Capturing The Research Advantage

Capturing the Research Advantage Inforrnation reduces uncertainty, but good exeottion is essential to investrnent

success.

Mark Edrvards and-Wayne H. Wagner

Berween the idea A -,.1 ,Lrr.v --^l;-, r!|lsL' rulu Berween the monon And the act Falls the Shadow T.S. Elioc, "The Hollow Men"

-

*adages

to eter Bernstein [1997] cites rwo captLrre the major lessons of capital market theory: "Nothing ventured, nothing gained," and "Don't put ali your eggs in one basket." The market rewards risk-cakers for invesung in uncertain sicuations, and diversificasion spreads the risk to

avoid disasters. What is the risk that Bernscein refers to? Capital market marhemaricians equate risk with statistical variabiliry. At best, this is an after-che-fact solurion to a

IvIARK EDWARDS the Plexus Group

(cA

in

rs

drrector

o[

Los Angeles

e0064).

WAYNE H. WAGNER is chairmrn of the Plexus Group rn Los Angeles.

1

8

CAPTUR]NG THI, RESEARCH,\DVAN'f AGE

betbre-the-fact problem. Robert Arnott [199a] by conuast adopts a more psychological vielv: "Il isn't risk if rt isn't unconfortable." Uncertainry about future outcomes creates discomfort, and dealing with dtscomfort and r.mpertect information is the essence of rcdvc invesrment managemenc. If the market conlLrms a decision, uncertai.nry is reduced. The decisions are then to impl.emenr. but potenrially expensive timely

-

-

IMPACT OF COMIVIISSION REDUCTION ON RISK Exhibits 1 and 2 summarize a srudy of potenrirl SPRING I999

EXHIBIT

effective stock selecrion.

1

COSTLESS RETIIRNS FROM

The first impulse of concerned managers is to identiflz excessive execution costs as the source of per-

SECTIzuTY SELECTION Round-Trip Buys

6-Wk. Decisron Retum

Absolute

I Hrrlrc

-

\Fll(l

+5.47% -2.21%

+3.20%

Market-Adjusted. +3.28% 4J2%

+3J6%

- Decision Retum 6-Wk. -

Seils

*Market-adjusted

return subcracts the market movement over the

srx weela post-decision.

EXHIBIT 2 COSTS TO CAPTIJRE RESEARCH ADVANTAGE

formance leakage. Sponsors and regulators have 3umped on this bandwagon, pressing managers, brokers, and exchanges to lower commissions and intraday impact. ln response, desks have mrnrmrzed these direct costs to where they now account for only 20% of the total shorrfall. Yet che performance gap persists.

TFIE "ICEBERG'' CFIARACTERISTIC OF TOTAL TRANSACTION COSTS The visible and hidden costs of trading resemble an iceberg. Ac the "tip of che iceberg" iie commissions

the smallest component. In order of

-obscuriry

increasing

lie rntraday brobut no iess punishing Round-Trip - and opporturury - interday tinring delays, ker impact, (Buys - Sell$ Buys Sel]s 'We represent this' idea in costs due to missed trades. Total Execution Cost -2.44% -4.74% -3.28% Exhibit 3. Absoiute Return Exhibit 4 decomposes the total shortfall (the ice+2.97% -2.95% Net of Costs -0.08% berg) for this study. The first column of numbers shows Market-Adjusred Return experienced costs for the traded/untraded shares. +0.84% 4.86% Net of Costs -0.01% Multiplying the raw numbers times the completion rate gives the weighted contribution to total cost. When traders and managers recoil from deciand actual rerurns over a six-week period. Six weeks is suficient to measure the short-term value of a firmt sions due to high commission and impact costs, capdecisions while accuraceJy assessing implementation tured returns drop as a direct result. In fact, the costs cost (see Wagner and Edwards [1998]). In this study we have not been avoided, buc merely slufted to the more measure decision value and costs from desired date of insidious cost categories of delay and opporrunity (the -W'e then com- shaded components in Exhibit 5). decision through the subsequent trades. Commissions account for only 5% and broker pare end-of-period price ro the price at the time of decision and to the actuai trade price, revealing both impact only 14% of cotal implementation costs. Togerher, delay and opporrurury costs accounc for more potentiaL and captured return. and sell 1 shows the potential buy Exhibit returns (excluding any trading costs) for over forry investrnent manager clients for the second and third EXHIBIT 3 IMPLEMENTATION COST "ICEBERG" quarters of 1997 . Exhibit 2 shows the captured rerurns for the same buys and sells, comparing decision price to net trade price. it shows that the cost of getting ideas into portfolios offsets the research advantage. 'We see chis pectern repeetedly each quercer.

These runaway execunon cos$ were largely ignored until the 1990s. All the star appeal in active management concerns stock selection (picllng winners) and not green eyeshade cost accountrng. Yet, with index funds grabbing a24o/o market share of 1997 pension equiry assets, managers are becoming aware that efici.ent cost control is as importrnt co net rerurn as SPR]NC I999

fHEJOURNAL OF PORTFOLTO M NACEMENT 19

EXT{IBIT

4

ACTUAI INTPLEMENTATION COSTS x Completion

Cost

Broker Impact Postponement Costs

Timing Delay Opportuniry

-l

t1

= Net Cost -R" hn "r

-R- hn 'r -?-l hn

Commission

Execurion Cosu

Rate

94%

hn

-106

bp

6%

-383 bp

-160 bp

Tocal Implementation Cost

than 80% of the shortfall. These costs are not due to trading badly; they are dLre to not trdding when it would but more cosdy to do so. be advantageous The raw opporturury cost is the forgone return for the unexecuted shares. The missed rrades would have returned 383 bp over six weeks had they been traded at the decision prices, whjle the shares acrually traded returned only 147 bp. Emphasizing low-cost brokerage worsens performance lvhen "safe" low-cost decisions are emphasized, while more expensive but more valuable decisions are deferred or abandoned. Expanding on Berruteint first saying, tailing co

adverse selection: an executed ser oftrades chat lags the returns of the desired trades. To wrn the active management game, then, the trader's 6rst goal nust be to to trade all the orders. place all the manager's bets 'W'hile decisions add value on auerdge, not every decision will be a winner. According to our date, only

"venture" cheam the porrfolio of gains, but aiso adds hidden costs, thus hurting performance in wvo ways.

feedback timrng strategi.es to emphasize those stock picks

THE PERSISTENCE OF PRICE MOVEMENT

suit of

55% of decisioru add vah-re. Professional money managers put great falth in their research processes, yet reaP benefit only when che overall market comes to agreemenc with rheir logic. To overconre the ftustration of markets thai are "out of sync," managers may implement positive price

for which the market is concurring. Traders, in conEast, are drawn to cheap liquidiry which leads to purnegariue pricefeedback sitr-rations.

This places man-

agers and traders ac cross-purposes.

This brings us to Bernstein's second concept: diversification. Managers make many investment deciyet neither manager nor trader knows a priori lvhich ideas are winners, even thoLlgh the overall seleccion process demonsrrabiy adds value. Trading only when the market provides low-cost execucions leads to si.ons,

The result is that many of the

manager's best decisions either become underweighted or are never whi.ch are executed. Meanwhile the weakesr ideas end up firlly executed, thereby the easiest to trade

-

dominating returns and leading to underperformance.

PRICE PERSISTENCE EXHIBIT 5 RELATIONSHIP AMONG IMPLEMENTATION COST COMPONENTS

We show that rimely manager decisions embody sufficient pri.ce persistence over a slx-week lvindow to validate a posirive feedback approach. Then rve shorv that market confirmacion srgnals can be used to help

manager and crader identify l,vhich decisrons to emphasize or co avoid. Finally, rve sho.,v that traders all too often fail to use chis confirming intbrmation, focusing instead on di.scovering cheap liquidiry and reducing trading costs. The result is an operational drrg on Perforrnence' posidve feedback sraregies rely on price persis-

2

0

c^pTuzuNc TllE

R-ESEARCH .\DVANTAGT

SPRING 1999

tence. Although we cannot argue in favor of technical models for stock-picking, we find that rinring overlays add value to fundamental stock selection. 'While large

the trade. Conversely, falling stocks rhat conrinue ro fall

trades impact stock pricing for a short period, inevitably prices reflect the composite of ail investors' judgments. Thus persistence reflects the flow of information as i.nvestors seek a valuation consensus. Over time, fundamentals have to win. But in the slx-week horizon used to analyze executions, trends dominate. Exhibit 6 summarizes a more detailed srudy of the same 1997 data.* We group trades according to strong, neutral, and weak price momenrum categories based on rwo-day price returns prior to decision. Buys rising or sells falling more than 4o/o ere iabeled strong, while buys falii.ng or sells rising more than 4Yo are labeled weak. We then calculate market-adjusted thirryday returns for each momentum group.

As a practical rool, market confirmacion can point managers and traders in the direction of the best deci.sions. Indeed, we find that momentum and growth managers often use market confirmation to add to existing orders. Thus, the trade desk plays a critical feedback roie in helping rhe managers decide which orders to pursue aggressively. The interesting parrs of Exhibit 6 are on rhe exlremes: the relative handful of trades that show strong or weak price signals. Managers wili want to pursue stocks showing price strength and deemphasize stocks with weak prices. As we will show, rrading problems arise in deairng wirh these extremes, noc in dealing with the bulk of the orders rhat show less volatile price movement. 'W'hy do these trends persisr? Managers often say that good news begets more good news. 'We believe a rnore likely explanation is rhat trading inrerest begets more trading interest. This is not an argurnent for momencum management. Rather, it is an encouragement for better manager-trader coordination. Tiaders can provide addicional information thar helps steer managers away from the least successfi:l stock selecrions.

The numbers in the firsr coiumn in Exhibir 6 show refurns and the percent of observarions &om the total set, based on these inirial momentum groupings. The strongest orders result in the best eventual returns, while the weakest orders continue to lag.

Next, we subdivide each momentum group based on price changes for the rwo days Jollowing the initial order. Buys into srrengrh rhat conrinued to rise ar Ieast 2Yo and sells inco weakness rhat dropped atleast2Yo are viewed zs confrmed. Reversions of greater than 2o/o are flagged ts reuersals. Al1 orher orders (che majoriry of the observations) are labeled as flat. We rhen calculare thrrry-day marker-adjusted rerurns on each subgroup. Managers Seek Positive Feedback

This study shows that the best performance, exclusive of inr.plementation cost, comes 6-om selecdng stocks for which the market currently appreciates the news. The odds of selecting winners increase when the

favorable

tend is confirmed in the wvo

days

following

show the weakest returns; these are the losing bets. (But they will be cheap ro rrade.)

Traders Seek Negative Price Feedback

How do rising buys and falling sells afi-ect the trading process? The evidence shows a conflict of interest. Whiie manasers are oriented toward ma.xim:zing returns, trading displays higher sensitiviry ro cosrs. Where managers seek positire price feedback, rrading displays a negatiue priceJeedback parrern.

Exhibit 7 shows the development of rerurns

as

large institutionai orders are worked for several days. The bars represent raw returns (rhe difference berween

EXHIBIT 6 THIRTY-DAY RETIIRNS FOR LARGE.CAP BUYS 30-Day Market-Adjusted Return (% obsewati.ons) Srrhcenrrent

Inidal N{omentum Buying rnro Srrength Neutral Momentum DUylng lnto Weakness SPRING 1999

Retum

(% obs.)

e.6% 2.1.%

-8.0%

(7%) (8670) (7%)

Price Tr'end Condicion 2 Davs After Order

Confirmed

17.1%

ss%

-rs.8%

(37%) (15%) (31%)

Revened

Flat 5.6%

2.7%

-4.8%

(41%) (67%) (47%)

4.s% -3.3% -3.8%

(22%) (17%)

(22%)

THEJOURNA]- OF PORIFOL/O MANAGEMENT

21

EX}IIBIT 7 DAY-TO-DAY TRADING EMPHASIZES CHEAP STOCKS 150 17q

^

100

8ru icu ;25

-o

tr10t5 f!

Tmded

Retum I

Unerecuted Share Retum

% freded

-

the price when the desk recerved rhe order and the closing price) for borh rraded shares (wtr-ice) and as-yec unexectrted shares (black) as days move from the release dace (0) to fifteen days posr-decision. The modesr rraderi return implies rhat rhe bulk of trading acriviry occurs in response to neutral or weak price movement, lvhile the much stronger tmexenfted share returns reflect reluctance to trade if the pocenrial cost is high. This pactern is represenrative of most desks: Orders with strong price moves are often deferred in hopes of price reversion, .,vhile rhe flat (cheap) trades are quickly execnred. Exhibir I shows rhe impact of rhis behavior on costs. As rhe stronger residual shares are execured, che

mareinal cost per sh:ire (black) more rhan quintupJ.es. Irorucally, these trades appear cheap in terms of inrraday market impacr by the dme they are execured, buic performance has been eroded by the hidden delay cosrs. A mrnager lve knolv once described his trader as someone who was "congerutally unable to step up."

EX}IIBIT

This telling statement sholvs horv dift-erenrly rnanagers and traders perceive risk. To a manager, rhe risk in a stock is the possibiliry of nor achieving expecred price growth based on fundamenrai research. The trade desk, holvever, gauges risk reladve to recent tradino r:rnqec a much shorter horizon. When i.nformarion flow is low, prj.ces often move r,vithin a narrolv band, and lead to lorv-cosr trading opportunities. Danger arises when prices move oucside normal ranges. When rrend takes over, trading needs to becorne more aggressive. The crader lvho delays compromises the most valuable decisiors. Unfortunately, rraders' all-coo-human terrdencies are abetted by sponsor pressure to dance to the cos[-minimization cune.

IMPACT OF MANAGEMENT STYLE ON TRA.DING COSTS Exhibic 9 shorvs that grorvth managers issue many more strength-oriented orders than value managers, 16% versus I0% for lrrge-cap and 23% versr.rs 1.4Yo for small-cap. Nore that value management often doesn't mean contrary even value managers ftequently seek price confirmarion. Value traders often believe they have much higher percentages oFweak orders, justifiTing a tendency to trade with minimal impact. Exhibir 9 relis a diff'erenr story for weakness-oriented orders. Value managers issue only a siighciy higher proporrion of conrrrry orders than gror,vth managers, 8% versus 7% for largecap mxnagers and 10% each for small-cap managers. Both groups are equally rempred ro rake advantage of price breaks co iruciace accion.

8

GROWING INCREMENTAI COSTS INCREASE

TOTAI

EX}IIBIT 9 \VHAT PERCENT OF ORDERS

COSTS

REFLECT MONIENTUI{?

SCVaI

i

SCGro LCVaI

LCGro

Erdo.dcd

Itudrcq -rlrdd

?,

CAPTURINC TH€ RISEARCH ADV,\NTAGE

SPRING 1999

Fxhibir 10 compares the trading and opporruniry costs for growth and value managers under conditions of buying strength (left columns) or buying weakness (right coiumns). 'We see that growth funds rrade strong orders aggressively, resulting in large impact but more modest trmrng cost. Value funds should show high timing and opportuniry coscs offset by minimal impact. Both growth and value desks show high costs in strength situarions, in contrast co much smaller trading gains in weak situations. Gains for borh sets are less than

half the size of the costs at the other end of the spectrum, again suggesring that traders of all stripes are drawn to rapid trading when Lrquidiry is cheap. Th c,rmmo',ry. the data ShOw that manager deCi-

ment ideas against 1) the uncercainry rhat the information is correct, and 2) the cosr of executing che trade. Managers, we observe, are very good at the first skill. Coming to terms with cost is the craders' chal.ienge. The cradert Job is to capture rhe value of rhe investment idea at minimal cost, balancing three addrtrve cost sources: 1) comm,rssion and price rmpact on crade compledons, 2) timing efrects generated from delayed trading, and 3) opporturury cosrs from abandoned trades. Traders are good at conrrolling visible costs, but excellence in the other rwo requires good communication with the manasers.

INCORPORATING TR,q,DING INF ORMATION INTO THE DECISION PROCESS

sions add value- However, much

and frequently all of the decision value is iost due- to lack of coordina-tion befween trader and manager. Our study shows chat price trends persist, and traders r,vho routinely shy away from high cost while seizing cheap liquidiry work counterproducrively to their managers. Irt been said many times before: The trend is your friend.

CAPTIIRING THE RESEARCH ADVANTAGE just that: acliue/y seeking and securing superior stocks, end actively removing from the porrfolio stocks wirh no particular appeal at current prices. Thus, buy and sell decisions are the drivers of active rerurn. Exhibit 2 shows that active decisions add value, but not enough to overcome the cost drag. Capturing the potential shown in Exhibit 1 is the challenge of active management. Implementation coscs nr-ake sense only when weighed against the benefir of enhanced performance. Active managers balance the expected value of investActiue investment management

EXHIBIT

is

10

GRO.WTH AND VALI-TE COST SOURCES

a

1oo

;30

slrcng

-zoo

J

re

Weak

-

Otde6

|

@

Grgwth ITiming SPRINC 1999

FIq '

g.* j

E

ode6

Vdw tr

Growth

lmpact

lopportuniq

The most ualuable cornmodity in the market is inJor-

mation that reduces uncertainty. Traders need to watch how che market responds ro rcti.vity. If the response is inconsistent with the manager's information, the manager may want to pursue more research. Thus, trading cost i,nformation is part of the research thrt gives a manager relative advantage. Failure to assess the motivation of the other market partrcipants is a primary source of active performance trouble. Acti.ve managers often fail to consider why they are able to trade some stocks easily, whrle orhers simply never get inro che porrfolio. Exhibit 1 shows that managers and models do the hard work idendfiTing stocks wirh better-rhan-average prospects. Talcrng only che easy trades lers the market become the "trade-don'r trade" differentiator. Thrs is running the process backward. Market dynamics ensure that trades are easiest when the gross research advantage is most negatrve. Thus, rfte hard trades are orten the least uncertain, but also the most proftable. The trader's prioriries musr be primanly ro capture the vaiue of the best ideas and secondarily to minimize cost. Traders who avoid impact ler rhe market rather than the manager choose rhe bem (see "How Tlading Can Distort Invesrmenr Selection" [1993]). No rrranager knows in advance whrch ideas will be the true

winners, but letting the opponen[s choose the bets makes no sense at all. The lesson: Managers must 6.rst measure and develop confidence in the value of their research, and then incorporate feedback from che market. Only managers who practice a disciplined implementation r--"n"ch L.ur rrn rrrin lino 5--.-,. mma appruaLrr wur rh. Lrrc S,CtIVe mana.gement CfaL,.-b THEJOURNAT OF PORTFOLTO M^NAGEMENT

23

Thus, we can add a third saying to Peter REFERENCES Bernstei.n\. Itt not, "A penny saved is a Penny earned,"

Robert' "Quantitative lvlanagement: Its Roll .in the because overzealous cost conrrol snufii out ,irty oppor- Arnott, Coming Decade.'' Financial Analysts Jouma!' May-June 1994' tunities. A better one is "'Tis many a rUp '**J."it*a

|ip." Good execurion is essential to investment success. Bemstein, peter.

Speech de|ivered

Conference, LPnl 2'

ENDNOTES

at rhe 6fth Plexus Group

1997 '

,,How Trading Can Distort Invesrment Selecrion."

Plexus

Grouo. 1993. "'We eliminate all small "cash" decisiors (less than 20,000

'Wayne H., end Mark Edwards. "lmplementing shares for large-cap [> 2.5 $bil] and less than 10,000 shares for Wagner, small-cap [<2.5 $bil]), and equal-weight the remaining decuions. Investmenr Srrateges: The Art and Science of Invesdng." In Frenk We analyze over 40,000 large-cap decisions and nearly 60,000 J. Fabozzi, ed., Handbook of PortJolio lvlanagemen| New Hope, PA: small-cap

24

decisioru.

CAPTURTNC TH-E RESEARCH .\DVANTAGE

FrankJ. Frbozzi Associates, 1998.

SPRINC I999