In Harris' book "Trading and Exchanges" he says the market makers make money off of uninformed traders through the spread which compensates them, and they lose money to informed traders. However, when I look at US equities traded on NASDAQ where the bid-ask spread is typically a penny, almost all of the traders are (>90%) "informed", in that the "true price" tends to be higher a few milliseconds to seconds later after a trade has actually been printed. In other words, if the trade touches NASDAQ (the ECN specifically), the price almost always moves that way, at least in terms of the next quoted market. So this leads me to ask the following questions regarding penny-spread stocks : Does this imply that market makers cannot expect to make money from the bid-ask spread in US equities, without 1) using internalization or payment for order flow schemes, or 2) all of the market makers essentially have some sort of short-term alpha model and do not make money from the spread in any way? Even with the rebate, the rebate is only awarded in the event of being incorrect short-term, meaning there is some slightly longer-term alpha in place. I guess I'm trying to figure out in a way whether rebate traders on exchanges like NASDAQ are effectively forced to play a reversion to the mean game -- and whether the GETCO/Knight/Barclays type operations only are able to exist in US equities as market makers because they have secured gobs of uninformed flow through side agreements unavailable to Joe Public Proptrader. I'm trying to reconcile Harris' claims with my observations about NASDAQ trading and the spread in the thick and liquid stocks.
good point. I read the book awhile ago and thought about the same thing. MM have pref order execution at most exchanges. My guess - dunno for sure - is that if their bid is hit, they get order preference to either sell at the higher price or sell immediately at the same price. When the market is highly volatile, I would imagine MM don't have that great of a day even when imploying various combination of strategies to hedge positions. If there is no price volatility - same price level and spread through the day, MM will just place liquidity in the bid/offer all day long and make the spread. Its how they react to volatilty - based on their exchange agreements and their technology - allows them to stay in the game. At the end of the day, mm have to fill orders under their mm obligations. So my bet is all of this comes down to rebates, hedge executions, and what type of preferential treatment they get from the exchange in terms of order pref.
Market makers can't jump the queue on NASDAQ (as far as I know); however, your answer hints at the fact that someone trying to be a market maker on NASDAQ may only be parking orders there to get a rebate exit on the other side of their position in hopes of extracting an inefficient fill or a rebate on positions they acquired elsewhere where they DO have preferential treatment. But this would suggest that in a fragmented market with very informed traders, you can't possibly make a market solely there and expect to make money solely on spread extraction.
Good question. I think you are looking at it in a complicated way. I think he says the obvious, i.e. that MMs make money from spreads, which is true, and they lose to informed traders, which is also true. But MMs make much more from spreads then they usually lose. If they do not, then they let the market flash crash and rip off the informed traders. You get the point? What I find more interesting is that the self-proclaimed experts in this forum, who usually attack me in other threads, only get involved when the keywords "tradestation", "portfolio testing", etc. you know what I mean, come up and they otherwise hide when interesting questions like yours are asked. They probably get paid by number of hits and there is nothing to be made here.
IMHO, many MMs (those of the HFT variety certainly) are systematic traders (or perhaps "systematic scalpers"!); they follow automated mechanical strategies that tell when to place limit orders, and where to place them relative to NBBO. Their PnL derives from trade profits + rebates being greater than trade losses + fees. Often, the same MM may not be quoting both sides of the "1 cent" spread you see; it depends on what position they have, and what their target is for that position. But the net effect of all the many quotes from different players is a "1 cent" spread.
This is what I also believe; that the "true spread" in the market maker's mind is something different than what the true market is actually quoting (i.e., he quotes 3 x 5 when market is 4 x 5), and that he wants to get his fills where he thinks the true bid-ask spread should be. However, this doesn't change the fact that his entry on NASDAQ has to be adverse, what, some 95% of the time. Even if he is quoting 3 x 5 on NASDAQ, that fill on 3 is going to go against him some 95% of the time. I'm not doubting that there are some non-adverse NASDAQ fills, but there are far, far more informed traders on NASDAQ than uninformed traders. This is why I'm skeptical that most market makers have to rely on non-adverse fills from NASDAQ to make money, which is what it feels like Harris' implication is in his book. So this is why I'm looping the concept back around: 1) Where does he get his non-adverse fills from -- internalization?, or 2) do almost all MM operations absorb the toxic flow because they have a different long-range valuation? All of the big names in the equities spaces also happen to have some serious side operations with regard to uninformed flow. Also, it seems like the consolidated futures markets would be a much more "fair" playing ground with regard to uninformed/wrong flow, but I'm not qualified to make that comment.
Are you confusing "adverse fill" with "losing trade"? After all, in a sense, ANY filled limit order will probably be an "adverse fill" at some point after it's filled (i.e. if price ticks beyond the limit price). But that doesn't mean that price is certain to keep moving in that direction. MMs mainly use limit orders, which yes are often "adverse fills" in your sense. But this is not the same as "losing trade".
In terms of definitions of "adverse": If the market is 3 x 4, and I see a limit fill on 4, I consider that fill adverse if the next market is 4 x 5 without any opportunity to see an exit on 3 for a spread extraction, and really adverse if it goes to 5 with no chance to get a reversion+rebate (for US equities) on 4. Meaning that there is no opportunity to extract the spread whatsoever before some negative (with respect to the MM) activity happens, even if the trader did everything perfectly in terms of technology and algorithm. I mean I see your point in that the first fill has to be "adverse", but if you're a the top of a 100k share queue and get 1000 shares off the top, and then the price sort of lingers there, then that fill is not exactly adverse, although my distinction is arbitrary. And if the trade went from 100k down to 100 shares, I'd consider that rather adverse at least in terms of some momentum. If an MM were able to repeat the ability to get 1000 shares off of the top with minimal adversity all the time, he'd get filthy rich. So conceptually, maybe all I'm saying is "manageably adverse" vs. "umanagebly adverse" or loser. I think Harris uses a V or V0 variable to represent this and then a true V, but I don't have the book in front of me to pull out the diagram I saw.
I ran a test that suggests (1) and/or (2) is likely to be true. On a large sample of NASDAQ trades I calculated the average effective spread as 2.5 bps and the average 1-minute price impact as 3.6 bps (both relative to the pre-trade midpoint). The price impact measures the change in midpoint from just before the trade to 1-minute afterwards and measures the adverse selection. So this means on a liquidity providing round trip trade a market maker would earn 5 bps - 3.6 bps, or 1.4 bps. This is the effective spread twice minus the price impact (which is lost on the entry but not the exit). This is what a naive market maker would make by providing liquidity on both sides of every trade, before rebates. There's not much in the spread these days. Internalizing would make the adverse selection cost go down, but internalizers have to offer a competitve spread and don't get the rebate. I'm guessing (2) is more likely, some kind of alpha model that helps decide when to quote agressively and when to back off.
In this example, the actual spread will be more like 1 x 6; once the program senses someone wants to get filled say 1k at 4, the market instantly goes 4 x 6 once the initial 100 or so shares are taken, and forces the buyer to pay up for the rest. The adverse fill is only temporary.