The Emperor’s New Clothes

This is Part One of Two:

The one year anniversary of the approval of the ironically named “Investors Exchange” has triggered more media coverage and I feel like the little boy the classic Hans Christian Andersen story: The Emperors New Clothes. Sadly, like the “clothes” worn by the emperor, the marketing claims made by IEX are similarly transparent. Equally sadly, others see it, but like the ministers and townspeople in the story, wont say so for fear of provoking the ire of the emperor and their supporters.

The key is to evaluate IEX as an exchange, however, I must clarify what an exchange is first, so consider the following definition from Investopedia:

An exchange is a marketplace in which securities, commodities, derivatives and other financial instruments are traded. The core function of an exchange is to ensure fair and orderly trading, as well as efficient dissemination of price information for any securities trading on that exchange.

This emphasized the primacy of displayed prices for exchanges as did the SEC, when they adopted Regulation NMS:

limit orders are the building blocks of public price discovery and efficient markets. They (commenters) stated that a uniform rule for all NMS stocks, by enhancing protection of displayed prices, would encourage greater use of limit orders and contribute to increased market liquidity and depth. The Commission agrees that strengthened protection of displayed limit orders would help reward market participants for displaying their trading interest and thereby promote fairer and more vigorous competition among orders seeking to supply liquidity.[1]

Therefore, for this commentary, we will examine IEXs marketing claims, based on how their claims apply to displayed limit orders and trades which interact with such orders. On their website, IEX claims to be a Fair, Simple, Transparent Exchange. We will start with Fair

“Fair” – The most appropriate definition of fair is free from bias[2], meaning that IEX is asserting that at their exchange, no class of investor has an advantage over other classes due to their architecture. They cite two features to back up this assertion: First, they do not offer co-location services, which they claim provides advantages to HF firms, and second their speedbump slows down the market which helps their clients vis a vis HF traders. Lets look at each in turn:

Co-Location – IEX is wrong; Their lack of co-location makes IEX less fair. While IEXs connection from their Point Of Presence (POP) in the Equinix datacenter to their matching engine is the same for all clients, HF firms are free to find the fastest connectivity to the POP itself and there is no one to regulate that pursuit. In the era before exchanges standardized connectivity in their co-location facilities, there used to be large variations in speed from particular ports and network locations to the exchange matching engine. The result was that HF firms, with the engineering expertise and economies of scale to do so, continually monitored their connections and optimized their network topology to gain an advantage. In reaction to this, Nasdaq, Bats, and ICE standardized the connectivity within each tier of service within the datacenter. Since IEX has no control over the datacenter environment of its members due to their no co-location policy, it is extremely likely that HF firms will act to optimize their network topology to gain advantage. That means that IEX is the ONLY exchange where HF firms CAN gain unregulated speed advantages. I want to clarify one point about co-location: I will not defend the exchanges practice of having multiple levels of service at increasing costs within their co-location facilities, as that is driving up costs for all trading firms. Thus, I agree with IEXs position on having only one level of service. I will, however, defend the fact that standardization within the datacenter, which colocation enables, is a much fairer set up, making IEX the exchange most vulnerable to HF firms.

SpeedBump – IEX is wrong; As I have previously written, the speedbump does absolutely nothing to protect displayed orders since it is applied between the POP and the IEX matching engine, and IEX is a price/time priority exchange. As a result, it impacts all orders/cancel messages the same, and, therefore, the faster firm still gets priority. Simply put, in all price/time priority exchanges, the only thing that matters is relative speed and IEX does nothing to change that. If that were the only issue with the speedbump, it would render their claim of fair misleading, but there is a larger issue. Due to the recent improvements of the SIP, IEX is now the exchange that provides the largest speed advantage to HF firms. Here is an example comparing IEX and Bats, which both have their connectivity for clients in the Equinix datacenter in Secaucus, to illustrates the issue:



Client posts a displayed order to sell at 10.50 in a Nasdaq listed stock.

(The SIP for Nasdaq stocks runs in Carteret)

After order goes thru speed bump, IEX posts displayed order to sell at 10.50 and immediately reports to the SIP. IEX clients that connect to their POP to see the quote, wait for the speed bump to see it, but firms co-located in Carteret with microwave links back to Secaucus could see the new quote faster despite the added distance.

Bats posts displayed order to sell at 10.50 and immediately disseminates quotes to own clients and SIP simultaneously. Bats clients that connect directly see the update much faster since they do not have to wait for transmission times from Secaucus to Carteret and back.

(later) a different client sends a buy order at 10.50

After order goes thru speed bump, IEX will match the buy and sell order & then report the trade & quote change to the SIP. The trade report to the client waits for the speedbump (meaning if that order was part of a routing strategy, that may need to go to other exchanges, it is frozen waiting on that report). Meanwhile, firms co-located in Carteret will see the execution faster there and may see it faster in Secaucus as well. This frees the HF firm to act on the information before IEXs client receives the execution report.

Bats will match the buy and sell order & report the trade and changed quote to the SIP at the same time they report the trade and quote updates to its own clients. Bats clients, therefore, receive the report at the same time or before other firms.

For more details on this critical flaw, along with nice charts, see the Meanderful blog[3], which brought this to my attention. For readers who want to know why the author of that blog and I both feel like the child in the title story, consider that this factual depiction completely reverses the Flash Boys narrative with respect to displayed orderson the Investors Exchange and there has been no press coverage of this fact.

“Simple” – To be fair, IEX, handling of displayed orders, is simple. It is only their dark order functionality that violates this precept. That is because the new IEX crumbling quote algorithm is arguably the most complex order type ever deployed by an exchange. That bears discussion, but is a topic for a different commentary.

“Transparent” – The most appropriate definition of transparent is easy to perceive or detect[4], meaning, in the context of stock exchanges, the display of orders or the ability to execute trades against displayed orders. By that definition, IEX fails to be transparent, as their own data shows, since more than 75% of the trades that take place on their exchange are in the dark. Most trading on IEX is, therefore, completely opaque and occurs when neither counterparty to the trade reacts to displayed liquidity or is willing to display their order. This is an extremely important point, because it nullifies their recent claim[5] that their low effective spreads prove their value, as an exchange, to long term investors. My logic is that liquidity seeking investors have two modalities:

  1. Smart Routing vs Displayed Liquidity – This is when SORs survey the displayed quotes from all exchanges and route orders to try and interact with that liquidity. This is the most typical form of smart routing, where SORs use either costs or execution certainty to route based on the displayed quotes at exchanges.
  1. Dark Seeking – This is usually accomplished by either probing at the midpoint in the dark or by sending Immediate or Cancel (IOC) orders at the NBBO sequentially to markets that are not displaying quotes. IEXs effective spread numbers are dominated by this type of activity, which includes a lot of midpoint trades.

I believe that when evaluating exchanges, it is appropriate to consider only #1. IEX, however, uses data that includes dark seeking activity, but fails to include dark pools in their comparison since those pools are not identified in the market data stream. Thus, I propose to use 605 data for all marketable orders, as a fairer analysis as it excludes midpoint-only, liquidity seeking orders, which, while valuable, can be measured separately as part of a Dark Only study.

On that basis, using April data from BestXStats, a provider of Rule 605 and other regulatory reporting, IEX underperforms all three Nasdaq exchanges and both Bats exchanges, with an effective spread divided by quoted spread (EQ) of 97.4 for all stocks compared to an EQ range from 91.3 to 95.3 for the other exchanges. More important, however, is the comparison of IEXs EQ of 97.4 to UBSs market leading dark pool, which posted an EQ of 71.6. Even if one assumed that 25% of the marketable orders in IEXs 605 report were executed against displayed quotes at the NBBO on IEX, that would mean that the fills at IEX against dark orders was at an EQ of 96.5, which is almost 35% inferior to the statistics UBS posted. Meanwhile, if one looks only at market orders, which is the most used order type by retail investors (which IEX has directly solicited as I have previously written), IEXs EQ in April was 142.4 which compares to a share weighted industry average (composed of market makers) of just under 50. While this is on a small % of IEXs flow, it is consistent with past months and represents the purest use of 605 data.

In conclusion, from the perspective of displayed orders, the analogy between my criticism of IEXs marketing and the little boy from the Emperors New Clothes seems apt. In the story, the people and those in power saw that the emperor was wearing nothing, but kept that to themselves. In the case of IEX, reporters and commenters have failed to challenge their assertions of fairness and transparency, despite the obvious incorrectness. As I have shown, both reasons that they provide to prove that they are fairer than other exchanges are not only wrong, but make them less fair. Abdicating responsibility for overseeing colocation fairness in the Equinix datacenter, where their POP resides, creates the opportunity for HF firms to gain advantage. Worse, the IEX speedbump, along with speed improvements to the SIP, creates information advantages for HF firms who are also co-located in Carteret (and perhaps Mahwah). IEX is certainly not transparent, with most of their trading occurring in the dark. While their dark execution quality is good, it is not better than other dark pools, and more relevant statistics show that they are not better than other exchanges. Despite all of this, I still wonder if people, including the press, will continue to accept all IEXs claims at face value. If so, then I will continue to be one of the only ones (publicly, at least) to see the emperors clothes for what they truly are…

Stay Tuned for Part Two Next Week