The mysterious case of the SPX spike

April 23, 2009 in Finance,Math

Get out your tin foil hats, there's something sinister afoot!

Zero Hedge, a blog which has been getting an astounding amount of press lately (in particular because of this rumor) posted the following picture this afternoon "without commentary" (please note I recreated their image to use EST times):


This is called a volume-at-price chart, or a histogram in which each bar represents number of trades at that price (prices are listed on the x-axis), in this case for the S&P 500.  You may recall that at the end of the day the market quickly rallied from being slightly down to close at the day's high of 851.92.  Zero Hedge's glass is always half hidden behind a conspiracy novel, and the implication here is that the spike is caused by heavy-volume market-manipulating fiends, aiming to mask... something evil (it's unclear what exactly, as you'll recall the post was "without commentary"). To drive the point home, ZH also posted an image of the SPY (the tradeable S&P 500 proxy) which, notably, did not exhibit such a spike.

At this point, level heads need to prevail.  Here is the exact same graph, but only including trades made at or before the close of trading at 4:00:

spx-vwap-423-4001Obviously, the conspiracy-driving spike is absent. Why? Here's Bloomberg's view of the S&P 500 over the course of the day, from half an hour before the open to half an hour after the last datapoint arrived:


Yes, the datastream continues after the close! In fact, there is one after hours tick every 15 seconds from 4:00 to 5:00, a grand total of 240 "trades" (there aren't really S&P 500 trades, only ticks) at the closing price, after the close. The key point is that these extraneous ticks are, for some reason, included in the VaP histograph if it is not limited to trades before the close.  When Zero Hedge took their picture, 39 minutes had elapsed since the close, or 156 extra ticks. Unsurprinsingly, the height of the "conspiracy bar" is exactly 156. Had ZH waited until 5:00, they would have seen the mystery spike grow to 240.  I know this because it's true.

So there's no conspiracy here, only a lagging data feed and a misspecified screen.  Here's the same comparison for yesterday, first as it appeared at 4:00 (click to zoom if you so desire):

spx-vwap-422-400 And then at 5:00 (it should be obvious that the index closed at 843.55 -- note the spike height is unsurprisingly 250, equal to the 240 after hour trades plus the 10 trades already at that level):spx-vwap-422-500

In fact, this phenomenon happens every single day.  There's no conspiracy revealed by this data.  In fact, it happens on every index Bloomberg tracks.  However, it does not occur for actual equity tickers like SPY; presumably Bloomberg's software distinguishes between securities which actually trade (and thus have actual "after hours trades") and those that are merely composites, ticking across until the data feed ends.

The first rule of statistics is to know your data.  You may now put your tin foil hats away.

{ 5 comments… read them below or add one }

Adrian April 24, 2009 at 1:38 am

Nice catch.


Procol April 24, 2009 at 1:41 am

Exactly why I posted on Zerohedge that this was pure nonsense. For someone who seems to know all the quantspeak, and technobabble, Durden seems woefully (or is it willfully) ignorant of these most basic mechanical quote issues.


Clear as Mud April 24, 2009 at 9:37 am

Thank you! I noticed that the VAP screen defaulted to 17:00 but would have never connected it to the GIP screen and the after hours ticks. It did seem to be a bit of fear mongering at ZH.


Arm April 24, 2009 at 1:12 pm

This actually proves nothing or actually disproves your point. Ok, so it is a lagging feed, kudos. But the spike still ocurrs on the consolidated data!

It is eve WORSE, all the buying ocurred literally in the last minutes as the index was closing and was not consolidated until later. The 250 trades all came in at the inflated higher price.


J April 24, 2009 at 6:07 pm

Arm, I think you may be misunderstanding me. When I say the 240 trades are lagged, I don’t mean they are real trades that are showing up after the close (which is a common occurrence with traded securities but not with indices).

Instead, they are just phantom prints – one every 15 seconds for an hour, at exactly the closing price – which are not representative of actual trading. The datafeed is lagging in the sense that every single day, instead of just cutting off at 4:00, it spits out a closing print every 15 seconds like clockwork and then cuts off at 5:00.

Maybe this is so that if there were ever a need to restate the closing price, Bloomberg would have an hour from the close to adjust the feed. I don’t know. But the regularity and constancy of the prints, plus the fact that indices simply don’t change after the close, makes me absolutely confident that they are not in any way representative of trading.


Leave a Comment

Previous post:

Next post: