20.11.2012 Views

recent developments in high frequency financial ... - Index of

recent developments in high frequency financial ... - Index of

recent developments in high frequency financial ... - Index of

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

94<br />

market order distribution that changes with the state <strong>of</strong> the book. One only has to<br />

base the computation <strong>of</strong> the conditional upper tail expectation on a vector <strong>of</strong> state<br />

variables F, i.e. calculate bE½mm j Qþk;<br />

FŠ.<br />

For the purpose <strong>of</strong> this study we focus<br />

on the unconditional market order distribution and leave model<strong>in</strong>g the conditional<br />

market order distribution as a topic for further research.<br />

3.2.2 Average pr<strong>of</strong>it conditions<br />

S. Frey, J. Grammig<br />

To justify the marg<strong>in</strong>al zero expected pr<strong>of</strong>it assumption, one implicitly assumes a<br />

repetitive two phase trad<strong>in</strong>g process. In phase one, agents submit and cancel limit<br />

orders until the book is free <strong>of</strong> (expected) pr<strong>of</strong>it opportunities and no agent wants to<br />

submit, revise or cancel her order. Limit orders are sorted by price priority and,<br />

with<strong>in</strong> the same price tick, by time priority. When the book is such an equilibrium<br />

the order book should display no ‘holes’, i.e. zero volumes <strong>in</strong> between two price<br />

ticks. In phase two, a s<strong>in</strong>gle market order <strong>of</strong> a given size arrives and is executed<br />

aga<strong>in</strong>st the equilibrium order book. After this event we go back to phase one,<br />

dur<strong>in</strong>g which the book is replenished aga<strong>in</strong> until equilibrium is reached and another<br />

market order arrives and so forth. Can this be a reasonable description <strong>of</strong> a real<br />

world trad<strong>in</strong>g process? The descriptive statistics on the trad<strong>in</strong>g and order<br />

submission activity reported <strong>in</strong> Table 1 <strong>in</strong>dicate a dynamic trad<strong>in</strong>g environment.<br />

For a large stock, like Daimler Chrysler, we have on average over 3,000 trade<br />

events per day, about 19,000 submissions <strong>of</strong> limit orders, <strong>of</strong> which over 80% are<br />

canceled before execution. One could argue that such an active limit order trader<br />

behavior <strong>in</strong>dicates a thorough monitor<strong>in</strong>g <strong>of</strong> the book which elim<strong>in</strong>ates any pr<strong>of</strong>it<br />

opportunities. This is quite <strong>in</strong> l<strong>in</strong>e with the theoretical framework. However, with<br />

on average 10 seconds duration between trade events (for Daimler Chrysler) the<br />

time to reach the new equilibrium after a market order hits the book and before a<br />

new order arrives, seems a short span.<br />

The marg<strong>in</strong>al break even conditions can also be challenged by the follow<strong>in</strong>g<br />

reason<strong>in</strong>g. The conditions imply nonzero expected pr<strong>of</strong>its for limit order units that<br />

do not occupy the last position <strong>of</strong> the respective price ticks. On the other hand, this<br />

implies that the whole book <strong>of</strong>fers positive expected pr<strong>of</strong>its for traders act<strong>in</strong>g as<br />

market makers. If market mak<strong>in</strong>g provides nonzero expected pr<strong>of</strong>it opportunities,<br />

then this would attract new entrants and the competition between these would-be<br />

market makers ultimately elim<strong>in</strong>ate any pr<strong>of</strong>it opportunities.<br />

These considerations lead us to consider an alternative to the marg<strong>in</strong>al pr<strong>of</strong>it<br />

conditions which does not rely on the assumption that limit order traders<br />

immediately cancel or adjust all their orders which show negative expected pr<strong>of</strong>it<br />

on a marg<strong>in</strong>al unit, and that also acknowledges the effect <strong>of</strong> market maker<br />

competition on expected pr<strong>of</strong>its. For this purpose we reta<strong>in</strong> most <strong>of</strong> the<br />

assumptions <strong>of</strong> the Glosten/Såndas framework. However, <strong>in</strong>stead <strong>of</strong> evaluat<strong>in</strong>g<br />

the expected pr<strong>of</strong>it <strong>of</strong> the marg<strong>in</strong>al pr<strong>of</strong>it for the last unit at each quote k, we assume<br />

that the expected pr<strong>of</strong>it <strong>of</strong> the whole block <strong>of</strong> limit orders at any quote is zero. The<br />

marg<strong>in</strong>al zero pr<strong>of</strong>it condition is thus replaced by an ‘average zero pr<strong>of</strong>it<br />

condition’. This assumption allows to differentiate between two types <strong>of</strong> costs<br />

associated with the submission <strong>of</strong> a limit order, a fixed cost component, like order<br />

submission and surveillance costs, and marg<strong>in</strong>al costs (per share), like execution or<br />

clear<strong>in</strong>g fees and opportunity costs <strong>of</strong> market mak<strong>in</strong>g. In the appendix we show that

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!