An Empirical Analysis of 130/30Strategies: Domestic and Inter-national 130/30 Strategies AddValue Over Long-Only StrategiesGORDON JOHNSON, SHANNON ERICSON, AND VIKRAM SRIMURTHY

GORDON JOHNSON

is a senior portfolio man-jgcr, incermitional invcst-meius, it Lcc MundtrCapital Group in Boston.MA.gjohnsnn^ Iceniundcr.com

S H A N N O N E H J C S O N

IS a porttulio nuinat^er,

international investments, at

Lee Munder Capital Groupin IJosron. MA.

sericson@leemunder.com

V I K R A M S R I M U R T H Y

is a pt)rtfoli<t manager,

international investments. Jt

l-oc MmultT Capital Gronp

in Boston. MA.

vsrimurthy @'leemurder.com

There is growing interest in the mar-

ketplace for enhanced active equity

strategies as investors search for

higher alpha in an era of lower

e.xpected returns. Evidence ot jus t how pop -

ular these strategies have become is apparent

in the tremendous growth in assets over the

last year. Investment industry periodicals esti-

mate that as.sets in these strategies may be as

high as $60 billion by mid 2007 and could

reach S5()0 billion wi th in five years. ' An

enhanced active strategy partially removes the

long-only constraint, al lowing managers to

short Li portion of their portfolio to enhance

active returns. Enhanced active strategies, also

referred to as constrained loti^'^-sliort, active exten-

sion, and short enabled, are often identified bytheir long and short target weights. Forexample, a portfolio that can short 20% and golong an additional 20% for a total of 120%long is known as a 120/20 portfolio. Similarly,a portfolio that can short 30% is called a130/30, and so on. Although the percentagethat can be sold short typically ranges from20% to 50%, 30% seems to be emerging as themost widely used in the industry. Conse-quently, in this article, we focus specifically onthe 130/30 variation.

A traditional long-only strategy invests100% in securities and is managed to a bench-mark. Because a long-only strategy is prohib-ited from shorting, the most a portfoliomanager can underweight a stock is its weight

in the benchmark. With a 130/30 strategy,100%! is invested in equities and the manageris allowed to short up to 30%) of the portfoho.Proceeds from, the shorts are then used to pur-chase 30%) of additional securities, so that thenet market exposure of the portfoUo remains100% as in a long-only portfolio." The abilityto short allows managers to underweiglit unat-tractive stocks in a more meaningful way thatmore accurately refiects negative return expec-tations. In contrast to a long-only strategv', anenhanced acti\'e sn'ategy levels the playing field,allowing managers to overweight and under-weight positions by the same magnitude. Evenmodest amounts ot shorting can dramaticallyimprove performance results since a managersexpectations for outperformance as well asunder performance are now better representedin the portfolio. In fact, not only does anenhanced active strategy increase expectedreturn, it is achieved without dramaticallyincreasing active risk.

Conversations with Wall Street firms indi-cate that between 60% and 80% of the 130/30strategies currently in the miirketplice aa' qu.m-titatively run. Tliis is largely because such strate-gies are particularly well suited to quantitativemanagement. Quantitative models rank stocksfi-om best to worst on a variety of factors. Thelower tail of unattractive stocks is .ilre.idy beingidentified as part of a quantitative long-onlyprocess. However, the inforni:idoii in the lowertail is not flilly incorporated into the portfolio.

2(X17 Ti-iE J Q U I I N A I OF ALTERNATIVE iNvrsTMtiN i^ 3 1

Removing the long-only constraint allows a manager tounderweight the lower tail of stocks by amounts that arecomparable to portfolio overweights in the upper tail. Inaddition, quantitative managers are often heavy usei-s of riskmodels and portfolio construction tools. Thus, they arealready well equipped to run this type of strategy in ahiglily risk controlled manner. Conversely, traditional man-agers may require additional analysts to identify short rec-ommendations or may need to revise their investmentphilosophy or existing tools to look for short opportuni-ties. This article focuses on 130/30 portfolios from thequantitative managers perspective, as quantitative man-agers dominate the 130/30 marketplace and the strategylends itself weU to this style of management.

In this article, we provide further evidence that anenhanced active strategy can add value over a long-onlystrategy, but examine it from an empirical perspective.We develop a quantitative alpha (stock selection) modeland use it to test historical performance over a 13-yearperiod for 130/30 and long-only large-cap strategies inboth domestic and international universes. The 130/30strategics, using the illustrative alpha model, substantiallyoutperform their respective long-only portfolios. Theportfolio tracking errors also increase versus their respec-tive benchmarks but proportionally by much smalleramounts.

Finally, we provide details on calculating perfor-mance attribution for the long and short portions of a130/30 strategy given that no industry norms have yetevolved. One of the key elements of our perspective onperformance attribution is based on the belief that a130/30 strategy is more similar to a long-only strategythan a hedge fund strategy. Thus, its performance shouldbe measured and evaluated versus the benchmark againstwhich it is managed. We compute long and short per-formance attribution for the 13-year historical 130/30tests. We find that both the longs and shorts contributeto perform.mce relative to the benchmark in most of theyears of the test period.

PREVIOUS EVIDENCE ON THE BENEFITSFROM RELAXING THE LONG-ONLYCONSTRAINT

Grinold and Kahn [2000] show that a portfolio'sefficiency, measured by the information ratio (IR),increases wben the long-only constraint is removed.They fmd that fully leveraged long-short strategies offer

the most improvement over long-only strategies whenthe universe of assets is large, the asset volatility is low,or the strategy has high active risk. Their analysis alsohighlights that the long-only constraint induces negativesize bias because the manager can only underweightlarge-cap stocks by a meaningful amount. This size biasatlects active long as well as short positions. They con-clude that the most important argument for long-shortinvesting is the enhanced portfolio efficiency that resultsfrom being able to short stocks where the manager hasnegative views.

Clarke, de Silva, and Sapra [2{)04| evaluate infor-mation loss related to portfolio constraints relative to thebenchmark on: market capitalization, industry, sector,and stock positions—as well as tbe long-only constraint.Tliey measure the impact on performance from relaxingeach portfolio constraint. They measure the impact bythe change in portfolio transfer coefficient, where thetransfer coefficient is the degree of information transferfrom a security-ranking signal into active portfolioweights.-* Their analysis indicates that the long-only con-straint is the most significant in terms of information loss.In addition, they find that 130/30 strategies achieve 9t)%iof the transfer coefficient that can be obtained with afully unconstrained 200/100 strategy—where a 200/100strategy is 200% long, 100%. short, and has 100%, netexposure to the market. Thus, much of the benefits ofrelaxing the long-only constraint can be achieved withmoderate levels of shorting. They note that, in practice,the investor will need to make trade-offs between thetracking error, the level of shorting, and the improve-ment in transfer coefficient as not all can be controlledsimultaneously.

Sorensen, Hua, and Qian |20()71 examine the addedcosts associated with running constrained long-short port-fohos, including leverage costs related to borrowing sharesand higher transaction costs generated from increasedturnover. They conclude that the enhanced performancethat can be obtained from these strategies outweighs theadded costs under reasonable assumptions. They note thatportfolio mandates will have different tracking error tar-gets and benchmarks, which suggest that the optimal IRcan be quite varied, depending on the mandate and theassociated costs of implementation. Jacobs and Levy [2006]illustrate the mechanics of the prime brokerage structureunderlying 130/30 strategies as well as operational issuesand portfolio construction.

32 AN EMPIR[I:AI. ANALYSIS OF 130/30 STUATECI FAU. 2(107

HOW A 130/30 PORTFOLIO WORKS

Exhibit 1 gives an inustration of the mechanics ofa130/30 strategy. For a $1UO iavestinent, the long-onlymanager purchases $ 1 Ot) worth of stocks that are expectedto outperform based on an alpha model. The net invest-ment is SlUf), the portfolio beta is at or close to 1.0 andthe market exposure is 100%. For a 130/30 strategy, themanager purchases $100 of stocks expected to outper-form, but then shorts S30 worth ofstocks that are expectedto underpertbrm. The proceeds from the shorts are usedto purchase $30 ot additional stocks that are expected tooutperform. The net investment is still $100, the betaremains at or close to 1.0 and the net market exposure is100%. The gross market exposure, however, is now 160%(130% long + 30% short). Despite this higher gross invest-ment of 160'X), the 130/30 portfolio maintains portfolioand risk characteristics similar to those ofa long-onlyportfolio. This is of particukr interest to investors since a130/30 strategy can be used in an overall plan allocationmuch like a long-only strategy because it is managed toa benchmark and provides 100% market exposure.

WHY THE ABILITY TO SHORT IS SO CRITICAL

One of the key benefits of the 130/30 strategy isthe ability to short securities in order to underweight

them by a meaningful amount. In a long-only strategy, themanager is not able to short stocks so the only way tounderweight a stock versus the benchmark is not to holdit. This is an important point because the typical bench-mark weight for a given stock is generally quite small, asshown in Exhibit 2. The columns labeled "All" in theexliibit show the counts of all tlie stocks in the Russel!1000 and MSCl EAFE mdices by GICS sector. Thecolumns to the right of the totals are the counts of stocksin each bencltniark by sector that are greater than 0.25%and greater than 0.50%. For the Russell 1000 and MSCIEAFE indices, respectively, oniy 37 out of 987 and 35out of 1174 stocks have weights in excess of 0.50%. Hence,a long-only manager's ability to underweight stocks bymore than 0.50% is limited to just 5% of the stocks ineither index. In addition, some sectors have fewer stockswith large benchmark weights than others. Therefore,the long-only manager's ability to underweiglit securitieswill ako vary across sectors depending on the stock weightsin that sector.

Exhibit 3 further delineates the benchmark data byshowing weights for stocks in a particular sector, sortedby alpha into quintiles. The sector chosen, for illustrativepurposes, is the Russell 1000 Materials Sector. Stockswith lower alphas (Quintile-5) are the ones that a man-ager would most want to underweight because they areexpected to underperform the most. Stocks that a

E X H I B I T 1How a 130/30 PortfloHo Works

Buy attractivestocks

Buy additionalattractive stocks

Short unattractivestocks

130/30 Portfolio hasmore exposure to

alpha model

-40%-Long-Only Portfolio 130/30 Portfolio Net Exposure

FALT, 2007 THEJOUKNAL OF ALTKRNAnvt INVESTMENTS 3 3

E X H I B I T 2Benchmark Counts by Sector

All Wt>0.25% Wt>0.50%

Energy

Materials

Industrials

Consumer Discretionary

Consumer Staples

Health Care

Financials

Information Technology

Telecommunication Services

Utilities

RIOOO

67

60

12217053

102

204

13219

58

EAFE

41

116

245

213

90

55

240

98

3244

RIOOO

4

2887

12

18

14

42

EAFE

6

72786

32

5

68

RIOOO

3

0

1456

9

630

EAFE

52

11

1

5

16

12I

Total 987 1174 79 87 37 35

llie Russell 1000 (R WOO) and EAFE benchmaHes are as oJDecemlyer 1, 2006. Counts are the numher of total stocks (Ail) in each sector as iivU as thosewall benchmark weights greater than 0.25% and 0.50%.

manager would most want to overweight have the highestalphas (Quintite-1) and are expected to outperform themost. Stocks with the largest benchmark weights inQuintile-5 have weights of ().2(")%, 0.13%, and I"). 14%. Notowning those three stocks results in very modest under-weights equal to their benchmark weights. Not owningany of the stocks in QuintiIe-5 results in a cumulativeunderweight of—0.77% to Quintile-5 stocks, which isunlikely to have mucli impact on portfolio pertbrnuuice.

This example shows that the long-only manager's abiUtyto underweight stocks in a particular sector depends notonly on whether there are large stock weights in thatsector but also where those stocks rank on alpha. Hence,the long-only managers ability to reflect negative returnexpectations by substantially underweighting poorlyranked stocks in the portfolio is severely restricted.

Exhibit 4 continues the Russell 1000 MaterialsSector example, except that it compares active weights

E X H I B I T 3Russell 1000 Materials Sector Weights by Alpha Level

0.40

gO.3O

0.20

0.10

0.00

Quintilo-S Quint ilc-4 QuintUe-3

[••HIMULB.1 •U L

Quintile-2 (Juinlile-l

L l l l l l . • • • l l i_ l . l l l l . . l lLower

Alpha• Higher

34 AN EMPIRICAL ANALYSIS OF 130/30 STRATKGIES FALL 2007

E X H I B I T 4Russell 1000 Materials Sector Active Weights

1.25100

y 0.50

? •̂•̂ ^> 0.00u.£ -0.25 ^5 -0-50

-0.75-1.00-1.25

1.25LOO

^0.75S? 0.50^0.25>0.00•5-0.25^-0.50

-0.75-! 00

-1.25

Long-Only Portfolio

Ouintile-5

- - • | • - - - •

Quintile-4

- "- • " •

Quinttle-3 QuintiIe-2

•

Quintile-1

130/30 Portfolio

QuintiIe-5 Quintile^ Qumtile-3 Qtiinttle-2 Quintile-1

HillLower - ^ Higher

Alpha

tor a long-only portfolio versus a 130/30 portfolio. Thelong-only portfolio has a total of 110 holdings, resultingin typical individual active weights of 0.80% for stocksheld. For the Materials Sector, active stock weights of thisniiigiiitude amount to four holdings, as shown in the upperpanel, by active weights of the four highest ranked stocks.Clearly, the positive active weights for the stocks held inQuintile-1 dwarf the negative active weights that can beachieved from the stocks not held in Quintile-5 in a long-only strategy.

Active weights for the Materials Sector of an equiv-alent 130/30 portfolio are shown in the lower panel. Theworst ranked stock is shorted, resulting in an underweightposition of—0.82%. The proceeds are used to purchasean additional long holding. The two new positions areshaded for easier identification. In this case, the manageris able to increase the portfolio's exposure fix>m a cumu-Litive underweight of-0.77% of Quintile-5 stocks notheld in the long-only portfolio to —1.59% cumulativeunderweight in the 130/30 portfolio. Even though onlyone stock has been shorted in this particular sector, theexposure to the bottom tail of the alpha model has been

doubled. By employing Uiis strategy; poorly ranked stockshave the opportunity to contribute more to performanceif they underperform because they make up a greaterproportion of the entire portfolio. On the long side, addi-tional attractive stocks can be purchased for a gross invest-ment of 130%, which provides greater diversificationamong top ranked stocks and more exposure to the alphamodel.

HISTORICALLY 130/30 STRATEGIESOUTPERFORM EQUIVALENT LONG-ONLYSTRATEGIES

Of considerable interest to investors is the magni-tude of outperformance that can be expected from a quan-titatively-based 130/30 portfolio over a comparablelong-only portfolio. In this section, we present resultsfrom historical back tests for long-only and 13()/3t) port-folios in the large-cap domestic and international uni-verses. An alpha model with six widely used factors isused for the portfolio back tests. Full details and perfor-mance for this alpha model are in Appendix A.

FALL 2007 THE JOURNAL OF ALTERNATIVE INVESTMENTS 3 5

Back tests cover the period from January 1. 1994,through December 31, 2006. Returns are before fees ortransaction costs."* This period includes the Technologyand Telecom bubble of the late 1990s and rhe subsequentburst in nnd-2(l()(J. This should not hinder our conclu-sion.s, since there are a reasonable number of years includedin the sample period before and after the bubble. In addi-tion, this time period provides a good test of 130/30strategies and the effectiveness of shorting during a marketbubble and its aftermath.

Strategies are rebalanced at the end of each monthwithin their respective domestic and international uni-verses. The strategy returns are compared to those of theactual Russell 1000 and MSCI EAFE benchmarks. For the130/30 portfolios, weights are targeted to be 130% longand 30% short and are rebalanced back to those weightsat the end of each month. Sector, industry, and countryactive weights are all kept within 5% of their respectiveuniverses. Predicted beta is targeted to be as close to 1.0as possible, relative to the respective universes. Strategiestested are for reasonably active portfolios with the long-only portfolios holding an average of 110 stocks and130/30 porttbHos holding approximately 143 long posi-tions and 42 short positions.

Long positions are selected from the top quintile ofstocks within each GICS sector. A sell discipline is usedthat requires a stock to be liquidated when it falls to thefourth or fifth quintile and no longer ranks well versus itsGICS sector peers. Proceeds from the sale are rotated upinto a Quintile-1 stock in that sector that is not alreadyheld. Shorts are selected from the lowest ranked stocks,or the bottom quintile of each sector. The cover disciplinerequires that a short position be covered when the stock'salpha rank imprtwes versus its peers and ranks in the secondquintile or above, or when the stock experiences strongup performance in excess of+20% since being shorted.Proceeds from the cover are used to initiate another shortposition among the lowest ranked stocks in the sector notalready held short in the portfoUo.

Exhibit 5 shows cumulative returns for the Russell1000 Index, the MSCI EAFE Index, and the corre-sponding long-only and 130/30 domestic and interna-tional strategies. For both asset classes, the long-only and130/30 portfolios outperform their respective bench-niarks over the 13-year test period. In both cases, thedomestic and international 130/30 portfolios outperformtheir respective long-only portfolios by a considerablemargin. These results provide additional evidence that

allowing a reasonable amount of shorting and reinvestingthose proceeds in additional long positions can substan-tially enhance investment returns. The results also indi-cate that an enhanced active strategy works well in aquantitative framework.

Exhibit 6 shows annualized excess return and var-ious risk and performance summary statistics for thedomestic and international long-only and 130/30 strate-gies relative to their beiichmarks. The domestic and inter-national long-only strategies generate similar annualizedactive returns of 7.6% and 7.3%, respectively. The 130/30strategies also produce similar active returns of approxi-mately 11%. which is roughly 1.5x the return of theircorresponding long-only strategies and is shown as a ratio.The proportionate increase in risk of the 130/30 strategyis substantially less. Risk, as measured by tracking error,for the 130/30 strategies is only 1.2x for the domesticand l.lx for the international, relative to the corre-sponding long only portfolio tracking errors. The infor-mation ratio, a measure of the excess return relative toactive risk, also moderately increases—t.2x for thedomestic 130/30 and 1.3x for the international 130/30.This exhibit shows that the 130/30 strategy in bothdomestic and international asset classes provides consid-erably higher return without substantial added risk. Thebatting average column shows the percentage of monthsthat the strategy outpei-toriiis its benchmark. For both thedomestic and international 130/30's, the batting averageremains very close to that of the long-only strategies. Theturnover column represents annualized turnover and showsthat a 130/30 strategy generates roughly double theturnover of a long-only strategy.

PERFORMANCE ATTRIBUTION

Since 130/30 strategies are new to the market, stan-dards have not yet been developed for performance attri-bution. For a long-only strategy, the portfolio is managedto a benchmark, so clearly the portfolio s performancesiiould be evaluated versus its benchmark. Regardless ofwhether the absolute R'turn for the portfolio is positiveor negative, if the relative performance is positive, theportfolio is considered to be outpertbrming. For an equityhedge fund, the universe of stocks that the fund can investis often broad and undefined. Therefore, hedge funds aregenerally not evaluated versus an equity benchmark andabsolute return, rather than relative return, is most rele-vant when measuring performance.

36 AN F.MI'IRK:.U ANALYSIS or I.10/.30 STRATEGIES FAIL 2007

E X H I B I T 5Cumulative Returns of 130/30 and Long-Only Strategies

1,400

^ 1,200aI 1,000« 800>

600 ^

400

200

0

I

Dom 130/30

Dom Long-Only

Russell 1000

'94 '95 '96 '97 '98 '99 '00 '01 '02 '03 '04 '05 '06

1,000

800

600

•Inti 130/30

Int! Long-Only

•EAFE

'94 '95 '96 '97 '98 '99 '00 '01 '02 '03 '04 '05 '06

We argue that a 130/30 strategy is more like a long-only strategy because it is managed to a benchmark andlias 100% exposure to the market. Therefore, a 130/30strategy's performance should be evaluated similar to along-only strategy and compared to its benchmark. Sincethe strategy incorporates some amount of shorting,investors will want to know how the long and short por-tions ofa 130/30 portfolio contribute to overall perfor-mance versus the benchmark. To measure this properly,the performance attribution between the longs and shortsneeds to account for the fact that the long and short por-tions of the portfolio are 130% and 30%, respectively, ofinvested capital. It is also important to understand that ifthe long and short portions produce the same excessreturns versus the benchmark, the active contributionfrom the shorts will always be less than the longs becauseit represents only 30% of the portfolio rather than 130%tor the longs. Thus, the contribution to the overall port-folio's excess return fix>m the long positions would be

their excess return multiplied by 1.3. Similarly, the con-tribution to the overall portfolio's excess return from theshort positions would be the nc^iuii'i' of their excess return(which we calculate as the benchmark return iniiius thereturn from the shorts) nuiltiplicd by 0.3. The point isthat the long and short portions ofa 130/30 portfolioshould be evaluated versus the benchmark like a long-only portfolio, not on an ab.solute basis as for a hedgefund. However, a slight modification is required to scalethe performance of the long and short portions of the130/30 strategy by their proportions of invested capital.

We will use the domestic 130/30 strategy resultsfrom Exhibit 7 to illustrate how to calculate the perfor-mance contributions for the long and short portions ofthe portfolio. The domestic 130/30 strategy returns acompounded average of 22.3% for the entire 13-ycarperiod versus the Russell 1000 return of 11.0%. There-tore, the strategy outperforms the benchmark by 11.3%over the period.

FAI.1. 2007 TunJoURNAt Of AiTtRNATlVE lNV£.'.TM£l>rrS 3 7

E X H I B I T 6

Strategy Perfonnance Summary

Domestic

Long-Only

130/30

Ratios

International

Long-Only

130/30

Ratios

HoldingsLong

no143

in143

Short

42

41

Avg.Long

0.88

0.85

0.880.86

Act Wt %

Short

-0.76

-0.78

ExcessReturn

7.6111.27

1.48x

7.2811.06

1.52x

Info.Ratio

1.25

1.51

1.21x

1.25

1.64

1.31x

IVackingError

5.43

6.48

1.19x

5.32

5.99

1.12x

BattingAverage

61

65

1.06x

67

68

l.Olx

"nirn-Over

52

106

2.04X

56

108

1.93x

Exce.->s Hiunis arc the arnmalizeA compouiukd return of the smuegy minus ihe annuiilizfd coiiipomitU-d rcturti of the resfWliif hftichmiirk. hifonitatioii nniosarc thf arithiticlic nivati c.xicss resum divided hy the tracking error irrsm ihf hcmhrmrk. Holdings are ihe iimnhcr of slocks held loui; or .̂ 7((>i7- Arcriij>c iUiitviit'(^/)(5 are the wrreipottditij; mivr tm^tts for hu^ >iiid short holdings. Stocks mt held are not included in the jverafic .<.hort aaife ircrtj/ir. Rdtios areportfolio stiitiitics divided hy the ivrrespondiiij; loii^-only portfolio statistics. All ponfolw statistics are annuahzcd. Returns rcpresnit model-driven results.

We define jR ,̂ R^, and Rj^ as the returns for the longportion of the portfolio, short portion of the portfolio. ;indbenchmark, respectively. Similarly, we define iV^ and IV^as the proportions of the portfolio that are long and short,respectively. The contribution to portfolio excess returnfrom the long portion, C^, is calculated as follows:

= 9.1%

The contribution from the short portion. C^, is cal-culated as follows (note the reversal of benchmark andportfolio return to obtain the negative of excess returns);

= 2.1%

Of the total 11.3% outperformance versus the bench-mark, 9.1 % comes from the long portion of the portfolioand 2.1% comes from the short portion. The remaining0.1% is caused by the periodic rebalancing between thelong and short portions of the portfolio in order to main-tain them at 130% and 30%, respectively, of invested cap-ital. We reter to this remaining portion as the "long-short

interaction." See Appendix B for further explanation ofthis interaction term.

In this example for the domestic 130/30 attribu-tion, the long portion of the portfolio had a positive returnthat exceeded the benchmark return, so it contributedpositively to the overall portfolio excess return for theperiod. The short portion ot the portfolio also had .i pos-itive return over the period, but it was substantially lessthan the benchmark return. We argue that it also addedvalue to the overall portfolio excess return in a 130/30framework since performance should be measured versusthe benchmark, as in a long-only strategy, rather than inabsolute terms. This is no different from a long-onlystrategy adding value when it beats its benchmark, eventhough its absolute return was negative. If it is a downmarket period and the benchmark has a more negativereturn than the long-only portfolio, the long-only port-folio is said to outperform regardless of the t̂ ict that itdeclined in value.

SHORTS AND LONGS CONTRIBUTETO PERFORMANCE IN MOST YEARS

Exhibit 8 presents the annual performance for thelong-only and 130/30 strategies and their respective bench-marks. Annual active return.s. long and short contributions,and long-short interactions are also given. Both 130/30strategies outperform their long-only counterparts in 11

38 AN EMPIMCAI. ANAIYS 130/30 STUATECIES FALL 2U()7

E X H I B I T 7Benchmark and 130/30 Strategy CompoundedAverage Annual Returns

Domestic Intemational

Benchmark

130/30 Strategy

Longs

Shorts

n.o

4.1

7.8

18.9

15.0

2.5Returm arc for the 1/1994-12/2006 period and represent model-driveti

of the 13 calendar years. In addition, both the long andshort portions of tbe 130/30 strategies generate positiveperformance contributions in most years. For the domesticstrateg)', long contributions are positive in 11 years andshort contributions are positive in 10 years. For the inter-national .strategy, the long positions contributed positivelyin 10 years and shorts contributed positively in eight years.There are no years where the contribution trom the longsand the shorts are both negative in either the domestic orinternational universes. This illustrates the diversificarion

potential for adding value from an alpha model both onthe long side as well as the short side.

The 130/30 long-short interaction is small for mostyears. There are years, however, when it is large. Theinteractions were large in 1998 for the domestic strategyand in 1997 and 2002 for the international strategy. Thelong-short interaction is caused by the turnover requiredto keep the long and short portions of the portfoho at130% and 30% of capital, respectively. Not surprisingly,the turnover required to maintain the 130/30 weights fortlie strategies is also high tor the same years. High turnoverbetween the long and short portions of a 130/30 strategy-does not necessarily result in large interactions asinteractions can cancel out within a given year. The yearwitb the highest turnover between the long and sliortportions was 2001 for both the domestic and interna-tional strategies. Yet, the interaction terms in that yearwere not particularly high.

The latter years of the technology and telecombubble as well as the Russian Financial Crisis occurred in1998 and 1999. The aftermath of the technology andtelecom bubble took place primarily in 2000 and 2001.Interestingly, the contributions trom the longs were neg-ative in 1998 and 1999, but the contributions fixim the

E X H I B I T 8130/30 Long-Short Performance Contributions

DomesHcRussell 1000 Return

Long-Only Retum

no/30 Retum.Aclive Reiunis A CimirihuiionsLong-Only Active Return130/30 Active Return

Long ContributionShort ContributionLong-Short Interaction

InternationalEAFE RetumLong-Only Retum130/30 ReturnActive Returns & ContributionsLong-Only Active Retum130/30 Active Retum

Long ContributionShort ContributionLong-Short Inleraction

1994

0.4

3.48.1

3.07.7

4.13.30.3

7.817.015.0

9.2

7.26.60.6

0.0

199S

37.841.346,6

3.58.8

3.05.40.4

11.29.19.7

-2.1

-1.5-4.0

2.9-0.4

1996

22.527.233.2

4.710.8

7.92.8

O.I

6.0

9.210.0

3,13,98,0

-4,0-0,1

1997

32.9

35.942.1

3.09.31.57,0

0.7

1.8

0,012.3

-1-810.61.4

7.91.3

1998

27.026.138.0

-1.011.0-0.79.9

1,7

20.014.5(8.5

-5.5-1.5-6.45.0

-0.1

1999

20.9

24.023.8

.VI2,9

-2.45.2

0,1

27.027.331.9

0.44.9

-1,15.70.3

2000

-7,810,816.9

IS.624.7

22.7

1.30.7

-14.2-1.7-2.5

12.411.712.3-0.60,0

2001

-12.5

2.37.5

14.820.024.2-3.6-0.6

-21.4-12.2-lO.l

9.311,3tl.O0.20,1

2002

-21.7

-16.0-13.7

5.67.96.41.6

-0,1

-15.90.08,5

16,024.420.9

2.41.1

2003

29.949.1

49.5

19.219.623.7-4.00,0

38.663.767.8

25,129.232.6-3.50.0

2004

11.421.7

17,1

10.35.7

8.8-3.30.2

20,231,735,2

11.414.916.7-180,0

200S

6.315.217.5

8,911.210.10.7

0,5

13.521.226.4

7,612.813.0

-0.20.0

2006

15.5

17.119.2

1,63.72,5

1.3-o,i

26.336.143.1

9.816.8

11.64.30.9

Avg

ll.O

18.622.3

7.611.39.02.10,1

7.815.118.9

7,3tl.I

9.31.60.1

Returns represent model-driven results.

FALL 2007 THE JOURNAL OK ALTERNATIVE INVESTMENTS 3 9

shorts were positive in both the domestic and interna-tional 130/30 strategies. In the aftermath period of 2000and 2001, the short contributions in both the domesticand international strategies were nominal or slightly neg-ative, whereas the long contributions were substantiallypositive in both universes.

SUMMARY AND CONCLUSION

Our findings provide ftjrther evidence that a 130/30strategy is a viable strategy that can substantially add valueover traditional long-only strategies without adding muchincremental risk. We use an empirical approach to showthat this strategy lends itself particularly well to quantita-tive management. For the time period tested, and in bothlarge-cap domestic and international universes, 130/30strategies perform considerably better than long-onlystrategies when using the illustrative alpha model. Thus,our empirical results indicate that removing the long-onlyconstraint allows a quantitative manager to capitalize onboth tails of their alpha model and substantially enhancestlieir ability to outperform. Both the short portion andthe long portion of the 130/30 strategies contribute to theoverall outperformance over equivalent long-only strate-gies. There were no years where both the long and shortportions of the portfolio detracted from performance.Iniportantly, the performance attribution shows that nei-ther portion drives performance for the entire period ordetracts trom it over time. Hence, the results provide evi-dence that a quantitatively-based 130/30 strategy addsvalue from both the long and the short sides.

The enhanced active strategy's claim to fame is thatit allows a manager to short a portion of the portfoho to

enhance returns. We examine this from the context ofthe quantitative manager who ranks stocks on a varietyof factors from best to worst. Historically, long-only quan-titative managers have purchased the best stocks and havenot held stocks that rank the worst. In an enhanced activestrategy, a manager can now underweight stocks withnegative return expectations by the same magnitude thatstocks with positive return expectations are overweighted.

We provide a framework for how to approach per-formance attribution for an enhanced active strategy. Per-formance of an enhanced active strategy such as a 130/30should be measured versus its benchmark, rather than onan absolute basis. The relative performance is decom-posed into contributions from the longs and shorts as wellas il long-short interaction term. This decomposition ofactive returns will allow investors to evaluate whether a130/30 manager is able to add value on both the longand short sides.

In summary, 130/30 strategies should be viewed asactive extensions of a long-only strategy in a particularas.set class and, therefore, may provide an alternative to along-only product in an investors portfolio. 130/30 strate-gies are managed to a benchmark, which allows them tobe evaluated in a traditional long-only framework. Theyprovide investors with equivalent exposure to the marketas a long-only strategy, but with greater potential for excessreturn. In essence, 130/30 strategies are not that differentfrom long-only strategies in terms of portfolio and riskcharacteristics. Therefore, investors should evaluate themin a similar fashion to long-only strategies.

40 A N E.MPIRICAL ANALYSIS OF 13(1/30 STRATEGIES FALI 2(107

A P P E N D I X A

In this Appendix, we present performance results andconstruction details for the alpha model used in the historicalhack tests. The factors chosen for the alpha model are widelyused and well known by quantitative practitioners. For example.all factors except Sales/Price and Share-Decrease were used inIJriish [2007]. We chose to use a straighrtbrward equal-weighted.ilpha model with well-known factors, since details of factonand their weighting?; could not be revealed for a proprietaryalpha model. That said, the illustrative alpha model containsmany factors that are widely used in proprietary alpha modelsand provides realistic relative performance analysis for a 13(1/30portfolio compared to a long-only portfolio when the samealpha model is used.

Security level intbrmation used to calculate the Indiv-idualalpha factors are obtained from I/B/E/S, Compustat, World-scope, and IDC. The alpha model is constructed using the fol-lowing SIX factors:

• Estimate IJ^evision is defined as the number of analystestimates revised up over a month, minus the numberrevised down, divided by the total number of analysts.Estimate Revision for the trailing three months areweighted together 50/30/20 with the highest weight onthe most recent month.

• Long-Term Momentum is ll-monrh total return indollars, lagged one month.

• Cash-Flow/Price, Book/Price, and Sales/Price are eachcalculated from Compustat data for domestic stocks andWorldscope data tor international stocks. In each case,the numerator is divided by shares outstanding and thendivided by price per share.

• Share-Decrease is the 12-month change in conmionshares outstanding, from Compustac for domestic stocksand Worldscope for international stocks.

Each month, the factors for each stock are normalizedwithin their respective domestic or international universe versustheir GICS sector peers. The normalized factors are equallyweighted to obtain an overall score, or alpha, for each stock.

The domestic universe consists of the largest 1000 stocksin the U.S. by market capitalization and is meant to approxi-mate the Russell lOOO Index. The international universe is aproxy for the MSCI EAFE Index and consists of the largestI (K)() stocks in the non-U.S. developed markets by market cap-italization. Both stock universes are formed at the end ofjuneeach year.

Exhibit A shows the performance of the domestic andinternational alpha models over the entire 13-year test period.

E X H I B I T A

Alpha Model Quintile Performance

DomesHcOuiiitile-1

Quintile-2

Quintile-3

Quintile-4

Quintile-5

International

Quintile-1

Quintile-2

Quintile-3

Quintile-4

Quintile-5

ExcessReturn

7.83

3.12

-0.34

-2.34

-8.26

7.94

2.83

0.58

-4.03

-7.23

InfnrmatiunRatio

1.68

0.94

-0.14

-0.67

-1.58

1.94

1.02

0.23

-1.27

-1.87

BattingAverage

75

62

51

41

33

73

61

53

38

30

Quintile-I contains the highest alpha stocks eacli month, wliileQuintile-5 contains the lov\est alpha stocks. Excess returnsshown are calculated from that quintiie's annualized equallyweighted return minus the annualized equally weighted uni-verse return for each period. The information ratio is the excessreturn divided by the tracking error versus the universe return.The batting average is the percentage of months that each quin-tile outperforms the universe return.

Exhibit A shows that the domestic and internationalmodels work well over the test period as excess returns for bothmodels monotonically decline from Quintile-I to Qumtile-5.In addition, the model is equally effective in the upper audlower tails as both outperform by appro.vimately the same mag-nitude. The top quintile in hoth the domestic and internationaluniverees outperforms 75% of the months whereas the bottomquintile outperforms roughly 30%. In other words, the bottomquintile underperfbrms ix)ughly UK)'Xi—3t)% or HYVn of the time.Given the positive results for the alpha model, we feel it is usefijlfor the purpose of illustrating the performance differencesbetween long-only and 130/30 portfolios.

A P P E N D I X B

Exhibit 8 in the performance attribution section of thearticle shows that the contribution of the longs and shorts toactive return do not add up to die entire outperformance of the130/30 portfolio relative to its benchmark. We call thisremaining portion the "long-short interaction." In order tomaintain the value of the longs and shorts at 13()'X) and M

FALL 2007 THE JOURNAL OF ALTERNATIVE iNVESTMEfrrs 4 1

respectively of the portfolio, some turnover is required at theend of every period. We assert that this turnover causes thelong-short interaction and prove this in the followingparagraph.

Assume the long and short portions of the ! 30/30 strategyare run separately and available as the two ETF's. L and S respec-tively. The 130/30 strategy can be achieved by holding a two-stock portfolio {long L by 130% and short Shy —30% of investedcapital). However, to maintain the ratio at e.xactly 130/30, theportfolio would need to be rebalanced periodically to returnthe weights of L and 5 to 130% and -30%, respectively. Oncethe long and short portions of the portfolio have been rebal-anced, the compounded return of the two-stock portfolio ceasesto be the weigh ted-average compounded return of L and .S,while the active return also ceases to be the weighted activereturn of L and S. This non-zero difference between the activereturn of the two-stock portfolio and the weighted active returnof L and 5 is the long-short interaction.

ENDNOTES

'See "130/30 Strategy Payday," Pensions & Investments(May 2S, 2007).

-In practice, managers of enhanced-active products typi-cally invest less in additional longs than they short. For example,the manager may short 30% but re-invest only 29% of die pro-ceeds in additional long holdings. One of the reasons for this non-symmetric ratio of shorts to longs is that the ratio should adjustback toward the targeted symmetric rado of 130/30 it the strategyworks. When shorts underperform, the longs become a smallerportion in the portfolio. Conversely, when the longs outper-form the shorts, they become a larger portion in the portfolio.For simplicity, our back tests keep the portfolio very close to the130/30 target ratio. Some managers may allow the ratio tochange, depending on market conditions. For example, a 130/30manager following a very tight adherence to the target ratiomight keep the ratio very close to 129/30 at all times, while amanager with a less stringent policy may allow wider swings inthe ratio depending on market conditions.

^Clark. de Silva, and Thorley [20021 develop the transfercoet^cient, which is an extension of the fundamental law ofactive management articulated by Grinold [1989]. It is calcu-

lated as the correlation between the risk-adjusted expectedreturns and the risk-weighted active exposures of securities inthe portfolio.

""Back test results are not adjusted for transaction costs orborrowing costs for the shorts. However, as Sorcnsen, Hua.and Qian [20071 demonstrate, under reasonable assumptions,the benefits of a 130/30 type strategy outweigh the additionaicosts. Adjustments have also not been made for whether a par-ticular stock could have been shorted at a given point in theback test period. Historical data on short availability is notreadily available and could not be included in our analysis.

REFERENCES

Brush, John S. "Value and Growth, Theory and Practice." TlieJournal of Portfolio Management (Spring 2007), pp. 22-30.

Clarke, Roger, Harinda de Siiva, and Steven Thorley. "I'ort-folio Constraints and the FLmdamenta! Law of Active Man-agement." Financial Analystsjournal, Vol. 58. No. 5 (September/October 2002), pp. 48-66.

. ''Towards More Information-EfHcient Portfolios." TheJournal of Portfolio Management (Fall 2004), pp. 54-63.

Grinold, Richard C. "The Fundamental Law of Active Man-agement." Tlie Joumal of Portfolio Management, Vol. 15, No. 3(Spring 1989), pp. 3IK37.

Jacobs, Bruce I., and Kenneth N. Levy. "Enhanced ActiveEquity Strategies: Relaxing the Long-Only Constraint in thePursuit of Active Returns." The Journal of Portfolio Management(Spring 2006), pp. 45-55.

Sorensen, Ronald Hua. and Edward Qian. "Aspects of Con-strained Long-Short Equity Portfolios." lite Journal oJ PortfolioManagetnent (Winter 2007), pp. 12-20.

To order reprints of this article, please contact. Dewey Palmieri atdpalmieri@iijournah.com or 212-224-3675

42 AN EMi'ikicAi ANAIYSISOK 130/30 STRATEGIES FAIL 2007