Research Question

Who benefits from a speed bump on marketable orders that provides a speed advantage to some limit orders?

Answers

1. HFTs liquidity providers: larger profits on the speed-bump venue.
2. Aggregate liquidity worsens.
3. Liquidity providers: lower profits on remaining venues.

Speed Bumps

The famous case: IEX's Speed Bump

• main idea: orders into IEX are slowed down by 350 milliseconds but updates SIP data (=NBBO) faster.
• => NBBO pegged orders trade at "fresh" prices

• Chicago Exchange proposes “Liquidity Taking Access Delay"
• Incoming marketable orders get delayed by 350 microseconds, limit order cancellations do not

Speed Bumps

• all orders send to TSX Alpha Exchange are randomly delayed, between 1-3 milliseconds
• the exception are "post-only" orders which are not subject to the delay

Here: TSX Alpha Exchange

Cost of using post-only?

How do you benefit from post-only?

before

How do you benefit from post-only?

after

Comment 2: Endogeneity of Adverse Selection measures in the paper

General Idea:

• the speed bump protects liquidity provider against adverse selection
• => avoid posted orders to become stale and can cancel ahead of market orders

Defining an informed order

• Step 1: Collect strings of trades
  • = trades within 50 milliseconds
• Step 2: Informed = displaced/absorbs entire NBBO (i.e. across multiple markets)
  • accounts for
    • the trade itself
    • plus trades by others
    • plus cancellations!
  • What if the NBBO is only 100 shares?
  • Creates an endogeneity when speed bumps are introduced!

Comment 4: Beware of retail orders
Retail Myth 1:

(not this paper, but generally)

"retail trades are small"

average volume for single-market marketable orders

Retail Myth 2:

"retail trades are single-market and small"

average volume for multi-market marketable orders

Smaller Comments on Results and Interpretation

1. "[The change] increase[s] profits for liquidity providers on [...] Alpha but [...] liquidity suppliers’ profits [are] reduced across remaining venues."
   • Often these will be the same entities: do they just cross-subsidize and costs/benefits net to zero?
2. NBBO Computation for Alpha trades - does it include Alpha?

• would explain quote-fade/fill rate result

overall I like many of the results, and I believe that they make intuitive sense, but I think the construction of measures requires work

Results and Interpretation

1. informed=consume all liquidity
   Tight spreads: 50-80% of trade-strings consume all liquidity
   Wide spreads: significantly lower levels of informed. 
   • I am sympathetic to the measure 
   • BUT: Way too much informed trading! If depth is low, too much will be absorbed
2. The fill-rate measure has to be explained more thoroughly, e.g., by using examples.

overall I like many of the results, and I believe that they make intuitive sense, but I think the construction of measures requires work

Broker Activities

Source: RBC Marketstructure Guidebook Jan 2016

Comment on Results on Adverse Selection

• Result: Adverse Selection on Alpha declines 
  • => makes perfect sense to me
  • hard to show in the data, but apparently one retail broker send all its flow there first.
• lower realized spreads for non-HFT liquidity provision on Alpha also makes perfect sense 
  • => only get to trade once the HFTs stepped out of the way

Effects on other venues

• "Section 4 establishes that Alpha’s systematic order processing delay against marketable orders enables the segmentation of uninformed order flow"
  • statement much too strong!
• Question: with minimum post size, what is the ratio of depth Alpha/depth rest when Alpha at NBBO?
• Feature of TSX Composite:
  • around 180 of 250 have 1-2 cent spreads
  • rest has very wide spreads
  • Suggestion: split sample!

my view: take with huge grain of salt due to the many simultaneous changes