Dollars and Jens
Tuesday, May 26, 2015
the modern stock market
There's a paper on SSRN about the modern stock market with its high-frequency traders and dark pools and so on, and I recommend it. I'd like to offer a couple remarks on some small bits of it:
- They give an example of a high-frequency trader placing orders to sell a stock on multiple exchanges at the same time; when one of the orders executes, it cancels the other. A large trader, with less speed, attempts to lift both offers, and finds that it can't. I have trouble intuiting that I should be sympathetic to the poor buyer here; it seems to me that one very appropriate response to market fragmentation is to get very fast access, post the same order at multiple venues, and sell wherever it sells first. Certainly the HFT is running the risk that the offers will be executed at both venues before even it can respond, and I wouldn't be especially broken-hearted for it if that happened, but it seems perfectly reasonable for it to do this when it can. It's worth noting that the authors' proposed regulatory change to address controversial behavior by HFTs — not to let agents have order-book information until the official NBBO has been updated and disseminated — would not impair this behavior unless they are also delaying execution transmission (or unless that takes longer anyway to go out than inside quote information does); you couldn't cancel your order in response to somebody else's execution on the other exchange, but you could still manage your own attempts to sell a quantity wherever you can without undue risk of executing more than you intended.
- The "maker-taker" fees seem a bit fraudulent in conjunction with the NMS rules and fiduciary duty rules for brokers. If a bid sitting at a venue at 45.00 will actually result in the buyer's receiving $45.002 per share executed, and the seller's paying $45.0025, it feels wrong to me for the tape to reflect an execution at $45.00; it doesn't feel quite as fraudulent to me for the "best bid" to be disseminated as $45.00 (I'm used to the idea that commissions are in addition to the quoted price), though possibly it should. If another venue is "paying for order flow" — I'm not talking here about the Citadels "internalizing" executable orders, but the practice that elsewhere seems to be called "customer priority" — and a bid resting at $45.00 means the seller pays a net of $44.999, calling that $45.00 again feels more fraudulent, and an agent required to route an order to the "national best bid" shouldn't seem to be able to call that a tie. This little irregularity also makes me wonder whether any high-frequency shops try to make much of their net revenue from payments for liquidity provision by e.g. having an order to buy at $45 on the platform where that would net $45.002 and responding to an execution by trying immediately to hit a bid at $45 on the platform that pays for marketable orders. It might be that the fraction of a tick per share would be insufficient to compensate them for the risks, especially in light of other opportunities for profit.
- There's some level on which I regard a lot of the controversial actions of the high-frequency traders as tying the different exchanges together into more of a true national market, and I view the objections to them as a mixture of denial and regret that the pretense of a single market is being blown.
- More broadly, I'm not sure the model of "informed" vs. "uninformed" traders is never misleading. I don't know whether there's work on what the environment of the no-trade theorems looks like when agents have different stochastic discount factors, but I'd like a little bit of a deeper understanding of ex-ante gains from trade in which it is common knowledge that some agents have different information from others. The general dynamic that traders motivated by needs for liquidity and traders motivated by superior information will both want to look like the former is presumably true, but where efficiency is being evaluated in terms of providing returns to cultivating information it seems like nuance here might be important — that in an important sense the liquidity traders may not lose as much from trading with informed counterparties as is being supposed in their models, and that conversely the informed buyers will ultimately be interested in having a liquid market into which to sell as well.