FinancePy: From Physics to Wall Street

Dominic (00:00)
I remember back in Salomon in 1994, our risk management system

crashed one night because the level of interest rates was like 0.2 % or 20 basis points and we were doing scenarios where we bumped it down by 25 basis points, the interest rate actually went negative, which was something that our models said could never happen.

Michael McLaughlin (00:36)
Hello and welcome to the Stack to Street Podcast. I'm Michael McLaughlin, and on this show I meet the independent developers and small teams who are building software worth watching. My guest today comes from the world of Wall Street derivatives. Dominic O'Kane is a physicist and former head of quant research at Lehman Brothers Europe, and for nearly two decades a professor in finance. He wrote the book on credit derivatives, and he built FinancePy

A free open source Python library for pricing them. If you're a quant or a student thinking about a career in finance, this episode is for you. Here's my conversation with Dominic O'Kane

Michael McLaughlin (01:12)
Welcome Dominic, and it's great to have you here

as our guest on the Stack to Street podcast. I think our audience would be very interested in hearing directly from you before we get into the FinancePy, but really hearing from you about your career development starting from university and making your way into Wall Street.

Dominic (01:34)
Okay, well, in a way it was a bit of a, not quite a random walk, but almost a random event because

I actually wanted to be a theoretical physicist and my first degree was in physics at Imperial College. And then I went to Oxford to do theoretical physics and there I did a particular area of theoretical physics called statistical mechanics and actually applied them to AI back then. we look at multi-layer neural networks.

that was my intended career path. And I then moved to Imperial College for a postdoc and looking forward, you know, those in those times, you know, everyone was cutting back on, you know, blue sky research. The Cold War was over, defence department weren't spending money. So theoretical physics was feeling very poor in those days and I was feeling extremely poor living in London.

I saw a job offer in the careers department from Merrill Lynch. It was old by the time I saw it, but it listed a whole lot of things about statistical, Monte Carlo simulation, C++, all this stuff. And I thought, that's me. So I was emboldened. then I through a headhunter and stuff. started talking to people in the city and

between deciding to leave academia and finding a job was about three or four weeks. So it all went very fast. I bought this book that everybody knows, everybody buys, called Options, Futures and Other Derivative Securities by John Hull,

And then I interviewed at two places. I interviewed at LTCM and didn't get the job, unfortunately. And.

Michael McLaughlin (03:14)
Well maybe well maybe fortunately at some level.

Dominic (03:19)
I remember they asked me in the interview whether I played golf and unfortunately I didn't. So I think that was my, that was a red flag for them. And

then I interviewed at Salomon Brothers and I did get the job there. yeah, so within a couple, within a month or two of deciding to leave academia, I found a job and it was remarkable how close it was in terms of what I was actually doing on a day-to-day basis.

how similar it was to what I was doing as a physicist. Because I'd moved in, so my job there was working in an internal hedge fund where we were doing a lot of fixed income, relative value trades and managing client money as well. And I was put in charge of the risk management system and they really dropped me in at the deep end. And it was pretty tough at the start, but

the computers were the same, the maths was the same, the Monte Carlo simulations were the same. And

the only difference was I was really, I had to get up two hours earlier in the morning and I had to wear a tie. but that seemed like a compromise I was willing to accept because it was like the beginning of a career.

Michael McLaughlin (04:29)
So you made the point about how maybe there was only a subtle difference between being a physicist and and being a a Wall Street quant. but it would seem to me that there was a a sort of a big difference insofar as you had

constituents, you had customers, you had traders, you had salespeople, you had risk managers. so so talk me through how that part of your career evolved where you were truly answerable to people for big time economic decisions.

Dominic (04:57)
So the first role was basically being in charge of the risk management system. So I had to basically, you know, be responsible for the overnight runs of the risk management system, making sure all the traders got their risk in the morning, that they understood the risk of their issues about positions not being there or all sorts of issues. They would come to me first.

But what I didn't get to do there was to do real like

hardcore derivative valuation because we were basically a price taker rather than a price maker and I thought you can't really be a real quant like a pricing quant unless you specialize and you you do that. So I joined Salomon in 94 and then in 97 I went to Lehman Brothers and I there went to be on the the trading desk on interest rate derivatives.

So in those days, was a a different world again to what it is now. mean, in those days, proprietary trading was a huge thing. And, you know, anyone who's read, who's read Liar's Poker and things like that knows about kind of the, Salomon influence. And a lot of the people I worked with were actually people from ex-Salomon people. And they're, they're putting on enormous relative value trades, you know, especially around Euro convergence and,

They needed models to help them work out what's the best way to implement these particular trades. So I was building some of these models.

then we had the Russia defaulted on its debt and that caused a huge, panic in the markets. And then LTCM blew up and all of a sudden the idea of all this big relative value stuff.

became a bad idea. And at that point, I thought, maybe now would be a good time to switch to do something different. So I switched to the credit derivatives world.

So

from about 1999, I started working on credit derivatives, wrote, started writing client facing research because we need to educate clients about how these contracts work. And

I was also heavily involved in building all the internal models that are used for valuing these trades.

Michael McLaughlin (07:05)
you mentioned the Russian default and the collapse of LTCM and I guess around 2001 we had Enron

So tell me what your life was like as a sell side quant when things were coming off the hook.

Dominic (07:23)
Yeah,

for the first two years, of working credit, nothing really defaulted, especially on the corporate side. And then suddenly it stopped being quiet. So we had Enron, Worldcom, Parmalat.

We had just endless numbers of defaults that are happening. as a quant, you're slightly removed from what happens on the desk because it's the role of the trader to manage that risk. So as a default is approaching, you know, we calculate risk measures, value on default and things like that, and recovery rate sensitivities.

The role of quant is that you have to spend hours and hours every day building models. So you need to be able to focus and to do that sort of stuff. If you get distracted every time there's a market thing going on, it affects your productivity.

So I don't know if it's worth saying a few words about the whole framework of the sell side quant.

Michael McLaughlin (08:19)
Yeah, absolutely. I think that'd be really helpful.

Dominic (08:22)
So, I mean, for many people, I

mean, when they hear about Wall Street quant, you know, they think of people like Jim Simons and Ken Griffin and people like that. Those are what I call buy-side quants. They're obviously using quantitative models to put on trades from which they hope to make money. The sell-side quant is kind of much less well known, but pretty important because...

The role of the sell side quant is to build the models that price the products that the banks want to sell. And this is a very interesting area of finance. And I don't think it's that well known outside of the very quant world. But it all goes back to this thing called Black-Scholes. So if you go back to the early 70s, traders were selling

contracts called options you could have a call option which says if the stock of IBM say 100 today you could say I have the option to to buy IBM stock in one year for $100 okay now there's another contract called a forward where you could do a forward say I have I want to buy IBM stock in one year at $100

The difference between them is that if you buy the forward, you have to basically buy the stock at 100 in one year's time. You don't have any choice. So if the stock goes to 150, you're very happy because you can buy it at 100.

but if the stock falls to 40, then you're very unhappy because you're buying it at 100, even though it's trading at 40. So this is what's called a linear derivative because the prices can go positive and negative. But the option said you have the option whether or not you want to buy.

If the price of the IBM stock goes up to 150, you exercise the call option, buy the stock for 100 using the call option, and you can immediately sell it in the market and make $50 of profit. If however the stock is trading at 50, okay, if the stock price has fallen, you can decide not to exercise your option.

Now the problem with that is that from a pricing perspective, it's a much harder problem than pricing it forward because you have what's called a non-linearity in the payoff. Because below 100 you don't exercise, above 100 you do exercise. And therefore pricing this derivative was very complicated. People didn't quite know how to do it. They had back of the envelope, rough ideas of how to price these things. So it wasn't until

Fischer Black, Myron Scholes and Robert Merton studied this problem and they realized that you can set up what's called a dynamic hedging strategy, every day what you do is you assume you have a value for the option. And every day you buy enough of the underlying stock,

so that a change in the underlying stock price is exactly offset by a change in the option price and vice versa. So you have one goes up and the other one goes down by exactly the same.

And they showed that if you do this dynamically over time and you do it very, frequently, you can actually hedge this option almost perfectly. And they produced a a beautiful second order partial differential equation that expresses this. this is the world that the sell side quant is living in. We're not worried about the...

where the stocks are going up and down. In fact, the trader who's hedging using this strategy, which is called Delta hedging, is indifferent as to whether the stock price is going up or going down. So instantaneously, that portfolio of the option and the hedges has got no sensitivity to the direction of the stock. And this was a huge leap forward because it meant that two traders

One trader thinks the stock's going to go up and other trader thinks it's going to go down, can agree on the price because through the Delta hedging mechanism, the direction of the stock doesn't matter.

They sell the option to the customer and then they do this dynamic hedging on the other side. so hopefully if they do this properly, they have no exposure to whether the market goes up or down. So it's a very kind of...

strange world and it took me a couple of years to really get my head around it. But this is how sell-side works. This is how banks were able to sell complicated derivatives to their clients. So the role of someone like me is to build that model for the particular product that we're selling and to produce all the sensitivities that the trader needs in order to do this delta hedging.

And that's the role of the sell-side quant.

Michael McLaughlin (13:15)
That explanation of Black Scholes options pricing, hedging of of option positions with with Delta on the underlying stock that was really well articulated. And I think it's actually a pretty good jumping off point to to get into FinancePy, your open source library of of Python-based models

would you take us through how you started to become the creator of FinancePy and also how you orchestrated the development of of the library with the help of the open source community.

Dominic (13:44)
I left Lehman in 2006, probably the best market timing decision I've ever made. And then I went into academia. So I've been working at a business school in France called EDHEC who has a very good financial master's of finance program that I've been teaching on. so over the years, over the last 20 or so years, I built a lot of kind of code.

that I use in my teaching and also I wrote a book on credit derivatives so I use it there. And actually it was during lockdown in 2020 that I thought maybe I should actually put something back by maybe releasing this code to the general community. And the issue that I'm really solving was that there is no kind of easy to read library

of financial derivative code out there. There is another very advanced derivative pricing library called Quantlib, but it's written in C++. So,

I wanted something where it can be done within Python and where the student or whoever is using it could just look down in Python, look down through all the layers and there is only at most two, if not three layers and see exactly where the calculations are being performed. So my aim was to create transparency so that I could use this in a teaching setting. So I could say to people, okay, you want to know how to price this?

This is the code

I also wanted to close the gap between the models that you find in textbooks.

and the models as we use them in the real world. Because in the real world, if I'm pricing an FX forward, a very simple trade to buy a currency forward. There's things like the spot FX has a settlement period of one or two days depending on the currency. So things like that are never mentioned in the books. Okay, it's a minor detail.

But if you're sitting on a trading floor, that minor detail has financial value. so I wanted to close the gap between the theory that you find in the textbooks and the practice of, you know, what you actually find if you're sitting on a trading floor and you have to price an FX forward and you're wondering why is the price that because it doesn't quite agree with what you saw when you read, you know, the formula in in Hull's book. Okay.

And then you'll discover that there are issues about calendars. There are issues about weekend adjustments and things like that. So you can see all of that within FinancePy. I've implemented the calendars for 20 different countries. So you can see all of the tedious and boring detail. But it's important detail because on a trading floor, if you get that wrong, you can lose several thousand dollars on a trade the minute you do the trade.

in terms of usability, the whole library is written in Python, which is notoriously slow. But around the same time I started it, there's a library called Numba, N-U-M-B-A, which has been released. And it means that if you put a little tag at the start of a function,

it will actually compile that Python function into machine code so that when it's run, you get speeds roughly equivalent to C++. So you get the best of both worlds. You get a transparent Python library. Python is a very easy language to read and to understand. And you get the speed of a compiled language as well.

So I think that makes it

very useful because then you can start doing heavy duty Monte Carlo simulations and other work.

Michael McLaughlin (17:25)
tell me why transparency is important in financial modelling.

Dominic (17:31)
Transparency is important because you need to get the price right.

So if you're working on a derivative on a trading desk and you have the wrong model or you price something incorrectly, then you'll get arbitraged.

So just to give a real example, say I'm working in Lehman Brothers before the crash and I priced the option at

25 and somebody goes out and they see that if they do the bid across the bid offer spread, they can sell that option to Goldman Sachs for 30. Then they're going to come to us and say, I want a billion of that at 25. And then they sell it to Goldman Sachs for 30 they make a huge profit. And then you'll discover that you've mispriced.

and then your trading book is showing an enormous loss. Okay, let's say five million, you'll have a five million loss on your trading book. So you want to avoid that. So before we could use a model officially on the trading desk,

it had to be approved by the model validation team. So they go in, they look at the code, they do all sorts of tests. So by having FinancePy as a sort of

a source of underlying models that everyone can see. And if there's a mistake or a bug there, it gives a a benchmark for these models. know, people can go and say, well, my model is producing the same number as that model and that number has been seen by many people.

I can assume that it must be correct. So, yeah, so that's essentially why having transparency is important.

Michael McLaughlin (19:07)
how many models are in are available in FinancePy?

Dominic (19:12)
well, there are lots, but we have to separate. into three kind of concepts. have products, models, and market.

Michael McLaughlin (19:18)
And give me an example of of one of

Dominic (19:20)
Okay, so a product might be...

an equity digital option. So that's an option that pays one dollar

if the stock crosses over some barrier. Okay, pays zero if it doesn't go through that barrier. So that can be used by, you know, investors to speculate on a particular stock move, stock price move. Now that will be priced generally within the Black-Scholes model.

And it's a fairly powerful model and it's surprisingly good. So underlying the Black-Scholes process is Black-Scholes model is the assumption that

The price of things follows what's called geometrical Brownian motion. Okay, sounds very complicated, but what it actually means is that in a stock price world, it means that the daily return of the stock is drawn from a bell curve distribution. Okay, it's normally distributed. Okay, so the daily return is normally distributed. The price is log normally distributed.

The only thing it doesn't have are large jumps. So if you want to tell the difference between the stock market and geometrical Brownian motion is you've got to look for jumps. If you find a big jump, then that's the real world because the geometrical Brownian motion doesn't have big jumps. So this model, geometrical Brownian motion, Black Scholes is what's used as the kind of the standard model.

But it's not usually sufficiently precise for most large investment banks. So what they will do is they will build more advanced models. models which have jumps in them, because I said that the Black-Scholes geometrical Brownian motion doesn't have jumps.

Well, people add jumps.

So there's a whole zoo of models that are being used. Within FinancePy, I have implemented probably 80 or 90 % of the different models.

Michael McLaughlin (21:11)
we talked about credit derivatives in the beginning and you you just mentioned, you know, stock jumps. with a credit, the big risk is a jump to default. Where it's sort of fine one day and in default the next day. And I presume you've addressed that in FinancePy through modeling.

Dominic (21:34)
Actually, that's a very good question. the CDS, okay, for CDSs, for example, credit default swaps, which are basically like insurance contracts against default. I have interest rate risk, spread risk, I have recovery rate risk, but I actually don't calculate the loss on default. So I probably should add that in, but that's a very simple calculation to do

Michael McLaughlin (21:56)
Yeah. And I remember in my credit derivative days, the Gaussian copula

Dominic (22:02)
Yeah, so the Gaussian Copula, okay, so now we're getting into the world of default correlation. So

default correlation was a world I was heavily involved in in the early, from about 2002, 2003. So it really took off. So the whole idea was that it started off with what's called default baskets. So the problem was in those days, before everything started blowing up.

spreads were not very high. So an investor who took credit risk didn't get paid a lot of spread. So people wanted to get paid more spread. So one way to get paid more spread is to expose yourself to not just the default of one asset, but say to the default of five different assets at the same time. And you'd like to lose money on the first default out of five. And...

without going into all the details, you find that the pricing of that first to default depends on the tendency of the different credits in the basket to default together or by themselves.

And this in the world of finance, this goes outside the Black-Scholes world because the Black-Scholes world assumes that for every risk, you can find an instrument that you can use to hedge it. So if I sell an option, I'm only exposed to the price of the option changing. I can hedge that by Delta hedging with the actual underlying stock. So if you can't hedge the correlation risk, what can you do? Well, there are a couple of things you can do. The first thing is,

to very conservative. So when you speak to your internal model validation, they say, okay, we think that the correlation is probably around 20%. We price the trade at 25%, okay, but we could be wrong. And we think that we could be wrong by plus or minus 10%, okay? So the true correlation could be between 15 and say 35%.

So what we would then do is the trader would do the trade and think he'd made a certain amount of money. Say he thinks he's made $500,000. We could then calculate what is the change in the value of the trade if we were to bump the correlation to those two extremes. And we'd find that in the extreme in which the model lost money, that the trade in the worst case would only be worth, say, $250,000.

So then what you say to the trader is, you can't take that $500,000 into your P &L, okay, and use it to get a bigger bonus this year. You have to put aside reserves of $250,000 until we actually know what the market is, okay? But then you can do a second thing. A second thing is to say, the trader will say, okay, but what I'm gonna do is,

I am going to find trades that I can do with my clients, other trades, which will reduce my correlation sensitivity. So if you sell a first to default basket, one way to reduce the correlation risk on that is to sell a second to default basket. That means that the client gets paid or loses, well, loses after the second, after the, so one has the default, then he loses on the second default.

And if you can actually sell the second, third, fourth and fifth to default, you've actually completely hedged the risk. And each time then you do that, you can say to internal model validation, I've reduced my sensitivity to correlation. Therefore I can take some of that P &L from reserve. So you're out of that 250,000 that's now in reserves. You can maybe grab 50, another 50, another 50.

So

this is kind of how people within the sell side derivatives world think about things. you never read about this in finance books. You never read about this in John Hull's book.

Michael McLaughlin (25:59)
Yeah, no one no one talked about how how bonus

pools were increasing because of correlation reserves being released.

Dominic (26:05)
Yeah.

So a copula is kind of a recipe for correlating different things together

the problem with this particular product was that computationally, was incredibly computationally heavy to calculate

The Gaussian Copula model is a surprisingly simple model. I'm not going to go into the details now, but it's surprisingly simple, but...

Fortunately, it was simple enough that we could actually calculate our risk overnight because the other problem on the trading floor is that if you're pricing something very complicated and the model takes, it's a bit like the weather forecast. If you've got the Met Office in the UK trying to predict tomorrow's weather, but their computer takes a week to calculate tomorrow's weather, then it's not very useful. It's the same on a trading floor. If your risk management system takes a week to calculate all your risks hedges,

then it's not very good for the trader the next morning who wants to know what he has to do in order to Delta hedge. So whilst we want to model it's powerful and captures the risks, it also has to run quickly enough that we can actually run it overnight.

So, what's often what you're often doing as a quant is you build a very sophisticated model, but then you have to have to figure out very clever approximations. What can I approximate? What can I cut? What can I reduce in order that the model can complete in time for the trader to get their risk?

So sometimes people think the quant is like, you know, they're thinking about this great model or this new stochastic process or this new kind of way of modeling stuff. A lot of the time we're thinking about the practicalities of using this model in a real world environment, you in an industry. So we're thinking we have a great model, but how can we simplify it so that we can get our risk in time?

And that's

something that I don't think people who haven't been on the trading floor don't really realize, the sort of constraints that the models are operating within.

Michael McLaughlin (28:04)
You

the world's changed a lot since the pre-Lehman and pre-GFC days

Dominic (28:09)
the main change I think that's happened since I left Lehman is that the amount of innovation has reduced substantially. In the early 2000s, people were coming up with all sorts of new structures, new products that they wanted to sell to investors.

Obviously, after 2008, everybody got a lot more conservative. Proprietary desks were pretty much banned. And therefore, I think the degree of innovation has gone down a lot.

So I think that life and also the amount of regulation has gone up. I'm not an expert on all of the regulation, but there's like, you know, the way that Basel and the other things operate, you know, are also very quantitative. So I think that the role of a quant these days has become not just a pricing quant, but also there's a regulatory quant role as well because...

the way the trading book is managed, how reserves are calculated has become much more quantitative than it was.

But I think there are perhaps new things coming along.

Obviously, the last 20 years, mean, quant have had to deal with very low interest rates. And very low interest rates is something that we've had to deal with the fact that interest rates in some places went negative. And if you're pricing an interest rate option, a call or a cap or a floor, you have to have a model that takes into the fact that for a period 10 years ago,

interest rates were trading at levels close to zero. I mean, I remember back in Salomon in 1994, I remember that our risk management system

crashed one night because of Japanese interest rates going below, I think they were going below 25 basis points because we used to perturb our risk, do perturbative risk. if the level of interest rates was like 0.2 % or 20 basis points and we were doing scenarios where we bumped it down by 25 basis points, the interest rate actually went negative, which was something that our models said could never happen.

Okay, and they would just kind of stop. the interest, so the risk management system failed to run. So we had to do an emergency fix to allow negative interest rates. Okay, that's just kind of like the first time I ever saw that problem, but we never thought we'd ever see it again.

But then obviously, you 10 years ago, we had Swiss interest rates going negative. We had other interest rates very close to negative.

Michael McLaughlin (30:52)
but to your point, we we talked about negative negative rates or sort of these anomalies the models weren't built for. I think the oil markets experienced

negative prices at at at one point around COVID.

but the opposite occurred in LME nickel. it went up by a factor of four in three or four trading days.

so really my my point was anomalies that occur in the market where one goes radically negative and the other one sort of went radically positive. you have to make sure that your model has a parameter that allows for the world to behave differently than you expect.

Dominic (31:31)
There's a of a double-edged sword there because you want to allow that, but at same time, if somebody inputs a negative price for a futures contract, you kind of want to think that that's a mistake.

Michael McLaughlin (31:45)
Well you can have a you can have a are you sure warning.

Dominic (31:48)
A lot of time, you know, people make mistakes in these libraries. You know, they type in the wrong, you know, they put in a coupon as a 5 % is five instead of 0.05 or something simple like that. So it's nice to have all these little validation checks to stop people from making mistakes. So, yeah, but yeah, there are exceptions that you have to handle.

Michael McLaughlin (32:07)
why don't you help us finish up with a more discussion around FinancePy and what's available to the user and what you hope users might get from it and also the the typical kind of user that you think should be attracted to the to the FinancePy github repo.

Dominic (32:25)
Okay, so I think that the user who will find it useful is, okay, first of all, the student. And I should say, I should sort of plug in advance my book, which I hope will be out next year, because I'm writing a book on financial engineering, which uses FinancePy So.

I'm about 600 pages in and it's probably another 200 pages to go. So it's going to be a bit of a beast of a book, but it covers the whole range of products. But I think a student who wants to see how these models work, how they're implemented, because the implementation is important too. It's not just the model, but you have to implement it right. It has to be fast and efficient and stuff like that.

So I think students will find it useful. I think that...

If you're on the buy side and you're an analyst and you want to buy an option and you want to test check the price, then it's almost an industry level quality library that you can use to check the price against. Obviously you're going to need market inputs. Now you can probably do that if you have Bloomberg. But Bloomberg is a black box. With FinancePy, you know...

You could see if you can reproduce Bloomberg and then you can play around with it. It can also be used by model validation. If they're validating a model, they can look and they can use this model as a reference and they can compare the prices and stuff like that.

And I think academics who are working on valuation, there are lots of academics who are studying the pricing of better pricing models for FX derivatives. So you may think, well, if I extend that model, I can do a better job. And...

rather than begin it yourself, you can just go and take my code. You can even implement your own model and then issue what's called a pull request and push, you know, put your model into the library. Obviously I have to kind of okay it and stuff like that, but if it's fine, then I would, I would have no problem with that. So, I mean, that's the kind of thing. So, and there are people already using this. know that Vladimir Piterbarg, who's the head quant, used to be the head quant of Barclays Barcap.

and it's now RBS, he's a professor at Imperial College. He teaches finance there, so he uses FinancePy. And I think there are other people around. To be honest, there's been a lot of downloads and I really don't know who's using it where and when. It'd be interesting to have some more feedback from people. it'd be interesting to find out more about how people are using it.

And the interesting thing is I'm now starting to get an increasing number of people who are coming in and submitting pull requests and who are participating in the

development of the library.

Michael McLaughlin (35:13)
Yeah, that's great. more people looking at it, more chance of finding inconsistencies or or bugs or even just chance of improvement. So that's that's excellent.

Well hopefully from the result of this podcast, there are more participants in the repo

Dominic (35:30)
the main

types of person in a repo who are doing work in a repo are probably quants or aspiring quants who may already be working in the finance industry

So if you are an aspiring quant, then this would be an excellent thing to put on your CV.

Michael McLaughlin (35:51)
we look forward to hearing about the completion of your book. I presume that people can be in can be in touch with you via the GitHub repo, FinancePy.

Dominic (36:00)
Yeah, you can raise issues there if you have specific issues. And I have an email address there as well.

Michael McLaughlin (36:10)
Well, Dominic, I I've really enjoyed the conversation. I think your work with FinancePy is is is gonna be very useful to many people out there, people who are trying to pursue a career evolution from theoretical quantitative disciplines to the world of Wall Street, both buy side and sell side.

Dominic (36:30)
Well, thank you for giving

me the chance to talk about it.

Michael McLaughlin (36:33)
Dominic, I really appreciate it. Thanks so much for your time.

Dominic (36:37)
Okay, thanks a lot Michael.