Persistence Of Price Movement

9rorrID

ocToBER 1 998

COMMENTARY 56 The Persistence of Price Movement

Managers make many investment decisions, yet neither manager nor trader knows v'hich ideas are winners, even though the overall selection process clearly adcls value. Plexus data show: (l) timely manager decisions ernbody sutficient price persistence to validate a positive f-eedback approach; (2) market confirmation signals can help managel and trader identity which decisions to emphasize or avoid: (3) traders very often fail to use this confirmin-{ information, focusin-e instead on discovering cheup liquidity and reducing tradin-e costs. The result is an operational drag on performance that places mana-qers and traders at cross-purposes.

While large trades impact stock pricing tor a short period, inevitably price: follow the cornpositc valuation. Over time. funclamental s have to win. But in the 6-week horizon used to analyze executions, trends dominate. The table groups trades accordir-rg to strong, neutftrl. and weak price rnornentLrm cate-qories birsed on tuo-du1 price rnorement p_ipl: to decision. Buys rising or sells falling rnore than 4c/c were labeled s/rzrg while buys falling or sells risin-e more than 4ch were label.edv,eak. We then calculate market-adjusted 30-day returns for each momentum group. The numbers in the leti column show retums and the percent ofobservations on these initial momentum

groupings.The best decisions

turn out to be those where pre-tracle con-

ditions indicated strength.

30 Day

Price Trend Conditiotr (2) Days After Order

Initisl Momentum

( 7tt<

)

7a ( 86Vo

)

9.6ct

Ncatral Mafiefrt.rfi

2.1

Itr)iiftg into Weukt

es:

ConJinned

(Vc obs.)

Bu!ingiilto Streilgth

.8n ( 7'r

)

I7.l

(

.17

F

,,

5.5sc (tso/o 1.t.,1

MANAGEfrS SEEK PAyT IVE FEEDBACK

of from implementation cost. comes selecting stocks for which the market currentiy appreciates the rrer.n s. The odds ol'selectirre wittnet's

Tl-re best pertbrmance. exclusive

-1

I

(t'(

)

lat

Ileversed

s.6 o/c ( 1l sc

)

2.7qa ( 67c/a

)

-1.8

(

7t

17

a/e

)

4.5'/a (

--1.8 Vc (

)

22 E( )

displav hi-eher sensitivity to costs. While mrnescr'\ seek po:ilive plice f-eeclback, tradin-e displays a negative

plice lecdback petterrr.

1: f)ay to Day Trading Emphasizes Cheap Stocks

Figure

125

100

759

Converselv. falline stocks that contin ue to fall show the

pre-trade conditions indicated strength.tt

22 Sc )

-3.3% ( 17qa

increase wl-ren the 1-avorable trenci is confirmed two days following the trade.

"The best decisions turn oat to be those w.here

Nexi, wc sub-

llarket -Adju\lcd Relurn

o E

5UO f

u'eakest returns; rhese are the losing

i:':,i:.$

j|i,;:jil ffi

f,ua"o Return

I

L]nex. Share Return

e.r,,pru.,t.altoot. rnarket confimation cetn clirect

Fi-eure I shorvs the devclopment of returns as large institutional orders

manatgers and trad-

are worked tor several <1ay,s. The bars

clivided each rnomentum group based on price changes fbr the tvn,o dttys follov,ing the initial order.

ers toward the best decisions. lndeed. we find thzrt morl-rentun-i and -trowth managers ofien use market confirrra-

Buys into strength thart continued to rise at least2a/c and buys into weakness that dropped at least 2c/o were viewed as confirmed. Reversions of greater than 2c/o were fla-eged as reversals. All other orders (the majority of the observations) were labeled as flat. We then calculate 30day market-adjusted retllms on each

tion to add on to existin-e orders. Why do these trends persist'? Mana-eers often sery that good news begets more good news. We believe a more likely explanation is trading interest begets more trading interest. .........WH I IE TBADEBS SEEK N EGATIVE PRI C E FEED BAC K

sub-eroup.

How do rising buys and falling sells affect the trading process? The evidence shows interests in conflict. While mana-qers are oriented towards maximizing returns, traders

repl'esent raw retllrns (the diff'erence betu,ecn the price when the desk re-

ceived the order ancl the closing price) for both traded shares (blue) and as-yet unexecuted shares (pink) as days move fiom the release date (0) to I-5 days atler decision. The modest traded return implies that the bulk ol'tlarling uctivity oe e urs in re-

sponse to neutral or weak price rnovement, while the much stronger unexecuted share returns ref'lect reluctance to trade if the cost is high. This pattern is representative of most

desks orders willr strong price moves are often deterred in hopes of plice reversion, while the f-lat (cheap) trades are quickly executed.

Figure 2. But Growing Incremental Costs Increase Total Costs

Figure 3. Growth & Value Cost Sources

The News! Weak Orders

100

Product Releases: We're updating our Website, adding more information and creating a platfonn for the de-

'100

6 o o

O

;-

tuv

F -200

300

ffi

Toral Des<

Cost

I

tncrenenral

Cosr ssooTradeo

I

Timing ffi lmpact mopportunity

on pending transactions. The

Figure 2 shows the results of this behavior on costs. As the stronger residual shares are executed, the marginal cost per share more than quintuples. Ironically, these trades appear cheap in terms of intra-day market impact by the time they are executed. However, performance has been eroded by the hidden delay costs.

In fact, both Growth and Value desks

A manager we know once described his trader as someone who was

To summarize, the data show that manager decisions add value. However, much and frequently all decision value is lost due to lack of coordination between trader and manager. Our study shows that price trends persist and traders who routinely shy away from high cost while seizing cheap liquidity work

"congenitally unable to step up." This telling statement shows how differently managers and traders perceive risk. To a manager, the risk in a stock is the possibility of not achieving expected price growth based on fundamental research. In contrast, the trade desk gauges risk relative to recent trading ranges - a much shorter horizon. When trend

livery ol some exciting interactive Plexus services currently under development. Our first interactive product, Instant PAEG/L, is a realtime facility fbr investors to estimate equity trading costs

show high costs

in strength

situations, accompanied by much smaller

trading gains in weak situations. Gains for both sets are less than half the size of the costs at the other end of the spectrum. again suggesting that traders of all stripes are drawn to rapid trading when liquidity is cheap.

counterproductively to their managers. It's been said many times before: the trend is vour friend.

data answers the queslion: "Baserl on histot'ical crpet'ience of c:omparable trades, v,hat shoultl this trade cost?" Plexus Additions

& Promotions: We welcome Matt Anderson and Douglas McRae to our

Systems Development team,

Uncomins Conferences: Security Traders As:oc i ation October 3-6,.Boca Raton: " Irttermediaries: Here Tr.tday, Gone Tontorr-on'.'" (David Hall)

AIMR Best Execution October 13, Philadelphia: " M ea sttri n g Transttc (Wayne Wagner)

ti

c.t

rt C osts"

IIR Best Solutions fbr Best Executlon

takes over, trading needs to become

November 9-10 NYC:

more aggfessive. The trader who delays compromises the most valu-

(Wayne Wagner)

IMN Superbowl

ol- Indexing December 6-9. Palm Springs " Managin g Trading C osts

able decisions.

;

A

Figure 3 compares the trading and opportunity costs for growth and value managers under conditions of strength (left columns) or weakness (right columns.) Growth funds trade strong orders aggressively, resulting in large impact but modest timing

cost. By contrast, Value funds should show minimal impact and high timing costs offsetby large timing gains in weak orders.

but's i,le'.tc

IIs

idc

i

tt r t'

r'acr i ve

Reprint any portion of this material with credit given to:

11lexrrs grorrp

Blvd., Suite 900 Los Angeles, CA 90064.

11150 West Olympic

Also available at http://www.plexusgroup.com' or contact us at

PH: 310.312.5505 FAX: 310.312.5506