Thu, Jan 19, 2023
SUMMARY KEYWORDS
Performativity, finance, markets, models, , risks Black-Scholes-Merton, derivatives, High-Frequency trading, model limits, statistical arbitrage
Meyrick Chapman
This is the second Exorbitant Chat and today I'm very pleased to talk to Professor Donald McKenzie. Professor McKenzie holds the personal Chair of sociology at Edinburgh University. A post he's held for many years since the early 90s. His work looks at the sociology of science and technology, and to explain their role in shaping the modern world. His particular focus that I have been interested in has been finance. And that's where I came across his work. And he's also recently done work in online advertising. So, Professor Mackenzie, Donald, thank you for joining me.
Donald MacKenzie
Well, once again, thank you for having me in your series.
Meyrick Chapman
Thank you. I've looked forward to talking to you for a long time. I was especially influenced by your book, 'An Engine, Not a Camera', which was published some time ago now, 2006 But I think is still highly relevant. The book is a comprehensive description of the symbiosis, if you like, that exists between the financial participants, financial markets, and the theorists that closely matches my own experience in finance. I wondered as we start this conversation, whether you could give an outline of the main thrust of your work in general, and perhaps some explanation, in particular, the subject of that book, 'An Engine, Not a Camera', which revolves around 'performativity'.
Donald MacKenzie
Thank you. Yes. Thanks for the kind remarks about the book. As you say, I'm a sociologist of science and technology, so I naturally focus on the more technical so to speak aspects of whatever topic it is that I'm studying, and I've done various things throughout the career. One, slightly bizarre, perhaps was studying the guidance systems of nuclear missiles, a topic that suddenly seems more relevant today than it's done for for several decades. Alas. So when I started getting interested in the financial system at the end of the 1990s it was kind of natural step to start looking at the sophisticated mathematical models that had become prominent in finance over the past 20 years. And the idea of performativity is in a sense, a very straightforward one. There's a difference between a statement such as 'It's sunny outside', and a statement like for example, if I'd been late for our online appointment and I started by saying, 'I apologize, I'm sorry'. The two statements are quite different. It's raining outside. Well, it's sunny outside, it's either is or it isn't. So to speak, if I in the right tone of voice, say, I'm sorry, I apologize, then that simply is an apology. And the application to financial models is that one way one could think about our financial model is as a description of a market and the market process. In other words, analgous too it's sunny or it's raining, but a model once it starts being used by traders, by regulators and so on, is no longer simply an external description of a market it becomes part of market processes and to put it very simply and crudely one form that that can take is the use of the model can push those processes that can for example, push patterns of prices, towards the postulates of a model, but it can also go in the opposite direction. So the use of a model under certain circumstances can be in my terminology counter performative. That's to say, it can shift the market it can shift market processes away from what the model postulates. So in the one scenario, the model is self reinforcing, so to speak in its use, in the other scenario, the counter performative scenario, it is self-undermining.
Meyrick Chapman
That that's absolutely central. I think, to the work that you've done in 'An Engine, Not a Camera', which is not there actually, not certainly not the only work I've read of yours, but I would say it was most closely aligned to my own work. And in that you focus a lot of attention on the emergence of Black Scholes Merton formula, which is a model for pricing options, a development, which shaped the way that options were priced and traded. Nassim Taleb, a famous writer and thinker of financial markets. And somebody I actually overlapped with, many years ago at Bankers Trust said that 'Most everything that's been developed in modern finance since 1973, is but a footnote on the Black Scholes Merton equation.' Were you also of the view that Black Scholes Merton was so fundamental to modern finance. And if so, are there any wider conclusions to be drawn?
Donald MacKenzie
Yes, I mean, I probably wouldn't put it quite as strongly as Nassim does and that's probably why his books outsell mine. He goes for the sharp and punchy formulation. I tend to have the traditional academics rather more guarded way of putting things but yes, Black Scholes Merton, perhaps particularly in Merton's version of it, in Merton's derivation of the model did become fundamental to modern finance because the the basic methodology is price a derivative, according to the cost of hedging that derivative. So it's a kind of formula so to speak, that can be applied with of course, various degrees of difficulty to a wide range of derivatives. And of course, the 1990s in particular, and the mention of Bankers Trust is relevant here. The 1980s-1990s saw an explosion of the use of derivatives in the financial system, a proliferation of the forms of derivatives that were available, and in the absence of a systematic way of pricing and hedging the risks of those derivatives, it will be difficult to see the explosion happening at quite the rate at quite the rate, that it did. And among the ways in which I think that worked, is that the hedging and the risk management of derivatives now that it's no secret, of course, to say that that was never anything like perfect, but the fact that you could do it mathematically and as it were, decompose the risks involved meant that a big player such as an investment bank, or a big options market maker, for example, did not have to hedge each and every derivative individually, which is, of course, a very expensive thing to do. You could, as it were look at your buckets of risk across your portfolio of derivatives. And in many cases, of course, if you're a big player, the risks kind of cancel themselves out. So what you're left with what you're left needing to hedge is a lot smaller, so to speak, than your overall portfolio. So that makes it possible for a big participant in derivatives markets to operate as if it had low transaction costs, because it's actually it's not having to hedge the whole thing. It's only having to hedge a proportion of the whole thing. And that in a certain sense is a performative effect, because of course, famously, the Black Scholes model presupposes zero transaction costs. So the use of the model in the way that I've just described reduces the transaction costs that parties face so in that sense, the model made itself more true.
Meyrick Chapman
Yeah, yeah. That's that's so interesting. And it's, it's something which financial market participants now take completely for granted the notion of netting, which is effectively what what you're describing so that you're hedging the individual risks rather than the gross risks. And you know, it reminds me of Robert Merton, the contributor to the Black Scholes model, who frequently described financial theorists as engineers. So you've talked about breaking down the parts of the derivative into, into constituent parts. Merton said he was seeking like an engineer to solve a real world problem by breaking things down. And a quote from Merton's is: 'In the real world, the financial system and the real economy are inextricably linked. What is more tangible, the machine or the software in it? If you don't have a well functioning financial system, you can't build the machine.'
Meyrick Chapman
And I like I like this quote. Because it explicitly links technology and implicitly communication to finance and the economy. And of course, by machines, he may have meant not just tools to make cars and planes but tools to make financial connections and financial products around the world. So I think that's more or less what you've just described. This way of looking at financial theory as engineering means it's also a practical solution, and therefore, it evolves. Yet it's based on mathematical precepts, often quite rigid precepts. And your own work, as described, can be described as shedding light on the way that mathematical models interact with our world. We have a situation where we're increasingly dependent on parameters specified in models, yet the world seems always to be as messy as it always has been. And I'm reminded of the warning from an Edwardian philosopher, which is going back same way, I admit, Alfred North Whitehead who talked of 'the fallacy of misplaced concreteness' or as a later phrase has it 'the map is not the territory' or the model is never correct. There is another there are other versions of this. And your own work seems to address this difficulty when we're looking at security of computer systems, financial models, or more recently in high frequency trading as well.
Donald MacKenzie
No, and Whitehead is of course correct. And it's important to say, however, that those who work with models in finance, for example, those who develop the models, those who trade using the models, and so on, are in my experience, typically fully aware of that. They know that the map is not the territory they know that the model is based on assumptions that are abstractions from the real world. Economists themselves know that, you know, part of the art of being an economist is working out just how much to abstract from the real world. Not abstracting so much that your model becomes trivial, not abstracting too little so that your model cannot be solved. So it's, this is not just an external insight, from people such as me, this is something that is known to financial market participants.
Donald MacKenzie
But and there's a very big but that needs said here. In many situations, there are pressures to set that kind of concern aside. So for example, regulators can de facto force market participants into courses of action that market participants know have risks and I think when the full story of the UK and liability driven investment by pension funds comes out, we will see that the regulators actually played a role in encouraging the large scale pursuit of that strategy.
Donald MacKenzie
But it's also the case that sometimes those pressures are internal to markets. So one of the models that I've looked at is the famous Gaussian copula used to price the collateralized debt obligations that were that were central to the 2008 global banking crisis. I interviewed people in that market, not just after the crisis, but also before it. And even before it, they would often say very scathing things about the Gaussian copula model, but it was very hard to stop using it because it had become something of a market standard. And if something's a market standard, if you don't use it, and then you have all sorts of problems even with, even with your accountants and how you know how they're going to value your different derivatives portfolios, and the like.
Donald MacKenzie
And then the other thing, final thing I'll say on this is in a sense of reference back to your to your to your comparison with or Merton's comparison with engineering. where of course exactly the same things apply to models in engineering, in the sense that the model of a bridge is not the same as the physical bridge. And the engineer knows that perfectly well. And of course, what engineers do in their everyday practices is to build in safety margins. They don't build in the flimsiest bridge that the model says will still stand up. They build a somewhat stronger bridge than that. There can also be in the financial system, an economic incentive to reduce the safety margin, because you can make more money by reducing the safety margin. I mean, think of banks operating with only a tiny sliver of equity and of course, there's a sense in which equity is the safety margin for a bank. There were perfectly understandable economic reasons why banks wanted to do that prior to the, to the 2008 crisis. But of course, this smallness of a sliver of equity was an important part of the story of why some relatively limited problems in the US subprime market suddenly became a threat to the stability of the entire global banking system.
Meyrick Chapman
Yeah, yeah. Well, that's, that's true. I lived through that, as well as lots of people lived through it. But but actually sitting on the trading desk quite close to the money market desk was was really something I was interested in you you mentioned there the LDI episode and I, I wrote a piece for the FT which played exactly on the points you make, in fact, it was called 'When Regs Bite Back'. And that looks at how the regulatory framework which has been created actually introduces its own risks, which I suppose is exactly the it's exactly the same model is the same performativity that you're that we were talking about earlier, and you have highlighted in the Black Scholes Merton model. I think, perhaps, moving on a little bit, that it's, it's natural to sort of bring things up to date. And perhaps you could talk about some of your more recent studies in high frequency trading, which, in my view, were or are a natural progression from your studies in the 'Engine' book, and I'm thinking in particular about your discussion of technology and mathematical models as key variables in high frequency trading, and how the high frequency trading industry in some ways is an arms race to stay at the cutting edge. And that, in effect, you've just described an arms race where equity gets reduced in banks, that's kind of an arms race to prove that you are as profitable or more profitable than your competitors. And in the in the case of high frequency trading. Does that just does that make it another technologically advanced form of asymmetric information for insiders versus outsiders? Which is been a characteristic of financial markets for a long time?
Donald MacKenzie
Yes. I think the answer to that is yes, yes, yes and no but let's start with a parallel with human traders. Specifically the specialists on the New York Stock Exchange, who were the trader, the designated traders, who coordinated the trading in any particular stock. The specialist was at least in theory, in actual practice, it was a little bit fuzzier than than this, the specialist was the only person or rather the specialist and their clerk or clerks were the only people who could see the order book. In other words, could see the list of bids to buy and offers to sell that had not yet been executed. And so the specialist had, in that sense, privileged information. If you were a broker on the floor of the New York Stock Exchange, you might sometimes be able to glimpse the order book because originally Of course, it was literally a book a set of pieces of paper and the clerk had to open it to write down your order. And when it was open, you could perhaps glimpse what the other orders in the order book were, but the issue there was, who can know that?
Donald MacKenzie
In a modern electronic market, it's typical, not universal, but typical for the order book, to be electronically visible to all participants, not just designated people such as specialists, and the issue there becomes not who can see the order book. But as it were, when can they see it? How fast do they receive the information about changes in the content of the order book? And so that's the kind of asymmetry in information that becomes relevant to high frequency trading.
Donald MacKenzie
And in a certain sense, high frequency at its core, high frequency, high frequency trading is pretty simple. The models that are directly employed in the trading can't be super complicated computationally, because if they are they take too long, too long to run. So as it's at its core, at its fastest core, high frequency trading, is about some pretty simple signals.
Donald MacKenzie
So for example, the ES as as it's called the Chicago Mercantile Exchange futures contract that tracks the S&P 500 index, if that moves in the data center in the outskirts of Chicago, and an algorithm is trading shares in the share trading data centers (which are in the United States they're all in northern New Jersey) you need to get that signal as fast as possible from Chicago to your algorithms in the data centers in New Jersey
Donald MacKenzie
And the very specific nature of the speed race above all, I mean, there are other components to it, but this is the this is the key one. The specific nature is that you can in a rough sense, divide the algorithms of high frequency trading into two categories. There are market-making algorithms that keep bids and offers in the order book so that the participant that comes along will have a bid or an offer against which they can trade so the there's a sense I mean, to call them passive is actually in some senses quite wrong. But the the market-making algorithms orders sit in the order book, waiting to be executed against.
Donald MacKenzie
There's another category of algorithm the market-taking algorithm sometimes called aggressive algorithm, and what that can do is as it were waiting the wings not doing anything most of the time. But when an opportunity arises it can pounce and the classic opportunity is the stale bid and offer. So prices have moved in Chicago. But a market maker system has not yet updated its bids and offers to reflect that. So the market-taking algorithm can profit by executing against the stale bids or stale offers. That means that there's a race every time one of those simple signals arises market making algorithms rush to cancel their stale bids or offers and taking algorithms rushed to execute against those offers before they are canceled. The wonderful work by the Chicago economist Eric Budish, with a couple of co-authors who then worked for the UK Financial Conduct Authority, actually trace the frequency of those bids and offers in the markets for the FTSE 100 stocks in London as traded on the London Stock exchange, and found essentially one race per minute on average for each of the symbols of the each of the stocks of the FTSE 100. So those those races are completely everyday events. They are, so, you know, you see them, if you have access to the right data, you can see them all the time in finance, in financial markets in financial market data, but you need the right data. Because one of the troubles with the standard order book data is that to identify the race, you have to identify the actions of those who lose the race as well as those who win the race and the wonderful and methodological contributions of Budish and his co-authors is to use London Stock Exchange messaging data so as to be able to spot those races.
Meyrick Chapman
That's yeah, it is it is amazing. It kind of reminds me of a conversation I've had recently about velocity about the velocity of money and how a monetarist, to take a famous school of economics, the monetarist school. The monetarist school in some ways must be outmoded because a quantity of liquidity doesn't have to be money per se, but it's financial liquidity observed at some point, could be monthly, could be weekly could be daily, but what you're describing is liquidity which ebbs and flows multiple times a second potentially. And therefore, the liquidity which we're really concerned about can change, the conditions could change very abruptly, but in between policymaker observations Have you have you had conversations or have thought Have you had thoughts about how policymakers? I thought you've mentioned the FCA so they are policymakers? Do they have any solution for for this ebb and flow, this expansion of credit if you like, and then simultaneous, perhaps a minute or two later a contraction of liquidity?
Donald MacKenzie
Yes. I mean, I don't think that policymakers are thinking about this in in the in the terms of monetary, monetary economics. But, you know, for sure there is potential policy interest in the question of the arms races because of course, the arms races and speed are expensive and ultimately they are paid for by the end investors in financial markets in the form of the bid offer the bid offer spread. You know, so there are some you know, so there is interest in the policymaking world about reforming trading, so as to reduce the incentives for speed races. For good or ill, though, so this is the policy makers, if you're a policy maker, you have to focus on the issues that are prominent on your desk right this week, this month, maybe this this year. This is more of a background issue. So it never gets centrally addressed, but perhaps it should be.
Meyrick Chapman
Yeah, yeah, I think this sort of overlaps with other areas I've been looking at for the last well, multiple years. But yeah, that's for another time. I mean, what you've been talking about here is an arms race. So if you like an acceleration of speed, and leaping ahead of competitors, and also the average customer, end customer in the financial markets. But there's another area which I find fascinating is that you also highlight the inherent limits of the high frequency trading approach. In fact, it veers into communication theory, whereby there are actual limits as to how much bandwidth can be used and how fast the messages can actually translate so or transmit. I wonder if you know that we can actually move that's from the kind of the, this is moving into almost the quantum mechanics of high frequency trading. But it does take us back to Black Scholes, Merton in a way because that model was a constrained model. It didn't properly describe the real world, and it was found out in the 1987 crash, we've addressed this a little bit now. And in the same way, that the Black Scholes model bumped up against its own limits, namely that markets are not normally distributed, as it had assumed, that in the same way, the high frequency techniques are bumping up against the limits of limits of physics, as I just said, Do you think there is some similarity here? I think I know the answer already. But it'd be interesting to talk about
Donald MacKenzie
There is a similarity, but there's also a very important difference, which is to say the limits against which high frequency trading is bumping up our as you just said, inherent limits of physics. Above all, Einstein's postulate, that the fastest that any signal can travel is the speed of light in a vacuum. And that's not just a theoretical issue now for high frequency trading. That's a real practical issue. You want to get as close as possible to the speed of light in a vacuum, which means that transmitting data from for example, Chicago to New Jersey, via fiber optic cable is just not fast enough, because glass of the core of a fiber optic cable slows light to down to about two thirds of the Einsteinian limit. So you want to therefore transmit through the atmosphere, which is almost as fast as as transmitting through a vacuum which means you want to use microwave. But as you just said, the bandwidth of microwave is limited by comparison of light with light and the fiber optic cables. So there is a kind of trade off there that's involved.
Donald MacKenzie
My feeling about the limits of Black Scholes, as revealed by the 1987 stock market crash and the role of portfolio insurance and as you know what portfolio shoot what the portfolio of insurance essentially was, was synthesizing a put a put option and therefore a floor to the value of your portfolio, according to the Black Scholes model. The way I would tend to think about it and this is this is not in any sense provable, but this is my instinct, that if if only one firm had been implementing portfolio insurance, you know, only one institutional investor, let's say had been implementing portfolio insurance, it might very well have been a perfectly sensible thing for that firm to do. The trouble in 1987 is that portfolio insurance was a widely used technique, and also I think was known to be a widely used technique. So traders were able to anticipate what portfolio insurers would have to do because of course, it's a mechanistic kind of of strategy. There is a rule as to how you synthesize a put, so you have to follow it and traders can work out what you're going to have to do. So I think, one mode of breakdown of financial models is that they can become less reliable, so to speak, the more that the are used, and the more therefore, that major participants in financial markets all have to do the same thing. under the same circumstances. It's very different being a big seller when you are the only big seller in the market to be a big seller when everybody else in the market is also a big seller.
Meyrick Chapman
It what I'm reminded when you were talking there, I'm kind of reminded of Minsky, that there is an inherent cycle the cycle is both benign, in which case the engine, the camera becomes the engine and it formulates its environment, the theory formulates its environment until imitation or pervasiveness or herding whatever you want to call it introduces fragility into the system. This reminds me quite closely of of Minsky's work, and it does also absolutely reminded all the time of how much imitation there is an investment strategies. You know, the stock trading platform eToro offers clients the ability to replicate the most successful recent strategies. Clients can simply request their their portfolio imitate these strategies. It seems that everywhere you look there is a tendency towards homogeneity in finance. The performativity that you describe in the engine book is obviously present in high frequency trading and the eToro imitation strategies. It leads to a form of trend following or herding, which means that the financial system as a whole is prone to runs like bank runs when the circumstances, particularly leverage, changes, and that's a we've got what I think is quite a good example of that recently in in the cryptocurrency world, which is another area you have you have looked at extensively. Is it possible to guard against the negative effects of this instinct, you mentioned earlier about making engineering robust in terms of bridge making? Is that is that is it is even conceivable in such in such a world as finance?
Donald MacKenzie
I think it's conceivable to a degree perhaps the most important degree is it regulators need to be very cautious about creating regulations that are going to force or at least given incentive to all market participants to implement the same kinds of model because of course, that is what that that is what creates imitation and as you've nicely put it, the run.
Donald MacKenzie
In a certain sense were it's not for that process, Liz Truss might still be Prime Minister of the United Kingdom, but it's for sure it's it's hard to guard against it in a more generic sense, because of course, people are typically in financial markets to make money. And if you see someone successfully making a lot of money, there's a natural temptation to try and work out what it is that they are doing and to do it as best you can, yourself.
Donald MacKenzie
The example of that, that I've been most struck by was the famous hedge fund, Long Term Capital Management, that spectacularly blew up in the end of the 1990s. Not in a certain sense because its investment strategies were stupid because they weren't I mean, they there was a lot of Schadenfreude about it because Merton and Scholes the economics, Nobel laureates were among the partners of LTCM. So people who did not like modern financial economics pounced on Long Term Capital Management as a sort of way of tarring, tarring the theory with the brush of this near catastrophic failure. But again, I think it was a case not of LTCM doing inherent stupid things. I mean, I'm told, for example, that after the banks that recapitalized LTCM, liquidated its portfolio a lot of that portfolio was bought by the Chicago hedge fund Citadel who made plenty of money out of LTCM's positions. So it wasn't that the positions were inherently stupid. It's because other people have been tantalized by how much money LTCM was seeming to make. And they started to put on very similar trades themselves.
Donald MacKenzie
And so in a sense of quite minor event, Russia's default on its ruble denominated bonds, which was surprising to people because countries don't usually denominate don't usually default on bonds in their own currency. So an inherently really rather limited event, triggered a kind of avalanche of forced liquidations of positions, adverse price movements, causing further forced liquidations, and so on. And that was all made possible by the fact that people were seeking to copy what LTCM was doing. And so that's, that's a kind of danger that is very hard to avoid. Now, an experienced hedge fund manager will of course tell you well, we you know, we're wary of crowded trades. As these as these would would, be called, these will be called. I mean, there is awareness of this. But it's one thing knowing that a crowded trade is dangerous, and it's another thing deciding to go against the crowd and not engage in the crowded trade. And sometimes you can lose your job by taking the second strategy.
Meyrick Chapman
I'd love to have you back because I think that this is this is a topic which I've missed out all sorts of questions. I'd love to ask you, but you've been very generous. With your time and it's been a great pleasure talking to you today. And I heartily recommend your work, particularly the work I'm familiar with, but it sounds like I need to brush up on some of your more modern work as well. And I hope we can chat again sometime in the future.
Donald MacKenzie
Thank you very much Meyrick and the pleasure has been entirely mine.
Exorbitant Chat: episode 2 - Donald MacKenzie