Here is a simple model for how a market maker trades stocks on IEX, or the New York Stock Exchange, or wherever:
- The market maker bids $40.00 to buy 100 shares, and offers to sell 100 shares for $40.02.
- A customer comes along and sells 100 shares for $40.00. Then another comes along and buys 100 shares at $40.02. Then another comes along and sells 100 shares for $40.00. Etc.
- The orders balance out, and the market maker earns $0.02 a share.14
- Except that sometimes the orders don't balance out: Sometimes someone wants to sell 1,000 shares. So the customer sells 100 shares at $40.00, then another 100 shares at $40.00, then the market maker gets nervous and changes his price to $39.99, and the customer sells 100 more shares at $39.99, and now the market maker is really nervous, and the customer sells 100 more shares at $39.98, etc., until the customer has sold 1,000 shares and the new price is, say, $39.89 bid / $39.91 offered.
- The market maker now owns 1,000 shares of stock for which he paid an average of, say, $39.95 per share, but which are worth only $39.90 per share — a loss of 5 cents a share. He bought a lot of stock, and the market moved against him.
That is bad! It is called "adverse selection." One way to interpret it is to imagine that the market maker sometimes trades with informed customers: When Bill Ackman buys stock, he has reason to think the stock will go up, and when it does go up he is happy and the market maker who sold him the stock is sad. But you don't need that interpretation: When a big index fund buys stock, it's not doing that because it "knows" the stock will go up. But when a big index fund buys stock, it is buying a lot of that stock, and its buying alone is likely to push the stock up. Supply and demand is enough to explain adverse selection.
So if you are a market maker, sometimes you make a little money on the spread (buying at the bid and selling at the offer), and sometimes you get run over by a big trade and lose money. This is pretty predictable, and you try to make sure that those two things balance out: You want to get paid more trading on the spread than you lose to adverse selection. The more big institutional investors you trade with, the worse the adverse selection will be. On a market like IEX, which is aimed at and supported by big institutional investors, you might be particularly worried about adverse selection.
Here is a simple model for how a market maker (say, Citadel) trades stocks when a retail brokerage sends it orders:
- It's the same, except
- There is no adverse selection!
If you are Citadel, and you know with 100 percent certainty that the customer on the other side of you comes from E*Trade, then you know with 100 percent certainty that that customer is not Bill Ackman. (I mean, not Pershing Square, anyway.) It might be Martin Shkreli, or Sarah Meyohas – some retail investors can move stocks, so there is some risk of adverse selection — but the risk is considerably reduced. Retail investors, as a class, have much less money and much less information than institutional investors. Their trades are much less likely to be predictive. What that means is that, if you charged a 2-cent spread in the public markets — bidding $40.00 to buy, offering $40.02 to sell — then when you remove the risk of adverse selection you can charge a smaller spread, say 1 cent. So you bid $40.005 to buy, and offer $40.015 to sell15 – pretty close to the number Citadel gets. (This is called "price improvement.") You charged the 2 cents to compensate you for adverse selection risk; once that risk is gone, you can charge less.
Actually you can do even better than that: Not only can you tighten your spread from $40.00 / $40.02 to $40.005 / $40.015, saving retail investors a half-cent a share each time they trade, but you can also pay their brokers something for the privilege. This is called "payment for order flow," and it makes people really mad, but it also helps those retail brokers to offer their services so cheaply. Trading stocks is basically free! It costs under 10 bucks a trade at the big discount brokers, and is literally free at Robinhood. Brokerage offices and research and computer systems and apps don't pay for themselves; payment for order flow pays for some of them.16
These basic facts of market structure — that trading with retail-only flow reduces adverse selection risk, and that market makers are willing to pay for it — suggests that Citadel is probably right that it offers better execution to retail investors, at least on small orders. (On larger orders, where "front-running" is at least a theoretical possibility, there is a bit more of a debate, though not much evidence that Citadel is wrong. 17) Because retail investors tend to be, for Citadel, considerably more harmless than IEX's institutional customers, it can offer them a better deal.
IEX should be approved as an exchange. As an individual investor I am at a disadvantage when my orders are front-run by Hft shops out to add cost to me. IEX does a service to the small investors, the ones who actually add liquidity to the market.
Citadel's letter is a non sequitur, but this is nonsense. This hasn't even the slightest connection to reality. No one is front-running this guy's orders, no one is trying to add costs to him; IEX's service is for big investors, not small ones, and the small ones don't add liquidity anyway. This guy knows nothing, except that he doesn't like "HFT shops" like Citadel, even if he can buy stocks cheaper from those shops than he could from an exchange like IEX.
And he isn't alone. Fights over IEX and high-frequency trading and exchange structure are mostly about institutional investors. Those fights matter, because institutional investors matter; if IEX makes trading more (or less) efficient for those investors, then that will be good (or bad) for millions of people whose retirement savings are managed by those investors. But these fights have nothing to do with most individual investors. If you are buying 100 shares of Apple from time to time, the evil high-frequency traders, with their colocation and their latency arbitrage and their front-running, are not your enemy, and IEX's application to be an exchange won't affect you one way or the other. That so many retail investors are so upset about high-frequency trading is just weird, a symptom of an American public that is unshakably convinced that the system — any system — is always rigged. "WE ARE TRYING TO HELP YOU," Citadel shouts, but no one is listening. No one believes them. Of course they are unhappy!
- Citadel itself, for instance, gets about 30 percent of Fidelity's order flow, 18 percent of E*Trade's, 15 percent of Charles Schwab's, etc.
- That's my Bloomberg View colleague Michael Lewis, of course, whose "Flash Boys" is about IEX.
- I would also read a popular book by a celebrity writer lionizing Citadel, though I am not holding my breath.
- E.g.:
IEX's adverse impact on retail investors does not stop with these statistics. The profoundly negative features that IEX trumpets as beneficial would actually have a uniquely bad and unprecedented impact on retail investors. IEX is designed to post prices that are stale as a result of IEX's intentional time-delay. These stale prices would harm retail investors because the wholesale market makers that execute retail orders would be compelled to pursue those stale prices. Ironically, these stale prices would have little impact on the trading of the "high speed" proprietary trading firms that IEX so aggressively smears because, as explained below, proprietary trading firms would be permitted to largely ignore stale IEX market prices. Broker-dealers executing or routing customer orders, on the other hand, cannot simply ignore IEX's prices if the displayed (and likely stale) price on IEX appears to be the best price. We've talked more about the speed bump here and here.
After all, Citadel pays some retail brokerages to execute their retail orders for them! It must think it has some advantages. As Liquidnet put it in its comment letter Tuesday:We encourage the SEC to look at the vast variety of comment letters received on this issue in aggregate. A peripheral analysis of the opinions that have been submitted on this application is in itself revealing. Let's look at the pattern of the type of organizations that are generally in support of the application and those that are against. In the "for" column: individual investors, asset managers, and now, with this submission, an independent agency-only dark pool. Opposed: competing exchanges and high-frequency trading firms. That alone presents the clearest case on where an IEX Exchange would fit into the current market structure ecosystem. Disclosure: Most of my money is in mutual funds, much of it at Vanguard and Fidelity (though not, as far as I know, Calpers). No one agrees with me on this, least of all the SEC, but I am right. (As a Vanguard/Fidelity/etc. customer, I am also biased!) Here is a graphical representation of that from Eric Scott Hunsader of Nanex LLC. It includes a random dig at me, because Hunsader likes including random digs at me, and because I tweeted about Citadel's letter with the comment "ooh." I suppose the "ooh" was ambiguous. They might become a bigger business when IEX becomes a public exchange? I don't know. Less than 1 percent of the New York Stock Exchange's volume comes from things called "market orders," though some amount of NYSE's immediate-or-cancel orders probably represent, in some sense, "market orders." The idea of the market order is somewhat elusive. It's 48 percent and 45 percent for NYSE and Nasdaq-listed stocks, respectively, atFidelity; 51 percent and 43 percent at Schwab; 52 percent for both NYSE and Nasdaq stocks at E*Trade. Others are lower: 23 percent and 22 percent forScottrade, 6 percent or 7 percent at Interactive Brokers, and just 5 percent at TD Ameritrade. Broader statistics are weirdly elusive. Markit has found that "retail executed 56% of their liquidity seeking orders via market orders," that is, as opposed to marketable limit orders, but that doesn't count non-marketable limit orders. Still "about half" seems roughly plausible. This is descriptive, not prescriptive. As a prescriptive matter … I try not to give investing advice but I will say: If you are a retail day-trader, consider using limit orders! You can use marketable limit orders — like, bid up to $41 for that $40 stock if you want — to make sure that you'll get an execution, but it's a scary world out there, there are a lot of flash crashes, and you don't want to put in a market order at the exact second that the stock spikes to $999.99. There is even evidence the other way. For instance, in a comment letter generally critical of IEX's execution quality, Markit found that IEX's execution quality for marketable limit orders (buy orders with a price limit above the current market offer, or sell orders with a limit below the current market bid) was better than that of wholesalers like Citadel. There is a broader debate about this. Here is a Barron's study finding that wholesalers are generally good for retail investors, because they offer price improvement on market orders. Nanex LLC points out that the price improvement is much less impressive on marketable limit orders — though still positive. 14. I think most people would call this $0.01 per share (that is, half the spread per trade): Buy 100 at $40, sell 100 at $40.02, you've traded a total of 200 shares and made a total of $2.00. But this is quibbling. Also in a realistic model the market maker might update his prices to reflect executions, which I have ignored for simplicity. Yes I know you are not allowed to quote sub-penny prices on public exchanges, but you are allowed to execute at sub-penny prices as a wholesaler. The quote in the text is not a literal lit quote, it is an illustration of how the market maker interacts with retail orders. Though actually not at Robinhood? You can think of PFOF as analogous to Amazon Affiliate links, where people can post links to Amazon products on their websites/blogs and Amazon gives a small rebate if a purchase is made through those links. The ultimate purchase price is the same regardless of whether the user got to the product through Amazon directly or through affiliate links, but Amazon splits the profits for products bought through affiliate links to incentivize more websites/blogs to link to Amazon products and broaden their overall reach. Prior to launching the product, we had some language saying that we intended to generate revenue through PFOF. The reason we removed this was because currently we execute all of our orders through our clearing partner APEX Clearing; as a result, Robinhood does not directly receive PFOF and any revenue we indirectly generate from it is negligible. Here is a paper from Themis Trading called "What Every Retail Investor Needs to Know" about how, in Themis's view, wholesalers front-run retail orders. The first part (labeled " Market Orders") isn't especially compelling; it involves the wholesaler selling to the retail investor at just inside the offer because the wholesaler "knows, according to his own faster version of market data, that the market for the stock is actually, or will soon be," lower. Sure, fine, but how often does a retail order come in at the exact millisecond that the wholesaler has updated his price but the market data feed has not? The second part ("Marketable Limit Orders") is more interesting, and more complicated. Basically Themis argues that on a 2,000-share retail order, the wholesaler can route a portion of the order inefficiently to public markets, driving up the public-market price, and then fill the remainder of the order internally at an "improved" price versus the now artificially high public price.
How real or prevalent is this? Apparently not very, at least for market orders of under say 10,000 shares. Consider Citadel's disclosure that, when it gets market orders of 2,000 to 4,999 shares, almost 94 percent are executed at or better than the national best bid or offer at the time of receipt (and almost 77 percent are executed inside the NBBO). That's pretty good! Even for orders of up to 9,999 shares, almost 90 percent are executed at the initial NBBO or better. On the other hand, orders of more than 10,000 shares (pretty big for a retail order!) are not tracked in this report, and neither are limit orders. (Thus Themis's section head, "Marketable Limit Orders.") Perhaps those are executed less efficiently and in a way that could be characterized as "front-running."
