You have been granted access to this page through First Click Free. Subsequent use of TabbFORUM will require logging in. If you don't have an account, registration is free.

Videos

  • Rail_thumb_screen_shot_2014-07-28_at_3

    Euronext Carves its own Path

    Euronext is independent again. In June, the move was confirmed when it listed on the main French, Dutch and Belgium markets. FT’s Peggy Hollinger asks chief executive Dominique Cerutti about his ambitious ...
     
  • Rail_thumb_ace

    T. Boone Pickens: Ace Greenberg Was a Great Man

    BP Capital Chairman and CEO T. Boone Pickens reflects on Alan C. “Ace” Greenberg's legacy. He speaks with Trish Regan on Bloomberg's "Street Smart." ...
     
  • Rail_thumb_theo

    Accelerating Intelligence

    There is a tradeoff between latency and the intelligence you can build into algorithms, acknowledges Software AG’s Theo Hildyard. But streaming analytics can enable firms to incorporate intelligence ...
     
 

More Video | Podcasts

Advertisement
Author_tdxme
Cristian Zarcu

TradeDynamiX

More From
Cristian Zarcu

07 March 2014

Inside a Half Second of Market Inefficiency

For 500 milliseconds, the spread in the SPY ETF became inverted, raising serious questions about the efficiency of our markets and the effectiveness of Reg NMS.

We all assume our markets to be efficient, but just how efficient are they really? On March 3 at 12:13:36, market inefficiency caused the spread in the heavily traded SPY ETF to become inverted, as illustrated in the charts below. While the negative spread only lasted for a bit more than 500 milliseconds, it raises serious questions about the efficiency of our markets, and it also puts the implementation of Reg NMS into question.

Should the markets ever experience negative spreads in the age of Reg NMS and HFT? Since quotes, and sometimes trades, occur at the microsecond level, one would expect our technologically advanced markets to be extremely efficient and such conditions to be corrected in a matter of milliseconds, if not microseconds.

So how do we then explain the chart below, which depicts the above-mentioned instance when the market in SPY got crossed and stayed crossed for more than 500 milliseconds, reaching at its widest a negative spread of nine cents – Best Bid $184.24/Best Ask $184.15? The market not only gets crossed, but the Bid and the Ask appear to be moving independent of each other, with the Bid climbing higher and higher, from $184.14 to $184.24 in 10 different moves, while the Best Ask seems stuck at $184.15. The Best Ask finally starts moving up but still stays below the best Bid for the approximately 500 milliseconds in question.

To clarify what's in the charts – the white triangles pointing up represent the Best Bid, while the white triangles pointing down represent the Best Offer. The red lines represent the Bid/Ask sizes (negative and positive numbers, respectively, on the left axis). I broke down the 500 milliseconds into two separate charts for a better viewing experience. 

So how do we explain this situation? Wouldn't we expect all other exchanges to route all their marketable Buy orders to LavaFlow, the venue displaying the lowest offer of $184.15? Wouldn't Reg NMS require them to do so? A closer examination of the rules shows that a crossed NBBO triggers an exception to the Trade-Through Rule of Regulation NMS – see Rule 611(b)(4) of Regulation NMS, 17 CFR § 242.611(b)(4). What this means is that when the NBBO is crossed (the best bid price is higher than the best offer price), exchanges do not have to route orders to away markets and are free to execute trades locally, regardless of whether better prices exist elsewhere. 

So should we then still expect to see instances where the markets get crossed, in spite of all the recent advances in technology? Given our example above, the answer seems to be yes. So the question then becomes: What length of time should be considered acceptable for market participants to recognize the pricing inefficiencies and correct them?

We would expect an exchange to route orders to another exchange offering better prices as soon as that superior price appears. But just how fast should exchanges be able to recognize that a superior quote exists somewhere else? Should it be microseconds? Milliseconds? Or maybe seconds? What should the level of specificity be, and is that currently defined somewhere?

We would also expect market participants benefiting from the advantages of co-location to recognize that crossed market condition and exploit that opportunity in microseconds, buying those shares at $184.15 Ask and selling them at the $184.24 Bid, helping the market re-price itself in the process. We certainly didn't expect the market in SPY to be inverted at all, let alone for more than half a second.

Comments | Post a Comment

13 Comments to "Inside a Half Second of Market Inefficiency":
  • Missing
    Bill Harts

    07 March 2014

    Cristian,
     
    Thanks for this thought-provoking article.  You have to ask yourself, "Why would a trader intentionally cross the market to that extent? Why didn't someone correct the inversion through arbitrage?" 
     
    While it's possible that there was a "real" crossed market, it's also possible that this was just a technological anomaly (I hate the term, "glitch").  So why not add some information about where the problem actually happened, i.e.,
     
    1. At the market data provider feed layer.  I think you mentioned that you cross-checked the Activ feed data (the gold standard) against the Bloomberg feed data and found identical prices so that means the problem happened further upstream.
    2. At the SIP level, aggregating the exchange feeds.
    3. At each individual market center (ADF/NASDAQ/NYSE) market data dissemination layer.
    4. I believe that Lava Flow has only recently begun quoting through the FINRA ADF, could there be an interface problem there?
     
    Also, you might want to check with each market to see if an ISO or other order was sent and rejected during this time period, or if "self help" was declared.
     
    Bill

  • Comment_tdxme
    czarcu

    07 March 2014

    Bill -- thanks for your comments and suggestions...we did reference the data with two different sources -- we were quite surprised by the 9 cents inverted spread and wanted to make sure it was indeed accurate. It was.  We also checked for ISO orders as well as any other exceptions to the Trade-Trough rule and we found the following:some 300,000 shares traded in those 500 milliseconds and almost 80,000 had no exemption of any kind.  
     
    Tech issues are always a possibility and we are actively trying to get some answers...not an easy proposition though -- folks are not all that eager to discuss such.  Interestingly enough, while in this instance the market was crossed by an excessive spread, we have come across negative spreads quite often since we started looking, leading us to believe that they are a more common occurrence than most of us ever suspected...we continue to monitor that particular type of behavior and we'll provide an update on our findings.

  • Comment_steve-wunsch
    stevewunsch

    10 March 2014

    It may be worth mentioning that the world did not end in that half second of crossed markets. Back when Reg NMS was being debated (2004-2005), there were those who pointed out that letting markets right themselves on their own, especially given the strange liquidity incentives for posting that had taken hold, with their seeming distortions of "efficiency" and maker-taker fee structure, might be better than going through all the regulatory rigmarole that became Reg NMS. My impression at the time was that the SEC said no to that approach primarily because it would have been embarrassed to see its grand electronic creation look so obviously flawed as to have crossed markets, even briefly. So instead we have Reg NMS, the flash crash and glitches galore characterizing an environment in which it is now possible, actually, for the world to end in 500 milliseconds. 

  • Comment_tdxme
    czarcu

    10 March 2014

    Very well put Steve.  The world did not end in that half a second nor did it end in the many other instances when the markets stayed crossed for seconds at a time!  Yes, we've come across quite a few of those too.  Reg NMS didn't help in any way in the Flash Crash incident -- quite the opposite I would argue -- and it is definitely not helping the crossed markets correct themselves, as Reg NMS exempts those instances from its Trade-Trough rule.  More regulation is definitely not the answer.  Markets will correct their own inefficiencies eventually, even if it does take a few seconds...better surveillance tools and detection logic can help speed that up.

  • Missing
    Bill Harts

    10 March 2014

    Steve, I think a lot of the push to eliminate the possibility of crossed markets came from the exchanges themselves.  Why bother to post a quote if other markets were going to ignore it?  Remember, we were coming from a market model centered around the Intermarket Trading System where trade-throughs were a routine occurrence (as well as a profitable trading strategy).  No one wanted to go back to that world.

  • Comment_tdxme
    czarcu

    10 March 2014

    It would appear that trade-throughs still occur quite a bit Bill -- much more routinely than one would expect...

  • Comment_steve-wunsch
    stevewunsch

    10 March 2014

    Bill, I don't think it was "the exchanges themselves" that were bothered by trade throughs, but only the exchanges that were not the NYSE, who had been frustrated by the Big Board's ability to sidestep their competing quotes via the ITS dodge, which was created by the NYSE. The SEC, always and forever the champion of killing off the incumbent monopoly via competition, created Reg NMS and its trade through prohibition in order to side with the likes of Archipelago so as to bring the Big Board down. It succeeded. But, as Christian said, trade throughs are still happening. So it is worth asking: What was accomplished, other than to give the SEC its monster role of killing the NYSE? And how are the markets better now than they were then? Cheaper, yes. But better? In any case, the stated rationale for the SEC's attack on the NYSE -- the elimination of trade throughs, as if they were the greatest evil in the western world -- has not happened. Going back is not possible or something I would recommend. On the other hand, eliminating the trade through rule, i.e., Reg NMS, might start the useful process of eliminating the SEC. Regardless of all the water that has gone over the dam now, the question we should be asking is: What good is the Commission doing at this point in terms of market structure? Now that the NYSE is no longer a monopoly, who needs the SEC?

  • Missing
    John Harris

    11 March 2014

    Unfortunately, the SEC still has quite a constituency: all of the firms it protects from competition (too-big-to-jail banks, clearinghouses, national ratings agencies, and exchanges, among others), securities lawyers, large issuers, and public accountants all come to mind. They are its protectors and beneficiaries.

  • Missing
    sambirnbaum

    11 March 2014

    Cristian, you mentioned that 300,000 shares traded during that period, at what price and how many actual trades ?

  • Comment_tdxme
    czarcu

    11 March 2014

    Sam - the image below shows all the prints in that specific second, marked by the red diamonds on the the top chart.  As you can see, there are quite a few.  For clarity, the green triangles pointing down represent the Best Asks while the yellow triangles pointing up represent Best Bids...it should be fairly easy to recognize the tick prices, as well as the relationship to the Best Bid/Ask at the time -- see left axis.

     

  • Missing
    sambirnbaum

    11 March 2014

    Cristian, sorry but the image is a little fuzzy and was hard for me to read.

  • Comment_tdxme
    czarcu

    12 March 2014

    Sam -- my previous message had a chart attached to it, which doesn't seem to appear on the website for you.  There were 380 trades in that second -- while most of them traded outside the NBBO, 38 of them actually had no exemption to Rule 611, the Order Protection Rule.  Hope this helps.

  • Missing
    sambirnbaum

    12 March 2014

    Thanks. It does.

You must log in to comment.