ASSET ALLOCATION: MANAGEMENT STYLE AND PERFORMANCE MEASUREMENT

An Asset class factor model can help make order out of chaos William F. Sharpe*
Reprinted from the Journal of Portfolio Management, Winter 1992, pp. 7-19.

This !p"ri#h$e% ma$erial has &ee' repri'$e% (i$h permissi!' )r!m The Journal of Portfolio Management. C!p"ri#h$ * I's$i$+$i!'al I',es$!r- I' .- .// Ma%is!' A,e'+e- Ne( Y!r0- N.Y. 12233a Capi$al Ci$ies4A5C- I' . C!mpa'". Ph!'e 63137 33.89:;;.

It is widely agreed that asset allocation accounts for a large part of the variability in the return on a typical investor's portfolio. This is especially true if the overall portfolio is invested in multiple funds, each including a number of securities. Asset allocation is generally defined as the allocation of an investor's portfolio among a number of "major" asset classes. Clearly such a generali ation cannot be made operational without defining such classes. !nce a set of asset classes has been defined, it is important to determine the e"posures of each component of an investor's overall portfolio to movements in their returns. #uch information can be aggregated to determine the investor's overall effective asset mi". If it does not conform to the desired mi", appropriate alterations can then be made. !nce a procedure for measuring e"posures to variations in returns of major asset classes is in place, it is possible to determine how effectively individual fund managers have performed their functions and the e"tent $if any% to which value has been added through active management. &inally, the effectiveness of the investor's overall asset allocation can be compared with that of one or more benchmar' asset mi"es. An effective way to accomplish all these tas's is to use an asset class factor model. After describing the characteristics of such a model, we illustrate applications of a model with twelve asset classes to analy e the performance of a set of open(end mutual funds between )*+, and )*+*.

ASSET CLASS FACTOR MODELS &actor models are common in investment analysis. -.uation $)% is a generic representation/

0i represents the return on asset i, &i)represents the value of factor ), &i1 the value of factor 1, &in the value of the n'th $last% factor and ei the "non(factor" component of the return on i. All these values are $potentially% un'nown before(the(fact, as indicated by the tildes. The remaining values $bi) through bin% represent the sensitivities of 0i to factors &i) through &in. A 'ey assumption ma'es a model of this sort more than simply an e"ercise in data description/ The non(factor return for one asset $ei% is assumed to be uncorrelated with that of every other $e.g. ej%. In effect, the factors are the only sources of correlation among returns. An asset class factor model can be considered a special case of the generic type. In such a model each factor represents the return on an asset class and the sensitivities $bij values% are re.uired to sum to ) $)223%. In effect, the return on an asset i is represented as the return on a portfolio $shown by the sum of the terms in the brac'eted e"pression% invested in the n asset classes plus a residual component $ei%. &or e"pository convenience, the sum of the terms in the brac'ets can be termed the return attributable to style and the residual component $ei% the return due to selection. Indeed, a 'ey contribution of this approach is the separation of return into these two main components.

E<ALUATING ASSET CLASS FACTOR MODELS The usefulness of an asset class factor model depends on the asset classes chosen for its implementation. 4hile not strictly necessary, it is desirable that such asset classes be )% mutually e"clusive, 1% e"haustive and 5% have returns that "differ". 6ragmatically, each should represent a mar'et(capitali ation weighted portfolio of securities7 no security should be included in more than one asset class7 as many securities as possible should be included in the chosen asset classe7 and the asset class returns should either have low correlations with one another or, in cases in which correlations are high, different standard deviations. 4hile the appropriate measure of the efficacy of any specific implementation depends on the uses to which the model is to be put, factor models are typically evaluated on the basis of their ability to e"plain the returns of the assets in .uestion $i.e. the 0is%. A useful metric is the proportion of variance "e"plained" by the selected asset classes. 8sing the traditional definition, for asset i/

The right(hand side of e.uation $1% e.uals ) minus the proportion of variance "une"plained". The resulting 0(s.uared value thus indicates the proportion of the variance of 0i "e"plained" by the n asset classes). It is important to recogni e that this measure indicates only the e"tent to which a specific model fits the data at hand. A better test of the usefulness of any implementation is its ability to e"plain performance out(of(sample. &or this reason it is important to consider not only the ability of a model to e"plain a given set of data but also its parsimony. !ther things e.ual $e.g. 0(s.uared values%, the fewer the asset classes, the more li'ely is the model to represent continuing fundamental relationships with predictive content1. To evaluate the e"posures of funds to changes in the returns of 'ey asset classes, the appropriate measure is the collective ability of a set of such classes to e"plain the time( series variability in the returns on a typical fund $e.g. mutual fund or separately(managed institutional account%. 9ote that this criterion differs from that often applied in evaluating factor models designed to describe specific portions of the overall capital mar'et. &or e"ample, when constructing an e.uity factor model, one might consider the ability of the selected factors to e"plain the time(series variation in the returns of a typical stoc'. :ost stoc' mar'et models include factors representing returns on industry groups and;or economic sectors (( factors that account for much of the typical security's return. If most managers diversify across industries and economic sectors,however, inclusion of factors related to differences in industry and sector returns will add little if any e"planatory power to a model designed to e"plain fund returns.

A TWEL<E ASSET CLASS MODEL The model we use has twelve asset classes. The return of each is represented by a mar'et capitali ation weighted inde" of the returns on a large number of securities. &or reasons that will become clear, it is important to note that each inde" represents a strategy that could be followed at low cost using an inde" fund. The composition of each inde" is specified in sufficient detail by its provider to enable an investor to trac' the returns with little error through a passive $inde"(li'e% investment strategy. Table ) describes the twelve asset classes and the indices used for the associated return series. :ost are widely used inde"es that re.uire no further description. The four less well('nown are those employed to represent 8.#. e.uity classes.

TA5LE 1 Asse$ Classes

Bills Cash-equivalents with less than 3 months to maturity Index: Salomon Brothers' 90-day Treasury bill index Intermediate-term overnment Bonds overnment bonds with less than !0 years to maturity Index: "ehman Brothers' Intermediate-term overnment Bond Index "on#-term overnment Bonds overnment bonds with more than !0 years to maturity Index: "ehman Brothers' "on#-term overnment Bond Index Cor$orate Bonds Cor$orate bonds with ratin#s o% at least Baa by &oody's or BBB by Standard ' (oor's Index: "ehman Brothers' Cor$orate Bond Index &ort#a#e-)elated Se*urities &ort#a#e-ba*+ed and related se*urities Index: "ehman Brothers'&ort#a#e-Ba*+ed Se*urities Index "ar#e-Ca$itali,ation -alue Sto*+s Sto*+s in Standard and (oor's .00-sto*+ index with hi#h boo+-to$ri*e ratios Index: Shar$e/B0))0 -alue Sto*+ Index "ar#e-Ca$itali,ation rowth Sto*+s Sto*+s in Standard and (oor's .00-sto*+ index with low boo+-to$ri*e ratios Index: Shar$e/B0))0 rowth Sto*+ Index &edium-Ca$itali,ation Sto*+s Sto*+s in the to$ 102 o% *a$itali,ation in the 34S4 equity universe a%ter the ex*lusion o% sto*+s in Standard and (oor's .00 sto*+ index Index: Shar$e/B0))0 &edium Ca$itali,ation Sto*+ Index Small-Ca$itali,ation Sto*+s Sto*+s in the bottom 502 o% *a$itali,ation in the 34S4 equity universe a%ter the ex*lusion o% sto*+s in Standard and (oor's .00 sto*+ index Index: Shar$e/B0))0 Small Ca$itali,ation Sto*+ Index 6on-34S4 Bonds Bonds outside the 34S4 and Canada Index: Salomon Brothers' 6on-34S4 overnment Bond Index 7uro$ean Sto*+s 7uro$ean and non-8a$anese (a*i%i* Basin sto*+s Index: 9T0 7uro-(a*i%i* 7x 8a$an Index 8a$anese Sto*+s 8a$anese Sto*+s Index: 9T0 8a$an Index

In effect, the institutional universe of 8.#. e.uities has been divided into four mutually e"clusve and e"haustive groups5. The first two represent a partition of the stoc's in #tandard and 6oor's ,22 stoc' inde". -very si" months the #<6,22 stoc's are ran'ed according to the ratio of the most recently published boo' value per share to a previous month(end price per share. A dividing line is drawn so that appro"imately half the total value of the ,22 stoc's is placed on either side. #toc's with high boo'(to(price ratios are placed in the "value" stoc' inde"7 the remainder are in the "growth" stoc' inde". A similar procedure is followed in constructing the medium capitali ation and small capitali ation stoc' inde"es. 9on(#<6,22 stoc's are ran'ed on the basis of total

outstanding mar'et capitali ation every si" months, and a dividing line drawn so that appro"imately +23 of the total value is above the line and 123 below it. :ost of the stoc's in the first group are placed in the medium capitali ation inde" and most of the remaining stoc's in the small capitali ation inde". To avoid e"cessive turnover in the composition of these inde"es of relatively illi.uid stoc's $and an associated high cost for inde" trac'ing%, any stoc' that has recently "crossed over the line" a relatively small distance is allowed to remain in its former inde"=. :any of the differences in returns of 8.#. e.uity mutual funds can be attributed to differences in their e"posures to these four asset classes. In effect, there appear to be two 'ey dimensions along which such funds differ. !ne may loosely be termed "value;growth"7 the other "small;large". &igure ) illustrates the composition of the four domestic e.uity asset classes. -ach inde" can be considered a capitali ation(weighted "center of gravity" of the securities in its associated class, as the dots in &igure ) indicate. 9ote that any combination of the four indices with non(negative holdings can be represented by a point in the area defined by the inde" locations $in this case, a triangle%,. &I>80- ) C!:6!#ITI!9 !& &!80 ?!:-#TIC -@8ITA CBA##-#

:uch has been written about both the small(stoc' and the value;growth phenomena. 4hile the terms "value" and "growth" reflect common usage in the investment profession, they serve only as convenient names for stoc's that tend to be similar in several respects. As is well 'nown, across securities there is significant positive correlation among/ boo';price, earnings;price, low earnings growth, dividend yield and low return on e.uity. :oreover, the industry compositions of the value and growth groups differ $e.g. companies with high research budgets tend to have low boo' values relative to their stoc' prices%.

Those concerned with these distinctions have focused most of their research on long(run average return differences7 that is, they have as'ed whether small stoc's or value stoc's "do better than they should" in the long(run. Bess attention has been paid to li'ely sources of short(run variability in returns among such groups. &or present purposes it suffices that such variability is substantial. &igure 1 provides relevant evidence/ The variability in returns across the four asset classes from year(to(year is far greater than would be encountered if groups with similar numbers of securities had been formed randomly. &und e"posures across these dimensions vary greatly. As a result, much of the variation in fund returns in any given period can be attributed to the combined effects of their e"posures to these asset classes and the reali ed returns on those classes.

DETERMINING FUND E=POSURES The traditional view of asset allocation assumes that an investor allocates assets among $potentially many% funds, each of which holds $potentially many% securities. 8ltimately one is interested in the investor's e"posures to the 'ey asset classes. These are a function of )% the amounts of the investor's portfolio invested in the various funds and 1% the e"posures of each such fund to the asset classes. The e"posures of a fund to the various asset classes are, in turn, a function of )% the amounts that the fund has invested in various securities and 1% the e"posures of the securities to the asset classes. 4hile it is possible to attempt to determine a fund's e"posures from a detailed analysis of the securities held by the fund, a simpler approach typically provides more than enough

information for purposes of asset allocation. #uch a method uses only reali ed fund returns to infer the typical e"posures of the fund to the asset classes. #ince only easily obtained information is re.uired, the approach may be considered "e"ternal", in comparison with methods that rely on information that may be available only from sources internal to the fund. Inspection of e.uation $)% immediately suggests a procedure that might be used in this connection. >iven, say, si"ty monthly returns on a fund, along with comparable returns for a selected set of asset classes, one could simply employ a multiple regression analysis with fund returns as the dependent variable and asset class returns as the independent variables. The resulting slope coefficients could then be intererpreted as the fund's historic e"posures to the asset class returns. Table 1 provides an e"ample for Trustees' Commingled 8.#. 6ortfolio $an open(end mutual fund offered by the Canguard >roup%. :onthly returns from Danuary )*+, through ?ecenber )*+* are used for the dependent variable, with the corresponding returns for the twelve asset classes serving as independent variables. The column entitled "8nconstrained 0egression" shows results obtained applying -.uation $)%. The first twelve rows show the resulting slope coefficients $bij values%, e"pressed as percentages. The coefficient total is shown ne"t, followed by the 0(s.uared value $also e"pressed as a percentage%. A substantial portion $*,.123% of the monthly variance in the fund's returns is e"plained by this e.uation. The coefficients, however, do not sum to )223. 4hat is more important, several are vastly inconsistent with the fund's actual policy $to invest in common stoc's with no short positions%.

Ta&le 3 Re#ressi!' a'% >+a%ra$i Pr!#rammi'# Res+l$s Tr+s$ees? C!mmi'#le% F+'% 8 U.S. P!r$)!li! @a'+ar" 1;/: $hr!+#h De em&er 1;/;

Constrained 8nconstrained 0egression 0egression

Bills Intermediate Bonds "on#-term Bonds Cor$orate Bonds !:4;9 -;94.! -54.: !;4.< :54;. -;14;: -5431 !.459

@uadratic 6rogramming
0 0 0 0

&ort#a#es -alue Sto*+s rowth Sto*+s &edium Sto*+s Small Sto*+s 9orei#n Bonds 7uro$ean Sto*+s 8a$anese Sto*+s Total )-squared

.4!9 !094.5 -<41; -:!413 :.4;. -!41. ;4!. -!4:; <54<! 9.450

:4.1 !!043. -1405 -:34;5 :<4!< -!431 .4<< -!4<9 !00400 9.4!;

0 ;941! 0 0 3040: 0 04!. 0 !00400 95455

The column in Table 1 titled "Constrained 0egression" reports the results of a multiple regression analysis similar to the first, with one added constraint/ The coefficients were re.uired to sum to )223. The reduction in 0(s.uared was slight $from *,.123 to *,.)E3%, but the inconsistency between the coeffficients and the fund's investment policy remains. The last column in Table 1 reports the results of an analysis where each coefficient is constrained to lie between 2 and )223 and the sum is again re.uired to be )223. As in the previous cases, the objective of the analysis was to select a set of coefficients that minimi es the "une"plained" variation in returns $i.e., the variance of ei%, subject to the stated constraints. An e.uivalently goal was to ma"imi e the associated value of 0( s.uared, subject to the stated constraints. &or this analysis, the presence of ine.uality constraints $ 2FG bijFG )223 for each i % re.uired the use of a .uadratic programming algorithm. The addition of constraints reflecting the fund's actual investment policy causes a slight reduction in the fit of the resulting e.uation to the data at hand $i.e., a decrease in 0( s.uared to *1.113%. 9ow, however, the coefficients conform far more closely to the reality of the fund's investment style, ma'ing the resulting characteri ation more li'ely to provide meaningful results with out(of(sample data. As Table 1 shows, the analysis suggests that the fund traditionally invests so as to obtain returns similar to those achievable with a portfolio with roughly H23 invested in a mar'et(representative portfolio of value stoc's and 523 in a mar'et(representative portfolio of small stoc's. ?uring the period investigated, over *13 of the month(to( month variation in the return on the fund could be e"plained by the concurrent variation in the return on this particular mi" of value and small stoc's.

STYLE ANALYSIS The use of .uadratic programming for the purpose of determining a fund's e"posures to changes in the returns of major asset classes is termed style analysis $see #harpeI)*++J%. The goal is to find the "best" set of asset class e"posures $bij values% that totals to )223 and conforms with rudimentary information concerning the fund's policies $typically, no net short positions in any asset class7 for funds 'nown to employ short positions, other bounds may be invo'ed%.. In this conte"t, the best such set of e"posures is the one for which the the variance of ei is the least. 0earranging -.uation $)%/

The term on the left can be interpreted as the difference between the return on the fund $the first term on the right% and that of a passive portfolio with the same style $shown by sum of the terms in the brac'ets%. The goal of style analysis is to select the style $set of asset class e"posures% that minimi es the variance of this difference. #uch a difference can be termed the fund's "trac'ing error" and its variance the fund's "trac'ing variance". 9ote that the objective of such an analysis is not to minimi e either the average value of this difference or the sum of the s.uared differences. Thus the method is not designed to find a style that "ma'es the fund loo' bad" $or good%. 0ather, the goal is to infer as much as possible about the fund's e"posures to variations in the returns of the asset classes during the period studied. 4e have noted that a .uadratic programming algorithm must be used for such an analysis. &or an e"act solution, one can implement :ar'owit ' critical line method)$see :ar'owit I)*+HJ%. This study uses the simpler gradient method described in #harpe I)*+HJ. Although in principle the latter produces only an appro"imate solution, differences between the results obtained with the two methods prove to be of no practical importance in this application.

MUTUAL FUND STYLES &igure 5 provides a graphical summary of the results shown in the final column of Table 1. The bar chart indicates the estimated style of the fund, and the pie chart the associated 0(s.uared value. In the latter the 0(s.uared value is identified as attributable to the fund's style and the remainder $) ( 0(s.uared% to selection.

It is important to note that the style identified in such an analysis is, in a sense, an average of potentially changing styles over the period covered. :onth(to(month deviations of the fund's return from that of style itself can arise from selection of specific securities within one or more asset classes, rotation among asset classes, or both security selection and asset class rotation . &or the sa'e of simplicity, we use the term selection to cover all such sources of trac'ing differences. It is sometimes helpful to e"amine the behavior of a manager's average e"posures to asset classes over time. To do so, one can perform a series of style analyses, using a fi"ed number of months for each analysis through time. &igure = shows the results from such a study for Trustee's Commingled 8.#. &und.

The point at the far right of the diagram represents the style described when the si"ty months ending in ?ecember )*+* are analy ed. This corresponds to the style shown in &igure 5. -very other point represents the results of an analysis using a different set of si"ty monthly returns $note, however, that each such set has fifty(eight months in common with its predecessor%. As &igure = shows, this fund's style appeared to remain .uite constant throughout the period analy ed. &igures , and E show the results obtained when the same type of analysis was applied to the returns of &idelity :agellan &und (( a highly popular open(end common stoc' fund. As &igure , shows, its style differed considerably from that of Trustees' Commingled 8.#. &und, with emphasis on growth rather than value stoc's and e"posure to medium( capitali ation stoc's in addition to smaller ones. The pie chart in &igure , shows that the fund is considerably more diversified $and;or engaged in less rotation% than Trustees' Commingled 8.#. &und. ?uring the period covered, over *H.53 of the monthly variation in :agellan returns could be attributed to the concurrent return on a passive portfolio with the style shown in the bar chart in &igure ,.

&igure E suggests that the :agellan &und progressively increased its emphasis on large growth stoc's and decreased its e"posure to small capitali ation stoc's during the )*+2s. This is not surprising, as the fund grew to appro"imately K)= billion by the end of the period, ma'ing substantial investment in very small stoc's increasingly difficult.

MUTUAL FUND TYPES &igures 5 through E show results for two particular mutual funds. Lere we provide a more representative view of the efficacy of the procedure, with style analysis performed for each of 5*, funds using returns from Danuary )*+, through ?ecember )*+*. Averages are ta'en for both the styles and 0(s.uared values of all funds classified as being of the same "type" by Daye C. Darrett < Company, Inc., the providers of the data used for this study. In all, seven such types are represented. The results are shown in &igures H through )5. -ach figure should be interpreted as representative of the style $bar chart% and variance due to style $pie chart% of a "typical" fund of the type. A portfolio invested in all the funds of a particular type would typically have a much higher 0(s.uared $percent of variance attributable to style% than is shown in the figure in .uestion. :oreover, there is typically considerable variation in both style and 0(s.uared values among the funds within each type group. >iven these caveats, the analyses provide useful illustrations of some of the features of the style analysis method.
U$ili$" S$! 0 F+'%

&igure H shows the results for a typical utility stoc' fund. #uch funds $atypically% concentrate their holdings in one industry. As a result, style accounts for an unusally small part $although still ,*.53% of the variance in return. Although such funds hold common stoc's, their returns behave more li'e a passive portfolio invested in both stoc's and bonds. That is, utility revenues are "stic'y" because to the regulatory process, causing shares of such companies to have features that are both stoc'(li'e and bond(li'e. The utility funde"ample emphasi es the fact that style analysis provides measures that reflect how returns act, rather than a simplistic concept of what the portfolios include. 9ote, finally, all e.uity e"posure is to value stoc's, relflecting the high dividend yields typical of utility shares.

Gr!($h EA+i$" F+'%

&igure + portrays a typical growth e.uity fund. Lere the most prominent e"posure is, as e"pected, to growth stoc's, although the typical fund of this type also responds to movements in the returns of other asset classes. 9ote the e"posure to Mills, which probably results from the actual cash holdings that many such funds maintin to meet li.uidity needs. !verall, the results illustrate the fact that few funds are "pure" in the sense of responding only to movements in returns of one asset class. The style analysis procedure can detect some of the subtleties that e"ist in practice, instead of classifying each fund by a single $pure% style. &inally, note that almost *23 of the monthly variation in return of the typical growth e.uity fund can be attributed to its style (( a result typical of common stoc' funds.

Gr!($h a'% I' !me EA+i$" F+'%

&igure * shows the characteristics of a typical growth and income e.uity fund. Lere too, style accounts for appro"imately *23 of the monthly variation in returns. The effects of a li.uidity reserve are probably at least partly responsible for the e"posure to Mills, although choices of stoc's with lower beta values than those in the asset class inde"es could also play a role. 9ote the almost perfect balance between value and growth #toc's, relecting an "#6,22( li'e" stance with respect to large(capitali ation stoc's. The e"posures to small and medium stoc's may reflect actual investment in such stoc's and;or a preference for e.ual weighting rather than capitali ation weighting within the large stoc' sector. In an

important sense, the source of a set of e"posures may not even need to be identified, as long as the e"posures are representative of li'ely future results.

Small S$! 0 F+'%

&igure )2 indicates that small stoc' funds do indeed buy small stoc's $as defined by the asset class used for this study%. Lowever, they also appear to buy somewhat larger ones. :oreover, there tends to be an emphasis on growth rather than value. This may reflect the actual purchase of large(capitali ation growth stoc's by some funds. It may also indicate a preference for medium(capitali ation stoc's with growth characteristics. As &igure ) suggests, a point lying to the right of the point rerpresentin the medium stoc' inde" can be represented by a combination of the large growth stoc' inde", the small stoc' inde" and the medium stoc' inde".

As before, the goal is to represent the behavior of the fund, not its precise composition. &inally, note that the 0(s.uared value is slightly lower $+H.E3% than for the other diversified funds (( perhaps reflecting the lower li.uidity of this sector of the e.uity mar'et.

5ala' e% F+'%

&igure )) shows that balanced funds are precisely that. 4hile any single fund may diverge substantially from the style shown in the figure, collectively balanced funds provide results similar to those obtained by holding all 8.#. asset classes and small amounts of foreign ones. As with other diversified funds, style accounts for roughly *23 of the monthly variation in the returns for the typical fund of this type.

Bi#h8>+ali$" 5!'% F+'%

&igure )1 shows that the method wor's well for bond funds as well as for stoc' and balanced ones. The typical high(.uality bond fund provides e"posure to corporate bonds, government bonds and mortgage(related securities, with style accounting for slightly over ++3 of monthly variance in return. In any given case, a mi" of, e.g., intermediate government bonds and corporate bonds might reflect actual holdings or the average .uality of the corporate bond portfolio. Thus a portfolio with a higher average .uality than that reflected in the Corporate Mond Inde" typically acts more li'e a mi"ture of corporate bonds $defined by the inde"% and intermediate government bonds. #imilarly, a portfolio of corporate bonds with a longer duration than that of the Corporate Mond inde" will "trac'" more closely with a mi" of

corporate bonds $defined by the inde"% and long(term >overnment bonds. As always behavior, not nomenclature, is relevant.

C!',er$i&le 5!'% F+'%

&igure )5 shows a case where an asset class not e"plicitly represented in a model can be represented well by the classes that are included. As shown, ++.+3 of the monthly variation in returns of a typical convertible bond fund can be attributed to the concurrent variation in the returns of a mi" of stoc's, bills, and bonds. This is not too surprising. A convertible bond has characteristics of both bonds and stoc's. !f course, as bond and stoc' mar'ets diverge, the relative sensitivities of any given convertible bond to the two mar'ets will change, giving such an instrument its distinctive non(linear characteristic. Interestingly, managers of convertible bond funds appear to

have preferred habitats, causing them to buy and sell convertible bonds so as to maintain fairly consistent e"posures to asset classes of the type utili ed in this study.

F+'% T"pe S+mmar"

As these e"amples show, a remar'able amount of information can be revealed from an analysis of the returns provided by the manager of an investment fund. This is especially gratifying since in the final analysis return is the product the investor buys from such a manager.

TBE IN<ESTOR?S EFFECTI<E ASSET MI=

!nce the styles of an investor's funds have been estimated, it is a simple matter to determine the associated overall effective asset mi". Betting 4i represent the proportion of the investor's portfolio invested in fund i, overall portfolio return $0p% will be/

As both e.uations $)% and $=% are linear, substitution of the former in the latter will provide another linear e.uation/

or/

where the bpj values are the portfolio's e"posures to the asset classes. As can be seen by comparing -.uations $,% and $E%, each bpj is simply a value(weighted average of the e"posures of the component funds to the asset class in .uestion, with the relative amounts invested in the funds used as weights. The resultant effective asset mi" $specified by the values bp), bp1, ...,bpn % will account for a large portion of the month(to(month variation in returns for the typical portfolio invested in many funds. 8nder the assumption that the residual ei terms are uncorrelated, diversification across funds will greatly reduce the variance of the final $non(factor% component and thus increase the portion of variance attributable to asset allocation. -ven if some of the residuals are correlated, the use of multiple funds will typically lead to substantial reduction in selection ris'. The effective asset mi" represents the style of the investor's overall portfolio. &or a multiple(managed portfolio, style is even more important than for an individual fund.

PERFORMANCE MEASUREMENT In a sense, a passive fund manager provide an investor with an investment style, while an active manager provides both style and selection. This statements can be used to define the terms "active" and "passive" management. 9ote that in this ta"onomy, the precise implementation of an asset class factor model play a crucial role. This suggests that one may wish to select a set of asset classes so that only superior performance relative to a static mi" of the chosen classes warrants the higher fees usually associated with "active" as opposed to "passive" management. This is the approach we ta'e, focusing on a fund's selection return, defined as the difference between the fund's return and that of a passive mix with the same style.

There are several desiderata associated with the selection of a benchmar' for performance measurement. A benchmar' portfolio should be )% a viable alternative, 1% not easily beaten, 5% low in cost, and =% identifiable before the fact. #tyle analysis provides a natural method for constructing benchmar's meeting these re.uirements. The return obtained by a fund in each month can be compared with the return on a mi" of asset classes with the same estimated style, where the style is estimated prior to the month in .uestion. 9ote that this differs from the use of the ei values obtained as byproducts of a style analysis, since the latter are in(sample, not out( of(sample values. To illustrate the approach for Trustees' Commingled 8.#. and &idelity's :agellan &unds, for each month t/ ). The fund's style is estimated, using returns from months t(E2 through t().E 1. The return on the resultant style is calculated for month t. 5. The difference between the fund's return in month t and that of the style benchmar' determined in steps )% and 1% is computed. This difference is defined as the fund's #election 0eturn for month t. &igure )= shows the cumulative sum of the monthly selection returns from Danuary )*+E through ?ecember )*+* for Trustees' Commingled 8.#. &und.H In such a graph, increases result from positive selection returns and decreases from negative ones. .

Average #td ?ev T$Avg%

(.2E 3 per month ).E* 3 per month (2.1,

The table below the graph in &igure )= summari es the results in a different manner. !n average, the fund underperformed its style benchmar's by (2.2E3 $E basis points% per month, with a standard deviation of ).E*3 $)E* basis points% per month. The t(statistic associated with the mean difference was, however, small in absolute value, suggesting that the average difference was not statistically significantly different from ero.+ &igure ), emphasi es the advantages to be gained by analy ing performance the way we have described. It compares the return on Trustees' Commingled 8.#. &und with that of #tandard < 6oor's ,22 stoc' inde" $commonly used to evaluate mutual fund performance%. The fund's performance, so measured, was over three times as bad as that shown previously/ the cumulative difference was ()23 and the average difference (12 basis points per month. Mut such a comparison includes results due to both style and selection. ?uring the period in .uestion the fund's style underperformed that of the #<6,22 $primarily because of its e"posure to small stoc's%. Indeed, this accounts for appro"imately two(thirds of the fund's underperformance relative to the #<6,22. An investor choosing Trustees' Commingled 8.#. &und could and should have 'nown that its style favored value stoc's and small stoc's. The choice to e"pose some of the portfolio to these asset classes should be attributed to the investor. 0esults $good or bad% associated with such the choice of a style should be attributed to the investor, not to the manager of a fund following that style.

Average #td ?ev T$Avg%

(.12 3 per month 1.)5 3 per month (2.E,

&igures )E and )H show the results of similar analyses for &idelity :agellan &und. As &igure )E shows, the fund provided a positive but statistically insignificant outperformance when compared with the #<6,22 over the period. Mut &igure )H shows that such a comparison mas'ed :agellan's truly outstanding selection performance. ?uring this period, the fund outperformed its style benchmar's by a cumulative amount of over 1,3. !utperformance averaged ,H basis points per month with a standard deviation of )2, basis points. The t(statistic of 5.HE shows that such differences were highly significant statistically. Two aspects account for the large t(value/ the relatively large average return difference and the relatively small variation in the difference from month to month.*

Average #td ?ev T$Avg%

2.)+ 3 per month ).=+ 3 per month 2.+=

Average #td ?ev T$Avg%

2.,H3 per month ).2, 3 per month 5.HE

MUTUAL FUND PERFORMANCE &idelity :agellan's performance from )*+, through )*+* is far from typical. 4hile only out(of(sample results can provide a definitive test of the collective performance of mutual funds, the average ei values obtained as a by(products from fund style analyses can provide at least some evidence on the matter. &igure )+ shows the distribution of the average trac'ing errors obtained from the style analyses of E5E stoc', bond and balanced funds. -ach value is the average ei value obtained from a style analysis using returns for one fund covering the period from Danuary )*+, through ?ecember )*+*. 9ote that the distribution is roughly normal, with a mean of (2.2H= $(H.= basis points per month%. This is roughly consistent with the hypothesis that the average mutual fund cannot "beat the mar'et" before costs, because

such funds constitute a large $and presumably representative% part of the mar'et. Annuali ed, the mean underperformance is appro"imately 2.+*3 per year (( an amount that, if anything, may be slightly less than the non(transaction costs incurred by a typical mutual fund.

Average G (.2H=

MEASURING AN IN<ESTOR?S PERFORMANCE In the paradigm utili ed in this article, an investor ma'es decisions that result in an effective asset mi" and a set of selection returns. In a sense, the investor selects a set of $passive or active% managers and a specific allocation of funds among such managers. >iven the managers' styles, this determines the investor's effective asset mi". The procedures described earlier can be applied directly to measure the efficacy with which the investor performs his or her functions. The performance of each month's effective asset mi" can be compared with that of a predetermined benchmar' asset mi" to assess the value added or lost due to asset allocation decisions $advertent or inadvertent%. The remainder of the investor's return is attributable to the joint effects of )% the fund

managers' selection returns, and 1% the investor's allocation of money among the managers. The investor selection return $#p% is simply/

where the #i values are determined out(of(sample, using procedures such as those described earlier.

CONCLUSION An asset class factor model can help ma'e order out of the chaos that often attends the investment process. It can provide a consistent view of investment decisions investors ma'e to economi e on information flows and e"ploit comparative advantages. The style analysis procedure described in this article allows such a model to be implemented economically. At the very least it can serve as a valuable supplement to other methods designed to help investors achieve their goals in cost(effective ways.