HFT, Price Improvement, Adverse Selection: An Expensive Way to Get Tighter Spreads?
CFA Institute recently commissioned a study to examine the effects of “trade-at” rules in Canada and Australia. In the first of two blog posts, I will discuss the practices these rules aim to curb. I’ll delve more deeply into the results of the study in a follow-up post.
As the debate on high-frequency trading (HFT) continues to rage, it is important to note that it is not only the “Flash Boys” that may threaten market integrity. A recent conversation with CFA Institute Capital Markets Policy Committee member Dennis Dick reminded me that there are more mundane market trends that may be undesirable for market integrity.
Some of you may be aware of the concept of “adverse selection,” but it is perhaps useful to provide an example of what this actually looks like in practice. In today’s market, the majority of “uninformed order flow” is internalized by firms that purchase retail order flow from brokers. By uninformed order flow we mean orders that are not seeking to correctly predict price movements in the next minute, hour, or even day (e.g., fundamental traders). Internalization is allowed under current market regulations as long as the price matches or beats the displayed National Best Bid and Offer (NBBO). Even in cases in which the price improvement is nonzero, it is usually only nominal but allows internalizers to “lean” on the quote.
What does this mean? Consider, for example, a stock trading at $20.00-20.01 at the NBBO. Internalizers fill all incoming retail marketable buy orders at $20.0099 (in front of the displayed offer) and all retail marketable sell orders at $20.0001 (in front of the displayed bid) — a nominal price improvement of $0.0001 in each case. They earn the spread minus the nominal price improvement. While this is ostensibly good for investors who receive a price improvement on their orders, it may be a case of penny wise, pound foolish.
And here’s why. For a retail order that gains $0.0001 in price improvement for 99 trades of 100 shares (with a 1-cent spread), that’s a 99-cent total gain from price improvement. But if the 100th trade is a limit order for the same 100 shares that does not get executed because internalizers are stepping in front of the trade, you would lose $1 because of the 1-cent spread on 100 shares you would have to pay in order to trade — wiping out the price improvement from the previous 99 trades! This example shows that tight spreads may not necessarily be economically significant.
Now, consider market makers who post visible limit orders. Their orders do not typically interact with uninformed retail order flow — this is always being internalized. Where does that leave them? They are trading with relatively more informed order flow and therefore are more likely to be on the wrong side of a trade — or increased adverse selection. But what does this actually look like?
This phenomenon can be observed in the consolidated tape data because trades will execute off-exchange when the quotes are stable and only interact with lit orders when the quotes change or “roll.” What’s happening here is that internalizers are causing trades to occur inside the quote by offering price improvement (as described above). When they see order imbalances that predict the quote is about to change or “roll,” they neutralize their net position by hitting the lit quote.
For example, if they have a net long position just before a quote rolls down, they will seek to sell at the bid knowing that they can imminently re-purchase at the lower future ask price. This manifests itself in the consolidated tape as a series of off-exchange trades executed at the NBBO, followed by an on-exchange trade just as the quote rolls. In this way, internalizers and HFTs are liquidity providers when quotes are stable, but liquidity takers when quotes roll — the posted quote is merely acting as insurance for this strategy.
This is a strikingly clean example of adverse selection in action — internalizers are mopping up uninformed retail order flow and only interacting with lit orders when the lit orders are on the wrong side of the trade. No wonder market makers are concerned about the toxicity of order flow in the market! For other market participants this is worrying because the incentive to post lit orders is compromised. This is the reason for the rise of maker-taker markets in which participants are paid a rebate to post limit orders and are charged a fee to access limit orders.
In summary, regulations like Regulation National Market System (or Reg NMS) and the Markets in Financial Instruments Directive (MiFID), along with payment for retail order flow, allow some trading firms to internalize uninformed order flow for no, or only nominal, price improvement. Meanwhile, algorithmic and high-frequency trading technology allows these same firms to only interact with lit markets when they are on the right side of a trade, and avoid being hit on the wrong side of a trade as well. The traditional human market makers, as a result, operate in lit markets that are full of toxic, informed order flow from the above-described firms, so their incentives to post limit orders are severely compromised — they cannot change their quotes fast enough to avoid falling on the wrong side of a trade by so-called algo traders — and many have become liquidity takers themselves.
To entice this limit-order liquidity back onto lit exchanges, some regulators have introduced “trade-at” rules, which are meant to limit the ability of firms to cheaply internalize uninformed order flow. In the next blog post I will examine the results of a CFA Institute-sponsored study that assesses whether these attempts have been successful.
If you liked this post, consider subscribing to Market Integrity Insights.
Photo credit: iStockphoto.com/halbergman
wholesalers cannot step in front of trades if they give sub-penny price improvement and have a resting limit order. NMS and Manning rules require a minimum of a penny of price improvement to prevent any resting limits from being executed. Also, factor in the exchange taker and access fees into any studies. In essence, a individual paying the offer for a stock on an exchange will also incur an additional $.003 per share. If they send orders to exchanges, they are also showing the market their “hand”. This may not matter for small retail investors but larger institutions will face HFT and ‘order prediction arbitrage’ which will cost them a ton of money. These instituions include mutual funds which are the favorite investment vehicle for most of the ‘uninformed investors’ you reference in the article.
Thanks for your comment.
I believe (and please correct me if I am wrong) that we are talking about slightly different things.
Rule 5320 (which supersedes the Manning rule) states (as you say):
“The proposed rule change establishes the minimum amount of price improvement necessary for a member organization to execute an order on a proprietary basis when holding an unexecuted limit order in that same security without being required to execute the held limit order.”
I understand this to mean that if a broker has a customer limit order to sell at 20, the broker cannot then go out and buy from someone else at 20 – they have to execute their customer’s limit order before trading for themselves. Unless they can get price improvement and buy from somewhere else at, say, 19.95.
However, what I am describing in this blog is a broker that is executing all incoming retail limit orders at 20 and thereby stepping in front of limit orders on lit exchanges (rather than stepping in front of its own customers). Market participants advise me that for internalization price improvement is not necessary as long as the price matches the NBBO.
Regarding your point about maker-taker, this is true, but the rebates are a band-aid solution to attract limit order liquidity that would otherwise not be there in the face of adverse selection caused by the dark trading. To get rid of maker-taker you’d have to put all the uninformed orders back on the lit exchange so that market makers would continue posting liquidity.