Who Made Money Yesterday During the Dow's Swan Dive? Smart Futures Traders - UPDATED

Ivy Schmerken published a nice overview piece on yesterday's market dislocation on the Wall Street & Technology Blog. A little background for those who have not been following the discussion:

As most of us know, problems began around 2 p.m. when the Dow Jones Industrial Average was already down 200 points –- in part a reaction to the sell-off in Chinese stocks, translated into concerns about the U.S. economy. I was in the car at 2:30 p.m. when CBS Radio reported that the market was crashing and the Dow was down 250 points.

The heavy sell-off in U.S. stocks caused a 70-minute lag in correctly calculating the value of the Dow Jones Industrial Average, according to Dow Jones Indexes, which issued a statement today.

In the release, Michael Petronella, president of Dow Jones Indexes, said the problem arose in the system responsible for feeding market data into the calculation system.

Recognizing the problem, Dow Jones Indexes switched to a back-up market data system for calculating the index, which immediately caused the index to drop by another 200 points at around 3 p.m. -– so the market was down over 500 points.

The switch over to the back-up systems caused all the prices to be processed at once, and this led to the downward spike in the reported index value, according to the release. When the Dow Jones index was adjusted to the true value and the market plummeted 200 points, this probably triggered algorithmic trading programs to generate buy and sell orders, which were concentrated in the Dow 30 stocks, rather than dispersed throughout the market, the source conjectures.

Ok, this makes sense. I don't know, maybe it makes you think that that Grasso guy wasn't so bad after all, but no matter. Bottom line - we had a little meltdown, the cash market systems couldn't keep up with the order flow and, therefore, the DJIA gapped down hard after the flood gates were opened. But what happened in between? Was there anyone, anywhere that could have made money from this debacle?

Of course. The answer: smart futures traders.

Please take a look at the chart below.

See the black line - that's the spot DJIA. See the blue line - that's the March DJH07 futures contract traded on the CBOT. So, let's walk through this together. In the morning the spot and futures markets pretty much tracked each other. Then look what happened - uh oh, the spot market is falling behind, while the futures market is reflecting the true market sentiment. They are starting to diverge, then wider, wider still, FOR ABOUT TWO HOURS, until BANG - the alternative cash system kicks in and the flood of sell orders drops the spot index like a stone. So this technical "glitch" was really, at its core, a timing delay. Then the futures market, as if it knew what was going to happen, ran up, after which the spot market followed with a significant lag. After a little sputtering and some continued dislocation late in the day yesterday, they tracking each other once again today. Whew.

So what does this mean? A savvy futures trader that saw the divergence could have positioned themselves to profit from the inevitable meltdown in the spot market, and the subsequent run-up after the futures rallied ahead of the spot market. AND HAD ABOUT TWO HOURS TO DO IT. It is pretty rare to find these profit opportunities in the real world, but they manifest themselves yesterday during the wild ride in the U.S. afternoon trading session.

Anyway, I generally don't write much about this stuff but felt compelled to share this after knocking around some thoughts with a couple of my colleagues. Interesting day, to be sure. Thanks to Eli for this thoughtful contributions and the chart. Enjoy.

ADDENDUM

In the wake of comments received by readers of this blog, I was made aware that there is an error in my chart - the plotting of the CBOT futures contract is not adjusted to take into account the NY/CHI time difference. First, thanks to those who commented. This is exactly what makes the blogosphere great. I'm usually the one imparting the checks-and-balances, but today I am the recipient of such!

All that said, I'd like to factually explain why the error occurred, and then to migrate to why I believe an arbitrage opportunity did exist, even after correcting for the time mistake. The only substantive difference in the analysis is that instead of a trader manually being able to enjoy the arbitrage profits that were available, the money was invariably captured by stat arbs whose programs picked up on the difference between the the spot DJIA (or the DJIA components individually or in ETF form via DIA) and the theotretical DJIA (as expressed through the CBOT futures). And I respectfully disagree with both Maoxian and my two commenters that arbitrage opportunities didn't exist, because they clearly did. My charting mea culpa and evidence of the availability of arbitrage profits follow.

Why the Error?

The chart from my post was taken from OptionsXpress, which uses Prophet.net for their charting application. Apparently, when comparing two instruments, Prophet.net does not adjust for time discrepancies, which means that clearly OptionsXpress doesn't, either.

From Prophet.net

From OptionsXpress

It also appears to be the case that Yahoo!'s new Beta charting application may also give rise to the same error:

   

These errors, not surprisingly, are not present when one undertakes the same charting exercise on Bloomberg:

So, for those who expect (as you should) factual correctness in all that I write I apologize. I am both embarrassed and humbled by this error. I'll work hard to ensure that it doesn't happen again.

The Argument for the Availability of Arbitrage Profits

Let me start with presenting a more granular decomposition of the Bloomberg data above, by importing the data into Excel and plotting the price movements in five minute time increments:

Ok, so do you see the divergence between the spot and the futures? It is not two hours, as I had mentioned orginally, but the divergence is in place for around an hour (which makes sense given the NY/CHI time difference). I'd like to use Chairman Mao's post as a vehicle for arguing my point that arbitrage profits did in fact exist:

Everyone knows about the Dow Jones Industrial Average calculation glitch during Tuesday’s slide. Here are a couple overlay charts: 1) the Dow futures with the DJIA, and 2) the Diamonds ETF with the DJIA. You can see that both the futures and the ETF were tracking the Dow component stocks correctly the whole time. There were no opportunities for arbitrage.

Well, I don't really agree. He is using candlestick charts, which are not sufficiently granular to observe the arbitrage opportunity in question. So no, you can't  "see that both the futures and the ETF were tracking the Dow component stocks correctly the whole time." Because they weren't and I think that point is pretty widely acknowledged. He continues on to say:

I don’t even watch indexes on a quote sheet anymore (who watches the DJIA intraday anyway?), but I do follow all the ETFs based on major indexes closely. I can trade the DIA; I can’t trade the DJIA.

Ah, this is the point. I find the chart comparing the DIA and the DJIA to be a red herring. One is simply the decomposition of the other. They should track each other, by definition. This is not the case with futures and cash, where divergences occur for a wide range of reasons, including technical glitches. But the comment that was the most telling: "I can trade the DIA; I can’t trade the DJIA." You may not be able to, Chairman, but stat arbs can and do, often thousands of times a day. How could a stat arb have made money? By their programs detecting the divergence between DJIA "real" (the average of the prints of the components that make up the DJIA) and the DJIA "theoretical" (the CBOT futures), executing program sells of DJIA-replicated single-stock baskets across any number of ECNs and simultaneous buys of the futures. While the stock trades wouldn't have printed due to the glitch, they would have been filled and settled after the backlog had been cleared in the 2:50-3:00PM time frame. Therefore, a stat arb who had sold the cash "rich" and bought the futures "cheap" could have captured the area between the two curves, ex-transactions costs. Not a bad return for a few CPUs.

I've read a lot of pretty virulent talk about how arbitrage conditions didn't exist. My charting error notwithstanding, I don't agree. And I think the facts - real as they are - are pretty compelling. But that is just my two cents. I'll leave it to the experts (those who write this stuff and pull down charts every day) to tell me otherwise.

Comments

[anon]

Dear Sir,

You need to do much, much more to validate the existence of an arbitrage opportunity. Most vendor cash feeds get quite slow so the prices you see are really [untradeable] prices in the past leading to a false divergence.

-Seasoned Stat Arb Player

[profittheory]

BillyW- It's true that as arbitrageurs we compute our own theoretical values, and that arb operations are continuously ongoing in the course of an average trading day; smaller opportunities always exist. However in this case, with (a) the theoretical price computed below the DJIA spot, and (b) the given wide spread, it was hardly one of those "normal" arbitrage opportunities you speak of.

Also, from a purely theoretical perspective, consider what it means when the futures price falls far below the published spot.

[BillyW]

There was arbitrage going on, as there always is - but there was no more opportunity than exists in a typical moving market, as every serious arbitrageur computes his own value for the index based off of underlying prices. The fact that DJ were temporarily publishing the wrong index value is a red herring - no serious trader would use that as a basis for trading.

[Roger]

Ryan, jeezus, I really must be doing a shitty job describing the opportunity here. The point you raise about the disparity (or lack thereof) between the cash and the ETF isn't the point. There was - and the chart proves it - a disparity between cash (in any of its manifestations - single stocks making up the DJIA, the DJIA itself or the DIA ETF) and futures. Why you say there wasn't I don't understand. I guess we'll just have to agree to disagree.

[Ryan]

usually your description of derivatives and trading strategies is right on... but in this case I think you are wrong. The key point is there is NO trade you could do (not with futures, not with ETFs not with the underlying basket of stocks) which would allow you to execute this "arbitrage"

There are two DJIA relevant here-- (a) the official number published by dow jones and (b) the theoretical number you would get if you follow the index definition and take the correct underlying stock prices.

(a) was incorrect because of a technology glitch-- dow jones company couldn't keep up with the volume of market data, so their published number diverged from what it should have been (b). However, you can't trade the DJIA-- its just a published number. The instruments you can trade (the underlying stocks, the futures, the ETFs) all traded consistently and in line.

[Roger]

jck, you raise good points. I'd say, however, that it is empirically difficult to validate your theory, with the exception of the collar. Certainly, collars we in effect starting at 1:03, well in advance of the divergence between futures and cash. This made trading large blocks very difficult. However, stat arbs that could break up small orders across a wide array of ECNs and access liquidity across "dark pools" (i.e., Liquidnet) had some shot at getting a meaningful amount of the stock side of the arbitrage executed.

That said, even if it was difficult to execute the stock side of the trade, it is possible that one could have profited by going long the cheap futures and covering later on, as the prices between cash and futures converged.

Anyway, thanks for your thoughtful comment.

[MING]

Chairman Mao's point is that since cash, future and ETF were all in line, there wasn't any arbitrage opportunity. But this efficiency can be explained as the result of arbitrage also. I think this eventually boils down to the question of whether there are traders using DJIA as a reference for Future or ETF trading. If these poor souls do exist, they would be trading at wrong prices and give more informed traders immediate profit opportunities.

[jck]

Roger:
first ,thx for the correction.
for the arb ,no doubt the basis was out of whack,but it would have been hard to execute the stock side given that bids were thin,melting away quickly,stops were hit etc..except perhaps in the early part of the "discrepancy"...also collars were in place.

[David Merkel]

The panic allowed me an opportunity to sell some TLT and buy some QQQQ in the dislocation. Wasn't that much, but I was able to clip a profit on both trades. I had no idea what was going on at the time, though.

[Yaser Anwar]

I think this chart maybe what you guys are looking for
http://tickersense.typepad.com/ticker_sense/2007/02/dow_calculation.html

[Michael]

Roger, you should check out Maoxian's post. HE says there was no opportunity for arbitrage -- http://maoxian.com/archive/when-index-calculations-lag/

[Roger]

JCK/Ming, thanks for the comments. I am looking into the issue and will add an addendum to the post a bit later.

Roger

[MING]

I think jck is right. The two lines match almost perfectly when shift by an hour. Even if they can predict a decline in stock, the pattern wouldn't exactly match.

Besides, if the NYSE order system is congested, I'm not sure how you would profit from this. The stock market effectively became illiquid.

[jck]

Complete nonsense.
The cash tape was delayed and the futures traded off the wrong cash price.Your chart is wrong,the futures price is on Chicago time and the cash on NY time.The march futures low was 12102 at 15:02 NY time.

[livepost]

I didn't have the DJI cash chart up so I didn't know there was a discrepancy until I heard CNBC say that the dow was down only 280pts, and I was watching YM and it was down around 400+pts, so I said, what the heck? Of course you see the 3pm reversal, when do you not see that these days on big trend days. Prop and hedge funds got to lock in those profits before the close.

[Yaser Anwar]

Michael-

It will be the usual top HF/Inst. guys who profit. When it was happening people didn't really have a clue until a little later when CNBC started talking about it. By then it was too late.

I had some troubles getting fills around 3ish as did quite a few traders I talked to. Some say it was the market makers manipulating it, some say the algorithms, we might never really know, could be both....

[Michael]

Good stuff Roger. I'd like to know who took advantage of that yesterday.