recent developments in high frequency financial ... - Index of
recent developments in high frequency financial ... - Index of
recent developments in high frequency financial ... - Index of
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Liquidity supply and adverse selection <strong>in</strong> a pure limit order book market 87<br />
Xetra/FSE faces some local, regional and <strong>in</strong>ternational competition for order<br />
flow. The FSE ma<strong>in</strong>ta<strong>in</strong>s a parallel floor trad<strong>in</strong>g system, which bears some<br />
similarities with the NYSE, and, like <strong>in</strong> the US, some regional exchanges<br />
participate <strong>in</strong> the hunt for liquidity. Furthermore, eleven out <strong>of</strong> the thirty stocks we<br />
analyze <strong>in</strong> our empirical study are also cross listed at the NYSE, as an ADR or,<br />
<strong>in</strong> the case <strong>of</strong> Daimler Chrysler, as a globally registered share. However, the<br />
electronic trad<strong>in</strong>g platform clearly dom<strong>in</strong>ates the regional and <strong>in</strong>ternational competitors<br />
<strong>in</strong> terms <strong>of</strong> market shares, at least for the blue chip stocks that we study <strong>in</strong><br />
the present paper.<br />
2.2 Data and descriptive analyses<br />
The Frankfurt Stock Exchange granted access to a database conta<strong>in</strong><strong>in</strong>g complete<br />
<strong>in</strong>formation about Xetra open order book events (entries, cancelations, revisions,<br />
expirations, partial-fills and full-fills <strong>of</strong> market, limit and iceberg orders) which<br />
occurred dur<strong>in</strong>g the first three months <strong>of</strong> 2004 (January, 2nd—March, 31st). The<br />
sample comprises the thirty German blue chip stocks constitut<strong>in</strong>g the DAX30<br />
<strong>in</strong>dex. Based on the event histories we perform a real time reconstruction <strong>of</strong> the<br />
order book sequences. Start<strong>in</strong>g from an <strong>in</strong>itial state <strong>of</strong> the order book (supplied by<br />
the exchange), we track each change <strong>in</strong> the order book implied by entry, partial or<br />
full fill, cancelation and expiration <strong>of</strong> market, limit and iceberg orders <strong>in</strong> order to<br />
re-construct the order book at each po<strong>in</strong>t <strong>in</strong> time. Our reconstruction procedure<br />
permits dist<strong>in</strong>guish<strong>in</strong>g the visible and the hidden part <strong>of</strong> the order book. The latter<br />
consists <strong>of</strong> the hidden part <strong>of</strong> the non-executed iceberg orders. To implement the<br />
empirical methodology outl<strong>in</strong>ed below, we take snapshots <strong>of</strong> the visible order book<br />
entries whenever a market order triggers an execution aga<strong>in</strong>st the book.<br />
Table 1 reports descriptive statistics <strong>of</strong> the cross section <strong>of</strong> stocks. The activity<br />
<strong>in</strong>dicators show an active market. Averaged across stocks, about 13,000 nonmarketable<br />
limit orders per stock are submitted each day. Among those, almost<br />
11,000 get canceled before execution. This <strong>in</strong>dicates that the limit order traders<br />
closely monitor the book for pr<strong>of</strong>it opportunities which is <strong>in</strong> fact one <strong>of</strong> the core<br />
assumptions <strong>of</strong> the underly<strong>in</strong>g theoretical model. The large trade sizes (on average<br />
over 40,000 euro per trade) <strong>in</strong>dicate that Xetra/FSE is a trad<strong>in</strong>g venue for<br />
<strong>in</strong>stitutional traders and not a retail market. Averaged across stocks, 2,100 trades<br />
are executed per day. Table 1 also reports average effective and realized spreads.<br />
Follow<strong>in</strong>g Huang and Stoll (1996) the average effective spread is computed by<br />
tak<strong>in</strong>g two times the absolute difference <strong>of</strong> the transaction price <strong>of</strong> a trade<br />
(computed as average price per share) and the prevail<strong>in</strong>g midquote and averag<strong>in</strong>g<br />
over all trades <strong>of</strong> a stock. Realized spreads are computed similarly, but <strong>in</strong>stead <strong>of</strong><br />
tak<strong>in</strong>g the prevail<strong>in</strong>g midquote, the midquote five m<strong>in</strong>utes after the trade is used. 7<br />
Note that <strong>in</strong> an open order book market like Xetra, there is no possibility to trade<br />
<strong>in</strong>side the bid-ask spread. Orders are either executed at the best quote or they walk<br />
up the book until they are completely filled. Table 1 shows that on average 15% <strong>of</strong><br />
the order volume walks up the book, i.e. part <strong>of</strong> the order is matched by stand<strong>in</strong>g<br />
limit orders beyond the best bid and ask. This implies that the effective spread<br />
is then, by def<strong>in</strong>ition, larger than or equal to the quoted spread. To ensure<br />
7 By choos<strong>in</strong>g a five m<strong>in</strong>utes lag we follow the previous literature, see e.g. SEC (2001).