Liquidity Provision and Market Making by HFTs

Transcription

1 Liquidity Provision and Market Making by HFTs Katya Malinova (UofT Economics) and Andreas Park (UTM Management and Rotman) October 18, 2015

2 Research Question: What do market-making HFTs do? Steps in the Analysis 1. Identify HFT and then market-making. 2. Characterize Behaviour: 2.1 How do they submit orders? 2.2 How do they react to trades? 2.3 How do they manage their inventory?

3 IIROC s regulatory data The Data and Sample All trades, orders, cancellations, quotes, etc. for all Canadian equity markets. Masked (consistently): market, broker, user-id. Our definition of a trader ID: broker ID + trader ID + account type. Idea: Treat, e.g., a liability trader s own account differently from the client accounts that s/he handles. Sample(s): 1. Classification and general order submissions: IIROC Highly Liquid TSX-listed stocks from Jan May 2013 (300+) 2. Reactions to trades and inventory management: TSX60-non-interlisted stocks March & May 2013 (17). 3. Also: General order submission behavior in TSXV-secs (2000+) and cross-asset class trading (e.g., stock to prefs, stock to debentures) for non-interlisteds (190+ pairs).

4 Identifying HFT Objective: Identify consistently without biasing towards trading behavior. E.g., order-to-trade or flat inventories would bias against certain behaviors. Possibly multiple IDs per trader/firm does not lend itself to inventories. Defining characteristic for HFT: Speed. Defining characteristic for market-making: post similar volume on both sides of the market. An aside: even if one knows the HFT companies, one should still treat everyone who shares defining characteristics as an HFT. If it looks like a duck [...] and quacks like a duck, then it probably is a duck. Speed criteria >85% of orders submitted <2ms of market events. order-to-cancel speed <250ms. fast reaction to first market-on-close imbalance publication at 15:40pm.

5 Identifying HFT Additionally: Overcome the multi-id problem by exploiting commonalities in key metrics (number of orders, trades, etc) in a cluster analysis to group similar IDs

6 Distribution of Types Note: in any given month, only about 80 HFT are active Distribution of Trader Types HFT MM buy side other retail

7 Trading Behavior Distribution of Transactions by Types Distribution of Value by Types active passive active passive HFT MM buy side retail others HFT MM buy side retail others

8 Order Submission Behavior: Aggressiveness Note: Data excludes unclassified TSX Stock Sample TSX Venture with HFT trading Distribution of Aggressive Orders by Types Distribution of Aggressive Orders by Types atbest better mktable worse atbest better mktable worse HFT MM buy side retail HFT MM buy side retail

9 Market-Maker Reaction to Trades Market makers don t want to be stuck with large inventories because of capital commitments; risk of adverse price movements. Market makers should/must worry about order-splitting algos. Move quotes/orders after trades. Complication: Markets are competitive not the only one quoting! Additional considerations: Stronger reaction for larger orders? Is order-splitting detectable? Do MMs become active and take out stale quotes? How do MMs manage their inventory if it leaves their comfort zone? How does inventory management affect markets?

10 Is there a reaction? Answer: Yes. Distribution of Opposite Size Cancelations per unit of time ms0002 ms0005 ms0010 ms0050 ms0100 ms0500 ms1000 buyside retail non MM HFT

11 Do they try to take out stale orders? Answer: Yes. Distribution of Same Side Aggressive Orders per unit of time ms0002 ms0005 ms0010 ms0050 ms0100 ms0500 ms1000 buyside retail non MM HFT

12 Do they react more to large orders? Cancellation Rate Aggressive Rate average opposite side cancelled volume Market makers cancellation vs. volume at order submission average same side aggressive volume Market makers aggressive orders vs. volume at order submission quantiles of vol quantiles of volume 2ms 5ms 10ms 2ms 5ms 10ms

13 Do they react stronger when the incoming trader has already traded a lot? Cancellation Rate Aggressive Rate average opposite side cancelled volume Market makers cancellation vs. trader inventory at order submission average same side aggressive volume Market makers aggressive orders vs. trader inventory at order submission quantiles of cumulative volume 2ms 5ms 10ms quantiles of cumulative volume 2ms 5ms 10ms

14 Inventory Management All traders inventories Market Markers Inventories

15 Order Imbalance All traders orders Market Markers orders

16 Do they manage their inventories? order imbalance cancel imbalance %lagged market *** *** maker inventory [-6.66] [-7.00] Observations 832, ,536

17 What does their inventory management do? Competition matters! time-weighted $-spread time-weighted bps spread midpoint range (= volatility) midpoint return %lagged MM *** ** 0.309** * inventory [-3.09] [-2.18] [2.46] [-1.77] Observations 832, , , ,536

18 Summary HFTs are classified as traders who react quickly to events in the market. Post aggressively, but make up a smaller percentage than their percentage of total orders. Very large factor in liquidity provision. Cancel quotes rapidly and take out stale quotes after trades. Stronger reaction the larger an incoming trader s position detectability? Actively manage their inventory with order submission behavior. Inventory management associated w. lower spreads & increased volatility.