The standard illustration of the efficient markets hypothesis is the thing about the economists and the $20 bill on the ground, but it is so old and stale at this point that Matt Yglesias had to invent a new version this week, and it's something like "if you find a penis-enlarging injection on the ground, don't pick it up, because if a penis-enlarging injection actually existed then Pfizer would already have picked it up, and so this one will kill you of exploding penis, QED." You could take this advice overly literally as an argument against all human effort, and perhaps you should, but in fact someone didn't take it literally enough, or at all, and so died of exploding penis.
"If it works someone's getting paid for it" of course doesn't imply the converse "if someone's getting paid for it it works" - particularly not in the penile-enlargement field - and I suppose neither does EMH; if anything it just implies "nothing works and nobody gets paid." Still, there is at least some weak intuitive support for the belief that if lots of sophisticated financial market participants pay for something, they're getting some value back in return.
"Improper early access to market data, even measured in milliseconds, can in today's markets be a real and substantial advantage that disproportionately disadvantages retail and long-term investors," said Robert Khuzami, Director of the SEC's Division of Enforcement. "That is why SEC rules mandate that exchanges give the public fair access to basic market data. Compliance with these rules is especially important given exchanges' for-profit business interests"
And here is Sanjay Wadhwa to Bloomberg:
“The NYSE chose ground-shipping for sending market data to the consolidated feed but used next-day air for its paying customers,” Sanjay Wadhwa, deputy chief of the SEC’s market abuse unit, said in an interview.
Stripping out the rhetoric, the SEC is mad because:
- The NYSE has an obligation to furnish its best-bid-and-offer and trade data to the consolidated tape - which ultimately makes its way to, like, the Google Finance feed you use to day-trade.
- It is also allowed to sell that data separately to high-frequency-trading firms who want to bypass the delay of the consolidated tape, but under Reg NMS Rule 603(a) it can only do so "on terms that are fair and reasonable" and "not unreasonably discriminatory."
- The SEC reads that to essentially mean "you can't give your data to HFT firms before you give it to the consolidated tape."
- NYSE has several programs to sell the data to HFT/algo firms: NYSE BBO,1 which basically gives HFT firms the same data that is sent to the consolidated tape, and Open Book Ultra, which gives HFT firms the entire NYSE order book rather than just the top of the book.
- But NYSE's compliance people weren't involved in the tech design of PDP and Open Book Ultra, and so the computers that sent data to HFT firms sent it a few milliseconds before the computers that sent it to the consolidated tape - with that "few milliseconds" rising to "several seconds" at various crunch times, including in particular the May 2012 flash crash.
So why is that bad? Well, arguably it is bad because HFT firms shouldn't get an unfair advantage, like Khuzami and Wadhwa say. Here is another reason that it's bad - from the SEC's order, not the press commentary:
The disparities in data transmissions that Rule 603(a) prohibits can have important consequences that risk undermining investor confidence and interfering with the efficiency of the markets. For example, a delay in the release of data to the consolidated feeds in contrast to the proprietary feeds can cause an investor relying on the consolidated feeds to make a trading decision based on a potentially stale picture of current market conditions. An exchange’s delay in sending its quotes to the consolidated feeds also can cause inefficient execution decisions at other market centers and, under some circumstances, create the appearance of a “crossed” national best bid and offer (“NBBO”), which occurs when the best bid exceeds the best offer. The appearance of a crossed NBBO can cause both uncertainty and the risk of a trade being executed at worse than the best available price.
That is less sexy and retail-investor-protection-y, but probably more important: if some market participants are trading on data that is several seconds older than others, then you get "inefficient execution decisions" and perhaps more concretely "people thinking that their trades should be unwound because they were priced in error." Witness the flash crash, but also the Facebook IPO. That destabilizes markets and is bad; arguably it's worse for algorithmic traders than it is for retail traders since no retail investor is dumb enough to, for instance, submit the same order dozens of times just because he hasn't gotten a confirm yet.
I suspect that this is the real reason the SEC is mad at the NYSE, and that Khuzami's "disproportionately disadvantages retail and long-term investors" is magic feel-good hand-waving, and you can tell from one really really simple fact, which is that NYSE is still selling both Open Book Ultra and NYSE BBO separately from the consolidated feed. NYSE BBO is $1,500 a month; the more detailed Open Book Ultra is a price-on-request deal.
If they're selling those services, and people (robots) are buying, then those robots are getting something out of them. One thing that you become aware of - or so I hear - if you measure time in nanoseconds is that data takes time to travel over wires. If you wait for the data to travel over wires from NYSE to the consolidated tape, and then from the tape to you, that takes more time than travelling directly from NYSE to you, even if the data is sent to you and the tape at the same time.2 So you pay your $1,500 a month to get it directly, and your unspecified extra amount to get the full-book rather than just the top-of-book data. And since data takes more time to travel over longer wires, you can pay the NYSE to rent a room next door to the data, so your wires can be short and your data fresh.
Google Finance is free and I'll go ahead and guess E*Trade runs considerably less than $1,500 a month. I'll also guess that the high-frequency trading firms that colocate their computers next door to NYSE and pay for the real-time complete order book can make decisions more quickly and based on more complete information than the average retail investor on E*Trade.
And none of that was banned or changed or discouraged or looked askance at by today's order! "The NYSE chose ground-shipping for sending market data to the consolidated feed but used next-day air for its paying customers" is a nice line but also purely silly; the paying customers have their own planes, and the NYSE built its distribution center at their airports.
Which is not to say the SEC did anything wrong by yelling at NYSE for building computers that broke easily. It's really important that NYSE's computers not break easily, and NYSE should be shamed and fined when they break and cause market havoc. But doing sensible market-stability or fraud-prevention things doesn't justify talking about them in terms of "we have shut down unfair advantages in the financial markets." If you're a retail investor, and you think that the SEC has shut down discrimination in data providing so that now you're getting your data at the same time as, say, UBS's algo trading operation, ask yourself: why is UBS paying so much more for that data than I am? Of course they're getting it faster. Of course they have an advantage - and, fair or unfair, no one's going to do anything about it. You'll just have to find other ways to outwit them.
NYSE Penalized by SEC for Giving Head Start on Trading Data [Bloomberg]
SEC Charges New York Stock Exchange for Improper Distribution of Market Data [SEC]
Efficient Markets Theory Can Save You From a Grizzly Death by Botched Penile Enlargement
1. The SEC calls this "PDP," which stands for "proprietary data product," but NYSE now uses that acronym for a bunch of proprietary data products; NYSE BBO (standing for "best bid and offer") is the one that the SEC is referring to.
2. Right? I am out of my depth here, and as far as I can tell the consolidated feed is also quite pricey, so if you are an HFT expert and think that today's order actually clamps down on people getting ahead by buying NYSE PDP data, let me know. But obviously it doesn't clamp down on people getting ahead temporally by colocating with NYSE servers, or informationally by buying depth-of-book Open Book Ultra data, etc.