Let's continue session two and talk about risk. So let's talk about idiosyncratic risk, so called alpha versus systematic risk, so called beta. These terms come from the single index model the SIM, Bill Sharpe 1963. And this is the model for the returns on a portfolio, just random portfolio. We're going to call that I asked the index and the universe of all portfolios. The expected return of our portfolio over the risk free rate is the alpha plus beta times the excess of the return on the market portfolio minus the risk free rate plus epsilon. So, alpha is this portfolios, abnormal return. Its idiosyncratic return, expected excess return. You'll hear all of those names and they all mean the same thing. Beta is the portfolio's, beta non diversifiable risk or systematic risk. And then the epsilon terms are residual terms and we assume that they are normally distributed with the standard deviation and an expectation of zero. So that's the SIM. And it's going to look awfully familiar and awfully close to the cath lab that we already talked about. And you all will have studied this so I won't spend much time on it. And how it converges to the CAPM is simply as the portfolio becomes larger and it has more assets in it, the alpha tends to zero. And those residual risks the epsilon tend to zero as well. There's the classic diagram on the right where we've plotted this line, and the intercept is alpha and the slope is beta, pretty elementary math. And why does this matter? The punch line is don't pay risk managers for market beta. This is so important. There's a temptation to just own a position in the S&P, and then when it goes up, the risk manager will say, well, you ought to pay me for that. And I would say no, I could have just put that in autopilot. I want to pay you for the alpha that you create. I want to somehow figure out the beta and take that away. And there's many, many ways to take it away. One of them is literally, to have algos automatically extract the S&P 500 risk out of an equity traders book. And automatically extract all the delta we'll get to that in a moment. And the book is just continuously flattened and so, there is no P&L associated with market data. That is one radical way to do that and, we have done that in many places. So let's talk about systemic risk different from systematic risk. And that's an annoying the terms are almost the same. We've already talked about in the first session of the fractional reserve banking system, where the bank deposits labilities exceed its cash in some scenario. And that's because customers are all withdrawing their cash at the same time. And there are fractional reserve systems. We have the famous dictum of Walter Badget. Economists column fame, but more famous for being the Victorian prophet of central banking lend freely again good collateral at a penalty, interest rate. Now, we have lived through, what happens when financial intermediaries, no longer trust one another. You have market liquidity failures, and then the financial crisis specifically, precipitated by the failure of Lehman and the near failure and rescue have AIG. We could talk about this all day, and we will talk more, in a subsequent lecture on this topic. But for now, I'll just tell you a story. When I first started working on Wall Street. I showed up and I had read the classic book Futures Options and other derivatives that everybody reads and I knew the black-scholes formula for pricing options. And of course it has the risk-free rate and that then I asked a colleague. What is the risk free rate? And that colleague said, use live war. It's not quite an answer. And in those days he actually didn't have Google. I'm dating myself. And so I went to the library. I goldman sachs and looked up the whiteboard. And it says London interbank offered rate, the interest rate at which banks lend cash to one another on an unsecured basis in the London InterBank Market and I thought, well that doesn't sound even remotely risk-free. That was really a puzzle. When I went back with more questions and I was told, please just use Libor as the risk-free rate. And indeed Libor and other measures, other risk-free rate. Closer for instance to what the federal government would borrow at were essentially for most times and most places the same thing. They varied by a few basis points up until the financial crisis there, the spread between the federal funds rate and the London interbank offered rate, blew out by almost 300 basis points. And so I learned the hard way. That that's what happens in a systematic event. That's what happens when financial intermediaries or banks no longer trust one another, and so everything is fine until the moment that it's not. Let's talk about something that's happening right now. Coronavirus 2020. Is this another financial crisis? I hesitate to make a prediction. I will make some observations, especially in the first few days of the very bad news of Corona Virus. As they affected the financial markets in the West, we saw very wide bid offers in treasuries. We saw the whiteboard is spread, very similar to the Fed Funds live or spread go up from 3013 on the 21st of February. So pretty tight 276 on the 13th of March, just to put that in perspective, that spread peaked at 365 basis points in 2008. And we see the Federal Reserve, committing trillions of dollars to repo and other markets. And so there's many ways to interpret this. First, I would say that a profound difference between the current crisis and 2008 is that, in 2008 the crisis began inside the financial system. When various deep and crucial and important aspects of the financial system weren't working smoothly anymore, or weren't working smoothly or working at all. And then it took quite a long time for many of those dislocations in the financial markets to work themselves. Through the real economy into unemployment foreclosures, other kinds of problems. This crisis is quite different, where the dislocations we're seeing in the financial system. Are a result of rather than the origin of problems in the economy, and so we see the financial system mirroring or reflecting dire problems in the real world. Of course, there's always a concern that those instabilities could then feed back. Into the financial system and cause all kinds of problems, notably operational problems of various sorts, that's a concern. And it's one that's got everybody on high alert and there's much expertise and wisdom and how to handle those kinds of challenges. I'll also observe that the Federal Reserve is using almost all at once, the playbook that was largely developed on the fly in 2008. So you could see the current crisis as the Federal Reserve becoming lender, and indeed market maker to the world. Now the details are important and complicated, the Federal Reserve is supplying money. Or funding to special purpose vehicles where the treasury namely the US taxpayer, is the equity owner. And so when can have concerns about this question close alignment of the Federal Reserve and the Treasury. Historically, we've always kept them quite separate in their function, one could also see. The Federal Reserve's actions as related to the regulations introduced in 2008. And so, if you think about the 2008 regulations and we'll cover them in a later lecture. One way you could think about them is that the Federal Reserve would in the limit to make banks completely safe. Require them to hold all of their assets in Federal Reserve deposit, that would be an extremely safe bank. It would also be a bank that isn't performing any of the functions that society needs banks to perform such as lending and making markets. And so in that very safe special case, banks would no longer be banks. And so you can see that the desire to make banks safer all requires all kinds of trade-offs. And so one could have a reasonable debate on the risks introduced by the regulations themselves. Making the banks safer, requiring them to have capital and liquidity, requiring them therefore to keep large portions of their balance sheet. In the form of cash and treasuries in unencumbered custody accounts. That's a really good thing but if the banks are required to keep so much in cash, they can't perform their desired function. Then there aren't buyers and sellers than the repo markets, which we'll talk about don't perform as expected. And so you find the Fed then having to step in and become market-maker lender to the world which entails its own rest. So, how will the Coronavirus 2020 pandemic unfold and mirror itself in the financial system, I bought an observation, no predictions. So let's talk about some of the canonical risk measures you all will have heard of value at risk or VaR. VaR has a deceptively simple definition, Var is the quantity v such that there is a probability p. That the mark to market value of the portfolio declines by more than v over some time horizon, which are called here, cow. And so there's lots of different VaRs depending on those quantities, and so typically, we think about the probability p being quoted for 1% or 5%. And we think of VaR typically is quoted and the time horizon of one day. So for instance, the one day 99% VaR would be the quantity v, such that there's a probability of 1%. That the mark to market value of the portfolio declined by more than the over the time horizon of one day. Now one interesting thing about this definition if you're a mathematician. Mathematicians in the room will have noticed it's a non-constructive definition. It says nothing at all about how actually to compute R and then we get into the very important detail, is parametric or non-parametric? In other words, are we making assumptions about how the underlying market variable that drive the past value of the portfolio are distributed. If we can make those assumption such as for instance that return of the underlying market price. Are normally distributed, therefore that the market prices themselves are log normally distributed. Well that might make the calculation of VaR relatively simple and straightforward. At least for some kinds of portfolios, portfolios containing for instance, only stocks or futures. On the other hand, my portfolio might be more complicated, it might have nonlinear instruments such as options. And also it may be the case and indeed it just is the case, there is no market variable that I know of that is actually in real life log normally distributed. And so far, the calculations that we make with these parametric assumption will be more or less inaccurate. And potentially useless and potentially even dangerous as we shall see. There are many critics of VaR, Nassim Taleb one of my favorite authors, author of The Black Swan fooled by randomness. One of my favorite books of all time, calls VaR very simply charlatanism, VaR claims to estimate the risk of rare events, but it gives false confidence. It emphasizes manageable risks near the center of the distribution and it ignores remote and fat tails. What we're seeing in the COVID-19 pandemic is clearly an example of a fat tail. It therefore by ignoring those things and giving false confidence creates an incentive to take excessive and remote risk. Otherwise known colorful on Wall Street as picking up pennies in front of a steamroller. So now let's talk about other risk measures, these are the ones that I grew up with. These are the ones that dominate on a trading desk and they're called The Greeks. And you can see several of my favorite Greeks up here on the screen. The definition of the Greeks collectively is they're the partial derivative in the sense of calculus. Of the Mark to market value of a portfolio with respect to a variety of independent variables. Those independent variables could be the price of the stock, it could be the dollar yen exchange rate. It could be an interest rate, it could be the passage of time and all of these Greeks have colorful names. Many of them are actually Greek letters, Delta, Gamma, Vega, Theta, Rho, Lambda, Epsilon, the others. Let's see, Kappa is a Greek letter Reasonably certain that Venna, Charm, Vomma, Veta, and Vera are not. And I, after having worked almost for 26 years, can't reside up the top of my head the definition of many of these Greeks. The big ones are Delta, Gamma, Vega and Theta. So, Delta is the partial derivative of the value of the portfolio, with respect to the market price, and that's assuming a simple portfolio, that only depends on one market price, for instance, the S&P 500. Gamma, is the second partial derivative with respect to the underlying market price. Vega is the partial derivative with respect to the volatility of that market price. And Beta, otherwise known as time decay is the partial derivative with respect to the passage of time. There's many other signs of stress testing that happen in banks. And that gives rise to the question of how exactly does one measure risk in complex interconnected systems such as large financial institutions. If we go back to the relatively simple case of the Greeks, well, once we know these partial derivatives, we can use various theorems such as black shells. And we can construct a portfolio that helps us hedge the risk of our underlying portfolio hence the importance and relevance to the Greeks. To a trading desk, but if you're the federal reserve, it's going to be a limited utility. The system of course contains trading desk but it's much more complicated than anyone trading desk, vastly more complicated and it contains many other activities, besides trading. And so, one thing that we've learned over time on Wall Street is to supplement those Greeks, which is where we started our work back in the 90s with many other kinds of risk analysis or stress testing. So we use constructed scenarios and we would give these scenarios, various names. And some of those scenarios were informed by events that actually occurred in reality. So we had cases credit spread widening or CSW, and that would be a particular kind of constructed scenario. There was a concern about Greece, leaving the euro and therefore re denominating out of the euro. And there's been concerns about other currencies potentially leaving the Euro as well. We've had the experience I mentioned the Swiss Flash Crash when a currency peg, that everyone had depended on suddenly go the way. And so based on that particular experience, we can ask about other currency pegs and what might happen if they go away. We have also over the years created scenarios where everything goes wrong. All market variables go to extreme places. However, that scenario while dire is still at least internally consistent. In other words the same market variable can't go up and down, and so that would be an example. One of the most important stress test, is the one known as the Federal Reserve's comprehensive capital analysis and review or sicar. We'll talk more about that, where the banks, the bank holding companies that are regulated by the Federal Reserve are required to simulate themselves. Their cash flows income statement and balance sheet, nine calendar quarters forward into the future, in the face of a severely adverse scenario given to the bank's once a year by the Federal Reserve. And then demonstrate even in that dire scenario, even at the worst moment in the unfolding of the crisis caused by that instantaneous shot at the banks still have enough capital, to perform the crucial functions of lending. So the evolution of modern risk management for complex financial institutions, has been quite a saga over the 26 years that I was on Wall Street. When I first got there at Goldman Sachs, there were atoms in Lotus or Excel but those were being rapidly decommissioned and and why? Well, for all of its virtues, there's some fundamental problems that Excel has got. Here's one, which is that it's very difficult to audit an Excel spreadsheet. It's very difficult to centralize and share, you end up with multiple versions. It's really easy, accidentally, or maybe purposefully to replace a formula with a value and then not notice that that cell is not changing anymore.. And of course when you later notice that you've almost certainly lost money hardly ever find that you made money, and so excels got all these problems. So, I joined Wall Street at a time when Goldman Sachs in a part of its business and the foreign exchange business had made the decision to banish Excel, build a platform. And we aspire to have that platform reflect everything about the trading business and, that convergence of the business and the software, so that they eventually became one in the same. Took 25 years, turned out to be powerful for a reason I've mentioned previously. Which is, once you have software and business converged, you can ask counterfactual what if questions of the software, where it's a lot easier to lose money in software simulation than it is to lose it in real life. Now, even with that success, I still have a long list of Desiderata and many of which we've built at Goldman Sachs and others are built in other places. But we're still a work in progress in many parts of the financial system. You want the platform to be immutable. You want it to contain everything, and you want to be able to roll back to the world as it existed at any prior moment in time. You want it to be global, you want it to be distributed. You want it to be transactionally protected, a trade either happened or didn't. You can never leave your model in some confused or partial state. You want a language, a scripting language in which you can ask in real time on the trading desk counterfactual questions. You want data flow models, so that if one parameter changes then everything else that depends on it automatically updates itself. Here's the big project that's happening right now. You want this platform to expand, in a seamless way the front, middle and back office. You want it to calculate not only profit and loss. And all the Greeks so that you can determine your market price risk, but you also want it to compute your counterparty credit risk. What if someone who owes you money doesn't pay you? Operational risk? Well, what if that trade didn't really settle, or you didn't confirm it properly or book it properly, reflected correctly in your risk management systems. What about scenarios that erode your capital or cause massive and destabilizing outflows of cash, so called liquidity risk. Now, there's another progression that's happened not so very long ago. Many of these calculations were performed monthly. Some of the most complicated risk calculations. Such as C car are performed yearly. And there's an inevitable process by which you want something that used to have monthly. Now you want it weekly, and then you want it daily. And then eventually you want it to be nearly instantaneous, no matter how complicated the the calculation is. So as out there as it seems, wouldn't surprise me if in the limit banks involved to the place where they can perform a proxy or a rough see car and then analysis, essentially not just at the end of every day, but in front of any potential transaction And therefore almost instantaneously as positions of all throughout the trading and investing day. So let's talk about some of the disruptions that are going on wall street. Well, first the rise of the quanta, I use the term Quanta don't care for it much, but it is the term that's evolved on wall street. The rest of the world has really settled on the term Data scientists. As I mentioned, Goldman uses yet another term called strat or strategist, they're really all the same thing. These are people with the math and software skill set working inside the various businesses and functions of the firm. Now, as we've had the rise of the quantum has been a huge capital investment in technology. And there's little question in my mind about that, the effect of that which is that it's dramatically increased liquidity and is reduced the bid offer spread. In many high volume markets, US equities being one of them. And you'll find that hedge funds and also the sell side firms are increasingly relying on software data and algorithms, not people to generate alpha can't resist talking about something I overheard once on the trading desk so there were there were guys like me unfortunately back then they were almost all guys, now it's much more diverse people like me who didn't look at all like a traditional trader and in fact that was not a trader and I've never been a trader. I had a different skill set the skill set of converging. The software the mathematical model with the business and the quote overhead was UK that kind of money to those kinds of gods but important to attract great people to compensation is a part of it. And so I've certainly seen over time, the convergence in compensation packages for people with a map and softer skill set to the levels of those with a risk management or trading or sales skill set. Something else you've seen, if you extrapolate this trend of hedge funds that rely on software data and algorithms not people, you get the systematic or the quant funds. And I will tell you a little bit about the requirements of the quant funds. The CEO of one of the large quantum funds once told me that here was his measure of success or his interaction with Goldman Sachs. He said, you and I, Marty, will have a lovely lunch once a year. And apart from that, no one at your shop will ever contact anyone at my shop on the phone or by email. Now, off course that was the Ethiopian, what he meant by that is he wanted everything to go seamlessly in software between his firm and our firm. And if ever there was a phone call and email it would because of a problem or a failure in that flow of break and he wanted the number of breaks to be zero. There has been a revolution in data science led by people with this map and software skill set working on the trading desk I might have been quote quoted once saying the traders you can't code will become extinct. For me, coding is like writing an english sentence. It's just something that everybody has to do. Most of us will not be professional coders, just as most of us will not be professional authors, everybody have ever needs to know how to do it. So how have we seen this play out? Well, we've seen the rise of traders who code and you could just as easily call them engineers who trade. Traders who make markets and eight to ten individual stocks have simply disappeared across much of Wall Street and why is that? Well that's because the core intellectual property for making markets now resides in software, not in brains or gods or any other organ of traders and computers. Software can see across all the stocks in the market and can see the factors that are driving stocks up and down together in a way that's very hard for individual traders who were each looking at 8 to 10 stocks a piece. Also does individual Traders will externalise risk into the market in ways that are inefficient in terms of in terms of crossing the bid offer and also paying exchanges commit Exchange Commission and currently. And computers can certainly optimize across that dimension as well. Here's something else we're seeing. Salespeople who pick up the phone and say, what would you like to buy or sell are disappearing fast. On the other hand, sales salespeople to deeply understand the client's pain points are always going to have a job. And they're getting more valuable and an extremely high demand, just like great enterprise software sales people. People who deeply understand pain points, not just people who are taking an order, something else we've seen is strap again. That's the GS term. They're everywhere. Bankers of course, we still need them for advisory work and that doesn't seem to be changing anytime soon. But certainly there are aspects of the day such as constructing common stock comparison, so that bankers can evaluate one company against all the other companies in its sector and a consistent way. We had Goldman and others Firm and people at other firms and heavily automated that workflow. And of course, as we talked about at the onset of the class, we see APIs everywhere. So let's put this in context and talk about the evolution of sell side equities. From 2012 to 2020, equity market volumes have been relatively stable ups and downs of course at about 7 billion shares traded per day. Over that period, revenues in the Goldman Sachs equities business have been stable as well, but margins and profits are up. By the way, these are US fingers. Headcount and the Goldman Sachs equities business is relatively stable, but straps are now the single largest group followed by sale and then a much smaller number of traders. And so, that composition of the roles within the head count have changed. And to put all of this in context, within that 7 billion shares traded per day, you can see that nausea is still the single largest followed by NASDAQ followed by CBO. We mentioned those three families of exchanges. And then that fourth 1.2 billion is of course I yaks. And then there's a large volume of shares, almost half of the total, a little bit under 2.7 billion that is traded off exchange or upstairs by firms such as goldman sachs and of course there are many others. Now we've also seen how the sell side business model has changed the sell side moving from closed and proprietary risk management to more open more interoperable tools and services. Goldman Sachs Marquee is one example. There's the Harvard Business School case study and our reading materials. And there, golden is made elements of sec DB data and services available to the client by a web user experience. You might have read some articles saying it was being given away for free. It was being open source. That's not how I would phrase it, rather would say we took sec dB, encapsulated it in API's and gave access to those API's to our clients. And there's a vigorous discussion discussion about the business model for those API's in some cases, such as for data. That only the sell side has provided by API that can be subscription models. And there's also regulatory drivers for a subscription model. In other cases, make the API's available to clients who are already coming to us for their market, making requirements. The whole goal of Marquee is to inspire clients to connect their systems directly to Goldman Sachs via API. In fact assign a success would be if Goldman clients tell their other liquidity providers, we want you to provide yours services to us as well, and we would like you to do it in a way that is API compatible with Goldman Sachs Marquee. So Goldman had been on a journey for several years to identify all the verbs, all the activities of the Global Markets Division. Encapsulate all of those activities as API's, some of the API's will be internal only, but many of them will be externalized. I'll pause there and then we'll go on to the last chapter.
