Trading at the Speed of Light: Electronic Communications

Networks and High Frequency Trading

The concept of buying and selling securities repeatedly for small gains and losses is not new but the emergence of the Electronic Communication Network (ECN) and electronic order-working systems, or so-called smart systems, now allow investors to buy and sell millions of orders in seconds. High frequency trading (HFT) refers to the trading of a large number of orders at very fast speeds. The trading occurs in small time intervals and is characterized by high turnover rates, high order-to-trade ratios, and access to the most recent financial information. Today, HFT firms uses proprietary trading strategies executed by computers to move in and out of positions in seconds or fractions of a second. What was once an hour or day in the world of trading has become an eternity in modern HFT markets. This report examines the evolution of ECNs and the rise of HFTs.

—Wikipedia, "High Frequency Trading"

The pneumatic tube—something that you have likely seen the last time you drove through a bank’s or pharmacy’s drive-through—was invented by William Murdoch in 1836. At the time, it was not considered much more than an interesting combination of plastic tubes and vacuums. For financial firms, this changed in 1853 when Josiah Latimer Clark created a pneumatic tube system spanning over 200 yards that connected the London Stock Exchange to a local telegraph company.[1] The transmission mechanism for accessing information and trading, in turn, became significantly faster—considerably faster than the pigeon system that was used in the 1700’s. Stock tickers were later introduced in 1867 for conveying security prices. This coincided with the creation of an expansive telegraph network across the United States.[2] In George Rutledge Gibson’s 1889 book, The Stock Exchanges of London, Paris, and New York: A Comparison, he writes that because of these advances there now were “scalpers” on the exchange floors who frequently and repeatedly traded small amounts of shares, holding their position for no more than a day—day traders.[3] Traders and speculators saw that these two technologies enabled them to access information even faster, and ahead of others, greatly increasing their chances of making a profit in a shorter period.

            These technological advances made it possible for organized exchanges to provide a continuous exchange market. A continuous market attempts to have constant trading in a security. To have such a feature, time discrepancies caused by different times when investors want to sell and when others want to buy are minimized by having dealers take temporary positions in a security. On the organized exchanges, these dealers act as either specialists or market makers, quoting a bid price to investors when selling the stock and an ask price to investors interested in buying, hoping to profit from the bid-ask spread. To facilitate such trading, investment companies began to purchase access to faster private telegraph lines, and the New York, American, and regional stock exchanges began charging telegraph companies fees as large as $1,500 per month to access their trading floors.[5] By 1930, specialists and pit traders were buying and selling positions at the physical location of the exchange, with high-speed telegraph services to other exchanges.

Electronic Communications Networks

From 1930 to 1990, stocks were discussed as trading on the organized exchanges, such as the New York Stock Exchange, the Nikkei in Tokyo, the London Stock Exchange, or the DAX in Germany, or as trading over the counter—the OTC market. Since 1990, electronic trading offered by electronic communications networks (ECNs), such as NASDAQ, Instinet, Wunsch Auction System (later to become the Arizona Stock Exchange), Tradebook, and Archipelago, transformed the core structure of the organized exchanges and the OTC market. By definition, an ECN is an electronic network that provides a security trading system that brings brokerage firms and traders together so that they can trade amongst themselves. ECNs were set up to compete with security exchanges, the OTC market, and other ECNs. Many of the early ECNs were created by regional exchanges. Regional exchanges have had a long history of being innovative. For example, the Cincinnati Exchange (which discontinued operations in 1995), developed the National Securities Trading System (NSTS), which allowed automated purchases and sales from the offices of member brokers and from the floor of the regional exchange.

            As ECNs were beginning to emerge in the 1990s, the SEC ruled that such alternative trading systems (called Reg ATS) could register as stock exchanges. As a result, ECNs such the Wunsch Auction System and Archipelago (after merging with the registered Pacific Stock Exchange), became competing electronic stock exchanges. In 2000, Jiway launched its electronic platform in which 6,000 European and American stocks were dealt amongst brokers. In addition to the SEC’s Reg ATS, the growth in electronic trading systems was also aided by the SEC mandate that the NYSE create an intermarket trading system (ITS)—an electronic trading network linking markets and facilitating trades. As the NYSE established the ITS, NASDAQ also created an ECN called Primex. The ITS and the Primex system, in turn, were accessible to any registered electronic exchange—a benefit that turned a number of electronic networks into competing exchanges very quickly.

            In response to the emergence of ECNs and electronic stock exchanges, the NYSE began to demutualize—going from owner/member associations to public companies. In 2006, the NYSE merged with Archipelago, and in April of 2007 the NYSE became part of NYSE Euronext, a holding company created by combining the NYSE Group, Inc. and Euronext NV. Today, the NYSE Euronext includes six equities exchanges and six derivatives exchanges, providing physical and electronic trading in stocks, bonds, and derivatives. It is a hybrid physical and electronic exchange. NASDAQ, in turn, has become a super ECN with internal order execution capabilities and a centralized order book.

Electronic Trading

In the 1980s, the NYSE developed the Super Designated Order Turnaround System, SuperDOT, to facilitate both small and large orders, as well as multiple-stock orders. Using this system, brokers would send orders via a computer directly to this system. The SuperDot was one of the early electronic trading systems and a precursor to today’s ECNs. In 2012, SuperDOT was replaced with the NYSE Euronext Arca platform that trades more than 10,000 exchange-listed equity securities, including NASDAQ listings. Traders that use this open, direct, anonymous platform are able to make speedy executions in multiple U.S. market centers. In the late 1980s, NASDAQ established its electronic system, the Small Order Execution System (SOES) for dealers to enter their trades. In 2005, NASDAQ acquired the Instinet platform from Reuters.

            Today individual investors through their brokers, institutional investors, dealers, and block traders can execute trades electronically through electronic trading systems provided by organized exchanges, NASDAQ, or an ECN. Electronic trading systems include electronic order-working systems, or so-called smart systems, designed to “work an order.” These systems work an order by gathering price information from many markets, breaking orders into smaller sizes, simultaneously buying and selling a large number of stocks comprising a portfolio, linking global markets, slicing orders to be traded at different times of the day, and evaluating different market-maker tendencies. Order-working systems are set up with data inputs about the securities that are traded, such as customer orders, brokerage firm orders, limit orders, and inputs about the markets where the securities are traded. With this information, the order-working system then searches markets where the whole transaction or parts of it can be executed. Often the system passes the order or parts of it to market makers, order-crossing networks such as Instinet, and other electronic order-working systems.

            The 1974 Securities Act Amendment mandated that the security industry move to a national market system in which all investors would have easy access to information and the ability to transact security trading quickly and efficiently. Over the last 40 years, financial markets have seen the expansion of NASDAQ, the development of the intermarket trading system, the emergence of ECNs, and the transformation of the NYSE as a physical exchange into the NYSE Euronext. These developments have made it possible for investors to obtain current information on security prices and to buy and sell thousands of securities anywhere in the world in seconds (see Exhibit 1).

   Exhibit 1

Exhibit 1

High frequency Trading

High frequency trading (HFT) can be defined as quantitative trading strategies characterized by short holding periods executed by computerized algorithms. HFT strategies involve simultaneously processing large volumes of information made possible by today’s ECNs and electronic trading platforms. In 2009, the major U.S. HFT firms (Renaissance TechnologiesChicago Trading, Virtu FinancialTimber Hill, KCG, IMC, and Citadel LLC) represented only 2% of all investment firms, but accounted for 73% of all equity orders. In 2009, the total assets under management for hedge funds with high frequency trading strategies was $141 billion. According to a study by the TABB Group, HFT accounted for more than 60% of all futures market volume in 2012 on U.S. exchanges.

            Although, there is not a set definition as to what is “high frequency,” there are some characteristics that “high frequency” traders do share that help paint a picture. The first is that the trading is algorithmic, meaning that the traders are employing sophisticated mathematics through computer programs to make decisions. Second, HFT’s investment decisions are made in much smaller time intervals with similarly small holding periods. Finally, HFT strategies share an emphasis on low-latency technology (the delay between the transfer of data and the moment that an instruction to transfer was given) with the purpose of lowering their response times.[6] In fact, the desire for lower response times for making algorithmic trading decisions has led firms to implement microwaves (satellites), lasers, relativistic atomic clocks, and even additional network nodes (connection points for data transfer) in the vast swaths of the ocean; all to gain advantages that can be measured in millionths of seconds (Exhibit 2).

            The HFT market can be segmented into three groups. The first are market makers who provide liquidity and earn a profit by capturing differences in bid-ask spreads. The second group consists of agency brokers that buy and sell large quantities of an asset on behalf of their clients. The last group consists of systematic traders and arbitrageurs who try to profit from price discrepancies.[7] Just like there are countless hedge funds and mutual funds with different strategies, the same goes for HFT firms (Exhibit 3).

   Exhibit 3

Exhibit 3

Technical Strategies: HFT firms often employ technical strategies. Technical analysis is based on the premise that all fundamental information is captured in the market price and that market statistics reveal all information— “the market is its own best predictor.” HFT programs monitor the flows of quotes and volume information, trend lines, upticks, downticks, short sales, option trades, and block trades (Exhibit 4). For example, when a stock’s price increases above the previous price—an uptick—it can be assumed the price movement was initiated by a buyer; when the price decreases below the previous price—a downtick—it can be assumed the price movement was initiated by a seller. An uptick-to-downtick spread or a ratio of upticks to downticks can be used as a measure of investors’ sentiment. HFT programs can monitor large amounts of stocks for significant or unusual ticks and changes in volume activity and then generate a buy or sell order depending on the nature of the event.

 Source: Johnson,  Equity Markets and Analysis , 2014.

Source: Johnson, Equity Markets and Analysis, 2014.

   Exhibit 4: Bloomberg Moving Average Envelope  Macy’s Price, 30-day Moving Average, 7/22/2013-11/1/2013  Source: Bloomberg, MBTR Screen

Exhibit 4: Bloomberg Moving Average Envelope
Macy’s Price, 30-day Moving Average, 7/22/2013-11/1/2013

Source: Bloomberg, MBTR Screen

Event Strategies: Certain recurring events generate predictable short-term responses for certain securities. When an event such as an unexpected good earnings announcement, IPO, macroeconomic information release, or merger announcement occurs, investors reassess the value of the security and by their subsequent trading cause the price of the security to change. HFT firms are often capable of extracting event information before it even crosses the news screen. News in electronic text is available from Bloomberg, public news websites, and Twitter feeds. HFT firms, though, use automated systems to identify company names and keywords to trade news before people can process it.

Arbitrage Strategies: Introduced to the financial vernacular in the 1980s, program trading refers to the use of computer programs in constructing and executing security portfolio positions. Program trading often involves using programs to: (1) monitor real time data of stocks, futures, and options to identify any mispricings, (2) define appropriate arbitrage strategies given mispriced portfolios and futures and options positions, and (3) execute orders so securities can be bought or sold immediately and simultaneously when arbitrage advantages exist. Today, HFT is used to implement index arbitrage where HFT firms take a position in stock portfolio highly correlated with an equity index and an opposite position in the index’s futures or futures options. Unlike the program trading of the 1980s, HFT technologies allow for more complex positions to be formed while using minuscule speed advantages when arbitrage opportunities appear. High frequency trading strategies also exploit classical arbitrage strategies, such as covered arbitrage that governs the relationship between forward exchange rates, spot rates, and interest rates in local currency and foreign currency.

Time and Less Uncertainty

Every analyst knows that forecasting the future price of a security is challenging. It is a truism to say that it becomes more challenging the further out in the future the analyst tries to predict. Even in sports, the odds of a certain score or outcome approach certainty when there is almost no time left in the game. The importance of this idea can be illustrated by looking at simple histograms of financial returns over different time intervals. Exhibits 5 shows four histograms of price returns on the S&P 500 since 1980 for the following intervals: yearly, monthly, weekly, and daily.

  Exhibit 5: S&P 500: yearly returns since 1980, 12 bins, normality present.

Exhibit 5: S&P 500: yearly returns since 1980, 12 bins, normality present.

  S&P 500: monthly returns since 1980, 30 bins, normality present.

S&P 500: monthly returns since 1980, 30 bins, normality present.

  S&P 500: weekly returns since 1980, 30 bins, normality no longer present, some skew and excess kurtosis visible.

S&P 500: weekly returns since 1980, 30 bins, normality no longer present, some skew and excess kurtosis visible.

  S&P 500: daily returns since 1980, 30 bins, no normality present, distribution is clearly leptokurtic and exhibits skew.

S&P 500: daily returns since 1980, 30 bins, no normality present, distribution is clearly leptokurtic and exhibits skew.

  Exhibit 6

Exhibit 6

            Notice that the bars get taller and more clustered around the center the smaller the period size. That is, they become peaked (kurtosis) as we increase the periodicity. In effect, what happens to the statistical description of asset returns as investors shorten the time intervals is that their average returns become smaller, but the variance becomes smaller with “peakedness” increasing.

            Exhibit 6 shows a histogram of returns on the SPY exchange-traded fund calculated over minute-long intervals; which clearly no longer resembles the familiar bell-shaped curve. The distribution’s kurtosis for the minute-interval graph is over 1,700 compared to a kurtosis of three for a normal distribution. The graph demonstrates that high frequency traders buying and selling in short periods are more certain of the return they will get than one who buys and sells over longer ones. By frequently trading in these short time intervals, HFTs have been able to generate sizeable returns— “Picking Up Pennies in Front of a Steamroller.”

Flash Crashes

At 2:30pm on May 6, 2010 the S&P 500 dropped 6% from 1,133 to 1,065 in fifteen minutes, before rebounding to close the day at 1,128 (Exhibit 7). The so-called “Crash of 2:45” is a flash crash: a fast fall in the market caused by program trades. In a join report, the U.S. Securities and Exchange Commission and the Commodity Futures Trading Commission concluded that the cause of the crash was a single sale of $4.1 billion in futures contracts by an investment fund in order to hedge its investment position. The report went on to conclude that HFT firms quickly magnified the impact of the investment fund's futures sale. Another Flash Crash occurred on August 24, 2015 when a number of hedge funds sold positions that led to the Dow Jones dropping from 16,312 at 11:30 to 15,651 at the close of trading (Exhibit 8). The amount of selling was simply too much, and the liquidity providers whose algorithms thrive on buying and selling quickly were switched off when there was no one to take the opposite side of a trade. Theses crashes have led to concerns about HFTs— whether they lead to efficient markets or whether they are destabilizing.

            With the emergence of electronic trading and ECNs, the fundamental question about HFT is: “What data or information matters to an investor that buys, holds, and sells a security in a fraction of a second?” At an investment symposium at Xavier University in February 2017, a member of the audience asked the panelists to explain how economic data is correlated to stock returns. One of the panelist responded by asking over how long a period of time the audience member meant. The clarification was clearly needed. Economic indicators and stock fundamentals are without a doubt important to the long-term performance of the general market, but does last month’s GDP release or a stock’s price-to-earnings ratio matter to an investor buying and selling in fractions of seconds? Probably very little, and that means that a profitable and growing segment of the market influences asset prices through buying and selling based on data mostly related to momentum and volume. In essence, because of the scale permitted by computers, a form of market timing has become a much larger force in markets than ever before. What this entails for investors’ returns and perceptions of financial stability is not yet clear, but one possibility of what this may mean for the future of portfolio management and finance is summarized by Blair Hull[8]:

“It’s been ingrained that timing the market is a bad idea. The combination of data and predictive analytics, these two tools, can make it possible. Just as in the past 30 years it’s been considered irresponsible to time the market, in the next 30 years it will be considered irresponsible not to time the market.”