Innovations for Fairer Markets in the Era of High Frequency and Algorithmic Trading

When you close your eyes and imagine the trading floor of a stock exchange, what may come to mind is a rowdy atmosphere of pits of traders shouting and flashing hand signals to one another amid a flurry of paper tickets. In today’s reality, this imagery is almost completely divorced from the reality of how equity markets — and trading — operate today. Upwards of 70 percent of all trades executed on a daily basis on Wall Street are not executed by humans. In fact, they are not even executed based on a human decision. They are executed by computer algorithms, and occur at almost incomprehensible speed, frequency and scale.

A Brief History on the Importance of Speed

When trading in any liquid market, speed of information acquisition and trade execution have always been of paramount importance.

In the late 18th century, the trading organizations with the fastest horses were the first to market with their orders. In the mid-19th century, before telegraph wires were used in Europe to send stock quotes between major exchanges, Julius Reuter (founder of the eponymous firm) used carrier pigeons to transmit quotes from Germany to Brussels because it was faster than horseback, according to the Reuters article “Chronology: Reuters, From Passenger Pigeons to Multimedia Mergers.” As time passed, the question of speed was governed by access to telegraph wires, then the telephone and subsequently access to the internet. Those with speed have always had a competitive advantage, even when they only had the fastest horses.

High Frequency and Algorithmic Trading

The proliferation of high frequency and algorithmic trading began in the late 1990s, according to capital market research firm The Tabb Group founder Larry Tabb, and it ushered in a revolution in equity market microstructure. The revolution began with a seismic change in how long it took to execute a trade. To illustrate, Tabb estimates that it takes a human being approximately 12 seconds to execute a trade. To contrast, computer algorithms can execute a trade in a microsecond, or one millionth of a second. Put differently, in the time it takes you to blink your eye once, a computer algorithm can execute 100,000 transactions. In the time it takes a human to execute a trade, a computer algorithm can execute 12,000,000 transactions. This is only possible, of course, because exchanges are now largely electronic and have been since the late 1990s.

The cleverness of current high frequency and algorithmic trading protocols is remarkable. Commonly used trading algorithms include:

• Those that send bursts of heavy activity here and there to test the market (i.e., whether there are nibbles on the other side). These algorithms send out bursts of buy and sell orders for various stocks across various exchanges, and when another computer bites on one of them, within microseconds the algorithm cancels the orders for which there were no nibbles.
• Those that exist simply to combat other trading algorithms, head faking them into driving a price up or down so the algorithm responsible for the head fake can profit.
• Those that trade based on microsecond delays in price changes across exchanges (referred to as “stale quote arbitrage”). This is possible because the same stocks trade simultaneously across multiple exchanges, and there are differences across the exchanges in how long it takes them to updates their prices for the same stock. In this case, the algorithm would learn the price of a liquid stock on one exchange, and buy/sell it on another before that exchange was able to update its price for that stock.

As Michael Lewis wrote in Flash Boys, “The U.S. stock market was now a class system, rooted in speed, of haves and have-nots. The haves paid for nanoseconds; the have-nots had no idea that a nanosecond had value. The haves enjoyed a perfect view of the market; the have-nots never saw the market at all.”

This is precisely the problem the founders of IEX Group — the protagonists of Flash Boys — set out to solve, and it’s the problem Darden Professor Marc Lipson examines in the forthcoming case “IEX Group Inc.,” created in partnership with Darden’s Richard A. Mayo Center for Asset Management and co-written with Mayo Center Director Aaron Fernstrom.

Put simply, the IEX founders discovered these speed-based inequalities in the stock market and noticed that large stock exchanges were benefitting at the expense of investors representing pensions and mutual funds. IEX created three market innovations to try to solve for these inequalities: the famed “speed bump,” the Crumbling Quote Signal and the D-Peg order type.

Leveling the Playing Field in an Era of High Speed Trading

IEX’s “speed bump” — a mechanism for delaying inbound orders by 350 microseconds — was created specifically to prevent stale-quote arbitrage, according to IEX.

It takes roughly 300 microseconds for IEX to learn about a price change for a stock trading on another exchange and to update its orders accordingly. The additional 50 microseconds provides a sufficient buffer, ensuring that even if a trader instantly learns of a price change anywhere in the market and immediately submits an order to IEX, all pegged orders1  on IEX update by the time the inbound order is processed.

However, even with the speed bump in place, IEX noticed that trades were executing at prices immediately prior to IEX observing changes in the National Best Bid and Offer (NBBO) that would have been favorable to the pegged resting order. Even though the speed bump was doing its job (i.e., preventing an algorithm from quickly reacting to an NBBO change and “picking off” a resting order on IEX), it could not prevent an algorithm from anticipating an NBBO change. This was done by another type of clever algorithm — one based on probabilistic models that predict NBBO price changes far enough in advance to circumvent the protection of the speed bump. While the predictions of these algorithms could not be perfectly accurate, those traders could likely gain an edge by predicting a price change 1 to 2 microseconds before the actual price change occurred.

Examining the Crumbling Quote Signal and D-Peg Innovation

One of IEX’s marvelous innovations is their proprietary Crumbling Quote Signal, which aims to solve the stale quote arbitrage problem. For example, say there is a stock with a stable NBBO that is currently $10.00 by $10.02. It would be natural that any buy interest at $10.00 would be spread over several trading venues. A seller or sellers might “arrive” at those venues one by one, first exhausting the buy interest at $10.00 in one place and then the next. IEX calls this a “crumbling quote” (i.e., the price change does not occur all at once, but gradually). Algorithms that observe this behavior need not wait for the final order at $10.00 in the final venue to be exhausted. Instead, they can make a reasonable guess earlier in the process as to what the short-term outcome is likely to be and adjust for it. This may allow them to exploit resting orders probabilistically, even with the speed bump protection in place.

Put simply, the Crumbling Quote Signal detects the moments when an NBBO change is likely imminent. IEX, though, had to determine how best to deploy this tool given their goal of creating fairer markets. They determined the best way to deploy the Crumbling Quote Signal was to embed it in an order type on their exchange. They named the order type the Discretionary Peg (D-Peg).

D-Peg orders passively rest on the IEX order book while seeking to access liquidity at a more favorable price up to the midpoint between the National Best Bid and the National Best Offer, except when IEX determines that the quote is crumbling (i.e., transitioning to a less favorable price in a predictable fashion). D-Peg orders rest on the book and can exercise “discretion” to trade. When the Crumbling Quote Signal is “on,” D-Peg orders behave less aggressively, seeking to avoid bad trades at the midpoint just before the NBBO moves in their favor. When the Signal is “off,” D-Peg orders have the discretion to trade at the NBBO.

The goal of D-Peg was to prevent orders from being “picked off,” and it appears to be working. Shares trading on IEX through D-Peg orders indeed experienced fewer instances of being “picked off” versus standard order types.

The Value of Fair Markets

While the slings and arrows of equity market microstructure have implications for large financial institutions, oversight organizations and regulators, they can be challenging for individual investors to appreciate or interpret.

To this end, IEX founder Brad Katsuyama offers a reflection to help us contextualize this very complicated ecosystem.

“Markets are the natural vehicle for the exchange of products, services and even ideas. We believe that everyone should have the opportunity to compete on a level playing field with transparent rules and processes — and that’s why we’re working each day to build fairer markets.”
 

The following article is based on the forthcoming case “IEX Group Inc.,” written by Darden Professor Marc Lipson and Richard A. Mayo Center for Asset Management Director Aaron Fernstrom, available soon on Darden Business Publishing.

• 1A “pegged-to-market” order maintains a purchase or sale price relative to the National Best Bid and Offer (NBBO). The NBBO is an SEC regulation requiring brokers to execute customer trades at the best available ask price when buying securities and the best available bid price when selling securities. These orders can be passive, or considered “resting,” depending on the width of the quote. To create a pegged-to-market order, the trader enters a limit price defining the worst price they are willing to accept, as well as an offset amount that computes the active limit price.