hans degryse , university of leuven and center mark van achter , university of leuven
DESCRIPTION
Frontiers of Finance Bonaire, January 13-16, 2005. Dynamic Order Submission Strategies with Competition between a Dealer Market and a Crossing Network. Hans Degryse , University of Leuven and CentER Mark Van Achter , University of Leuven Gunther Wuyts University of Leuven. Motivation. - PowerPoint PPT PresentationTRANSCRIPT
Dynamic Order Submission Strategies with Competition between a Dealer
Market and a Crossing Network
Hans Degryse,University of Leuven and CentER
Mark Van Achter,University of Leuven
Gunther WuytsUniversity of Leuven
Frontiers of FinanceBonaire, January 13-16, 2005
Motivation
Recently: “new trading platforms” coexist with “traditional markets”
New trading platforms - Alternative trading systems:– Electronic Communication Network (ECN)– Crossing Network (CN): “a system that allows participants
to enter unpriced orders to buy and sell securities. Orders are crossed at a prespecified time at a price derived from another market.” (SEC (1998))
Motivation
Crossing Network:– Lower costs (no spread), Anonymity– Uncertain execution– No price discovery– Examples: Instinet Crossing Network, ITG Posit, E-Crossnet
"A survey of fund managers shows an expected 90% increase in crossing volume over the next two years"
Motivation
Goal of this paper:
Investigate impact of interaction of a batch-type CN and a continuous dealer market (DM)
on
the liquidity and order flow dynamics in both markets
Main Findings
DM caters to investors with high willingness to trade whereas CN to those with lower willingness to trade
Introduction of CN induces “order creation”
Even with random arrival of buyers and sellers and despite the absence of asymmetric information, systematic, non-random patterns in order flow arise
Outline
Related Literature & ContributionsSetup of the ModelEquilibrium– Markets in Isolation– Equilibrium: DM and CN
Empirical PredictionsDifferent Informational SettingsConcluding Remarks
Related Literature & Contributions
Static
Dynamic
One market Interaction CN
Parlour (RFS 1998)
a.o.
H&M (JF 2000)
Dönges et al. (2001)
Many papers
X
We construct a dynamic model analyzing the interaction between a CN and a DM
We add a CN to the dynamic models analyzing an individual trading system.
Related Literature & Contributions
Setup of the Model
Based on Parlour (RFS, 1998)2 days in the economyAgents decide upon consumption on both days:
is the subjective preference or typeAsset which pays out V units of C2 on day 2
Trading takes place during the first day, claims to the asset are exchanged for C1
Setup of the Model
Trading day:– Consists of 1,…,T periods– One agent arrives each period (= trader)– Traders are characterized by
• Trading orientation: Buyer or Seller (probability b and s)
• Type: Willingness to trade
– Traders choose between submitting an order to the DM, an order to the CN (both have order size = 1) or no order
– Orders cannot be modified or cancelled
Setup of the Model
Dealer Market:– One-tick market with ask A and bid B => A-B=1– Dealers stand ready to trade at these quotes
Crossing Network:– Orders are stored in book (b=buy, s=sell):
– Cross takes place at T– Price of the cross is midprice of quotes at DM
Setup of the Model
Orders in CN-book
|sell| buy
Setup of the Model
Orders in CN-book
|sell| buy
matched at T (time priority)
Setup of the Model
Informational Settings
Transparency
Partial Opaqueness
Complete Opaqueness
Setup of the Model
Informational Settings
Transparency full information benchmark case
Partial Opaqueness
Complete Opaqueness
Equilibrium: Solution Strategy
Determine cutoff values between order submission strategies, taking execution probabilities as given
These values are levels of β at which the trader at time t is indifferent between two specific strategies
Equilibrium: DM in Isolation
Equilibrium order submission strategies:
Equilibrium: CN in Isolation
Equilibrium order submission strategies:
Equilibrium: DM & CNEquilibrium order submission strategies for given probability p:
Equilibrium: DM & CN
The cutoff points are dynamic:
Buy side CN/DM:
Empirical Predictions
Do there exist systematic patterns in order flow ?
What is the effect of a DM or a CN order on future order flow ?
Empirical Predictions
Order flow after a DM order at time t:
“The direction of previous DM trades does not affect subsequent order flow”
Effect of a CN order at time t to order flow to CN/DM:
"CN buys are more likely to be preceded by CN sells compared to other orders: CN sells ‘invite’ CN buys“
“DM buys are more likely to be preceded by CN buys compared to other orders”
Different Informational Settings
Transparency : benchmark >< reality: CN order book = Opaque
Different Informational Settings: Complete Opaqueness
&Partial Opaqueness
Different Informational Settings
Under opaqueness, traders are unable to condition their strategies on CN order book information
Complexity of model increases tractable 2-period model to compare cases
Main ResultSystematic patterns in order flow for transparency and partial opaqueness, but not for complete opaqueness
Concluding Remarks
Dynamic model: interaction between a CN and a DM
Order creation due to introduction of CN
For transparency and partial opaqueness cases: even with random arrival of buyers and sellers and despite the absence of asymmetric information, systematic, non-random patterns in order flow arise
Results are robust to introduction of uncertainty
Uncertainty
We now introduce uncertainty and time variation in the value of the asset V
Assume Vt follows a random walk:
Dealers set each period At and Bt around Vt
Traders forecast the final value of the asset VT, and the price of the cross (AT+BT)/2
Uncertainty
New cut-off betas:
Uncertainty
Cutoff values more time dependent, and reflect also uncertainty about V
Using the new cutoff betas, propositions remain valid and systematic patterns in order flow still exist