Andreas Park PRO
Professor of Finance at UofT
Sniping in Fragmented Markets
Katya Malinova (DeGroote School of Business)
Andreas Park (University of Toronto)
Western Finance Association 2020
-
computerized
-
multiple marketplaces
- forced linkage of markets by regulation
Modern Markets
Key ingredients of modern equity markets
1,000
Shares at Canadian Offer
300
400
investor sends buy order to broker
buy 1,500 shares
for trader: no-trade-through regulation => broker must split among three venues
Shares
1-tick off
400
100
2,000
Modern Markets
The fundamental challenge
Many questions in microstructure relate to multi-markets, for instance,
- relevance of fees
- emergence and role of HFT
- impact of fragmentation
- price discovery
- market making across many markets
- limit order placement and execution cost/quality
Why do multi-markets matter?
Physical distance cannot be overcome
buy 1,500 shares
buy 400 shares
buy 1,000 shares
buy 100 shares
fundamental problem for researcher:
how can we string together trades across many venues?
fundamental advantage that we have:
we can string together trades and we can differentiate (somewhat) what comes first and what follows
- Stylized facts of multi-market trading (MMT): who and how much?
- How does the market react to MMT?
- Does it matter whether people trade across markets?
- How does MMT affect what others do?
Our Contributions
Research questions and directions of inquiry
-
IIROC (Canada) "CAT"; same data as in
- Comerton-Forde, Malinova, Park (JFE 2019)
- Brogaard, Hendershott, Riordan (JF 2020)
- Korajczyk & Murphy (RFS 2019)
- Sample period: March 2013-May 2013;
Use market change event end-April 2013 - Sample stocks: 30 highly traded, non-interlisted firms
Background: Data
Sources and sample
Data on Investors vs. Traders
What the data does and doesn't contain
What we'd like to know:
What we do know:
trading venues
investors
brokers
trading desks & systems
Usage of multi-market orders
How often and for how much?
Note: NBBO depth (and depth at the largest market) is usually larger than 100 shares
Are multi-market trades different?
Market Reaction to Trades: Price Impact by Single-vs Multi-Market
Usage of multi-market orders
Is it only used by smart traders?
Use the same classification as Comerton-Forde, Malinova, Park (JFE 2019), loosely:
- retail: special order type
- institutional: per day, same direction, many trades
Fun Fact: retail trades are on average larger than institutional trades
Are multi-market trades different?
Market Reaction to Trades: Price Impact by Size
medium:
- >100 shares
- <90th %-ile per stock
Fun Facts:
- only 23% of trades move the price within 10sec
- retail trades move the price more often than institutional trades (24% vs 19%)
After a trade ...
Price Impact: something happens after the trade
1. Flurry of cancellations
2. Flurry of aggressive orders
Are multi-market trades different?
Market Reaction to Trades: Price Impact by # traders vs #markets
60%
10%
25%
5%
Key Observation
- 86% of post-trade $-volume comes from 14 (of ~2,800) traderIDs
- these IDs trade 80% of their aggressive volume right after other traders
- \(\Rightarrow\) specialized strategy
After a trade ...
Price Impact depending on post-trade activities: fast traders?
Observations
Price Impact depending on post-trade activities
Big question: how do we interpret the last point?
fast traders create the appearance of a more informed trade and generate a larger price impact
fast traders are better at predicting that a trade is more informative and act while they can
only 20-25% of trades move the price
multi-market trades have larger price impact
but: seems that multi-trader (!) is key
trades that involve the fast traders (snipers?) have the largest price impact
\(\Rightarrow\) Ideally: experiment such that trade information content not affected but fast traders' sniping ability curtailed (or improved)
How do you make physical latency disappear?
Market A and B move to the same data centre: April 29, 2013
Key Tech Change
Physical Merging of Markets A and B
What happens when?
Algos, Routers, Venues
What happens when? Post Change
Algos, Routers, Venues
- easier for aggressive traders/algos to hit both markets simultaneously
\(\Rightarrow\) more trades hit both markets - MM "over-"posting becomes riskier
\(\Rightarrow\) reduced liquidity provision/quoting - less posted liquidity makes it harder for snipers to pounce
\(\Rightarrow\) less "sniping" activity
Our Premise
Disclaimer:
three other concurrent changes
- same tech platform
- A's market-making program got canceled (had only minor uptake and even if it matters, the canceling would help our premise)
- Fee change for a subset of orders that were practically no longer used
What happens when? Post Change
Behavior Variables: regression results
Did it become easier to hit both markets?
same ms before: 22% same ms after: 31%
single-trader, multi-market \(\nearrow\) 2%
Did liquidity decline?
bid-ask spread: No.
quoted depth: Yes
(+ went down on the lower-rebate venue)
Are "snipers" less active?
Yes. Decline 1-2%
What should happen to price impact?
Behavior Variables: regression results
investors unlikely to know about or mind the system change:
\(\Rightarrow\) no change to intrinsic price impact
if "snipers" move the price too much
\(\Rightarrow\) price impact without "snipers" should decline
What should happen to price impact?
Behavior Variables: regression results
Findings
- price impact drops marketwide for all trades
- price impact drops for non-sniped trades
- no change for sniped trades
- Multi-market trading: large % of trades & $-volume; used by unsophisticated
-
Price impact: Multi-market >> single market
-
But key factor: "other" traders, not # of markets
- Of "others", small group of "snipers" is key
- "Snipers":
- likely good at identifying informative trades
- but some indication that they make price impact larger than it would be without them (\(\Rightarrow\) excess price volatility)
Summary
Answers to research questions
What happens when?
Algos, Routers, Venues
investors
routers
brokers
trading venues
desk
execution algorithm
- Long literature, including
- Joel Hasbrouck (e.g., "One Security, Many Markets: Determining the Contributions to Price Discovery", JF 1995)
- O'Hara & Ye (JFE 2011): good for mkt quality
- Degryse, de Jong, van Kervel (2014): visible vs. dark fragmentation
- Bernales, Riarte, Sagade, Valenzuela, Westhei (2017): fragmentation without competition
- most closely related:
- van Kervel (RFS 2015): over-posting exists
- Baldauf & Mollner (WP 2015) (theory): splitting of liquidity across markets
- Brogaard, Riordan, Hendershott (JF 2020): HFT/limit orders generate price discovery even in absence trading.
Literature
Work on fragmented markets
Observations
Price Impact depending on post-trade activities
HFT-MM detect informative trade and move the price before they get run over: HFT-MM contribute to price discovery with limit orders
fast (sniping) traders run over the HFT-MMs are better at predicting when a trade is more informative and act while they can
MMs get run over by informative investor trades
bid-ask spread is MM compensation
bid-ask spreads can be low but quotes move fast
bid-ask spreads are slightly higher because of snipers
Kyle 1985/Glosten-Milgrom
Brogaard, Hendershott, Riordan (2020)
Budish, Crampton, Shin (2016)
fast (sniping) traders react quickly on the possibility of a future price movement and create noisier prices
not clear how this affects spreads and risk/returns
Social Cost
Literature
Yang & Zhu (2018) ("Backrunning")
traditional view
premise
HFT-MM are better MMs
HFT-MM and HFT-snipers interactions
HFT-snipers interactions with the market
Are multi-market trades different?
Market Reaction to Trades: Price Impact
Fun Facts:
- only 23% of trades move the price within 10sec
- retail trades move the price more often than institutional trades (24% vs 19%)
Are multi-market trades different?
Market Reaction to Trades: Price Impact by # traders vs #markets
64%
36%
85%
15%
Sniping in Fragmented Markets
By Andreas Park
Sniping in Fragmented Markets
Presentation for the 2020 WFA in San Francisco (held virtually, of course).
