IEX has for years campaigned against phantom quotes (bids and offers that disappear right when investors want to trade) and exchanges earning SIP data revenues and “kickbacks” given to liquidity providers for posting lit quotes at the NBBO.
Therefore, it was a little surprising to see IEX file for SEC approval of its D-Limit order type, with quote protection, given it was designed to quote fade and cause routers to miss liquidity that they thought they might get.
It was also a surprise to see the SEC approve it, especially just months after denying Cboe’s speed bump (without quote protection).
As we see today, D-Limit has already changed market quality and incentives—possibly forever—in material ways that most experts predicted.
One of these things is not like the others
For an investor looking at how much liquidity is on the NBBO nowadays, IEX has gone from 0.6% of the NBBO size to almost 6% overnight, increasing advertised liquidity.
But that doesn’t mean market quality has improved. Markets without a speed bump are not able to fade away from incoming orders, and should be close to 100% fulfillment.
Importantly, the approval of D-limit hasn’t just created phantom liquidity, it’s created a new platform of economics, previously considered bad for market quality, which exchanges need to consider.
Price sliding for market makers
The whole point of D-Limit is to allow providers to avoid adverse selection. As IEX says in a March 18, 2021 blog, their “delay on all orders… allows IEX time to (use) data from other exchanges and update our quotes to reflect those changes... before trades are executed.”
Adverse selection occurs when a trade causes quotes to tick down (on a bid, or vice versa). That leaves the liquidity provider with a fill at a worse price (bought higher) than in the “new” (lower) market prices (Chart 2).
In short, IEX’s formula uses price-setting orders from other exchanges, copying them on IEX. They also slide their market makers’ orders based on updates from price-setting markets based on their proprietary data feeds. This moves limits before orders en route to those quotes that can trade at those quoted prices (Chart 1 and Chart 2).
D-Limit adds virtual rebates
Of course, “adverse selection” causes losses for market makers. Being filled on a firm quote at the old (higher) price will cause a loss if the market maker needs to hedge (sell) at the new market (bid).
Another way to offset adverse selection is to avoid those trades entirely by price sliding just after the market has changed (but before anyone can trade on IEX). That is what IEX’s D-Limit is designed to do.
You can see how similar the economics of liquidity provision are in Chart 1. While rebates offset the adverse selection that a market maker experiences, D-Limit avoids those trades, giving the market maker a “virtual” rebate and roughly the same economics.
Chart 1: From a provider’s perspective, the economics of rebates and D-limit are similar—both reduce the net costs of adverse selection for providers
Other studies show that the market very accurately prices the total costs of liquidity provision, combining less obvious opportunity costs with explicit costs like trade fees and rebates. So we should expect the value of IEX’s virtual rebate will add to around 30mils, the same as the rebates add to fills on maker-taker exchanges. Of course, one of the complaints about maker-taker markets is the costs to takers of explicit take fees. With D-limit, the costs are less obvious: It's the costs to takers of missing trades.
How do you value a quote that you can’t trade at?
One way is to look at all the IEX quotes at the NBBO, and assume that takers paid the new spread on the quotes that slid on them. Based on that, we estimate IEX D-limit fading adds to taker costs by $57 million per year (annualizing October 2020 data).
Chart 2: From a takers perspective, take fees and D-Limit both cost takers more than the lit quote being advertised
Just like real rebates, IEX’s virtual rebates for lit quotes attracted more market makers to their NBBO. Data shows that IEX lit quotes matching the NBBO improved almost instantly, from less than 5% before D-limit to more than 50% now.
That improvement in quoting is important because it means...
IEX is all-in on data revenues
We recently discussed how the SIP pays exchanges for their data: the proportion of time and the notional size on the NBBO both count for quote revenues. The "puzzle masters” clearly knew all this when they petitioned the SEC for a protected D-Limit lit quote, as that ensured the quotes counted for SIP quoting revenues 99.9% of the time, even if they overstate true liquidity available to takers.
SIP data shows that IEX has significantly increased its data revenues from these quotes. Based on Q4 2020 SIP reports, IEX looks set to earn more than $16 million per year more in additional quote revenues, thanks to D-Limit. This far exceeds the cost of proprietary data to IEX.
Interestingly, they are only on track to grow trading revenues by $2 million per year.
Chart 3: IEX is already receiving $16 million per year in additional data revenues from D-Limit
The gap between quoting and trading revenues earned also highlights how well IEX lit quotes fade at the moment of actual trades.
Before D-limit, IEX quote revenues were 23% of their total SIP revenue. With D-limit, IEX will earn 69% of its SIP revenues from quotes, even though most of its trading is still hidden orders (Chart 4).
Other data also confirms that IEX quotes are fading before they can be traded.
Post D-Limit, IEX liquidity is still overwhelmingly hidden. In fact, at-the-touch executions have only increased by a small percentage. Increased trading revenue is likely coming from hidden orders that are being hit by takers routing to the protected D-limit NBBO. Some might call this a “bait and switch,” especially as hidden orders also earn IEX much higher trading revenues of 9mils each side.
Chart 4: Despite all the quotes and quote revenues, IEX volume is still predominantly hidden
Shakedown in D.C. is complete
For most of the experts in the industry, none of this is a surprise. But the impacts are far-reaching.
The SEC’s approval of D-Limit, with protected quote status, shows the goalposts for acceptable market quality and incentives have shifted materially. Phantom quotes are OK. Virtual rebates are accepted regardless of the cost to takers. It’s fine for some exchanges to artificially delay sending some NBBO prices to the SIP—in fact, the NBBO doesn’t even have to be available to all, all the time.
Rather than make markets more efficient, regulators have added more free riding. Sliding prices based on price discovery from other exchanges are allowed.
Now, market makers setting public prices on lit exchanges help an unrelated exchange earn millions in SIP and trading revenue.