Let's continue in this chapter to talk about trading, flash crashes, and risk. Let's begin with trading. Now I'm going to talk about the traditional setup of roles in the trading desk. We have salespeople, who generally sell specific products and assets to clients. Of course they also buy assets and products from clients. You'll find that sales team on Wall Street on the sell side are generally organized according to the asset class. Which is very different from how sales forces are generally organized. For instance, in the rest of the country where you'll typically find a geographic organization of the Salesforce, sometimes there will be a client segmentation of the Salesforce, but usually geographic first, then you have the salespeople selling all of the products of the firm. That's what happens in most of the economy. Though on Wall Street generally doesn't happen that way. You've got a particular salesperson for instance, selling cash equities, and you might have another salesperson selling over the counter derivatives, where equities are the under liars. That'll be the main form of organization. There also be a geographic overlay. You've got the traders. The traders can be market makers. They can be proprietary or prop traders. Sometimes or historically at least, there had been a little bit difficult to tell the difference. There's generally a great deal of confusion in the land and a conflation of proprietary trading and principle trading. In both cases, traders are putting the firm's balance sheet and capital at risk. But in proprietary trading, that specific kind of principle trading, the trader is buying things because he or she has a hunch that they're going to go up in value, and selling things because he or she has a hunch that they're going to go down in value. Whereas market makers are doing something quite different, as we mentioned, which is they stand ready to buy or sell depending on what the client wants to do. If the client derives and the client wants to buy, then the market maker is going to quote a selling price. If the client derives, and the client wants to sell, the market maker is going to quote a buying price. That's the distinction where it can become a little bit fuzzy, if there are important decisions to be made by a market maker once the market maker has bought or sold at the client's behalf, of course, doing the opposite of what the client is doing. The client buys, market maker sells and vice versa. Once the market maker has done that, the market maker has some hedging, some risk management to do. It wouldn't be effective or prudent instantaneously to hedge all of the risk, in fact the only perfect hedge is selling it, so you'd just be buying something and immediately selling it, that obviously wouldn't work. You'd have to do proxy hedging of various kinds. Then the residual risk or the basis risk between what you bought or sold from the client and the hedge is of course, a kind of risk that the trader will carry on the books with the firms. There are choices to be made about the amount of that residual risk, there can be a little bit of a proprietary lean and we'll talk about that more. Then there's everybody else. There's back office operations, there's compliant, there's legal, there's information technology or IT. Then there are other engineers embedded in the trading business, who are golden facts, they're called strats, short for strategists. They're called quants. Really just think of them all as data scientists. I'd talked a bit about potential conflicts between client execution and prop. This is something that the regulator is working with the self side, had been very careful to govern and mitigate and limit. But if you just step back and ask yourself a broader question about the way the whole system functions. Imagine a company that had built a search engine, but instead of the search engine that many of us, perhaps all of us use, the Google search engine, where you just type your query to a web page. Imagine that you had to pick up the phone and call a salesperson. Then you would communicate your search terms to that salesperson. The salesperson would type that search query into an internal and proprietary Engine, and then get you back the results, and then read the result to you over the phone. That wouldn't seem to any of us to be a very effective way for Google to work, but yet it's a pretty good description of how much of the sale side works even to this day. Obviously a lot of that room for transformation there. Let's talk a little bit about one of the regulations that have brought a great deal of change to the equity's business in the US, and this is the National Market System Regulation from 2005 or Reg NMS. We're going to get into this in much more detail in session 6. The National Market System broke the duopoly of the NYSE and the NASDAQ that reigned in the 70s. NMS consolidated and expanded many rules that had been promulgated over the roughly 30 years between the 70s when the regulators began the National Market System journey up to 2005, the date of Reg NMS. Reg NMS requires that exchanges make bids and offers, sometimes also known as ask prices, visible to all the participants, retail and institutional. There are many provisions in Reg NMS, but the most important one is called the Order Protection Rule. It's also known as the Trade-Through Rule or Rule 611. That rule, and you can go over to the right, established the National Best Bid and Offer or NBBO requirements. Here's a summary of what the rule says. The rule, of course, has a lot more detail and legal leads in it. It says that brokers must execute trades at the best which is the lowest available price, or ask or offer price when buying securities on customers behalf, and at the best or highest available bid price when selling securities on a customer's behalf. It requires trading centers to establish, maintain, and enforce written policies and procedures that are reasonably designed to prevent the execution of trades at prices that are inferior to protected quotations, displayed by other trading centers. We'll get into an example of what happens when a regulator is dissatisfied with a particular performance of one of the institutions that it regulates. Another particular term introduced by Reg NMS, which is a Protected Quotation. That is the quotation as a bid or an offer that must be immediately and automatically accessible and it must be the best bid or the best offer in that venue. So that's Reg NMS and a little detail on the Order Protection Trade-Through, Rule 611. Let's talk a little bit about some of the criticisms of Reg NMS, many would say that it causes excess fragmentation, so instead of the duopoly of two exchanges, we ended up with 57, [inaudible] , which some would say it's about 50 more than would really be effective as a matter of national policy. It greatly increases complexity. It created the opportunity for high-frequency traders to capture latency arbitrage, which is extremely short-term differences between prices for the same security but on different exchanges. The unfortunate effect that institutional buyers and sellers of large blocks have to access small quotations, and that could potentially tip off the short-term or prop traders or principal trading firms. Just in parentheses, I'll mention while Reg NMS is particular to the US, there is a European Union equivalent which is called MiFID II. Other critics would say that instead of prioritizing stability and liquidity, what Reg NMS does is it prioritizes speed and low fees creating a race to the bottom. Because of this problem, I mentioned on the left that institutional traders of large blocks must access small quotations, there's an increase in dark trading, trading on alternative trading system. We'll talk a little bit, give you an example of how that works. Let's go through the life cycle of a trade. Now, trading is an extremely complicated business, and so I'm not going to talk about all possible trades and all possible market of all possible kinds of products. There are just way too many to talk through. So instead, for this [inaudible] video, I've chosen three very different kinds of trades to give you a flavor of the very different kinds of trading that can go on and depending on a particular market, jurisdiction, and product and exchange, you'll get variations on this theme. Let's consider the life cycle of a trade. We're only looking at the front office right now, we'll get to the back office in later lecture. We'll consider three very different trades, these the will cover all the different kinds of trades and all the ways to execute them. But it's a spanning and almost an orthogonal set of very different trades, very different ways to execute. Let's go to the first one. Imagine you are a Robinhood customer and you enter a market order to buy 100 shares of Apple, under payment for order flow or PFOF agreement. Citadel Securities, Virtu and other market-makers pay Robinhood to process some specified portion of all the orders that Robinhood sees from its retail customers, including you. Remember, if you're receiving a free service, then you are the product. Robinhood gains almost half its revenue from PFOF, hence, no need for commissions. Market-makers are required to execute customer trades at the National Best Bid and Offer, NBBO or better. They specialize in fighting the most efficient way for them to execute at the NBBO. There's information content in that flow of retail orders. Information that Citadel and Virtu monetize in their trading patterns, hence their willingness to pay for the order flow. Arrange all the protected bids. Remember, those are the immediately and automatically accessible orders to buy the highest price across all the exchanges from high to low. Arrange all the protected offers, the immediately and automatically accessible orders to sell at the lowest price across all the exchanges from low to high. That arrangement is called the Montage. Collectively, the protected quotes from all the exchanges are called the Top of Book. Then list all the other bids and offers from best to worst. Here is a hypothetical montage for Apple in which the NBBO is 324.02 at 324.10. In the simplest case, you receive a bill at 324.10 and the sell order that you lifted has its quantity reduced from 200 - 100. In reality, the market maker might route the order first to ATS, alternative trading systems. Some of them designed to internalize orders, others designed to be low commissioned venues and yet others where the participants have the opportunity to cross at mid, halfway between the best bid and the best offer. If the market maker finds an opportunity to fill your buy orders somewhere at the same or better price than all the protected offers in the market, that's called price improvement, it can execute the trade off-exchange. Robinhood is required to review the handling of each customer order or else to have a regular and rigorous review process of the aggregate orders to ensure that all customer orders are filled at the NBBO or better. Note that the NBBO can change rapidly and unstable markets. By the time you execute, you may not have obtained the current NBBO. The SEC says that as long as your price was the NBBO within the last second, the market-maker hasn't violated any rules. Robinhood receives a $1.25 million fine from FINRA, the self-regulatory organization in December 2019, for having an ineffective review process and agreed to commission an independent evaluation of its best execution or best-ex procedures. Let's consider a very different case. You're an institutional asset manager who wants to sell one million shares of the SPY, as in P500 exchange-traded funds or ETFs. First, choose your broker based on the strength of your relationships and the quality of the broker's APIs and other services. Let's say you choose Goldman Sachs. Second, decide whether you want high-touch execution, where you talk to a GS sales trader on the phone or in the chat screen or low touch, where you pre-specify on a web screen or through an API call your explicit instructions on how to trade. For example, I want to achieve today's volume-weighted average price or view up avoiding the open and the close. Third, how urgently do you want to get this sale done? If it's urgent and high-touch, a GS sales trader will likely propose, immediately filling your order by using GS capital and balance sheet to buy the million shares of SPY, that's called a Principle Fill. At a price equal to an agreed and explicit spread from the NBBO. The sales trader may also electronically merchandise the order on a variety of venues. Alternately, you might ask a sales trader to work the order so the broker is acting as your agent for a commission and not committing as berms capital. In that case, there isn't all that much difference between high-touch and low-touch. The broker employs a variety of tactics or outgoes and these of course in software to facilitate your sale. ALGOs make several decisions during an execution, including among other things, which venues and order types to use, how to break up and space out the submission of individual child orders created out of your big parent order to sell a million shares and the price at which to trade. Now let's go to the last case. You're an institutional asset manager who wishes to buy $250 million on the run 10 year US treasury bonds. Very similar workflow applies for interest rate swaps and other OTC derivatives. Indeed for block trades of equities, many broker dealers still operate this way. Here's the old school on the phone version. As of 2019, roughly 30 percent of US treasury trades happen in this old school way. The percentage is even higher for corporate bonds and derivatives. You pick up the phone, call your favorite market maker with whom you've established trading relationships and legal documentation, and you say you'd like to buy $250 million worth of, on the run 10-year treasury bonds. You ask the market maker for an indicative offer, which the market maker doesn't have to stand by and then shop around and ask other market-makers for their indications. In the next step, you ask one of those market-makers for a firm quote and you say, good till canceled, good for a specified number of seconds or minutes. That's the quote you're looking for. You get the quote, if you then say you're done, the market maker is reputationally bound to fulfill the trade. If you're, the client, develop a reputation for seeking firm quotes and then not trading, you'll end up having a hard time obtaining good firm quotes. Note to the wise by requesting quotes from several market-makers, you indicate to the market some degree of interest in buying. Each market maker will assess the probability that the trade actually occurs at some estimated size, as well as the probability that you give her the trade, and trade accordingly. The active requesting multiple quotes from multiple market-makers generally moves the market away from you, especially if the size you want to trade is much greater than the commonly traded size or the so-called social size for that instrument in that market. Now let's think of another way, and this is the more modern way to trade. As of the late 90s, electronic platforms such as Tradeweb, began to automate and electronify the process I just described of asking for quotes. These APIs are called request for quotes or RFQ APIs. Roughly 70 percent of US treasury volume now trades electronically using a variety of methods. In an increasingly popular trading method available on Tradeweb, it's one of the dominant electronic venues for secondary trading of US treasuries with about a $100 billion of treasuries traded each day, you can consume an API called STAQ or Streaming Access to Tradable Quotes. You define the market makers from whom you want to receive tailored quotes. The market makers continuously stream firm bids and offers to you, tailored to you, your credit, your relationship for a variety of sizes. You have the opportunity to receive relationship prices from your favorite market-makers rather than prices from anonymous counter-parties, without communicating in advance your intention to buy or sell and the size you want to trade. Think of the stack stream as a collection of montages, such as the one we showed before for buying a 100 shares of Apple. Standard software consumes the stack API, as well as select RFQ levels also obtained by API from trusted market-makers to construct an ever-changing stream of montages. Armed with all that information, you select the best or lowest offered to Lift. Now I am expanded. I also know that there's a lot of information in there. You'll see that the little video is on Vimeo and YouTube, where you can also play it more slowly and you can play it a few times. It is worth going through in understanding these trade execution lifecycle for three very different kinds of trades. Now let me move on to the next topic. Flash crashes, which you all have heard about. There had been many flash crashes. The big one that a lot of people referred to is the one that happened on May 6th, 2010, and I'll go into that in some detail. That's listed here is the Dow Jones. Generally people felt like, we talked about the Dow Jones much, talked a lot more about the S&P 500, just because the Dow Jones is not market cap weighted. In that sense, rather arbitrary, and the S&P is market cap rated. But there's been many other kinds of flash crashes. One that happened in 2015 when the Swiss overnight announced they were removing the peg between the Swiss Franc and the Euro. People didn't see that coming. There was one of the Pound a year later. One that began last year. Of course, more recently, the March 2020 COVID-19 pandemic that we're all living through. You can see a very dramatic percentage change in the NASDAQ composite and in the Dow Jones over very short period of time. Let's go back to really the original Flash Crash which happened on May 6th, 2010. This is the Greek political crisis. There was negative geopolitical news all day long, endless talks about BREXIT. Starting at 2:30 PM Eastern Time, the major US stock market indices collapsed and they rebounded rapidly. The entire episode lasted about 36 minutes. At the low point, the indices dropped that nine percent intraday, erasing $1 trillion of stock market capitalization before recovering. Some stocks, the one I remember most is Procter and Gamble, but there were others, traded it at bizarre prices of one penny. Why did it happen? There are many theories. One of them was the fat fingers theory, someone just put it in the wrong trade. It's amusing hypothesis but demonstrated pretty quickly to be false. Another one was high-frequency traders are to blame. However, regulators concluded that a practice called quote stuffing associated with the HFTs was not a major factor, quote stuffing is placing and then almost immediately canceling orders almost as a team. Another theory is that there were large directional bets. What Alan Reid and others were hedging their long positions because of all the negative geopolitical news. Then there's another one, one that's yours truly thinks is the more likely to be a powerful explanation, which is broken market structure, equity markets under Reagan and after chaotic complex systems, nobody really understands. That would be unintended consequences of a well intentioned desire to keep the various trading venues consistent with one another. Here's something, technical problems at NYSE on the day led to five-minute delays in reporting NYSE quotes to the consolidated quotation system. Traders who had direct access to the NYSE quote, noticed the stale quote off by five minutes and they pulled all their orders. That's really the smart thing to do if you've got software that's trading these market data that doesn't make sense, just pull all your orders out. For other reasons, that left only stub quote and the stub quotes were to buy it a penny and sell at a $100,000. Other algos noticed the inconsistencies going on all over the place and they attempted to flatten their positions, the positions they already had at that moment by putting in contrary market orders. But then those market orders traded against the stub quotes. That's my theory of what happened. Who got blamed? A 2010 report by the joint regulators, the SEC who regulated the stock exchange markets. The CFTC who regulated the futures exchanges. That report blamed the flash crash on an institutional asset manager Waddell and Reed it didn't mentioned specifically, but the things they said it was an inevitable conclusion that entity executed an algo to sell 75,000, what will call the E-Mini contracts, futures contracts on the S&P 500. The notional value of that order was $4.1 billion. That's where the SEC and the CFTC pointed the finger of blame. In a related happening, UK authorities and the CFTC charged Navinder Singh Sarao with market manipulation, commodities fraud, and spoofing. He spent four months in a London jail in 2015. You remember he's a London point and click day trader who was working from the basement of his parents home in a London suburb. He spent four months in the London jail. In November 2016, he was extradited to the US where he pled guilty, in a Chicago federal court. In January 2020, he was sentenced to one year of home incarceration in London. I'll stop there and in the next chapter, we'll go on to risk.
