Some smart – but conflicted – routers | FT Alphaville

Some smart – but conflicted – routers

A block trade, born in a US-based pension fund, is traveling the electronic execution highway.

The trade — let’s call him Benny — is sent to the fund’s broker for execution. That broker, now sitting in his office in New York, has a number of options when it comes to Benny. He can send Benny to an exchange like the NYSE or attempt the trade over-the-counter. Or he can send it to another division of the brokers’ own firm, to be filled using the company’s own inventory. He can internalise little Benny.

Which one will he choose? How will Benny’s execution destiny be fulfilled?

The proliferation of ‘smart’ routing options — in tandem with algorithmic and high-frequency trading — isn’t new. Brokers have internalisation or ECNs. Exchanges have things like matching.

All of these things look to route trade orders in the ‘smartest’ way possible — or so they’re usually marketed to institutional investors. Back in 2010, Morgan Stanley estimated that routing conflicts generated a $63m profit for brokers and $86m for exchanges per year, in a letter to the SEC.

And now there’s an even higher figure to mull.

A report from Woodbine Associates says that as much as $4.5bn in opportunity cost or ‘alpha-at-risk’ is wasted market-wide on sub-optimal order routing decisions. Themis Trading reckons that means market-makers controlling the routing decisions could be getting a $4.5bn boost from trades.

The report itself is quite interesting since it uses the Rule 605 data filed to the SEC market centers to show the extent to which they provide executions at prices better than the public quotes to investors using limit orders. And Woodbine’s wasted alpha-at-risk number is based on a few assumptions:

Assuming all order flow is purely interchangeable and could be executed at the exchange with the lowest effective spread, holding all other factors constant, the annualized, aggregate market-wide alpha-at-risk would approximate $4.5 billion. This upper-bound estimate is derived by simple multiplication: the approximate difference in share-weighted average effective spread between the highest and lowest value exchanges is multiplied by aggregate exchange printed shares at exchanges in 2010. Accordingly, the range in share-weighted average effective spread of approximately 3/10 of a penny was multiplied by 1.482 x 10 [to the ninth power] shares traded on exchanges in 2010.

A summary of the report — with some exchange comparisons — over here.

Related links:
After the trade is made –
What’s in your router? – Themis Trading
All eyes on broker-dealer internalisation – FT Alphaville