pick someone down the street. But you know, once you’ve reinforced your belief, it takes a while to change it. So I think delays are part of the human psyche as well.

Corey Hoffstein  47:59

One that comes to mind, for me is institutional mandates. Oh, yeah. And the delay and ability to really adapt those. And I think this is again, going to be an interesting year, whether inflation volatility persists or not, and how long the lag will be for institutions to adopt meaningful diversification away from just stocks and bonds. I certainly see it in the mandate space. I’d be curious if you’re seeing it, the adoption of tail risk hedging programs seems to have accelerated post March 2020, because it was the most obvious point as to how they could potentially be impactful for long term geometric returns.

Hari Krishnan  48:36

Yeah, the thinking behind this was very good, in a way because a lot of people had been banging on about how stocks and bonds could go down together. I think I talked about it for a while. And that’s what’s happened. And what’s your defense, then? Really, it’s long convexity strategies as trend following maybe, to some degree, at least, they will probably is and buying protection. And that’s about it, really. So it makes perfect sense. How it’s done is another question. But I can totally realize why people do this. I do feel and I’m maybe you’ll chuckle when you hear this, that the institutions that try and hire tail risk managers do require a lot of education, not because they don’t know markets will they do often, but because they believe that a certain rigidity of hedging can lead to the outcomes that they want. So I often get requests such as Can you give me a scenario payout profile? This is what I have on can you give me the payouts for this? And so I will give them a matrix and it will say over 510 3060 90 days, whatever, if there’s a move of x, and if I make an assumption about how much volatility will move in the given market, I’m hedging in this will be your pair can even that’s a bit noisy, but it often takes less or more time for these things to happen. And as you pointed out, your rolling position sometimes along the way, resetting you’re taking some profits, so really think that this is a business where too much rigidity in the process is not ideal. And as long as your manager can assure you that they’re not taking other tail risks in the attempt to hedge your tail risk. In other words, they’re not swapping one form of left tail for another than I think this hat will always be a fairly fluid space, as it should be.

Corey Hoffstein  50:24

I want to take a hard left turn here and actually go back to our discussion because we veered off and in a wonderful tangent, but I want to go back to our discussion a little bit on the more practical nature of tail risk hedging. And, and here in particular, in equity markets, one of the potential changes we have seen over the last couple of years is different dominant agents emerge in the option space. So there’s what I would call sort of the 1000 duck sized horses in the form of so socially levered retail traders, they’ve sort of disappeared recently, but certainly a potentially large dominant agent as a collective in the last couple of years. And then as maybe a counter example, the big horse sized duck would be the JP Morgan hedged equity fund that has very well documented role schedules and think has something close to $20 billion in both of these could potentially represent risks, or conversely, potentially represent structural edges that hedgers could take the other side of how do you think about monitoring for the emergence or disappearance of these dominant players and structural edges within the option space?

Hari Krishnan  51:33

Well, the retail example that you gave us a good one, you know, the game store type episodes in 2021, how does retail become powerful? It’s a bit like in 9019 84, out of the out of the pros rise up collectively, when none of them was very powerful individually. Well, you alluded to one, which is implied leverage buying options. And the second factor is the social network phenomenon, where if enough people cottoned on to the same trade, and they’re able to put down a small amount of premium for a dip out of the money call option, let’s say, collectively, that can be quite a position that dealers have to take the other side off, that’s something I don’t do, because I’m restricting myself to macro labs. For me, the central meeting place for risk is the s&p 500, as a central meeting place for macro is the yield curve in the US and outside. So that’s where I’m focused. But definitely, if you’re looking at individual names, retail is important, as you point out less so now, I’m sure are the other side you mentioned the large institutional, the JPMorgan hedged equity vehicle, which is huge, and which has telegraphed all of its moves as it rolls its color from a one quarter to the next. That’s stuff that sort of points back to what I was talking about when I talked about having bank contacts. I wish my bank contacts still as good as they were, many people have left. But that’s the sort of stuff where you can really mine information. And the thing is that the sell side is never, I misunderstood the whole setup. When I was starting out, I thought the sell side knew a lot of stuff, in terms of helping people to form investment opinions, never me sitting in my armchair I could read the best sell side research for the best firms. And I would be empowered as an investor. I no longer believe that there are exceptions. Of course, what I do believe is that just looking at all of the broker chats, and all the flows that I’ve looked at over the years, is a rich source of ideas for concentration risk. That’s where it’s coming from. Even on Twitter, you can find out some stuff about what people are talking about. It’s much less verse varied, as you pointed out, I think I think he pointed it out or implied it, people are generally very equity focused, which so you don’t get that much from that. Setup. But banks are a great place to find information about flow. It’s just knowing people in fixed income is very useful. Very, very useful.

Corey Hoffstein  54:03

For all the great mysticism and ink spilled on tail risk hedging in the past, you said to me, and I’m quoting you here directly, tail hedging is just a bag of tricks.

Hari Krishnan  54:15

I did say that. Yeah,

Corey Hoffstein  54:16

I don’t I don’t expect a magician to reveal their secrets. But can you elaborate on what you meant by that?

Hari Krishnan  54:22

What I meant to say is that tail hedging is, is should not be taking strong directional views on the market. If it were, you don’t need to tell which you just take that view. You know, if you think the markets coming off tomorrow, dump your lungs and go short. If your conviction sign off. Tail hedging isn’t really supposed to do that. So what can it be? If it isn’t that that’s the question I’m trying to grapple with now. If it isn’t that and it’s just going out and buying a put 15% out of the money put once a quarter and then rolling it, which is a piece maybe the JP Morgan thing roughly, I don’t see great value When that either, unless, of course, you’re doing things that have great value, aside from the puts, in which case you can eat the cost or the implied negative alpha doing that as an offset of the stuff you’re doing, that’s very good. And I know I know they pick stocks within that program. I don’t wish to give them any disrespect. But if you only took that leg, there’s no value in that you sit you quarry sitting at your, on your desk, or me sitting in my chair. What do we add? Well, we need to figure out ways to protect the downside as efficiently as possible. Given the current state of the world, that’s where the bag of tricks come in. I could put on any number of hedges at any given point in time, I could buy a one day to maturity put, I could buy a one days in maturity out of the money plus, I could buy a put spread, I could buy a put butterfly, I could, this is just in the s&p, I could buy a straddle and hedge it could buy an in the money option and hedge it. If I’m trying to make bets on realized volatility. visa vie what I paid for the option, I can go long datas I can play the call side to get bounded gains with maybe some edge, there are all sorts of things I can do. And across markets, I can do many, many more things. The bag of tricks idea is where do I think the surface should be priced? Where has it been priced? And where’s the value? There’s always value somewhere in the surface. Or at least there’s egregious overvaluation somewhere in the surface. And so you sell a bit of that, and you buy quite a bit more of other stuff to hatch. So the bag of tricks is simply that it’s what’s the best hedge given today’s environment. Now, in the first book, the second leg down, I use the well worn phrase Every dog has its day. And I figured this out. When I started back testing option strategies. There was some scenario for any option strategy, even the silliest one that pays out pretty nicely. Looks good. And even the best strategy has stuff that doesn’t work well. So I started thinking, well, options, strategies have to be regime dependent. So let’s say give me the simplest give you the simplest decision tree, by a put by put spread by a put butterfly, s&p 500 them, what are the conditions in which I would do A versus B versus C, that’s the bag of tricks approach. I’m gonna figure out what I can do. It might not be the be all and Antle if I buy a put spread instead of an outright put one to one, same size. Obviously, I won’t make as much in a crash. But it might be cheaper, it might be worth the trade off. Maybe I can widen the strikes, and make it a little better. If the skewer seat maybe I can do more. And one thing that a lot of people don’t understand, maybe they do, but being presumptuous is that payout ratios can sometimes go up when you try to spread. And I don’t just mean buying a butterfly and praying that it will land up maturity at the peak a lot referring to that. And the brokers love to sell stuff like that. You could make 20 to one on this thing, and it costs you very little. But of course you need to hold to maturity and absorb the gamma risk with the day to go and sweat over it to get that payout. Not really speaking of that, and we’re speaking of trades where you can do more units, if you trade a spread, instead of just buying a put out rice, you just cut the premium cost to the point where for a fixed budget, you can just do a heck of a lot more.

Corey Hoffstein  58:34

So in contrast to what you’re calling a bag of tricks, that sounds fairly simplistic on its surface, I know you also work with a dedicated machine learning team.

Hari Krishnan  58:46

I do I’m not I’m not the spearhead of this. So I’m piggybacking off it, I have a few theories and put them out there, which is that machine learning is great when you have large amounts of data, and you want to identify the fine structure of the data, machine learning often relies upon throwing away the tails. Because if the tails are too heavily impacted, impactful, then the machine learning will have some problems. So you kind of want to look at the belly of the distribution and find find structure. In that sense machine learning goes very well with tail risk management because if you can cover the tails, then you can go deeply into what happens most of the time and look at average returns and forget about compounding and winterize the tails, get rid of them and do quite a bit. And so there you know, there are lots of ways to make those sorts of systems a little bit less long. One of the real curses of machine learning, if you had run it up to this year, and maybe even including this year is that you’d be pretty much always long bonds and long stocks. Why? Because there’s a trade off in machine learning. You want to have low bias Yes, but you also want to minimize model variance. Which means if you chop the data up into parts into chunks, and you train the system on one chunk versus another, you don’t want the rules to be entirely different. So the way to minimize variance is just to be long from 2010 to 2020 21, by the market got same rule independent of the data that you input. Now, there are ways to improve this, of course, and this applies to bonds too, and the ways to improve it or to maybe force the system to be a bit less long, maybe to use constraints, maybe to wait choppy periods more heavily or down periods more heavily to train a bit differently this than the other. But I think it’s a very interesting complements. And what we’re trying to do, although we haven’t finished this project is to come up with bear market plays that are tail risk catches, using machine learning. So maybe looking at the 20% worst outcomes down to the 5% worst outcomes, where the base rate isn’t so low, there’s nothing you can do, and then hatch the tails, and use that as an adjunct to more just a maximum convexity bag of tricks or whatever type of program you might think. So I’m trying to use machine learning to create and construct more standard spread trades that profit from big moves that aren’t outrageously big.

Corey Hoffstein  1:01:18

Hurry. The last question I’m asking everyone this season, is to reflect upon their career and answer, what was the luckiest break you had, other than coming on the show? Well, this can’t be the luckiest break you’ve had?

Hari Krishnan  1:01:34

Well, it depends on what I consider to be luck. I mean, when I never planned to go to graduate school, I don’t think I’ve said this. But when I was an undergrad, I never used to go to the seminar, we had an applied math seminar. And one day, the professor called me and then I thought he was going to tell me, I was gonna get a horrible grade or something. But he said, you know, you really have done much better than I even would have expected, blah, blah, blah, I think you should go to graduate school, I’ll call up some places. And I think you’ll get it. I said, Okay, sure. A week later, I got a call. And they said, well, we’ll give you fellowship, blah, blah, blah. And so as a test, I thought, well, maybe I’ll apply somewhere else, that’s a lesser university without the call. And I got rejected. And so I pointed out how that that kind of pushed me in the direction of doing applied math in various forms as a career. And then when I was a postdoc, my first boss took pity on me because I hadn’t been making much progress in my research. And so I set up the AV equipment for a speech that she gave, at the end, she sort of saw that I was shrinking in the background, not wanting to take on any of the academic powerhouse, people, partly because I was cynical, but also because I was a bit more timid. So she hired me, she just called me up. And I’d never didn’t even consider working in that space. And I wound up doing that. And it was just a bunch of happy accidents. I’d meet people socially and my wife got hired through that, this that the other, so I couldn’t put my finger on one thing, but I think it was a series of events that compound it that led me to where I am now. And this is one of the unfortunate things for many people in life. And fortunate for others, that knowledge is highly compounding. And I don’t just mean book knowledge, I mean, meeting people learning about their perspective. The more people you meet, the more people you have a beneficial relationship with to weigh. The more you learn, the more you can leverage that learning, the more they can leverage it and introduce you to other people. And it just keeps growing. So I have not replied to your question. I’ve just rambled but it’s a series of compounding steps. Is my reply.

Corey Hoffstein  1:03:41

Around about as it was, I think it was a fantastic answer. So thank you for joining me on the podcast. Sorry. It’s been fantastic.

Hari Krishnan  1:03:47

You bet. It’s a pleasure. I hope to be back