Background on Internalization

Andreas Park

My Definition for Canada

internalization = a trade arranged within a brokerage without the participation of order flow that does not pertain to the brokerage

Specific Types

  • against broker's inventory
  • against another client's order

How arranged?

  • normal assumption: against unexpressed liquidity
  • Canadian specialty (and why we are here): on-exchange against standing, visible orders through broker-preferencing

Theoretical Concerns

  • broker discriminates:
    • keeps "desirable" = uninformed trades to itself
    • => cream-skimming
  • rest of market gets "undesirable" = informed orders
    • market suffers from adverse selection
    • traders less willing to post
    • market quality and price discovery deteriorate
  • internalizer "free rides" on price discovery
    • reduces incentive to post

Theoretical Benefits

  • allows to avoid exposing one's willingness/need to trade large quantity:
    • prevent market to move against large & benign order
    • reduce volatility and improve price discovery

With theoretical pros and cons
=> empirical question:
What does the data say?

 

  • challenging problem: 
    • markets differ => not all studies apply to Canadian environment
    • in most US-centric studies: internalization = wholeselling
    • times change
      • different players
      • different trading products
      • => markets evolve
  1. effect is benign

    • Battalio, Greene, and Jennings (1997); Hansch, Naik, and Viswanathan (1999); and Kam, Panchapagesan, and Weaver (2003)

  2. harmful to market quality
    • Battalio, Greene, and Jennings (1998); Chung, Chuwonganant, and McCormick (2004); Grammig and Theissen (2005); Chung, Chuwonganant, and McCormick (2006), Weaver (2011), Hatheway (2009)

A cautious view of the evidence on off-exchange internalized trades

most applicable, nuanced paper: Comerton-Forde & Putnins (JFE 2015):

  • "Dark Trading and Price Discovery"
  • studies non-displayed trading in Australia
  • dark in AUS = broker-internalized (they operate own dark pools)

Findings

  • price discovery deteriorates when %dark > 10% of $volume
  • but block trades up to 15% can be beneficial

Too much dark is bad

bad

good

Keeping blocks out of the lit market is good

bad

good

What does this mean for Canada?

  • Although similar, AUS is not Canada
  • Upstairs crosses, dark trading, and broker-preferencing have been here for years
  • Let's look at TSX stats, noting that:
    • same broker trades are is not necessarily broker-preferenced
    • broker-preferenced TSX marker captures trades that would have occurred without broker pref
Type %value
blocks/"intentional crosses" 11%
other dark  8%
TSX broker-preferenced 12%
Total of all 31%

Has there been a change?

Points of contention

uptick by 2% daily volume in broker preferencing

note: most is client-to-client

  • This uptick seems more than a blip.
  • Is it enough to have an effect? 
  • Main theoretical concern is related to cream-skimming by selective brokers
  • Looks like client-matching
  • => different form of filling a cross?

client-to-client

client-to-self

Arguments for broker-preferencing?

  • Historic concern (going back to 1997 UMIR rule changes) reg internalization:
    • undisplayed liquidity hampers price discovery
    • But: resting order portion is displayed => not dark
      • No first-order effect on price discovery
  • saves settlement fees
  • allows liability desks to unwind positions against broker's own flow
    • => may make blocks cheaper

Summary

  • Main conceptual question:
    • Concern: cream-skimming
    • Benefit: less market disruption and exposure of large-size traders
  • Empirically, results have been mixed (aren't they always!)
  • Data from TMX
    • suggests small uptick in broker-preferencing in late 2017
    • driven by client-to-client interactions
  • Final thought:
    • broker-preferencing generally involves posted liquidity that is available to all
    • but could be invoked without meaningful posting time
    • => maybe require that BP only applies after minimal resting time?

Internalization

By Andreas Park

Internalization

presentation at the TSX special roundtable on internalization

  • 1,016

More from Andreas Park