Technically Speaking, November 2016

Editor’s note: this was originally published by KCG at KCG.com. KCG offers services designed to address trading needs across asset classes, product types and time zones. The firm is one of the world’s largest independent market makers.

Crib Sheet:

The US Equity market is the most liquid and arguably the cheapest in the world for long-term investors. It is also a market where ETFs and hedge funds thrive. No single event got us here – it’s been a process of constant refinement and learning. Three consistent factors have contributed:

1. Innovation
2. Competition
3. Regulation

Since the invention of the telegraph in 1832, technology has had a disruptive influence on market structure. Often, regulators have had to step in to require incumbents to adopt technology, spend on productivity enhancements and break down exclusive access.

Now nearly all trading is done electronically, which has changed how trading is done as well as who can do it. Those who have thrived are the ones who have adapted: improved their processes, reduced their costs and improved their access to relevant information. New electronic markets mean venues and asset classes are better connected now than at any time in the past – and trade mispricing evaporates in nanoseconds, not minutes.

Exhibit 1: We estimate volumes accelerated consistently after WWII.

What’s the point of markets?

Efficient capital allocation

One of the core purposes of markets is to efficiently allocate capital and resources.

Primary markets help bring investors and corporates together, and help ensure that assets are sold for the highest price possible where (supposedly) they will add the most value. To achieve that, equity markets have evolved with a number of specialized participants with very different objectives. At a basic level:

• Corporates need to issue stock
• Investors want to buy stock
• Exchanges provide a listing venue

These participants are typically focused on collecting high quality and consistent information. This information is used to make valuation assessments more reliable.

Liquidity Provision

Secondary markets are critical to allow investors to reallocate funds to new opportunities. This makes trading a key function of markets.

But trading is not “free.” There are operational (explicit) costs and impact costs, which eat into the returns of investors.

Trading works well when more participants are involved, as different views bring participants to each side of a trade. Trading works even better for participants who can trade much more cheaply than the average investor – as they are then able to take on trades that long-term investors couldn’t afford to make.

That makes additional participants important:

• Exchanges or ECNs provide a venue for posting bids and offers, or issuing RFQs
• Brokers are important to help match buyers to sellers while retaining anonymity
• Hedge funds and traders, with higher turnover strategies, are important to provide medium-term stock trades which are more likely to facilitate large rebalances done by investors
• Arbitrageurs are important to ensure short-term relationships between stocks and indexes don’t break down, which also transfers liquidity between securities
• Specialists and market makers are important to facilitate trading when natural sellers (or buyers) cannot be found

These participants are typically transaction based. They benefit from investments they make in delivering productivity and economies of scale.

The earliest innovation

1792: A centralized market: lower latency, consistent rules

On May 17, 1792 twenty-four brokers signed the Buttonwood Agreement. Ironically, this agreement set a floor commission rate charged to clients and bound the signers to give preference to the other signers in securities sales. The agreement also established a centralized market, with consistent listing and trading rules, which made it much easier for information to be transferred and buyers and sellers to be matched. Over time, restrictions on manipulative trading were adopted as well as formal organs of governance.

1832: Electrification of messages impacts market share

The invention of the electric telegraph improved the speed and sharing of information. This in turn helped consolidate markets, and New York rose to dominance over Philadelphia after weathering some market panics better.

1872: Specialists bring intermediation and liquidity

Large buyers and sellers don’t always align – which, if a buyer (or seller) has urgency, creates volatility and, at the extreme, crashes. This is why a key requirement of the market is to also provide liquidity. The introduction of specialists, allowing intermediaries to facilitate large trades, was designed to smooth volatility and reduce the impact of large trades – theoretically a win-win for the investor and the intermediary.

1884: First index created

In 1884, Charles Dow composed his first stock average, which contained nine railroads and two industrial companies that appeared in the Customer’s Afternoon Letter, a daily two-page financial news bulletin, which was the precursor to The Wall Street Journal. This later became the Dow Jones Industrials average, in 1896. Because calculations were done manually, the index was computed as a simple sum of the prices of the included stocks (divided by a divisor to account for corporate events). The Dow remains a price weighted index today.

1924: Early mutual fund creates an ‘institutional’ investor

The first open-ended mutual fund in the US opened in 1924. This was critical to the democratization of share ownership for the masses. Open-ended funds allowed for small investors to pool assets in a way that afforded professional portfolio management. Over time, it led to a multi-trillion-dollar industry, which in turn saw funds so large that trades took days to execute. This ultimately created a new type of liquidity need in the market.

1933: Regulatory reaction

Not surprisingly, the result was a raft of new regulations – with comprehensive industry coverage – that remain mostly in force today, covering everything from IPOs to how exchanges work, to how investment funds protect client assets:

• 1933 Securities Act, covered IPOs. It required prospectuses with financial disclosures. It also included rule 144, which discloses onmarket sales by insiders.
• 1933 Banking Act, otherwise known as Glass-Steagall, separated banking and securities firms. It also created the FDIC deposit insurance as well as rules on speculative use of bank credit.
• 1934 Securities Exchange Act, created the SEC to regulate exchanges and secondary trading. Also included anti-fraud rules.
• 1938 Maloney Act Amendments founded FINRA (then NASD) to supervise members subject to SEC oversight.
• 1940 Investment Company Act, included rules to safeguard investor money from asset mangers, including restricting leverage. Modern ETFs still need exemption from parts of the 1940 act, although it specifically excludes hedge funds.
• 1940 Investment Advisers Act, was formed to regulate advisors. Does not mandate qualifications for becoming an advisor, but imposes fiduciary obligations and requires registration.

1960s: Inside information isn’t fair

By the 1960s, as turnover picked up, consensus changed on the value and rewards paid to specialists. In fact, even the SEC found that floor traders were using their informational advantages to fade orders or even to trade on.

Original available online.

The beginnings of electronic trading…

Electronic trading lowered the barriers to entry for alternative trading systems – leading rapidly to even more fragmentation – as investors unhappy with the restrictive status-quo designed better ways to trade.

1969: The world’s first ECN

Electronic trading was born in 1969 when Instinet, the first electronic communication network (ECN), connected institutional investors and promised disintermediation, anonymity and after-hours trading. In time, the traditional role of exchange as a trade handler would be blurred.

1971: The beginnings of an electronic stock market

The year 1971 saw game changers in two fields: the opening of Nasdaq and the invention of the microprocessor. Nasdaq was the first electronic stock market in the world. At first, no actual trading took place between buyers and sellers. Instead, the NASDAQ showed market wide quotes, which made the inside spreads easier to see for investors, effectively lowering spreads for informed investors.

Regulators spur innovation & connection

1970s: Exchanges are a little slow to adapt…

Electronic tools made it far easier to share data more equally. But investors who were not paying for direct market access were often still not aware if they were getting the inside (or best) quote. Regulators stepped in to adapt their rules to use computers to keep new markets linked and better disseminate trade and quote data to the masses.

1975: Regulations bring consolidation + competition

As the number of markets grew, the SEC was tasked to facilitate the establishment of a National Market System (NMS). Importantly, the Consolidated Tape System (CTS) was introduced to provide last sale information in a consistent way, despite the fact that trading was now happening on multiple different venues. In the National Market System, there was also to be competition, and the competition was to be fair. And for there to be price competition, NYSE’s fixed rate commission, a 183-year-old tradition, was finally abolished.

1976: NYSE launches DOT

NYSE went electronic by launching the Designated Order Turnaround system, and small orders would bypass floor brokers and go straight to specialists to be matched. A specialist, however, still had to manually approve each trade, which also gave them time to fade a stale quote.

1978: Exchanges connect enabling consolidated quotes

In 1978, the Intermarket Trading System (ITS) connected competing exchanges (not including NASDAQ, which was considered a quotation system for OTC traders), and orders could be routed from one exchange to another for better execution prices. In the same year, the Consolidated Quotation System (CQS) began disseminating all quotes from exchanges, to be joined by NASDAQ a year later.

Computers start to take over

1979: Machines are allowed to trade on their own

Only in 1979 did it become lawful for members of exchanges to trade their listed securities without being physically present, and that had great implications for OTC markets.

1980s: Investment & execution automated

Enabled by more readily available technology, program trading, where computer programs helped decide when to trade securities, rose in popularity in the 1980s.

1982-4: Nasdaq adapts differently for retail and institutions

In 1982 NASDAQ, the missing piece of a truly connected market, joined the ITS, and the computer linkage connecting exchanges and OTC markets was pitting securities markets against each other. In 1984, Nasdaq’s Small-Order Execution System (SOES) went live to, fittingly, execute small orders. It was designed to improve trading for small (retail) orders, not large institutional orders.

1986: The world’s first dark pool

Instinet launched the first dark pool in 1986.

1986: Redundancy becomes important

As it became clear that the health of the financial markets relied on hardware and software just as much as it did on people, Nasdaq in 1986 opened a fully redundant backup system.

1987: Crash and market makers ignore obligations

Black Monday occurred on October 19, 1987. With a backdrop of increasing military tension between the US and Iran, the Dow Jones Industrial Average (DJIA) dropped by 508 points (22.61%) in a single day. Some blamed selling by program traders, most notably the computerized selling of portfolio insurance hedges. Once again, market makers avoided obligations – on the floor (it was reported that traders wouldn’t pick up the phone) and on the Nasdaq. Retail investors were also hurt, many small orders were not responded to during the crash. Ultimately to ensure liquidity, participation became mandatory and beginning in 1988, SOES executed orders automatically

1988: Circuit Breakers to counter systemic risks

After Black Monday, regulators developed new rules, known as circuit breakers, allowing exchanges to halt trading temporarily in instances of exceptionally large price declines in some indexes.

The late 90’s rise of electronic trading

In hindsight, the period from 1995-2005 can be seen as contributing to a dramatic fall in trading costs. This is mostly due to automation.

1992: FIX protocol automates parent-order routing

The FIX Protocol was designed by Salomon Brothers and Fidelity in 1992 to standardize electronic communications in equity trading, so machines in different firms could understand the same instructions.

1993: First EMS empowers the buy-side

In 1993, Instinet launched Instinet OMS, an early example of execution management systems (which automate sending of orders and settlement instructions). The traditional role of broker as an order handler had been blurred; soon that role would become obsolete.

1995: First ECNs improve institutional spreads and speed

In March 1994, a study by two economists, William Christie and Paul Schultz, noted that NASDAQ bid–ask spreads were larger than was statistically likely. Around that time a number of ECNs became active – including Island (now owned by Nasdaq), Archipelago (now the NYSE Arca) and EDGE (now owned by BATS). ECN’s attracted liquidity inside the market makers spreads, primarily via a system or rebates to liquidity providers. These later spurred the maker-taker models of today.

Late 90’s: Early online trading for retail

In 1994, the first brokerage firm (now TD Ameritrade) offered online trading via its “WealthWEB”. However this really proliferated after 2001, with the development of dedicated online trading portals. Over time, many of the retail platforms decided to route orders to market maker firms to have their orders filled – circumventing the need for retail traders to understand the complexities of modern market structure.

2001: First algorithmic trading for the buy-side

Before buy-side algorithms were available, broker-dealers still had to work the orders. The delivery of algorithmic trading to the buy-side via their EMS’s helped automate trade execution process too.

Regulators spur competition

Although NYSE adopted the DOT back in 1976, and the ‘87 crash was blamed (partly) on program traders[1], computers really didn’t start to impact the market until auto-execution was implemented in the 1990s. This was critical to the development of buy-side EMSs and client-facing algorithms – which, combined with FIX protocol, changed the way trading worked.

1997: Order handling rules – an early NBBO

Order Handling Rules required that market makers’ best quotes be made public. To this end, if the best quote was on a private ECN, that ECN was allowed to communicate to the public system the best quotes they received, and market participants could access those prices through Nasdaq. Order Handling Rules had an immediate impact on official spreads which narrowed dramatically.

1997: Tick size reduction – to Teenies

The Order Handling Rules had another implication – ECNs and Nasdaq had to ponder over tick sizes. ECNs had finer increments (typically $1/64) than Nasdaq, so rounding was needed for display. In 1997, the tick size on Nasdaq reduced from $1/8 to $1/16, and NYSE followed suit.

1999: Reg ATS sets different rules for dark pool vs exchange

Rules for ATSs were still inconsistent, and they were regulated as broker-dealers despite functioning as securities markets. In an effort to bring ATSs into the national market system, Regulation ATS in 1999 prescribed that ECNs could choose to be regulated either as a market or a broker-dealer. But it also required adequate investment as well as setting reporting requirements. Dark pools and ECNs both fall under the definition of ATS. The differentiator is the lack of quote visibility on a dark pool. Consequently there can still be better prices off exchange, but you need to ping the venue to find them.

2001: Tick sizes go metric

In 2001, tick size reduction continued and fractional pricing gave way to decimal pricing, making the two-century-old Spanish system obsolete. In the next four years, even before Reg NMS protected NBBO, spreads collapsed 90%, arguably the most significant reduction in trading costs to date. This flowed back to investors not just with smaller spreads for their stock trades, but much more efficient index and ETF arbitrage – making hedging and cash equitization much cheaper. Importantly, this era represents a confluence of changes that enabled new participants, with higher turnover strategies, to thrive:

• Decimalization reduced costs for high turnover strategies like index and statistical arbitrage.
• Computers allowed for more, and more complex, trades to be calculated – faster.
• Electronification of the markets allowed computers to interact with the markets directly. Reducing operational costs and risk of partial fills and legged trades from computerized trading models.

Reg NMS: A solution to fragmentation and a source of new problems

2005: Investors still not getting best-ex

The emergence of new venues after Reg ATS together with a fundamentally more free-market and competitive system had increased fragmentation. Despite ITS, where a market-wide best bid and offer could be found electronically, it was obvious that some customers still weren’t always getting the best prices available. Reg NMS fixed all that by prescribing “best ex” to be the NBBO – regardless of other factors like fixed and variable exchange charges. This is in sharp contrast to the approach taken in Europe where critics say “too much judgment” is required.

2007: Forcing a continuously interlinked market

Regulation NMS was adopted in 2005, but not enforced until 2007 – giving participants time to adapt their systems to the wide-ranging changes it introduced. It took the idea of a consolidated NBBO further. Rather than creating a mechanism for the participants to see the quotes in the market – and hoping they’d do the right thing – it mandated that trades must occur at the best bid or offer. This put the responsibility for best-ex on the broker (and the exchange). In turn this meant that all participants needed to connect to each other electronically. Exchanges even have to route trades away to competitors if they offer a better market. This effectively automated away venue arbitrage.

Reg NMS created new issues

Reg NMS automated away many of the issues of the 90s, and created a truly integrated electronic market with equal access. But it unintentionally created new issues:

• Latency Arbitrage: Reg NMS automated away venue arbitrage. But it didn’t eliminate latency arbitrage – where traders who invested more in trading than exchanges did, could still grab mis-pricings across venues.
• New exchange subsidies: By guaranteeing price protection for any exchange showing an order at the NBBO, requiring that all participants take data feeds, and guaranteeing a share of SIP data revenues to all exchanges, the SEC made it much easier to launch a successful new venue.
• Maker-taker became critical for all exchanges to maintain market share. Without subsidizing market makers, the chance that a venue would show protected bids and offers was reduced.
• Order type complexity: decimalization reduced spreads to barely economic levels for market makers. The introduction of take fees made trading across the spread unintentionally, uncommercial. This led to increasingly complex new order types designed for rebate capture.
• Fragmentation: Combined, these supported even more fragmentation, at the exchange level and also in dark pools (to avoid paying expensive taker fees)
• Subsidizing liquidity provision encourages fade: maker-taker strategies and fragmentation combine to encourage overposting by market makers – which requires careful inventory management to avoid adverse selection. Although this means the depth of book is relatively unchanged despite decimalization, it makes book fade much more likely for slow routed orders.
• Broker conflicts: As most long term investors still pay all-in commissions, brokers must absorb whatever take fees accrue on client orders. Given the difference in inverted vs take fees is often a material percentage of execution only commission rates, routing incentives are misaligned. There is a lack of transparency into what the true tradeoff is between premium and low-cost routing, as well as how much low-cost routing has contributed to lower commissions.

A series of new crises

2007: Quant meltdown & computer feedback loops

In what may have been a sign that the market was frothy, August 9, 2007 saw what came to be known as the Quant Meltdown. Although the market was down a mere 380 points, the feature was the underperformance of quant funds. According to studies, initial losses triggered risk reduction, which created more trading, resulting in a vicious cycle for quant funds. Consensus seemed to be that the growth of quant funds – investment strategies powered by computers – had created consistent and therefor crowded trades. But it also reflected yet another way that computers could quickly react to new information and create a negative feedback loop that increased systemic risks.

2008: Credit crisis and great recession

In 2007, a series of credit related events unfolded and by 2008, powered by the collapse of an inflated property market, the US sank into a full-fledged recession. It seemed clear once again that loose lending and leverage had contributed to another dangerous asset bubble. Equities sank almost 57% from their highs, volatility climbed to levels not seen since the Depression. Some of the oldest securities and insurance firms in the US went broke and many others needed substantial government intervention to keep them solvent.

2010: Flash Crash

Just as the equity market seemed to be recovering, the US stock market plunged 9% in the space of less than 15 minutes on no new news. Many billion-dollar companies traded at stub quote (1-cent) prices and a majority of ETFs disconnected from their net-assetvalue. Now dubbed the Flash Crash, the blame fell squarely on computerized trading. An automated futures trade was thought to have started the crash, and HFT were later blamed for perpetuating it. There were a number of other contributing factors – the sheer volume of trading caused technical problems with the NYSE Arca and the consolidated quote system – that under Reg NMS was critical to ensure trading and compliant prices. The date in question was not without its own macro issues. The European sovereign crisis was unfolding, and ultimately dragged the market down even more than the low on May 6, 2010.

A series of renewed regulatory reactions

2008– now: Regulatory reaction in the aftermath of the economic panic, a number of rules were changed or introduced — all to address market volatility.

• Short selling of some stocks was temporarily banned as a regulatory coping mechanism. Seen by some as undue market intervention, the ban was the then SEC chairman’s biggest regret of his term.
• 2010: Stock circuit-breaker program introduced, and then enhanced across indexes
• 2011: Uptick rule re-implemented
• 2012: limit up and limit down price bands Perhaps most significant, Dodd-Frank was introduced in 2010 to ensure that banks are not too big to fail, and when they do fail, it is not because of proprietary trading.

2014: Reg SCI to reduce systemic risks

Many of the new crises had elements specific to the adoption of computers for trading. Specifically, because computers can process very fast, but not think for themselves (yet) there is a new systemic risk from the rapid formation of feedback loops.

• The quant meltdown was attributed to the rapid growth of quant funds, and the similarity of their data driven approaches and risk tools, which led to a concentration of factors across funds, and created a feedback loop as initial selling caused performance to degrade and model risk to spike.
• The failed FB IPO was caused by too much message traffic, and the inability of the matching engine at Nasdaq to process orders and cancellations fast enough.
• The flash crash was thought to be compounded by
  • a selling feedback loop caused by electronic market makers trying to exit long positions as the market cycled down
  • buyers exited the market as their models broke down under the extreme event
  • Exchange servers being bogged down by the sudden spike in message traffic
• In 2013, a software error in Nasdaq’s Securities Information Processor (SIP, part of the Consolidated Tape System) occurred and trading was halted. A year later, Nasdaq won the bid to keep running its SIP, with promises to speed it up and maintain a live, full-speed backup system, because without reliable technology, there would be no trading to even speak of.

Computers changed trading significantly

Of all the changes made in the last 100 years, the past 20 years has probably seen the fastest, and most dramatic, changes in trading. Trading has been revolutionized:

• Desktop EMS’s mean the buy side can control executions themselves.
• Computerized settlement systems mean it’s faster and cheaper to settle multiple trades
• Reduced trade sizes can minimize the signaling cost of every order sent to market

The impact of this can be seen in Exhibit 6 below. Average trade size has fallen 90%. To compensate, the frequency of trading has risen. However, trading frequency has increased far more, resulting in volumes today being around 10x higher than 20 years ago.

Progress stalled state since 2007

Interestingly, after a credit-crisis-induced spike, markets have returned to 2007 levels of average trade size, trade frequency and volume (Exhibit 6). In addition, value traded is flat (Exhibit 2).

It could be that the most revolutionary change was spurred by the rapid automation of trading around the late 90s – as EMSs, Reg ATS and decimalization combined to democratize trading.

Frictions may now be so low, that squeezing more returns out of the market for investors to keep might become much more difficult.

Investors have more responsibilities

In the early 1900s, the regulations reinforced a clear demarcation of the roles of intermediaries. Fund managers handed off trades to Brokers, who left the actual execution to specialists on the floor. Thanks to the advent of EMSs, electronic trading and algorithms (or online trading for retail investors), most investors now send orders directly to the market themselves.

This led to a dramatic increase in the level of responsibility, and required sophistication, of traders – just as the markets have become faster and more complex.

Everyone is a high-frequency trader

Whether you’re still calling in orders by phone or sending them from your desktop direct to a router, it’s likely that we are all trading electronically now.

The speed of light is constant, so the playing field is more level than you might think – especially with everyone using computers to trade, but:

• Being slow has always exposed investors to adverse selection.
• The selection of algo (or algo provider) is ever more important.
• Smart algorithms should outperform dumb algorithms.
• Fast routers with good queue priority should get better fills than slower routers.
• Using order types and limit orders to reduce your signaling is important.
• Even VWAP algorithms can take advantage of direct market feeds and forecast pricing.

On a positive note, computers are also very good at capturing data – so it’s also easy to measure if what you’re doing is better or worse than before. We find our market makers are always experimenting – with routing and flow and venues – to see what works best. That’s good advice for all of us.

[1] In the 1980s Program Trading represented the automation of stop orders, which would generate a ticket for the floor traders to execute. Today, program trading represents the working of baskets. Modern program trading benefits from the development of electronic routing and execution in modern markets – allowing computers to work thousands of stocks, and their child orders, at the same time.