Monday, April 13, 2009

Dark Pools' Role In Liquidity Provision

The same "liquidity providers" that have been discussed previously in Zero Hedge, are also in dark pools to make sure trading costs are shifted from observable to less observable or not directly observable. Observable trading costs are the difference from VWAP, open price or arrival price + commission and spread costs. Less observable are bid/ask drift before a dark pool print. For example if I put an order to buy 100k shares of IBM to POSIT, UBS, or GS (especially the latter two where prop traders are more than happy to get right of first refusal on the bid/offer if they so desire) the stock might move up about 10 cents (not an actual observation) and than print my block. Within 15 minutes after the print, the stock is back to being 10 to 15 cents lower. Example of non-observable cost is P&L reversion. In the IBM example, I put an order to buy 100k shares to POSIT. The stock does whatever it would have done anyway and within 30 minutes my block is printed during regular crossing session. In the next trading day, the first 30 minute VWAP will be about 10 to 15 cents lower. So if I buy IBM on POSIT on a regular basis, on average, I can pay 10-15 cents less just by buying the next day on open market. If we factor in higher crossing networks commissions (mostly invisible, charged via net markups), the promise of a free lunch become the reality of very expensive lunch.

Obviously, mutual and pension fund traders have high alpha, they need access to expensive liquidity provided by GS principal bids desk, crossing networks and other means where trading costs are allocated via the "back door" and do not affect buy side traders' bonus scheme payouts. And obviously, teachers, firefighters and the police along with 401(k) investors are on the receiving end. What's new?

In essence, dark pools do not facilitate liquidity provisions over and above what the high-vol players provide, and one can argue, do not lower transaction costs at all - these are merely contraptions that are designed to fool naive traders that immediate vol spikes can be avoided on large block trades, however the cost in the long run is prohibitive enough to where the entire model start cannibalizing itself. Potentially the only benefit from the pools is as an opaque conduit for "not so legitimate" transactions. Sphere: Related Content
Print this post

10 comments:

Anonymous said...

Not sure if this started today, but ZH is now picked up by Bloomberg.

Anonymous said...

In re: "Dark Pool's . . ." post -

I think you are saying that if you bid for 100k of IBM and there is no natural seller there who hiys you immediately, then a prop trader at a brokerage (you said UBS or GS) might buy stock in mkt, accumulate, and then sell to you in POSIT. In doing tihs rather than going directly to a broer with an order you will have saved the explicit commission the broker would have charged, but perhaps paid more on the actual execution price of the 100k than you otherwise would have.

So far so clear I think. But can you then explain the part about the next day's VWAP over 1st 30 minutes, and why you then say that: "So if I buy IBM on POSIT on a regular basis, on average, I can pay 10-15 cents less just by buying the next day on open market"

Many thanks

archimedes said...

What will happen when your "vanilla" money stops coming to feed the machines?

What if states declared that pension funds are forbidden from "investing" in the financial virtual world and instead should only focus to return principal + inflation to the pensionee? That goal shoudnt be too difficult for the pension funds, right? You dont need esoteric Game Theory and Quantology, running in an isolated system where thermodynamics dont apply (because it is virtual), to facilitate that.

In other words, what if Keynes and his evolved legacy -this monstruous machine world- was wrong?

Anonymous said...

"not so legitimate"

You mean, like "illegal"?

Great Mellon's Ghost! said...

BEHOLD VORTEX(sm)!!!!!!!!

http://tinyurl.com/djdt8j

brought to you by :

http://www.bnyconvergex.com/

(that's Bank of New York between you & me)

ranked #1 in NYSE Euronext volume for 10/08, the craziest month since 9/01!

currently serving on the Board of Managers of our holding company is the esteemed ex-Senator John E. Sununu!

http://en.wikipedia.org/wiki/John_E._Sununu

(also currently once again serving his country as a member of the TARP Congressional Oversight Panel)

BNYConvergex
We Got You Covered (Seriously)
Why Go Anywhere Else?

Disclaimer: Vortex would not exist without the generous $3B support from millions of unborn anonymous donors (you'll know who you are when the time comes).

Thank you from the bottom of our green shoots.

GMG$! said...

...and black hearts.

Anonymous said...

There are a number of issues with this post. First of all, you're lumping POSIT in with two broker sponsored pools (PIN and Sigma X). The flow seen in POSIT is quite different from that seen in any of the broker pools.

Second, every trading venue (lit or dark) has its share of participants who are looking to "game" large orders. If the accusation is that the brokers are the ones gaming their own pools, I would be extremely interested to see something that actually backs up your claim.

Third, where are you getting your reversion numbers? Every trade has some temporary market impact (which leads to reversion). Shifting your trade from a dark pool to an open exchange will not get rid of this. More likely than not, a block traded on the open market will experience MORE, not less, reversion.

There are many things wrong with dark pools, but I wouldn't be so quick to add global conspiracy to the list.

Anonymous said...

This post is waaay out of my league.

Anonymous said...

please learn that vol does not mean volume

MW said...

While your the article is true to some extent it does not convincingly show that there is an increased cost. The author's implicit assumption is that the block is printed at one time. That is way too simplistic as in reality oftentimes algo's are being used to fill orders. Since algo's can range from passive to ultra aggressive, I'm fairly certain that compared to the leakage when using using a prop desk or the specialist, using Dark Pools in conjunction with the appropriate algorithm will reduce your cost or at the very least reduce info leakage