How AI could see mistakes affect stock markets a million times faster

0:12Simoney Kyriakou: Hello and welcome to the latest episode of Margin Notes with me, Simoney Kyriakou, editor of FT Advisor. Now, regular readers and listeners will know we've been talking a lot about risk the past few weeks on Margin Notes. We've discussed AI risk, cybercrime risk, and now we're talking about how risk, financial engineering and model uncertainty are shaping today's markets. You could say that listening to this podcast is becoming risky business in itself. But as volatility and structural change become the defining features of the global economy, as well as the domestic UK economy, how should advisors and clients understand various risks and navigate these in portfolio construction.

0:57Simoney Kyriakou: Well, joining us today to talk through all these things and more is Professor Radu Tunaru. He is the Professor of Finance and Risk Management at Henley Business School. He's the head of the ICMA Centre for Finance, and he's the author of Real Estate Derivatives from Econometrics to Financial Engineering, and the author of Model Risk and financial markets, from financial engineering to risk management. So if anyone knows anything about risk, it is Professor Tinaru. So welcome to the podcast. Thank you, Simone. No pressure. No pressure. Now, let's just start actually looking at what risk is and how we define it, because it can be defined in so many different ways.

1:40Simoney Kyriakou: And sometimes it's just used as this nebulous catch-all term but right now everything around us feels quite uncertain so how can we really define risk in a in an era of persistent uncertainty well it is difficult to define risk when the boundaries are changing all the time but in essence risk is still related to potential loss Risk has many facets, and as we discover, there are new facets appearing all the time. And it is intrinsically related to uncertainty. So in financial markets, we cover uncertainty through the contracts in the financial markets. In other words, financial markets exist because uncertainty exists.

2:34now in we go through a period of great transformations so we have blockchain ai sustainability climate change geopolitics all coming at the same time this is pushing the boundaries of risk but at the same time is opening the door for further financial innovations so it could be a great opportunity as well.

3:02Simoney Kyriakou: Well I like that risk can present the opportunity to think of new things and and you know create new opportunities but at the moment we all seem to be looking at traditional models of risk and using traditional risk modeling to create portfolios or to understand what's going on now. So what are the pitfalls with using those traditional models if, as you say, everything seems a little bit more risky now? One major pitfall is model risk. This is an area that I researched for a long time. So let's assume that you know exactly the risk measure that you need to work with, which is either known from academic circles or prescribed by the regulator.

3:49You still need to find a model that will help you to compute that risk measure. So you need to select a model out of the myriads of models which are out there. Let's assume that you'll find the best model available. You still need to estimate the parameters from the data available. So this uncertainty about which model is the best model, how to estimate the parameters from the data, is introducing further uncertainty if you want further risk into estimating the risk for your purposes, for your investors. Now, this process can be compounded by computational risk. So through the development of AI and quantum computing, we are able to execute things millions times faster.

4:43And we will see in the next decade how this will evolve. Now, this is very good. is helping calculating risk faster, but it also introducing risk in the following way. Any mistake that you do, it will be executed millions times faster. So you can have an AI moment when all the investment banks or large investment banks are using AI models to calculate for their risk calculations. And if there is a mistake, which is there in the models, all AI models will take the mistake together and they will execute perfectly at the same time the wrong thing. The problem with that is that the regulator will not have the time to see it, to digest it, to react to it.

5:33So this problem can become a huge issue in the next decades, depending on how the AI is implemented in the financial system.

5:42Simoney Kyriakou: I mean, we're talking that could potentially move markets, right? It could cause stock prices to fall by millions, maybe even millions. I can give you an example. May 6, 2010, the flash crash. In just a few minutes, Dow Jones dropped by 9%, wiping out$1 trillion from the financial markets. It recovered in the next 20, 30 minutes. But that's already pointed out to the problem, to the issue. Since then, the regulators started putting some measures in place, but this is still a great risk related to the AI in financial markets. Well, that's right, because AI is presented to this wonderful opportunity and traditional risk models don't really account for that.

6:37Simoney Kyriakou: But when we're looking at the latency, extremely high push to improve latency and on stock exchange trades, we run the risk of creating significant, significant errors. That's really worrying. Indeed, the most successful companies lately are companies using algorithmic trading, using a lot of AI. so if something goes wrong there it will be done a lot faster and on a much larger scale oh gosh you know my dad gave me a coffee cup which says if i drink caffeine i do things stupid things faster and with more energy but this just seems like it's even worse you know faster more problematic much deeper and then how do we then unravel the mistake you say it will take the regulators naturally longer to note the problem and then move on.

7:34Simoney Kyriakou: But once that money has been lost, once those markets have moved, how do you then unravel that? Do you then just put a blanket stop and say, right, no more algorithmic trading? Or how do you then start to unpick the problem? The cycle always goes the same way with new innovation. There is the enthusiasm, the exuberance about the new innovation. then there is some type of crash that shows that actually is not so easy. But the survivors of that period will consolidate and they will show the way forward. So you need to go through this kind of process and the pain to have the game in the end. Well, that means that nobody will want to be the first mover with it.

8:22no actually there will be many that will be tempted to be the first movers because there's an advantage of the first movers into the new territory but if you remember the dot-com the dot-com was a disaster at the time but the companies that survived after uh became really you know the huge companies that we see today i mean that's certainly these are big risks then

8:47Simoney Kyriakou: that we have, both with traditional models that can't quite cope with the modern era and the AI models that represent the modern era could be problematic. But let's look slightly beyond investments because, again, you mentioned the GFC, you mentioned the dot-com crisis. There are other products that have been using financial engineering that could also carry a huge amount of risk that's not often discussed, things like mortgages, things like asset-backed securities, things like equity release. Could you talk about some of the risks that are sort of being put into the market through these kind of products?

9:27Yes, I think many investors perhaps don't realise that the mortgages do represent the market. So if you talk about UK, we are talking about 1.5 to 2 trillion sterling pounds in outstanding loans. globally$60 trillion outstanding loans. Now, mortgages are exposed to interest rate risk, which is managed through interest rate swaps. For the UK, interest rate swap outstanding loans are around$150 to$250 trillion, whereas swaps globally, outstanding loans stand at about$600 trillion. So the proportion of loans in the UK managed versus globally is a lot smaller, it's about 4%. But in terms of swaps, London is a huge market.

10:27Now swaps were financial engineering products that were only of academic interest until the 1980s when they were starting to be applied to manage the interest rate risk in mortgages. So this is how this market exploded. It's a very important product, and it shows how the innovation can drive financial markets. Equity release mortgages are a very interesting product because they solve two problems for governments in major economies. The cost of long-term care, which is spiraling out of control in all major economies, this is one big problem. But the second big problem is the pension deficit. So with equity release mortgages, you can help to solve both these problems.

11:18Now, to manage a portfolio of equity release mortgages, it's not so easy because you have a bundle of risks combined. You have interest rate risk, you have house price risk and mortality risk. And if I may, I would like to add that in the UK, you also have model risk. Why model risk? Yeah. Because model risk is only present in all these individual types of risks, but also because the regulator in the UK is unfortunately contributing to model risk. They specify the model, they specify the value of the parameter to be used.

11:57Simoney Kyriakou: Okay. Is that something that the regulator has directly been doing or is that just a reflection of the market in the UK? This is how they considered to take a more conservative position and to dictate the terms in terms of modelling and the parameters to be used. So they were very conservative, which is a good thing in terms of crisis, but it's not a good thing in terms of developing this market. So at the moment, the equity release mortgage market is only 1 % of the outstanding mortgages. So it's still a lot to grow. And given that this is such an important market for the government, I think you need to reach the right balance in terms of regulation and freedom to develop to allow this industry to grow.

12:47And it is also, in my view, unprecedented that the regulator will specify the model and the values of the parameters. I think this is for the market players to decide.

13:01Simoney Kyriakou: But it's a difficult thing for the regulator because it has such a multifaceted role anyway. It's now being tasked also to deregulate for growth, but it's also being tasked to promote competition in the market. It's also being tasked to protect consumers. So it has all these pillars which sometimes work really beautifully in tandem with each other. And sometimes they sort of almost conflict with each other. And I think the regulator did the right thing in imposing very strict controls on mortgages and strict affordability checks back in sort of 2014, 2015, just to try and make sure that people were not overextending their credit and protect the consumers.

13:42Simoney Kyriakou: but now it's opening up again to allow more lenders to lend and to allow more products to innovate that also presents a risk doesn't it because then you're introducing the human element back into it you're introducing human risk back into this mixture of risks that you were discussing earlier you know people just overextending their credit that's correct but first of all i like to start by saying that the disasters in the mortgage credit markets observed during the global financial crisis occurred mainly in the United States. Europe managed pretty well that period. And even in the UK, American banks that were managing UK mortgage portfolios did quite well.

14:30Simoney Kyriakou: We didn't have the Fannie Mae's. We didn't have the free banks. So the problem for UK and European banks was liquidity, which was impaired because the players who were supposed to provide liquidity were American banks that had their own internal issues. Now, since then, a lot of regulation, a lot of the regulatory environment changed quite a lot. And that's why private equity companies moved into this space, because they did not have to deal with regulation the same way like the investment banks had to do and they were supposed to use capital in a different way. Whether the same kind of mistakes will reoccur, time will tell but in my opinion if they do they will happen again in the US, not UK, not Europe.

15:27Simoney Kyriakou: Right, so there isn't that sort of contagion risk that perhaps we are afraid of? contagion will come through liquidity so i think bank of england and various other players learned from the problems related to liquidity at the time and they are better prepared to deal with that yes yeah it's not often said um professor that i come out in defense of the fca but i certainly came out swinging for them there and then. But again, let's look at sort of the mortgage credit markets still. Let's stay on that. Let's look at asset-backed securities. A lot of people are talking about asset-backed securities within investment portfolios, using those as debt proxies rather than fixed income.

16:16Are we perhaps kind of not learning the lessons from the GFC?

16:20Simoney Kyriakou: Is it different now? Well, it is different because you have a different regulatory regime. And I think many market participants learned from the mistakes. It's still vivid in the memory what happened then. This doesn't mean that securitization was a bad thing. I think securitization as a technology was a great thing, which just was misapplied. It's the same like you have with antibiotics. If you take the antibiotics the wrong way, many bad things can happen. If you apply securitization the wrong way, many bad things can happen. What is different now is the private equity companies, which are very opaque.

17:06The regulator doesn't know a lot about them. They don't have that capacity to look through the system and to understand the scale of the problem. Now, we've seen some problems in the US. I think UK and Europe in particular, because UK seems to line up on securitization more with Europe than with the US, I think they will not be exposed to large disasters like we've seen in the global financial crisis.

17:39Simoney Kyriakou: Well, that's certainly the hope. And keeping on private markets, The UK government has said we really want people to be opening up pension schemes and institutional schemes to invest more in private markets. Now, the UK government, I think it was the House of Lords, basically pushed back and said, we're not going to mandate investing in private markets because we want people, investment schemes, pension schemes to be able to say how much and whether or not to actually invest in private markets. But we are operating on this kind of deregulate for growth agenda in the UK. We are seeing the need to back British, to promote infrastructure investments, and it seems the government is very clearly making a push this way.

18:28Simoney Kyriakou: Given what you said about US private markets, and I appreciate there's a slight difference between the UK and Europe and the US, what sort of risks could there be in this kind of deregulate for growth agenda? One major risk is that investors may become risk tolerant. So there will be a behavioral change. And the second risk is that this regime change occurs at the same time with very high geopolitical risk that is difficult to measure, predict and control. so you can have good situations, you can have bad situations I think deregulation, we've been there before in the beginning of 2000s, there was a lot of deregulation Alan Greenspan put more faith in the markets to run themselves and we know what happened after that but there are also positive examples and I can give you two When investment banks were trading credit for swaps, they insisted on collateralization and they did not recognize credit ratings.

19:42This was nothing to do with the regulator. This was bottom-up from the industry, self-organizing, and it worked really well. The other example is BNP Paribas, which in August 2007, they pulled out of the ABS market realizing that their modeling was impossible to calibrate with what was going on in the market. So they could not understand what was going on in the market. So again, it was not a regulator or some other entity that will come and say the market is really skewed now, was an important market player. And they pulled out of the market starting the beginning of the global financial crisis.

20:29Simoney Kyriakou: It could work. in both ways. So a bit of self-regulation. I remember the BNP Paribas example very clearly, because at the time, people claimed BNP Paribas had suddenly adopted a boring approach. But as the GFC deepened and liquidity needed to be put back in the market, everyone turned to BNP Paribas and said, please, can you help kind of push this pipeline through? Because we need liquidity to be pushed through. And people turned to BNP Paribas in Paris and said, can you please help. So it actually seems that sometimes being boring or being accused of being boring is the best way. Again, the same with the dot-com boom in the early noughties.

21:11Simoney Kyriakou: Phillips and Drew was accused of being boring and it really worked against them, but actually they were proved right. Do you think there is therefore merit in advisors and investors just being a little bit boring? Yes, I think when you don't understand how things work, or maybe you see a huge change from the way you understood things were working, I think it's very good to take a step back and pause. investors often forget that you have three possible states you can be positive and act accordingly you can be negative and act accordingly and you can also pause in the you don't do anything you just retract and wait and i think the you know this is very important when markets change extremely fast now what is happening now is geopolitics which is is not something you are prepared to see every day, every year.

22:19You know, these kind of events occur maybe once every 50 years, 100 years. So it's more difficult to navigate when you don't have some kind of memory what will happen in the markets afterwards.

22:32Simoney Kyriakou: But then we've had three significant events since 2020. We've had COVID, we've had Russia's invasion of Ukraine, and now we've got the Middle East crisis. So it seems that perhaps these big events that only happen sort of once every couple of decades are happening a lot more quickly now and having a lot, a far greater and deeper geopolitical effect on everyone. We're just so interconnected now. Do you think that because of this interconnectedness, Now, we also have to accept that geographical risk is going to become much more prevalent. And then how do we protect ourselves against that? I mean, we can't completely just go all domestic and put all of our money straight into the domestic markets because it doesn't work that way.

23:24Simoney Kyriakou: But what could we do to kind of protect ourselves, protect our clients? um the examples you gave are really good because they they showed that diversification does not work um in in times of crisis diversification helps you to manage volatility which is not during the crisis is outside the time of crisis during crisis uh like you said covid war in ukraine etc etc., everything becomes very highly co-integrated. So everything starts to move the same. Now, how you manage that, in my view, you need to have to prepare for that scenario ex ante, to know that if you arrive into this scenario, you know exactly what strategy to use.

24:18And even more importantly, if you can, buy some financial instruments that will offer you protection and you buy those instruments in good times. So 2014, 2016, when it was a period of very low volatility, it was straight from Minsky theory book, you know, low volatility is destabilizing. That was the time when many governments perhaps would have bought very cheap options to protect themselves against this kind of difficult global crisis.

24:52Simoney Kyriakou: I mean, fix your roof when the weather's dry. Yes, exactly. But COVID is different. I think we should not put kind of two and two together because COVID is different. The geopolitics one, we've seen flashes of this before. The biggest, the most important event in financial markets was Black Monday, 1987. And that happened on a Monday, like the name says. But actually, it happened because the week before, there was this firing between U.S. Navy and Iran Navy, but very similar at a smaller scale, though, to what we are seeing now. So investors in Asian markets panicked, thinking Third World War is starting, and they dropped all the exposure to U.S.

25:45stocks. So it was just a panic from the Asian investors. Now, this could happen again, that, you know, if this conflict prolongs for a lot longer, I think we are going to see a lot more U.S. stock being done by various important financial market centers. It's difficult to react now, because now you don't have that much room for maneuver. Everything is expensive. Any protection of any kind is expensive. But investors could have thought about such a possible scenario a few years ago, maybe when seeing the invasion in Ukraine. They could have thought this could spread out and, you know, put their protection for their portfolios at that time at a much cheaper price.

26:39Sure.

26:40Simoney Kyriakou: I guess it goes to show the value of having a really good financial advisor to talk you through this and to actually be that pessimistic voice saying, this could happen, that could happen, this has happened before, this could happen again. We almost don't like to talk about all the negatives when it comes to putting a financial plan in place. But actually, as well, you've made a career out of it, we really need to be talking about risk a lot more, don't we? Indeed, that's correct. Well, Professor Tinaru, I'm afraid that is all we've got time for. And probably I'm going to be upsetting all my listeners because they're having to go back and look at risk again and read up on all the modelling.

27:23Simoney Kyriakou: But thank you very, very much for taking your time to speak to us. Thank you for having me. Thank you all for listening. Until next time, take care and be safe.

27:38Thank you.