How Big Is The Strike Zone?

Building

Belter

9rorrlt Perf

orm.rnGe

COMMENTARY 72

SEPTEMBER

2OO2

HOW BIG IS THE STRIKE ZONE? At the heart of Plexus Group's transaction cost measurement is the comparison of the "strike price" at the time when the trade order became actionable to the transaction prices received in the marketplace. The difference is the cost of implementing the order. In this spotlight, we discuss how Plexus establishes a strike price, and present some ideas on how to set more representative strike prices.

preponderance of buyers and sellers in the market: When buyers were dominant, the volume weighted price would likely be above the mid-

Strike Prices The financial management industry assigns great prominence to markei prrces. Yet we all realize that any historical price is no more than a snapshot or a flicker of a volatile, dynamic process.

Suppose a portfolio manager forwards an order to his trade desk on February 13th at forty seconds after 1:30 in the afternoon. The implementation shortfall methodology requires comparing price movements between this point of initiation (the strike price) and the ultimate execution prices. Should we accept the bid, the ask, or the last print (13:30:40 on 2113102) as the best benchmark? Did the portfolio manager really focus all of his intellectual power on that particular instant? ln contrast, would some appropriate sample of the concurrent stream of prices result in a more representative strike price? Years ago Plexus adopted a solution for this strike price question that applied a "representative time frame" approach. We knew - or could infer - the time at which the order was presented to the buyside trade desk, but we were concerned about the representativeness of these flicker prices. Our solution was to average prices surrounding the moment of order release. Somewhat arbitrarily, we chose a ten minute clock period during which the order was received on the trade desk. Our loqic was as follows: we minimize the effects of the prices bouncing of between bid and asked. 2. We selected the volume weighted price for this short time interval instead of the mid-spread price because would tend weigh the 1

. By using an average

it

to

spread price.

3.

Averaging dampens

the

pafticularly unstable

opening and closing prices.

4. The recording of time stamps into the

order management systems is often subject to small random delays. 5. Most investment managers base their selections on a threshold price, not on a specific momentary orice.

6. Organizing our databases on a ten minute time slice vastly accelerates data retrieval and reduces storage requirements.

While ten minutes felt "about right," not everybody buys into this logic. Some argue that ten minutes is too long an intervalwhen markets are moving rapidly. So we decided to re-examine our assumptions in the light of new data capabilities at Plexus. We looked into our new daily-updated databases of tick data on the 8000+ most widely traded stocks on the planet. We examined the dis-tribution of number of trades per day across all companies. We then expanded our investigation to focus on non-retail trades greater than 1000 shares or 5000 shares per trade.

The table shows the compiled information for the last trading day in June, 2001. Any of the first three columns can be read as a category. To illustrate, consider the last line in the table. One company (Cisco) recorded 52,455 trades that day, 874+ per minute, or on average one every seven-hundredths of a minute.

Suppose we were

to

rebuild our average price databases to include higher frequency data Elimi nating Retail Tradi ng "buckets." Instead of ten minute buckets, we could form one minute intervals on, say, every stock with One of the bothersome aspects of counting each more than '1000 trades per day. In the table, this and every trade is we overstate the liquidity on standard applies to the line describing the 1001-5000 companies such as the NASDAQ-Iisted Trades Per Day. 1000 trades per day implies a trade every 5 seconds on average, or 12.5 per minute.

Responsible statistical averaging requires

a

respectable sample size. Plexus distrusts any trading statistic based on less than ten observations. Thus twelve samples per minute, 5000 trades or more per day, corresponds to this lower time interval to build reasonable samples.

companies that command a great deal of retail interest but little trading in institutional size. To analyze the depth of institutional interest, we performed the same analysis, this time eliminating all trades less than 1000 shares.

The results clearly indicate how rapidly market liquidity thins out. Only nine stocks make the

twelve Traeies

per

lfraximum Trades

Average min:sec Between Trades

cutoff:

#of

Cum Pct

0.01

80:00

541

6.6

0.03

40:00

Cqt

r

1'1-50

0.13 0.25 1.25

8:00

rsia,+

2e.ri

328

4:00

934

+t.C

188

557

79.6 go.e

873

't280

365 297 36

231

Stocks

ittt

o.g

25.00

0;48 0:24 0:05 0:2.4

0001 -50000

125.00

0:0.48

qi

se.s gs.gs

cz4c3 (max)

874.25

0:0.068

1

100.0

10'1-500 501 -1 000 001

-5000

5001 -1 0000 1

NASDAQ NON.US

6-1 0

5'l-100

1

NYSE

Day

per Minute

2.50 12.50

870 odo 56

9S.3

The table shows us that exactly 99 stocks, I NYSE, 75 NASDAQ, and 15 non-US had 5000 trades or more per day. Only 12% of the institutional universe has more than 5000 trades per day.

91

407

65

2s3

38

1

005

199 12'

2

0

trades/min.

six

NYSE, three NASDAQ, and zero non-US. The names are Honey-

well (story stock of

the day), Exxon

Mobil, Citigroup, Nokia, Global Crossing (story stock) AOL

Time Warner, Intel, Oracle, and Cisco.)

The number

stocks with at 1000 trades drops from 785 stocks to 190. 1

of least

Our conclusion is thus even stronger than before. Focusing on only institutional sized trades over 1000 shares, there are only a handful of comCompanies with lower trading frequencies require a panies with sufficient daia to justify shoi-ter longer time interval (e.9. ten minutes or more) to gather enough data for averaging. However, averaging intervals. lf we set a standard of at least suppose that instead of cutting off at 5000 trades per 5000 trades per day to form rea-sonable day, we cut off at 1000 trades. In this sample there averages, 10 minute or longer trade intervals are are 785 stocks with more than 1000 trades per day; preferable for 99.9% of all stocks in our universe. 199 NYSE, 372 NASDAQ, and 214 non-US stocks. Against a standard of 1000 institutional trades per Even by this loosened standard, 90% of the stocks day, a minimum of a 10 minute interval is required trade so infrequently as to make an averaging for 9B.B% of the trading universe. interval less than ten minutes questionable. Reprint any porTion with credit given to:

Plexus News Plexus Group has been acquired by JPMorgan lnvestor Servlces. This relationship helps us to rapidly strengthen existing services and introduce new products, while maintaining our confidentiality, objectivity and trustworthiness.

''Iexug

9rorrp

50 W. OUmpic Blvd., i1900 Los Angeles, CA 90064 PH : 31 0. 31 2. 550 5 F AX : 3 1 0. 31 2. 5506 vwvw ple xusg roup.@m @ 2002 Plexus Group. lnc. 1 11