Main question: How should the pilot be run?

• background

• key guiding principles

• main features

• sample construction

• statistical measures

• questions for industry

• (maybe) tick size too wide in very competitive markets
• allow more aggressive posting and then attract market orders

• market making is costly => reward for posting
• attract posted orders to marketplace

multi mkts

TSX intro rebate

first inverted venue CX2

NEO speed bump

changes to OPR

Alpha speed bump

TSX

Omega

Lynx

Alpha

NASDAQ

NEO

Pure/CSE

CX2

Aequitas Lit

Nasdaq Dark

ICX

Liquidnet

MatchNow

Protected

Unprotected

Dark

• different fees for
  • special orders (dark, post only)
  • price levels
  • listing (TSX/TSXV/CSE/Aequitas)
• bulk discounts, market maker discounts, etc

• m/t split is irrelevant, only net fee matters
• bid and offer prices adjust to account for net fee

• m/t fees not paid directly by investors but brokers pay/receive 
• m/t target specific traders (HFTs) and brokers (retail)

• post at low rebate venue: trade faster
• post at high rebate venue: earn more fees
• US evidence (Battalio et al 2016): harms passive retail orders

perfect markets

conflicts of interest

imperfect markets

brokers' routing practices, measured by the execution quality of orders

different market participants (i.e., retail traders, institutional participants, liquidity providers)

the level of intermediation

standard measures of market quality

simplicity

all-encompassing

don't try to do too many things or to be too nitty-gritty

transparency

must pay to adapt, no exceptions for venues

design, methodology, procedure 

meaningful sample

avoid biases and intrinsic noise

simplicity

all-encompassing

rebate prohibition for approximately 120-160 securities

transparency

prohibition group = approximately half of highly liquid, all venues (inverted, unprotected)

design, sample selection & analysis codes publicly posted

meaningful sample

focus on highly liquid securities, split into 3 subgroups

timing

align with SEC pilot and coordinate on sample

TSXV

TSX

CSE

ETP

NEO

Concern 2: low-price 

• securities with low prices have wide relative spreads
• => noisy observations

Result for analysis

• often inconclusive
• will never know why there is a "non-result"
• undermines usefulness of the pilot

~200

~100

~100

• matching criteria corporates

  • firm size (market cap)

  • price level

  • trading $-volume

example for a subset of highly liquid securities

• trading costs
  • effective & realized spreads (w & w/o fees)
  • price impact
  • implementation shortfall

• execution quality
  • time to execution depending on book position

• posting activity
  • orders posted by position in the book

• price efficiency
  • return-autocorrelation
  • variance ratios
  • range measure

• posted liquidity
  • quoted bid-ask spread
  • quoted depth at the top of the book

Mathematically: we estimate

\[\Delta{DV}_{it}=\alpha\cdot \textit{pilot}_{t} +{\textit{other } \atop \textit{variables}_{t}}+\delta_i+\epsilon_{it}\]

and ask: is \(\hat{\alpha}\not=0\)? 

• even with prohibition, exchanges can  recreate the economics of today
  • e.g., one with m/t 0/5 and one with 5/0
  • post prohibition scope:
    • largest - smallest maker fee
• examples for m/t fees post rebate prohibition:
  • .5 net/.5 net, net/0, 0/net

• Scope for conflict for limit orders:
  • |largest maker fee+largest rebate|
  • currently: 42 mils

• limit of scope of conflict: net fee
• currently: 4-5 mils
• what will net fees be?

Question to industry: 

• Is our concern valid?

• Is symmetric pricing requirement acceptable?

• possible solution: staggered introduction of treated groups

• introducing all pilot changes on one date: risk that a major event creates a common shock that confounds the results

Question to industry: 

• Will firms adapt only once?

Example

• matched pair: 
  • treat iShares' XIU and control with Horizon's HXT
• suppose
  • spread for XIU \(\uparrow\) for HXT \(\to\)
  • investors choose the one that is cheaper to trade
  • => volume in XIU \(\downarrow\)
  • => XIU spreads may increase further

Question to industry: 

• Is this concern valid?
• If so, does it make sense to include ETPs at all?

• matching may create spillover effects that lead us to under- or overestimate true effects

• Any specific concerns regarding the SEC pilot vis a vis ours?

• Are our small sample and noise concerns reg TSXV and CSE securities valid?
• Which matching characteristics do you recommend?

• What do you believe is the most likely reaction by exchanges regarding fee innovation?
• What fees do you predict?
• Limiting the frequency of fee changes?

• Do you have suggestions to measure routing practices?
• Are we missing any important measures?