academic papers mutual fund risk-return profiles: a novel ... · henry i. silverman is an assistant...

Academic Papers

Mutual fund risk-return profiles:A novel use of triangulationReceived (in revised form): 19th November, 2007

Henry I. Silvermanis an assistant professor of finance at Roosevelt University in Chicago and director of an asset management firm based

in London. He lived in the UK for a number of years where he completed his doctorate and taught finance and risk

management at the University of London for executives at JP Morgan Chase. Engaged in research on investment

companies, their return objectives, investment strategies and risk-taking behaviour, Dr Silverman has conducted

numerous interviews with US and European fund managers and performed analytical work on their disclosure practices

and documents. Dr Silverman’s findings were presented to the Chairman of the British Government’s Financial

Ombudsman Service in 2003. Dr Silverman is author of the academic text ‘Theory and Practice of Fund Management’

and a member of the Financial Management Association as well as the New York Society of Security Analysts.

Walter E. Heller College of Business Administration, Roosevelt University, 18 South Michigan Avenue, Chicago, IL

60603, USA

Tel: +1 312 281 3319; Email: [email protected]

Abstract This paper triangulates with ethnographic content analysis and time series

data to discern risk-return profiles for active equity mutual funds. The paper identifies

and establishes associations between a variety of investment objectives and risk

factors disclosed or otherwise encoded in the prospectus and annual report, and

compares risk factors appearing in these documents with levels of risk observed in the

time series data. The findings are largely consistent with the predictions of portfolio

management models discussed in the literature; however, the triangulation process

also reveals critical gaps between what is disclosed in each of the primary narratives

and what is observed in the secondary data, ie between what is said and what is

done. Risk exposures measured ex post are not always communicated in the

documents. For example, the study finds instances of undisclosed pseudo-industry risk

in the form of concentrated technology holdings which may reflect a violation of

prospectus covenants. The paper also finds elevated levels of residual risk in the

secondary data which may be indicative of benchmark gaming.

Keywords: mutual funds, portfolio management, ethnographic content analysis,

risk, investment objective, triangulation

INTRODUCTIONStandard risk pricing and portfolioframeworks are largely based on theassumption that the principal investor isin control of security selection, assetallocation and rebalancing decisions

associated with the portfolio. Thisobviously follows for individualinvestors who manage their ownportfolio via direct equity and debtholdings. Similarly, large institutionalinvestors, such as endowments,

# Henry Stewart Publications 1752–8887 (2008) Vol. 1, 2 191–222 Journal of Risk Management in Financial Institutions 191

foundations and life insurance firms,often manage their own portfolios orhave the resources to contract with aprofessional investment manager whowill construct a bespoke portfolio forthe client based on the latter’s objectives,risk tolerance and constraints.1

Mutual fund investors, on the otherhand, are in a distinct position: they mustattempt to assemble an optimal portfolioby selecting from an assortment of over8,000 ready-made portfolios, each withits own unique risk-return characteristics,and managed by agents who, althoughregulated by federal agencies (in theUSA, for example, the Securities andExchange Commission (SEC)) andoperating under the oversight of a boardof directors, have no responsibility forensuring that the fund meets the specificneeds of participating investors. Intrinsicto the mutual fund industry then, is aprincipal–agent relationship thatprecludes both shareholder control overdecisions on security selection as well asany contractual obligation on the part ofthe fund manager to optimise theportfolio in line with individualshareholder concerns (investmentcompanies do bear certain obligationswith respect to disclosure of investmentobjectives and risks, as will be discussedlater in this paper). This heightens theexigency for investors, financial advisersand risk managers to be able to identify,via disclosure and reporting documents,the investment policies pursued and risksborne by each fund.

Mutual fund prospectuses and annualreports represent official communicationsissued by investment companies toprospective and existing mutual fundshareholders. They are, as the Presidentof the Investment Company Institutestated in 1995,2 the key disclosure

documents that must be provided to allinvestors. These documents aregoverned by regulations originallylegislated by the US Congress in theInvestment Company Act 1940 (ICA)and by subsequent amendments to thatAct. Among other provisions, the ICArequires investment companies to‘disclose their financial condition andinvestment policies to investors whenstock is initially sold and, subsequently,on a regular basis’.3 Further, the focus ofthe ICA is on ‘disclosure to the investingpublic of information about the fundand its investment objectives, as well ason investment company structure andoperations’.4 Specifically, subsection (b)of s. 8 of the ICA requires all investmentcompanies to file a registration statement(prospectus) with the SEC, whilesubsection (b)(2) of s. 8 of the ICAstipulates that the registration statementcontains ‘a recital of all investmentpolicies of registrant’. Althoughinvestment companies often includeinformation of this kind in newsletters,prospectus ‘wrappers’, advertising andsales literature, it is the prospectus andannual report which are typicallyaccepted as the most authoritativeaccounts of a fund’s risk-return profile.

There are, however, limitationsassociated with reliance on the mutualfund prospectus and annual report for thepurpose of apprehending the objectivesand risk-taking of the investmentcompany. The very fact that they areofficial documents, governed andenforced by federal legislation, suggeststhat while mutual fund prospectuses andannual reports may adhere strictly toguidelines set out in law, ie disclosing therequired information in the prescribedformat, they may reveal little beyondthis. While content and format

Silverman

Journal of Risk Management in Financial Institutions Vol. 1, 2 191–222 # Henry Stewart Publications 1752–8887 (2008)192

standardisation may facilitate analysis,they may also circumscribe theinformation value of the materials. Forexample, in 1998 the SEC adopted a newregistration form for investmentcompanies that permits, but does notrequire, funds to include disclosure aboutthe types of investors for whom the fundis intended or the types of investmentgoals that may be consistent with aninvestment in the fund.5 As investmentcompanies may face civil or criminalpenalties for making materialmisrepresentations in official documentslike the prospectus, it is not unreasonableto assume that they will be extremelycautious about disclosing informationbeyond what is required by law. Such aninformation gap would underscoreGarfinkel’s characterisation of recordsbeing primarily ‘contractual’ rather than‘actuarial’6 and may be reflected, asPatton points out, in documents that areoften highly variable in quality, withgreat detail in some cases and virtuallynothing for other ‘programmaticcomponents’.7 Absent a discriminatinganalytical framework, it is thereforefrequently problematic to detect theinvestment objectives pursued by fundmanagers, the precise nature of the risksto which shareholders are exposed andthe relationships between objectives andrisk-taking.

LITERATUREAcademics like Fabozzi argue that, as anintegral part of the investmentmanagement process, mutual fundmanagers will attempt to construct anefficient portfolio within a mean-variance framework.8 Jones too, suggeststhat, consistent with Markowitz9

portfolio theory, managers control riskby using optimisation technology to find

the portfolio with the least amount ofrisk for a given level of expectedreturn.10 Bodie et al. argue that anoptimal, actively managed portfolio isone that maximises the reward-to-variability ratio, that is, the expectedexcess return (relative to the benchmark)divided by the standard deviation ofreturns.11 Reilly and Brown suggest thatactive equity portfolio management is anattempt by the manager to outperform,on a risk-adjusted basis, a passivebenchmark portfolio where the averagecharacteristics of the benchmark(including such factors as beta, dividendyield, industry weighting and firm size)match the risk-return characteristics ofthe client.12 Similarly, Jacobs and Levypropose that the goal of optimisation foractive managers is to maximise portfolioreturn while tying portfolio risk to thatof the benchmark; specifically, that theportfolio’s systematic risk matches thatof the benchmark.13

Yet, these academic tenets are rarelypresented as matters of concern, muchless policy, by most active fundmanagers. In an interview with PeterTanous, the manager of the BrandywineFund with over US$4bn in assets, FosterFriess, offers this astonishing admission:

‘We do not relate to most of those tenets

of modern portfolio theory. Asking us

about standard deviation would be like

asking a plumber how many kilowatts he

wanted to plug into a lamp. It just doesn’t

compute with us.’14

One possible explanation for the relativeinattention paid to modern portfoliotheory (MPT) by active managers is thefact that many mutual fund portfolioshold dozens if not hundreds of securities,and idiosyncratic diversification will

Mutual fund risk-return profiles: A novel use of triangulation


occur to some extent even if it is not anexplicit objective of the fund. Evans andArcher, for example, find that about 90per cent of the maximum benefit ofdiversification is derived from aportfolio of just 12–18 randomlyselected stocks.15 On the other hand,concentrated idiosyncratic risk may, infact, be an overt objective of the fundwhen the investment company has beenchartered as a ‘non-diversifiedinvestment company’ under the ICAand is therefore permitted to allocatemore than 5 per cent of fund assets to asingle security (this must be disclosed inthe fund’s prospectus). The GabelliValue Fund is an example of such afund. Holland suggests that fundmanagers avoid implementingoptimisation routines to find the efficientfrontier and the optimum risk-returnportfolio under MPT because of theuncertainty implicit in forecasting stockrisk and return characteristics.16 Thisargument fails to account, however, forthe fact that fund managers face similaruncertainties when, for example,accounting-based fundamental analysis isemployed to select stocks for a portfolio.

Interviews with fund managers andshareholder communications issued byinvestment companies suggest the raisond’etre of active fund management centreson the objective of picking stocks thatwill deliver high absolute returns. Forexample, when Fidelity Investmentsasked Jason Yee, manager of the JanusGlobal Value Fund, what his immediateand long-term objectives were, hereplied, ‘Good returns. I guess I think asmuch about providing good absolutereturns in both up and down markets asanything else’.17

Similarly, in an interview with PeterTanous, Eric Ryback, manager of the

Lindner Dividend Fund, was asked abouthis investment objectives. He replied,‘My personal goal is to do 20 per cent peryear, if I can. However, some marketsdon’t allow me to do that’.18

The objective of producing highabsolute returns is what one would expectif, as suggested by the Boston ConsultingGroup19 andMinsky,20 fund managersare assessed on the total return theydeliver to shareholders. Further, highabsolute returns will contribute positivelyto the total amount of assets undermanagement via capital appreciation and(to the extent that high absolute returnsare consistent with shareholderobjectives) new share sales. This, in turn,will enhance fee income and market valuefor most investment companies.

If an active investment manager isbeing compared with a performancebenchmark, then the difference betweenthe manager’s portfolio return and thebenchmark return, ie active or excessreturn, is of crucial importance to bothmanager and client. As described byGrinold and Kahn, the consistency ofthe active return as measured bydispersion, ie tracking error, active risk orresidual risk, will also be relevant to theclient.21 According to Grinold andKahn, investment managers and pensionplan sponsors are much more averse tothe risk of deviation from thebenchmark than they are averse to therisk associated with the benchmark itselfbecause a high level of residual riskmeans that there is a high probability ofbeing among the worst-performingmanagers with the resulting possibilityof termination. Thus, they argue thatmanagers tend to reduce the amount ofresidual risk in order to avoid thebusiness risk inherent in being placedlow in fund comparison tables. Loftus,

Silverman


however, suggests that, as the traditionalfocus within the active managementcommunity has been oriented moretowards the production of excess returnsthan towards controlling tracking error,it is common to find tracking error inexcess of 4 per cent — a substantiallevel.22

The academic literature and variousempirical studies suggest that investmentmanagers’ objectives and attitudestowards risk and return will vary basedon the type of fund they run, theircontracting and fee arrangements withthe investment company and theprovisions of the client contract orprospectus. Horan, for example, findsthat pension funds are more likely thanother types of portfolios to have lowtracking error and market betas close toone.23 He attributes this to the risk ofpersonal legal liability faced by pensionfund managers if returns are significantlyless than index benchmarks. By way ofcontrast, Horan suggests that managersof foundations and endowments arelikely to be more focused on absoluteperformance than tracking error as theseprofessionals are typically compensatedbased on a percentage of assets undermanagement — as the value of assetsgrows, so does the manager’s fee.Brinson et al. argue that investmentmanagers for insurance companies aremore likely to be concerned with assetclass allocation than tracking errorbecause they must focus on the ability tofund future liabilities.24

In a study involving structuredinterviews with UK equity fundmanagers, Baker finds that the mostimportant risk across fund types(pension, life assurance, private clientetc) is the failure to meet fundobligations, that is, asset-liability

matching, with 70 per cent of the totalsample regarding this as important.25

This is followed by commercial risk,which Baker defines as the risk of poorperformance relative to competitors,with 40 per cent of the total sampleregarding this as important.Interestingly, only 29 per cent of allfund managers thought total variabilityof return (as measured by standarddeviation) was important and a mere 14per cent thought the systematic risk ofthe portfolio (as measured by beta) wasimportant. This is both instructive andsurprising as it suggests that, on thewhole, UK fund managers have notadopted the traditional risk measurespredicted for individual investors byMarkowitz26 portfolio theory and thecapital asset pricing model (CAPM).Indeed, one pension fund managerinterviewed reveals, ‘I can certainly saythat no share has ever been bought onthe basis of its beta or standarddeviation. We only do it for interest andfor the trustee’s entertainment.’27

When Baker segregates the samplebased on the performance benchmark,she finds, perhaps expectedly, that thegroup most subject to relativeperformance monitoring — pensionfunds — is the only group in which amajority of the fund managers considercommercial risk to be important.28

Baker further finds that fund managersin this group perceive externalperformance measurement assignificantly affecting their portfoliodecisions, with almost half of the samplereplying that it has a ‘considerable’ or‘extreme’ influence upon portfoliomanagement. One pension fundmanager interviewed reveals:

‘If your livelihood depends on the



retention of funds and if those funds are

being measured by WM [whole market]

every 3 months, and if you know that

three or four bad quarterly performances

are going to cost you your fund, then you

tend to reduce and reduce and reduce

your exposure to anything outside the

index.’29

The process of progressively reducingexposure to assets outside of the indexwill result, if followed to its logicalconclusion, in a portfolio consistingexclusively of benchmark assets, ie, aportfolio with zero active risk. Thus,Baker’s results strongly suggest that, forequity pension fund managers whoseperformance is being compared withsome external benchmark, minimisingtracking error will be a key objective.This is consistent with the statedpreference of the equities managerinterviewed by Modigliani30 and theargument that pension plan sponsors31

and mutual fund managers32 are risk-averse to deviation from the benchmark.If, however, performance benchmarkingis pervasive in the fund managementindustry, Baker’s results would runcounter to the suggestion by Loftus33

that the traditional focus within theactive management community has beenoriented more toward the production ofexcess returns than toward controllingtracking error. Moreover, a paramountdesire on the part of the fund managerto minimise active risk would constrainportfolio choices and therefore likely beincompatible with the fund objective ofmaximising absolute or total shareholderreturn discussed earlier.

According to Jacob, an investmentmanager’s tendency to ‘mimic’ thebenchmark, that is, deviating from it inonly minor ways so as to avoid

underperforming it, is a widespreadactivity in the fund managementindustry.34 Jacob argues that managersknow they will not be fired for simplymatching benchmark returns so theirself-interest is served by minimising therisk of underperformance rather thanmaximising the chance ofoutperformance. This behaviour,however, is suboptimal from theshareholder’s point of view as he/she endsup paying the high fees associated withactive management but receives onlybenchmark returns (less fees). Given thistheorised conflict of interests betweenmanager and shareholder, one mightexpect an active fund manager whopursues a strategy of ‘closet indexing’ toface an increased risk of termination, yetJacob suggests the opposite is true.35

By way of contrast to the strategy ofminimising deviation from thebenchmark, Jacob proposes that othermanagers may attempt to ‘game’ thebenchmark by investing in asset classesthat are specifically excluded from thebenchmark.36 Peter Lynch, formermanager of the Fidelity Magellan mutualfund, is cited by Jacob as an example ofan investment manager whodramatically outperformed hisbenchmark, the Standard & Poor’s(S&P’s) 500, in the late 1980s byinvesting roughly 25 per cent ofMagellan’s capital in foreign stocks. Sucha strategy, of course, reflects an increasein the amount of active risk taken byLynch but, according to Jacob,37 wouldalso have resulted in a substantial rise inthe amount of systematic risk borne byfund shareholders.

Gaming the benchmark may alsowork to reduce systematic risk in theportfolio. Bolster and DiBartolomeo,for example, note that during a period

Silverman


when he believed the US equity marketwas grossly overvalued, Lynch’ssuccessor, Jeff Vinik, elected to shift alarge proportion of assets from largecapitalisation, growth equities to bondsand cash equivalents.38 Bolster anddiBartolomeo suggest that Vinikundertook this strategy in an effort tobetter his performance both in absoluteterms and relative to his peer group ofgrowth managers.39 The Magellan fundwas, however, classified as a large-cap,growth stock fund at the time40 so thisreallocation of assets would have resultednot only in a reduction of the systematicrisk exposure of the fund, but wouldhave significantly altered its formalinvestment style as well. Such a radicaladjustment would clearly be suboptimalfor a shareholder who chose to invest inMagellan with the specific objective ofgaining exposure to large-cap growthstocks, factor risks identified in theFama-French Three-Factor model.41 Inan effort to improve his absolute andrelative returns therefore, Vinik’sgaming strategy may have been inconflict with the objectives of Magellaninvestors.

Fund management that deviates fromthe benchmark significantly enough toresult in a change of style, and thereforeaverage factor risk, has beendocumented by DiBartolomeo andWitkowski42 who employ a Sharpe43

style analysis of 748 mutual funds andfind that a significant proportion ofmutual funds follow strategies that aredistinct from their stated objective.Using discriminant analysis, Kim et al.44

report similar results, finding that thestated objectives of over 50 per cent ofmutual funds do not match theirattributes-based objectives. Further, theselatter authors find that more funds are

deviating into lower-risk objectives thanhigher-risk objectives (as in thepreviously discussed case of Jeff Vinik atMagellan).

Benchmark gaming clearly has thepotential to create severe agency issuesbetween fund management companiesand their shareholders. Sirri and Tufano45

show that the best performing mutualfunds subsequently receive large netinflows of new money, yet funds thatdemonstrate average or poorperformance do not experience aproportionate outflow of investor capital.As most funds receive fees as a fixedpercentage of assets under management,46

this asymmetry creates a nonlinear payofffunction similar to that of a call optionwhere the fund management firm has anincentive to engage in volatile investmentstrategies so as to maximise returns andfees, regardless of the client’s riskpreference.47,48 This may drive fundmanagers to reach for risky assets outsideof the benchmark as Peter Lynch ofFidelity apparently did in the 1980s, asdiscussed above.

There are indications that fundmanagers have become more sensitive tothe potential agency problems associatedwith benchmark gaming. In April 2000,Pensions and Investments magazinereported that the J. Sainsbury plcPension Scheme had become the firstUK plan sponsor to implement apenalty arrangement with its investmentmanager, Barclays Global Investors Ltd(BGI), where BGI agreed to reimburse aportion of its management fee if theportfolio exceeded agreed limits fortracking error.49 Subsequently, inJanuary of 2003 Risk Magazine reportedthat BGI had ‘radically upgraded’ itsapproach to risk management andanalytics with new portfolio



optimisation software designed to focuson the control of active risk.50 TheGlobal Portfolios Group of ChaseFleming Asset Management has alsocalled attention to its commitment tocontrolling active risk by announcingthat, as part of its asset allocationstrategy, portfolio tracking error wouldnot exceed 1–2 per cent.51

RESEARCH PURPOSE ANDMETHODOLOGYThe 1990s witnessed unprecedentedgrowth in the value of US mutualfunds. Total assets under management inthe mutual fund industry rose more than700 per cent from just under US$1trn in1990 to a peak of nearly US$7trn by2000, about half of this growth comingfrom investment returns.52 The samereport shows that, during the decade,the number of US households owningmutual funds increased from 23 millionto 50 million.53 Indeed, the 1990srepresented a landmark decade forinvestment companies in terms of assetsunder management, shareholderparticipation rates and overall capitalappreciation, as well as for stock marketreturns — between January 1995 andDecember 1999, the S&P 500 Indexincreased a record 220 per cent while theNASDAQ Composite Index grew by anextraordinary 440 per cent. In view ofthese historic gains for equities andmutual funds, it was decided to focus thepresent investigation on the investmentactivities of equity fund managersduring the latter part of this period, ie1996 to 2000. Specifically, the studyaims to discover the investmentobjectives and risk-taking of actively-managed equity mutual funds and tomap the relationships between objectivesand risks.

MethodologyThe study purpose suggests a multi-methods approach to investigation.While the task of documenting classesand levels of investment risk yieldsreadily to a traditional quantitativemethod in the sense that the relevantvariable is well defined and positivismhas been used for purposes of ‘measuringand quantifying phenomena’,54

apprehending a mutual fund’sinvestment objective(s) dictates a morequalitative and interpretive approach,relying critically on insight into theintentions of fund managers, for whichthere may be no easily discernible proxythat lends itself to pure quantitativeanalysis. Alternatively then, Patton notesqualitative methods are ‘ways of findingout what people do, know, think, andfeel’.55

For Patton, credibility dependscritically on ‘rigorous techniques andmethods for gathering high-quality datathat are carefully analyzed, withattention to issues of validity, reliabilityand triangulation’ [emphasis added].56

Triangulation, a concept originallyintroduced by Webb et al.57 and thenapplied to qualitative investigation byDenzin,58 captures the notion of theimportance of examining phenomenafrom a variety of vantage points so as toestablish the trustworthiness (validity) ofqualitative investigation. Denzin definesfour basic types of triangulation: (1)methodological triangulation, (2) datatriangulation, (3) investigatortriangulation and (4) theorytriangulation. The present study willemploy types (1) and (2) of Denzin’striangulation framework. The multi-methods approach to investigation notedabove — a mixture of qualitative andquantitative analysis — is the

Silverman


embodiment of methodologicaltriangulation. According to Denzin,‘between-method triangulation can takemany forms, but its basic feature will bethe combination of two or moredifferent research strategies in the studyof the same empirical units’.59 Therationale for this strategy is that ‘theflaws of one method are often thestrengths of another; and by combiningmethods, observers can achieve the bestof each while overcoming their uniquedeficiencies’.60 Data triangulation refersto the selection of dissimilar comparisongroups as a sampling strategy in order todiscover what concepts the groups havein common across settings. Within thedata triangulation construct, Denzindelineates subtypes of time, space andperson, in the sense that ‘all sociologicalobservations relate to activities ofsocially situated persons’ and ‘a focus ontime and space as observational unitsrecognizes their relationship to theobservations of persons’.61

In this study, the ‘empirical units’ arethe mutual fund management companies(person as collectivity62) and the datasources will be a combination of (1)mutual fund prospectuses and annualreports (primary data) and (2) historicalmutual fund risk-returns data. The spaceor ‘social area’ in the context of thisstudy is the medium within whichofficial shareholder communications takeplace, namely, the written documentsfrom investment companies specified in(1) above. The time subtype is addressedby examining mutual fund prospectusesfor indications of fund managementcompany investment objectives/risk-taking ex ante, while examining annualreports and secondary data for evidenceof fund management companyinvestment objectives/risk-taking ex post.

Silverman notes that ethnographyseeks to understand the organisation ofsocial action in particular settings63 andHammersley and Atkinson argue that,as written accounts are an importantfeature of many settings, ethnographersneed to take account of documents as areflection of the setting underinvestigation.64 These latter authors citethe extensive use of documents byGamst65 in his study of locomotiveengineers and in research emanatingfrom the early Chicago School. Pattonobserves that a particularly rich sourceof information about manyprogrammes is programme records anddocuments and that ‘in contemporarysociety, all programs leave a trail ofpaper that the evaluator can follow anduse to increase knowledge andunderstanding about the program’.66

Similarly, Silverman suggests thatpublic records constitute a ‘potentialgoldmine’ for investigation in large partbecause they reveal how agenciesaccount for and legitimate theiractivities.67

Hammersley and Atkinson offer anenthusiastic endorsement of the use ofdocuments in qualitative research:

‘The presence and significance of

documentary products provides the

ethnographer with a rich vein of analytic

topics, as well as a valuable source of

information. Such topics include: How

are documents written? How are they

read? Who writes them? Who reads

them? For what purpose? On what

occasions? With what outcomes? What is

recorded? What is omitted? What is taken

for granted? What does the writer seem to

take for granted about the reader(s). What

do readers need to know in order to make

sense of them?’68



Altheide suggests that several aspects ofan ethnographic research approach canbe applied to document analysis toproduce ethnographic content analysis(ECA),69 defined generally as thereflexive analysis of documents.70

Although Altheide71 acknowledges thatECA has been less widely recognised as adistinctive research method, he notesthat various facets of the approach areapparent in document analyses byhistorians, literary scholars and socialscientists. More recently, Altheide72

defines ECA as ‘an integrated method,procedure and technique for locating,identifying, retrieving and analyzingdocuments for their relevance,significance and meaning’, where ‘theemphasis is on discovery and description,including search for contexts, underlyingmeanings, patterns and processes’.Thematic categories using ECA areestablished both a priori and through acontent analysis of text. Once theresearcher has derived categories and acoding system from the ethnographicstudy data, quantitative content analysisdesign provides a set of procedures tosystematically code the categories withreliability checks to analyse, validate andreport the results.73,74 Within ECA,concept development, sampling, datacollection, data coding, data analysis andinterpretation are reflexive.75,76

Although creation of a mutual fundprospectus and annual report typicallyreflect the efforts of various agents —investment banks, lawyers, advertisingfirms etc — these documents are aproduct, ultimately, of the investmentcompany offering the mutual fund. Assuch, prospectuses and annual reports areessentially cultural artefacts77 or ‘patternsof cultural construction’78 unique to aspecific group of industry professionals.

Thus, notwithstanding the fact that theprospectus and annual report areostensibly written for investors and mustmeet certain standards by law forreadability and intelligibility, one mayexpect that these documents will containtechnical terminology, conceptualthemes and language generally, that areidiosyncratic to the investment companyand/or to the fund managementindustry. Meaning and significance (forthe outsider) may be obscured,ambiguous or heavily nuanced.Prospectuses and annual reportstherefore, are subject to rigorousdiscovery and interpretive researchprocedures, supporting an ECAapproach to analysis.

Data sourcesAs indicated earlier, the social area in thecontext of this study — the locus whereinvestment company and investor ‘meet’— is the mutual fund prospectus.Although the social mode of theprospectus is a monologue, itnevertheless represents the setting withinwhich the investment company and theinvestor interact, in the sense thatinformation is exchanged.

The primary data sample has beendrawn from US mutual fundprospectuses and annual reportsemploying a strategy that reflectspurposeful sampling. The sample ispurposeful in that funds have beenselected for inclusion based on a numberof criteria that are ‘congruent with thestudy purpose and that will yield data onmajor study questions’79 specifically: (1)age of the fund, (2) type of fund, (3)amount of assets under management and(4) representation in the secondary datasample. Criterion 1 is necessary toinclude in the data set only those mutual

Silverman


funds that were operating during thestudy window. Criterion 2 is necessary toinclude in the data set only those mutualfunds permitted by charter to (i) holdequities generally, (ii) hold equitiesoutside of a narrow class of securities, iesector funds and (iii) actively manage theportfolio. Criterion 3 is intended tocapture large active equity funds in termsof assets under management, ie thosethat, in the aggregate, controlled a largeshare of the equity market in the late1990s. Finally, criterion 4 facilitates acomparative analysis of mutual fundinvestment objectives/risk-taking ascommunicated to investors ex ante in theprospectus and ex post in the annualreports, with actual levels of portfolio riskincurred by these same funds over theassociated observation period. Thus,criterion 4 contributes to the datatriangulation aspect of the study (see theearlier discussion). A total of 40documents — prospectuses and annualreports — issued by four US mutualfunds over the period 1996–2000 wereselected for analysis.

The secondary data have beenprovided by Lipper, a Reuters company,and consist of the following measuresfor each fund for each calendar yearbetween 1995 and 2000: standarddeviation, beta and R2 against the S&P500 Index. The modern portfolio theory(MPT) statistics (beta, R2) are estimatedby performing a least-squares regressionof the fund’s return over US Treasurybills (excess return) and the excessreturns of the S&P 500 Index on amonthly basis. Beta and standarddeviation figures are annualised. Asecond data sample consists of sectorweightings for each of the mutual fundsappearing in the primary data sample.Sector weightings are reported once

annually for each year between 1996 and2000. These data have been provided byMorningstar, Inc.

ProceduresThe mutual fund prospectuses andannual reports were examined usingECA as discussed above. The materialswere initially reviewed to enhancefamiliarity with the basic organisationalstructure, conceptual framework andterminology across documents. Thisprocess was facilitated by previousexperience with similar documents inboth research and portfoliomanagement. The literature review andpreliminary questions for study helpedto identify some general conceptualcategories and descriptive themes in thematerials, which were noted. Categoriesprovide structure for grouping units ofanalysis into the same conceptual unitsthat have similar meaning,80 however, asAltheide points out, ‘although [certain]categories and variables initially guidethe [ECA] study, others are allowed andexpected to emerge throughout thestudy’.81 Thus, ECA is a process ofconstant discovery and constantcomparison of meanings and nuances.82

To avoid a claim of ‘forcing’ anoutcome, however, it is important toensure that caution is employed inrelation to how categories are definedand in assigning patterns of language useinto each category. In the present paper,this type of control was aided by thelegal obligation of the investmentcompany to disclose specific informationwithin the prospectus.

Employing a reflexive process, thestudy group worked across documentsin the sample and employed a process ofiteration between the data and theemerging classification system. Although



ethnographic analysis allows for data tobe allocated to more than onecategory,83,84 it was found that in thisstudy it was possible to constructcategories which ultimately met thecriteria of ‘internal homogeneity’ and‘external homogeneity’ as set forth byGuba;85 the former being demonstratedby the fact that the data within eachcategory dovetailed in a meaningful wayand the latter because there is a cleardifference between categories, that is, alack of overlap and duplication. It was,however, possible to establishrelationships between categories whichenabled questions to be refined andfacilitated analysis of the data withquantitative content analysis. AsAltheide suggests, once the categoriesand coding system have been derivedusing ECA, a quantitative content analysisdesign can provide a set of procedures tocode the categories systematically withreliability checks embodied in theframework to analyse, validate andreport results.86 The final categories anddefinitions were as follows:

. Investment objective: Statements or data

that indicate the objective(s) or goal(s) of

the fund. Statements were made either

by specific, named individuals, eg ‘John

Smith’ the fund manager, or by the

investment company as a collective, or

both.

. Risk exposure: Statements or data that

indicate the type of risks an investor in the

fund may be (ex ante) or were (ex post)

exposed to. As in the prior category,

statements were made either by specific

named individuals or by the investment

company as a collective, or both.

As indicated earlier, ethnographiccontent analysis is used to document and

understand the communication ofmeaning, as well as to verify theoreticalrelationships.87 Informed by thisapproach, the individual data elements ineach category were examined within thecontext of the various theoretical modelsof risk and return discussed in theliterature review, interpreted andsubsequently subsumed into broader,generalised categories for furtheranalysis. The data generated by thecoding system were analysed usingquantitative frequency counts as well asby qualitative narrative data from thetext.88 Thus, narrative and numeric dataare presented concurrently89 to support adescriptive interpretation. Validation issupported by the use of examples fromthe text itself to demonstrate claims90

and through data triangulation91wherethe coded data appearing in a mutualfund prospectus (ex ante) are comparedwith the coded data appearing in thecomplementary annual report (ex post)for the same year.92

Reliability and validity of resultsdrawn from ethnographic contentanalysis are enhanced via intercoderreliability.93 This procedure is derivedfrom quantitative content analysis suchthat categories are tested forreproducibility when coders, unaware ofthe objectives of the study, analyse thesame text and identify comparable datacategories and elements.94 For this study,a recent MBA graduate was trained inthe coding schema described above andasked to code text from a randomsample of documents that had previouslybeen coded. Interrater reliability wasassessed using a Scott’s phi95 to correctfor chance agreement. Krippendorff96

suggests that a phi reliability score above0.80 indicates trustworthy coding forsupporting category definitions and

Silverman


drawing inferences. The secondarycoder’s percentage of agreement withthis researcher was 0.82. When highreliability is achieved, the coding rulescan be applied to the entire text.97

As US mutual funds frequently fail todisclose their target or ‘best-fit’benchmark index in officialcommunications, changes in active/residual risk,98 that is, the risk associatedwith active management of a portfoliosuch that holdings diverge from apassive benchmark portfolio, wereassessed by including in the test sampleonly those funds with R2 measuresrelative to the S&P 500 Index of greaterthan 0.50 in the base year (1995) and inevery year of the study window (1996–2000). This ensures a good fit with theS&P 500 and contributes to themeaningfulness of the tracking errorvalues. The significance of R2 values isdiscussed by Field.99

RESULTS AND ANALYSISFigure 1 illustrates the data elements

found in the investment objective andrisk-taking categories. Workingdefinitions for the data elements havebeen drawn from the financial modelsdiscussed in the literature review; thus,for example, text referring to the riskassociated with investing in a specificsecurity was classified as exposure to‘idiosyncratic risk’.100

Frequency counts and descriptivestatistics for the data elements bydocument type and year for each fundare presented in Tables 1–4, respectively.

Observed levels of volatility (asmeasured by standard deviation), marketrisk (as measured by beta), residual risk(as measured by tracking error) andpseudo-industry risk (as proxied by theproportion of portfolio assets invested intechnology firms) for each fund by yearare provided in Table 5. Thepresentation of data in this formatparticularly facilitates the time subtype ofthe triangulation aspect of the analysis ascounts in prospectuses (ex ante, narrative)can be readily compared with counts in

High absolute returns High relative returns(non-risk adjusted)

High relative returns(risk-adjusted)

High relative returns(peers)

Investment objective

Factor

Idiosyncratic

Market

Interest rate

Absolute loss

Relative underperformance(non-risk adjusted)

Relative underperformance(peers)

Failure to meet objective

Volatility(unspecified)

Active/residual

Risk exposure

Prospectus/annual report

Figure 1: ECA data categories and data elements



Table 1: Vanguard Tax Managed Capital Appreciation Fund frequency counts and descriptive statistics


High Highrelative relative High

High returns returns relativeabsolute (non-risk- (risk- returnsreturns adjusted) adjusted) (peers)

Prospectus: 04/19/1996 4 1 5 1Annual report: 12/31/1996 3 2 2 4Prospectus: 04/18/1997 3 1 4 1Annual report: 12/31/1997 2 2 2 7Prospectus: 04/27/1998 3 1 4 1Annual report: 12/31/1998 6 1 7 8Prospectus: 02/22/1999 2 0 3 0Annual report: 12/31/1999 9 1 12 15Prospectus: 03/28/2000 3 0 4 0Annual report: 12/31/2000 2 1 6 7

Mean: All documents 3.70 1.00 4.90 4.40SD: All documents 2.10 0.63 2.81 4.61

Prospectus: 04/19/1996 4 1 5 1Prospectus: 04/18/1997 3 1 4 1Prospectus: 04/27/1998 3 1 4 1Prospectus: 02/22/1999 2 0 3 0Prospectus: 03/28/2000 3 0 4 0

Mean: Prospectuses 3.00 0.60 4.00 0.60SD: Prospectuses 0.71 0.55 0.71 0.55

Annual report: 12/31/1996 3 2 2 4Annual report: 12/31/1997 2 2 2 7Annual report: 12/31/1998 6 1 7 8Annual report: 12/31/1999 9 1 12 15Annual report: 12/31/2000 2 1 6 7

Mean: Annual reports 4.40 1.40 5.80 8.20SD: Annual reports 3.05 0.55 4.15 4.09

Risk exposure

Relativeunder Relativeperformance under Failure

Absolute Interest Active/ (non-risk- performance to meet Volatilityloss rate Market Idiosyncratic Factor residual adjusted) (peers) objective (undefined)

Prospectus: 04/19/1996 4 1 3 0 2 5 1 0 1 1Annual report: 12/31/1996 0 0 1 0 0 6 0 0 0 0Prospectus: 04/18/1997 4 0 4 0 2 6 1 0 1 1Annual report: 12/31/1997 0 0 2 0 0 5 0 0 0 0Prospectus: 04/27/1998 3 0 4 0 0 7 1 0 1 1Annual report: 12/31/1998 1 0 3 0 0 2 1 0 0 1Prospectus: 02/22/1999 4 0 2 0 3 2 0 0 0 1Annual report: 12/31/1999 1 0 1 0 1 1 0 0 0 1Prospectus: 03/28/2000 4 0 3 0 3 1 0 0 0 1Annual report: 12/31/2000 2 0 1 0 2 2 0 3 0 1

Mean: All documents 2.30 0.10 2.40 0.00 1.30 3.70 0.40 0.30 0.30 0.80SD: All documents 1.62 0.30 1.11 0.00 1.19 2.19 0.49 0.90 0.46 0.40

Prospectus: 04/19/1996 4 1 3 0 2 5 1 0 1 1Prospectus: 04/18/1997 4 0 4 0 2 6 1 0 1 1Prospectus: 04/27/1998 3 0 4 0 0 7 1 0 1 1Prospectus: 02/22/1999 4 0 3 0 3 2 0 0 0 1Prospectus: 03/28/2000 4 0 4 0 3 1 0 0 0 1

Mean: Prospectuses 3.80 0.20 3.60 0.00 2.00 4.20 0.60 0.00 0.60 1.00SD: Prospectuses 0.45 0.45 0.55 0.00 1.22 2.59 0.55 0.00 0.55 0.00

Annual report: 12/31/1996 0 0 1 0 0 6 0 0 0 0Annual report: 12/31/1997 0 0 2 0 0 5 0 0 0 0Annual report: 12/31/1998 1 0 4 0 0 2 1 0 0 1Annual report: 12/31/1999 1 0 1 0 1 1 0 0 0 1Annual report: 12/31/2000 2 0 1 0 2 2 0 3 0 1

Mean: Annual reports 0.80 0.00 1.80 0.00 0.60 3.20 0.20 0.60 0.00 0.60SD: Annual reports 0.84 0.00 1.30 0.00 0.89 2.17 0.45 1.34 0.00 0.55

Silverman


Table 2: Legg Mason Value Trust frequency counts and descriptive statistics




Prospectus: 07/14/1995 16 5 1 0Annual report: 03/31/1996 16 24 0 4Prospectus: 07/31/1996 16 20 1 3Annual report: 03/31/1997 36 9 0 1Prospectus: 07/31/1997 16 6 2 3Annual report: 03/31/1998 26 7 0 1Prospectus: 05/29/1998 16 6 2 3Annual rReport: 03/31/1999 27 7 1 5Prospectus: 07/30/1999 9 5 3 3Annual report: 03/31/2000 25 8 0 3




Annual report: 03/31/1996 16 24 0 4Annual report: 03/31/1997 36 9 0 1Annual report: 03/31/1998 26 7 0 1Annual report: 03/31/1999 27 7 1 5Annual rReport: 03/31/2000 25 8 0 3


Risk exposure


Absolute Interest Idio- Active/ (non-risk- performance to meet Volatilityloss rate Market syncratic Factor residual adjusted) (peers) objective (undefined)









Table 3: Vanguard Asset Allocation Fund frequency counts and descriptive statistics










Risk exposure









Silverman


Table 4: Brandywine Fund frequency counts and descriptive statistics










Risk exposure











Table 5: Observed levels of risk

Std Dev Beta Tech %* Trk Err Std Dev Beta Tech %* Trk Err Std Dev Beta Tech %* Trk Err

12/31/1994 12/31/1994 31/12/1994 12/31/1994 12/31/1995 12/31/1995 12//31/1995 12/31/1995 12/31/1996 12/31/1996 31/12/1996 12/31/1996

Fund name 12/31/1995 12/31/1995 31/12/1995 12/31/1995 12/31/1996 12/31/1996 12/31//1996 12/31/1996 12/31/1997 12/31/1997 31/12/1997 12/31/1997

Vanguard TM Cp App; Inv 6.90 1.12 N/A 1.25 12.08 1.09 22.50 1.35 16.34 1.01 18.70 1.51

Legg Mason Value Tr; Prm 7.72 1.21 N/A 1.5 11.69 1.02 15.00 1.31 18.89 1.11 26.80 2.39

Vanguard Asset Alloc; Inv 4.34 0.71 N/A 0.86 8.42 0.77 11.30 1.06 11.03 0.70 14.10 1.63

Brandywine Fund 15.34 1.29 N/A 4.23 15.64 1.25 40.40 2.80 19.24 0.86 35.40 4.22

Legg Mason Value Tr; Prm 7.72 1.21 N/A 1.50 11.69 1.02 15.00 1.31 18.89 1.11 26.80 2.39

Vanguard 500 Index; Inv 4.98 1.00 0.01 10.40 1.00 0.01 15.27 1.00 0.01

Fund name Std Dev Beta Tech %* Trk Err Std Dev Beta Tech %* Trk Err Std Dev Beta Tech %* Trk Err

12/31/1997 12/31/1997 12/31/1998 12/31/1997 12/31/1998 12/31/1998 31/12/1998 12/31/1998 12/31/1999 12/31/1999 31/12/1999 12/31/1999

12/31/1998 12/31/1998 12/31/1999 12/31/1998 12/31/1999 12/31/1999 31/12/1999 12/31/1999 12/31/2000 12/31/2000 31/12/2000 12/31/2000

Vanguard TM Cp App; Inv 24.82 1.21 23.20 1.50 15.30 1.16 30.60 1.47 19.83 1.12 23.70 2.48

Legg Mason Value Tr; Prm 27.62 1.29 21.20 3.11 19.64 1.28 34.40 3.54 19.31 1.13 13.30 1.93

Vanguard Asset Alloc; Inv 13.18 0.62 15.60 2.33 7.46 0.56 23.80 1.89 9.04 0.53 28.40 2.43

Brandywine Fund 27.41 1.27 48.20 3.70 19.38 1.22 60.70 3.64 24.06 0.88 30.90 5.88

Legg Mason Value Tr; Prm 27.62 1.29 21.20 3.11 19.64 1.28 34.40 3.54 19.31 1.13 13.30 1.93

Vanguard 500 Index; Inv 20.61 1.00 0.02 12.60 1.00 0.03 16.40 1.00 0.03

*As per Morningstar Principia database. Figures do not always correspond with those appearing in annual reports.

Silv

erm

an

Jo

urn

al

of

Ris

kM

an

ag

em

en

tin

Fin

an

cia

lIn

stitu

tion

sV

ol.

1,

21

91

–2

22

#H

en

ryS

tew

art

Pu

blic

atio

ns

17

52

–8

88

7(2

00

8)

20

8

annual reports (ex post, narrative) andmeasured risk exposures (ex post,observed).

A Spearman correlation coefficientmatrix for the data elements across alldocuments, years and funds appears inTable 6.

At a confidence interval of 95 per cent,the association between the investmentobjective of ‘high absolute returns’ and‘idiosyncratic’ risk is positive andsignificant; the associations between ‘highabsolute returns’ and ‘market’ risk and‘active’ risk, respectively, are eachnegative and significant. Similarly, theassociation between the investmentobjective of ‘high relative returns (non-risk-adjusted)’ and ‘idiosyncratic’ risk ispositive and significant, while theassociation between the same objectiveand ‘market’ and ‘active’ risk,respectively, is again negative andsignificant. Conversely, the investmentobjective of ‘high relative returns (risk-adjusted)’ is negatively associated with‘idiosyncratic’ risk and positivelyassociated with both ‘market’ risk and‘active’ risk, respectively. Finally, theassociation between the investment

objective of ‘high relative returns (peers)’and the risks of ‘absolute loss’ and ‘failureto meet investment objective’ are eachnegative and significant.

The empirical results suggest thatwhere disclosure and reportingdocuments reflect a focus on theinvestment objectives of delivering highabsolute returns or high relative returnson a non-risk-adjusted basis, mutualfund narratives orient around specificrisk associated with individual securities;conversely, there is a dearth of discussionon portfolio risk relative to either themarket or an appropriate benchmark.One manifestation of this tendency, expost, is that historical returns for thesefunds are typically presented in narrativetext and tables that omit performancecomparisons with relevant marketindices. These findings are consistentwith predictions in the literature where,in the pursuit of high absolute returns,active equity managers may be expectedto concentrate on ‘stock picking’ ratherthan portfolio optimisation within asystematic or mean-variance framework.This frequently results in a portfoliowhich contains a substantial amount of

Table 6: Spearman correlation coefficients

Investment Objective

Relative returns Relative returns Relative returns

Risk exposure Absolute returns (non-risk-adjusted) (risk-adjusted) (peers)

Absolute loss –0.033386708 0.057773846 –0.145599367 –0.420637349

Interest rate –0.035179222 0.111940175 –0.100295795 –0.075097683

Market –0.479938474 –0.442475398 0.374520252 –0.283995172

Idiosyncratic 0.646312095 0.481583042 –0.451428532 –0.081205730

Factor –0.148998854 0.019416100 –0.054517629 –0.139625453

Active/residual –0.497642902 –0.369582384 0.341463824 –0.223573262

Relative underperformance –0.075545522 0.271206415 0.253875415 0.127700234

(non-risk-adjusted)

Relative underperformance –0.034938954 –0.093313938 0.221780373 0.073968375

(peers)

Failure to meet objective –0.293840046 –0.263611605 0.056746469 –0.500775465

Volatility (undefined) –0.252929602 –0.110206568 0.153198095 –0.219585890



unique risk and where performance isdetached from an appropriatebenchmark.101

Conversely, the data suggest that,where a fund adopts the primaryobjective of producing high relativereturns (risk-adjusted), the investmentcompany draws attention to market andactive risk in shareholdercommunications while devoting aminimum amount of space to disclosureor discussion of idiosyncratic risk; theformer can be seen in investmentperformance that is typically comparedwith an appropriate best-fit index. Theseresults too are consistent with theliterature, which suggests that managerswho focus on risk-adjusted returns willbe inclined to construct well-diversifiedportfolios closely linked to anappropriate benchmark and whereexcess returns will be governed largelyby adjustments to levels of systematicand residual risk.102,103

Where the funds in the present studypursue the objective of outperformingtheir peers (competing funds), they areparticularly averse to a discussion of therisk of (1) absolute loss, ie loss ofinvestor capital and (2) failing to meettheir investment objective. Thedisinclination to warn investors on thefirst possibility may simply be due torelative indifference to absolute loss, arisk factor which is not directly relatedto a failure to deliver superiorcomparative performance. Reticence toadvise investors that the fund may fail tomeet its objective, however, clearlyreflects a failure to disclose a relevantrisk to mutual fund shareholders.

The following two sections provide adetailed exposition of the ECA resultsfor two of the four funds in thesample.104

Vanguard Tax Managed CapitalAppreciation FundThe Vanguard Tax Managed CapitalAppreciation Portfolio (VMCAX) waslaunched in 1994 and is one of over 200funds in the Vanguard family. VMCAXcurrently manages over US$4bn in assetsand is currently categorised as a ‘largeblend’ fund by Morningstar. Accordingto the most recent prospectus (20th July,2007), ‘the fund seeks to provide a tax-efficient investment return consisting oflong-term capital appreciation’ bypurchasing ‘stocks that pay lowerdividends and are included in the Russell1000 Index’.

Across all documents and years forVMCAX, the investment objectiveoccurring with the highest meanfrequency is ‘high relative return (risk-adjusted)’ (see Figure 2). On average,this objective had the highest number ofoccurrences across prospectuses,although it came second to ‘high relativereturns (peers)’ across annual reports.Typically, VMCAX identifies itsbenchmark index in the prospectus andsubsequently compares fundperformance for the year with thisbenchmark in the annual report. Anexemplary reference from the 1996prospectus reads:

‘The Capital Appreciation Portfolio seeks

to provide growth of capital with

nominal current income from investments

in equity securities; the Russell 1000 Index

is the Portfolio’s benchmark index.’

Across all documents and years forVMCAX, the risk exposure occurringwith the highest mean frequency is‘active/residual risk’. This exposure alsooccurs most frequently, on average,across prospectuses and annual reports,

Silverman


respectively; although in 1999, it wassurpassed by the number of references to‘absolute loss’ and ‘factor risk’ in thecombination of prospectus and annualreport and in 2000 by references to‘absolute loss’, ‘factor risk’ and ‘marketrisk’ in the combination of prospectusand annual report. In the 1997prospectus, a reference to active/residualrisk reads, ‘The Portfolio emphasizeslow yielding stocks; therefore, its returnwill vary from the return of the Russell1000 Index’.

Note that, while VMCAX is tied viabenchmarking to the Russell 1000Index, investors are effectively advisedthat they will be exposed to active/residual risk associated with the active

selection of securities.In Table 5, one can see measured

levels of risk for VMCAX for the baseyear 1995 and for subsequent years,corresponding roughly to the periods foreach prospectus and annual reportcombination. R2 measured against theS&P 500 Index is relatively high,fluctuating between a minimum of 0.65in the base year and a maximum of 0.99in 1998; so although the Russell 1000 isthe benchmark index reported incompany documents, the S&P 500 is areasonably good fit index for VMCAXand market betas and tracking errorsestimated against the S&P 500 aredeemed meaningful risk indicators forVMCAX.

Abs

olut

e re

turn

s

Rel

ativ

e re

turn

s (n

on-

risk-

adju

ste

d)

Rel

ativ

e re

turn

s (r

isk-

adju

ste

d)

Rel

ativ

e re

turn

s (p

ee

rs)

Abs

olut

e lo

ss

Inte

rest

rat

e

Mar

ket

Idio

sync

ratic

Fact

or

Act

ive/

resi

dual

Rel

ativ

e un

derp

erfo

rman

ce(n

on-

risk-

adju

ste

d)

Rel

ativ

e un

derp

erfo

rman

ce(p

ee

rs)

Fai

lure

to m

eet o

bjec

tive

Vol

atili

ty (

unde

fined

)

S1

3.7

1

4.9

4.4

2.3

0.1

2.4

0

1.3

3.7

0.40.3 0.3

0.80

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5Investment objective

Risk exposure

Figure 2: Vanguard Tax Managed Capital Appreciation Fund mean counts across all years and documents



As discussed above, ‘active/residualrisk’ is the most frequently occurringreference, on average, across alldocuments in the risk exposure category.However, references to active/residualrisk decline steeply beginning with theannual report in 1998 and remaininfrequent in both the prospectuses andannual reports for 1999 and 2000. In adivergence from the narrative data,measured tracking error figures forVMCAX rise in a near-linear fashionfrom a low of 1.25 in the base year to ahigh of 2.48 in 2000 (see Table 5). Bycontrast, the secondary data generallysupport the primary data for ‘marketrisk’, which occur second in frequencyto ‘active/residual risk’ across allprospectuses and annual reports. Withthe exception of 1997, for example,estimated systematic betas for VMCAXmeasure well above 1.00 throughout thestudy window. Thus, between 1996 and2000, fund managers for VMCAXappear to have taken on high levels ofmarket risk and generally increasinglevels of active/residual risk. In the caseof the former, this pattern was beingsubstantially communicated toshareholders in official documents; in thelatter, disclosure was strong from 1996to 1998 but then dropped off sharply.

Another notable failure tocommunicate exposure is in the categoryof ‘factor risk’. Whereas the proportionof technology stocks in the VMCAXportfolio was already relatively large inthe base year (20.2 per cent) and grew toover 30 per cent of the portfolio in 1999,there was an average of only 1.3references to ‘factor risk’ across allprospectuses and annual reports,although this figure did rise to a total offive references in the 2000 documents.This omission is particularly critical

from the perspective of compliance aseach of the VMCAX prospectusesindicate that the portfolio will notconcentrate its investments in aparticular industry. Indeed, the 1997–2000 prospectuses specifically state thatthe fund will not ‘invest more than 25per cent of its assets in any one industry’— a covenant that appears to have beenviolated during at least one period, in1999. While it may be argued thatVMCAX had over 25 per cent of itsassets in the technology sector (asopposed to industry), the individualholdings in this group are likely to havebeen similarly sensitive to many of thesame macroeconomic and/orfundamental factors and thus havehighly correlated returns.105–107 Animplicit acknowledgment of thisassociation appears in the 1999 annualreport where the chairman and chiefexecutive officer of Vanguard, John J.Brennan, reports:

‘Technology stocks rocketed higher

during 1999 and carried most market

indexes along with them. The Vanguard

Tax-Managed Funds benefited from this

advance, each posting double-digit returns

during the 12 months ended December

31, 1999.’

By formally classifying technology as asector rather than an industry, VanguardFunds may have escaped legal liabilityfor concentrating its assets in possiblecontravention of its prospectus;nevertheless, the data suggest VMCAXshareholders were subject to substantial— and relatively undisclosed — factorrisk during the study window.

As discussed above, an analysis offrequency counts suggests that VMCAXfocused on achieving high risk-adjusted

Silverman


relative returns as compared with eithernon-risk-adjusted relative returns orabsolute returns. In the documents,performance comparisons are oftendrawn against the benchmark index, theRussell 1000, and Vanguard informsshareholders that returns will typicallyvary from the benchmark as a result ofthe fund manager’s preference for low-yielding, ie ‘growth’ stocks.Interestingly, Vanguard fails to point outthat, as growth stocks have a muchlower risk premium historically than‘value’ stocks,108 expected returns forVMCAX will be lower than those ofthe more diversified Russell 1000 and,thus, realised returns are relatively betterthan reported when adjusted for risk. Infact, Vanguard actually misinformsshareholders by stating:

‘over time we expect the returns of the

Capital Appreciation Portfolio and the

Russell 1000 Index to be similar because

the short-term differences in returns

between the growth and value segments

of the market have tended to even out

over longer periods.’

The abovementioned French and Famafindings on risk premiums are based ondata between 1927 and 2001 — a verylong period to wait for the growthsegment to draw even with the valuesegment.

Two other points to emerge from theprimary data deserve note. First, theVMCAX documents contain noreference to idiosyncratic risk. This iswhat one might expect given a quasi-indexing management approachresulting in a well-diversified portfolioof several hundred securities. The lack ofattention to idiosyncratic risk contrastsmarkedly, however, with that seen in

the documents of Legg Mason ValueTrust to be discussed later in this paper.

Secondly, the investment objective of‘high relative returns (peers)’ occurs onlythree times in five prospectuses but is themost frequent investment objectiveacross the annual reports, averaging 8.2references. This is perhaps due, in part,to the nature of the annual report as amedium for communicating investmentresults; many of the occurrences appearin performance tables which typically donot comprise a part of the prospectusformat. Beyond this consideration,however, it is noteworthy that VMCAXexceeded the performance of its peergroup of funds in every year of thestudy window, with the exception of2000. In the annual report for this finalyear, references to ‘high relative returns(peers)’ decline by over 50 per cent fromthe 1999 annual report. It appears thatby attenuating, on an ex ante basis, theinvestment objective of superior relativeperformance with respect to peers,Vanguard is able to trumpet, ex post,excellent comparative results when theseoccur, while obviating potentialcriticism should such results fail tomaterialise.

Legg Mason Value Trust PrimaryClassThe Legg Mason Value Trust PrimaryClass fund (LMVTX) commenced in1982 and is one of about 40 funds in theLegg Mason family. LMVTX currentlymanages over US$21bn in assets and iscurrently categorised by Morningstar asa ‘large growth’ fund. The most recentprospectus (1st August, 2007) states:

‘The fund invests primarily in equity

securities that, in the adviser’s opinion,

offer the potential for capital growth. The



adviser follows a value discipline in

selecting securities, and therefore seeks to

purchase securities at large discounts to the

adviser’s assessment of their intrinsic value.’

Across all documents and years forLMVTX, the investment objective withthe highest mean frequency is ‘highabsolute returns’, occurring about twiceas frequently as the second-rankingobjective: ‘high relative returns (non-risk-adjusted)’ (see Figure 3). Onaverage, ‘high absolute returns’ also hadthe highest number of references acrossboth prospectuses and annual reports,respectively. Fund managers forLMVTX appear to focus heavily onproducing long-term growth of capitalby way of high absolute returns. This

focus includes a great deal of attentionpaid (particularly in the annual reports)to individual securities and sectorsdelivering high, non-risk-adjustedreturns. The 1998 annual report, forexample, reads:

‘The Fund purchased large positions in a

variety of technology companies such as

Dell Computer, Storage Technology, and

Digital Equipment over the last two years

when such shares were under severe

pressure due to concerns about earnings

prospects. When those concerns did not

materialize, the shares of those companies

rose sharply, adding materially to the

Fund’s returns. We also benefited from

our long-standing position in banks and

financial services companies, which

Abs

olut

e re

turn

s

Rel

ativ

e re

turn

s (n

on-

risk-

adju

sted

)

Rel

ativ

e re

turn

s (r

isk-

adju

ste

d)

Rel

ativ

e re

turn

s (p

ee

rs)

Abs

olut

e lo

ss

Inte

rest

rat

e

Mar

ket

Idio

sync

ratic

Fact

or

Act

ive/

resi

dual

Rel

ativ

e un

derp

erfo

rman

ce

(no

n-ris

k-ad

just

ed)

Rel

ativ

e un

derp

erfo

rman

ce(p

ee

rs)

Fai

lure

to m

eet o

bjec

tive

Vol

atili

ty (

unde

fined

)

S1

20.3

9.7

12.6

8.2

2.4

0.3

8.8

2.43.6

1.9

0.90.6 1.3

0

5

10

15

20

25


Risk exposure

Figure 3: Legg Mason Value Trust mean counts across all years and documents

Silverman


performed strongly in the past twelve

months.’

Note here the emphasis on the fundmanager having selected ‘winning’stocks (ex post) with no explicitcomment on the factor risk (ex ante)associated with large investments intechnology companies which had beensuffering from the probability ofdistressed earnings, risk factors identifiedin numerous studies.109

Similarly, frequent references to theinvestment objective ‘high relativereturns (non-risk-adjusted)’ suggest afocus on delivering high relative returnswhile omitting, in context, a discussionof the risk factors associated with theseinvestments. In the 1997 prospectus, forexample, the fund managers write:

‘The Adviser believes that the securities of

sound, well-managed companies that may

be temporarily out of favour due to

earnings declines or other adverse

developments are likely to provide a

greater total return than securities with

prices that appear to reflect anticipated

favourable developments . . . purchasing

securities depressed by temporary factors

will provide investment returns superior

to those obtained when premium prices

are paid for issues currently in favour.’

Moreover, LMVTX fails to identify aspecific benchmark index whenpresenting relative results. For example,the annual reports contain tablescomparing the fund’s performanceversus indices representing disparatesegments of the equity market, includingthe S&P 500, the Dow Jones Composite,the Value Line Index and the Russell2000, some of which appear in certainyears and not in others. There is no

unambiguous reference in any of thedocuments to a single, best-fitbenchmark index for the fund. Rather,as in the 1998 annual report, LMVTXsimply comments ‘The Fund had anexcellent year, significantlyoutperforming all relevant indices’.

Similarly, the 2000 annual reportreads ‘As you can see, during our fiscalyear ended March 31, our results laggedthose of the popular indices’.

The infrequent references to ‘highrelative returns (risk-adjusted)’ occurringin LMVTX documents appear almostexclusively in discussions of the use ofoptions and futures in an effort toachieve returns similar to those availableon the same securities in the cashmarket, albeit with less risk.

Across all documents and years forLMVTX, the risk exposure occurringwith the highest mean frequency is‘idiosyncratic risk’. This exposure occursmost frequently, on average, acrossannual reports; although it is surpassedby the number of references to ‘absoluteloss’ (ranked first), ‘active/residual risk’(ranked second) and ‘factor risk’ (rankedthird) across prospectuses. Given anintense focus on the investment objectiveof high absolute returns, it is notsurprising to encounter frequentreferences to ‘idiosyncratic risk’ and‘absolute loss’ as these embody thepotential failure to achieve high absolutereturns in a portfolio that is heavilydependent on the performance ofindividual securities. In every annualreport during the study window, forexample, LMVTX lists securitiesclassified as ‘weak performers’,highlighting their losses for thereporting period. The prospectuses areladen with risk disclosures on thepossibility of fund losses, overall. An



excerpt from the 1996 prospectus istypical in this regard:

‘Mutual Fund shares are not deposits or

obligations of, or guaranteed or endorsed

by, any bank or other depository

institution. Shares are not insured by the

FDIC, the Federal Reserve Board, or any

other agency, and are subject to

investment risk, including the possible loss

of the principal amount invested.’

Table 5 presents measured levels of riskfor LMVTX for the base year 1995 andfor subsequent years, correspondingroughly to the periods for eachprospectus and annual reportcombination. Although lower, onaverage, than that of the Vanguard TaxManaged Capital Appreciation Fund, R2

estimates for LMVTX against the S&P500 Index are still reasonably high,fluctuating between a minimum of 0.60in the base year and a maximum of 0.92in 1998; as such, the S&P 500 is areasonably good fit index for LMVTXduring the study window and marketbetas and tracking errors estimatedagainst the S&P 500 are deemedmeaningful risk indicators.

References to ‘active/residual risk’occur fairly frequently in the LMVTXprospectuses, coming second, onaverage, to ‘absolute loss’. Generallysupporting this pattern of disclosure isaverage measured tracking error, whichis substantially greater than the meantracking error of large-cap equitymutual funds during the five years of thestudy window.110 With the exception ofthe prospectus in 1999, however,LMVTX primary documents contain noreferences to ‘market risk’ in the riskexposure category, whereas thesecondary data reveal systematic betas

that measure above 1.00 in each of thefive years of the study window, peakingat 1.29 in 1998. Thus, with respect tolevels of active/residual risk, ex postobserved results for LMVTX generallycoincide with disclosures in the ex antenarratives. By contrast, in the case ofmarket risk exposure, there is a strikinggap between the ex ante and ex postperspectives.

As in the case of the Vanguard TaxManaged Capital Appreciation Fund,there are also interesting issues related tofactor risk exposure for LMVTX. Theprospectuses, for example, state thatLMVTX will not invest 25 per cent ormore of total assets in any one industry.An examination of the LMVTX semi-annual report dated 30th September,1997, reveals that the fund had 24.6 percent of its assets invested in ‘computerservices and systems’, 4.5 per centinvested in ‘telecommunications’ and 4.0per cent invested in ‘media’.Telecommunications holdings, however,included over US$65m (1.7 per cent oftotal assets) of Nokia, the manufacturerof mobile phones and media holdingsconsisted exclusively of over US$149m(4.0 per cent of total assets) of AmericaOnline, the internet service provider.Adding just these two securities to theinvestment value of the computerservices and systems category wouldhave brought total technology holdingsto over 30 per cent of fund assets in1997. The picture is similar in 1998 andeven more dramatic in subsequent years;total technology holdings representedover 40 per cent of LMVTX assets in1999 and remained over 35 per cent ofassets in 2000. Indeed, the 1999 annualreport reveals holdings in AmericaOnline, alone, totalled 18.9 per cent offund assets.111

Silverman


The industry classification data are ofcourse open to interpretation; however,notwithstanding the fact that AmericaOnline is formally assigned to the‘media’ sector in the statement of netassets appearing in the annual report, thecompany is frequently referred to as a‘technology’ holding in the narrativeportion of the primary documents.Consider, for example, this excerpt fromthe 1999 annual report:

‘Our six best performing stocks in the

fiscal year ending March 31, 1999 were all

technology based. America Online led the

way, rising over 750% in the 12-month

period. Also more than doubling in the

year were cellular phone leader Nokia,

biotechnology giant Amgen, Dell

Computer, and MCI Worldwide.’

This commentary is particularlyinteresting in view of the fact that, aswith America Online, Legg Masonofficially classifies Nokia, Amgen andMCI Worldwide as members of sectorsother than ‘technology’ in the statementof net assets (Dell is formally classifiedunder ‘technology’).

The LMVTX prospectuses do refer to‘factor risk’, averaging 4.8 referenceseach year and peaking with 10 referencesin 1999. In view of the discussion aboveon industry concentration and theprecipitous decline in the value of mosttechnology stocks in 2000, this finalprospectus contains a noteworthy andperhaps prescient caveat:

‘Value funds often concentrate much of

their investments in certain industries, and

thus will be more susceptible to factors

adversely affecting issuers within that

industry than would a more diversified

portfolio of securities.’

Notwithstanding the mischaracterisationof itself as a ‘value fund’, LMVTX hasclearly signalled to its investors thepotential factor risk associated with aportfolio heavily weighted in a singleindustry.

CONCLUSIONSThis study employed primary andsecondary data, arrayed longitudinallyand cross-sectionally, reported both exante and ex post, and examined using acombination of qualitative andpositivistic methods. Specifically,primary mutual fund documents weretreated with ECA112–116 to discover andcodify the investment objectives andrisk-taking of active equity fundmanagers and to establish associationsbetween objectives and risks. Secondaryrisk data were examined to validatefindings based on the primarydocuments.

The study has identified andcatalogued a variety of mutual fundinvestment objectives and risk factorsdisclosed or otherwise encoded in theprospectus and annual report. Positiveassociations have been found foridiosyncratic risk with two investmentobjectives: (1) high absolute returns and(2) high relative returns (non-risk-adjusted), while the same risk exposure isnegatively associated with theinvestment objective of high relativereturns (risk-adjusted). Market risk andactive risk are each positively associatedwith the investment objective of highrelative returns (risk-adjusted) andnegatively associated with the objectivesof high absolute returns and highrelative returns (non-risk-adjusted),respectively. These findings are largelyconsistent with the predictions ofportfolio management models discussed



in the literature. The investmentobjective of high relative returns (peers)is negatively associated with the risk ofabsolute loss and the risk of the fundfailing to meet its objective; this latterassociation reflects an absence ofappropriate risk disclosure.

The triangulation process revealsimbrications but also critical gapsbetween what is disclosed in each of theprimary narratives and what is observedin the secondary data, ie ‘between whatis said and what is done’.117 Riskexposures measured ex post are notalways communicated in the documents;the study found, for example, instancesof undisclosed pseudo-industry risk inthe form of concentrated technologyholdings which may reflect a violationof prospectus covenants.

The study also found elevated levelsof residual risk in the secondary data.This may be indicative of benchmark‘gaming’ as described by Jacob118 andlead to mutual fund style driftdocumented by DiBartolomeo andWitkowski,119 which frequently resultsin portfolios that do not match statedobjectives. Style drift is stronglyindicated in the study of Legg MasonValue Trust which describes itself as a‘value’ fund in primary documents butwhere, in 1999, technology holdingsaccounted for over 34 per cent of totalassets and the average price-to-bookratio for the portfolio was 12.4.

The findings suggest that thedisclosure process can be erratic, withperformance comparisons and the focusof fund manager commentaries varyingfrom prospectus to annual report andfrom one period to the next. This lackof consistency may be incidental or mayreflect a systematic effort on the part ofinvestment companies to position their

funds in the most favourable lightrelative to competitors and benchmarks,as results and market conditions change.Annual reports, in particular, frequentlyomit or obscure critical data and offer expost rationalisations of investment policyrather than a rigorous accounting ofrisk-adjusted performance. Themalleability of investment termsappearing in primary documents canalso be problematic; for example, thefuzzy distinction between ‘industry’ and‘sector’ may enable fund managers todisguise the amount of pseudo-industryrisk to which shareholders are beingexposed. These issues highlight the needfor regulators to continue in their effortsto improve and extend the disclosureprocess.

To date, ethnographic content analysishas been employed successfully in mediaand sociological-based empirical studies.The efficacy with which ECA facilitatesthe identification of mutual fundinvestment objectives and risk factors —as well as the power of the triangulationprocess to validate while also revealingomissions and inconsistencies traversingprimary and secondary data — suggeststhis discriminating analytical frameworkhas an emerging role to play in thefinancial risk management arena forboth researchers and practitioners.

References1 Hyperion Training Limited (1997) ‘TheChartered Financial AnalystExamination’, Hyperion TrainingLimited, London.

2 Fink, M. P. (1995) ‘Oral Statement ofMatthew P. Fink, President, InvestmentCompany Institute, Before theSubcommittee on Telecommunicationsand Finance Subcommittee US Houseof Representatives on the ‘‘CapitalMarkets Deregulation and

Silverman


Liberalization Act of 1995’’, H.R.2131’, in ‘Institute Testifies on State-Federal Regulation of Mutual Funds’,available at: http://www.ici.org/statements/tmny/95_house_fundreg_tmny.html (accessed27th November, 2007).

3 US Securities and ExchangeCommission, available at: http://www.sec.gov/index.htm (accessed 15thNovember, 2004).

4 Ibid.5 Investment Company Institute (1998)‘Mutual Fund Connection’, 20thMarch.

6 Garfinkel, H. (1967) ‘Studies inEthnomethodology’, Prentice Hall,Englewood Cliffs, NJ.

7 Patton, M. Q. (1990) ‘QualitativeEvaluation and Research Methods’,Sage, Thousand Oaks, CA, London andNew Delhi.

8 Fabozzi, F. J. (1998) ‘Overview ofportfolio management’, in Fabozzi, F. J.(ed.) ‘Handbook of PortfolioManagement’, Frank J. FabozziAssociates, New Hope, PA.

9 Markowitz, H. M. (1952) ‘Portfolioselection’, Journal of Finance, Vol. 7,No. 1, pp. 77–91.

10 Jones, R. C. (1998) ‘The active versuspassive debate: Perspectives of an activequant’, in Fabozzi, F. J. (ed.) ‘ActiveEquity Portfolio Management’, Frank J.Fabozzi Associates, New Hope, PA.

11 Bodie, Z., Kane, A. and Marcus, A. J.(2001) ‘Essentials of Investments’,McGraw Hill, New York.

12 Reilly, F. K. and Brown, K. C. (2000)‘Investment Analysis and PortfolioManagement’, Dryden Press, HarcourtCollege Publishers, Orlando, FL.

13 Jacobs, B. I. and Levy, K. N. (1998)‘Investment management: Anarchitecture for equity market’, inFabozzi, F. J. (ed.) ‘Active EquityPortfolio Management’, Frank J.Fabozzi Associates, New Hope, PA.

14 Tanous, P. J. (1997) ‘Investment Gurus’,

New York Institute of Finance,Englewood Cliffs, NJ.

15 Evans, J. L. and Archer, S. H. (1968)‘Diversification and the reduction ofdispersion: an empirical analysis’, Journalof Finance Vol. 23, No. 5, pp. 761–767.

16 Holland, J. (2000) ‘Fund management,intellectual capital, intangibles andprivate disclosure’, paper presented atAlternative Perspectives on FinanceConference, University of Dundee, 23–25 July.

17 Fidelity Investments (2001) ‘The peoplebehind your funds’, Forum, October.

18 Tanous, ref. 14 above.19 Boston Consulting Group (1994)

‘Meeting the Value Challenge,Shareholder Value Management’,Booklet 1, Boston Consulting Group,Boston, MA.

20 Minsky, H. P. (1996) ‘Uncertainty andthe institutional structure of capitalisteconomies’, Journal of Economic Issues,Vol. 30, No. 2, pp. 357–368.

21 Grinold, R. C. and Kahn, R. N. (2000)‘Active Portfolio Management’,McGraw-Hill, New York.

22 Loftus, J. S. (1998) ‘Enhanced equityindexing’, in Fabozzi, F. J. (ed.)‘Handbook of Portfolio Management’,Frank J. Fabozzi Associates, New Hope,PA.

23 Horan, S. M. (1998) ‘A comparison ofindexing and beta among pension andnonpension assets’, The Journal ofFinancial Research, Vol. 21, No. 3, pp.255–275.

24 Brinson, G. P., Randolph Hood, L. andBeebower, G. L. (1986) ‘Determinantsof portfolio performance’, FinancialAnalysts Journal, Vol. 42, No. 4, pp. 39–44.

25 Baker, M. (1998) ‘Fund managersattitudes to risk and time horizons: Theeffect of performance benchmarking’,The European Journal of Finance, Vol. 4,No. 3, pp. 257–278.

26 Markowitz, ref. 9 above.27 Baker, ref. 25 above.



28 Ibid.29 Ibid.30 Modigliani, L. (2002) ‘Real men don’t

use benchmarks’, Financial ManagementReview and Outlook, Fall, 15 October.

31 Horan, ref. 23 above.32 Grinold and Kahn, ref. 21 above.33 Loftus, ref. 22 above.34 Jacob, N. L. (1998) ‘Evaluating

investment performance’ in Bernstein,P. L. and Damodaran, A. (eds)‘Investment Management’, John Wiley& Sons, New York.

35 Ibid.36 Ibid.37 Ibid.38 Bolster, P. and DiBartolomeo, D.

(2002) ‘Mutual fund performance, riskexposure, and the compromise ofinvestment objectives’, working paper,November.

39 Ibid.40 Morningstar. PrincipiaTM database

(CD-ROM).41 Fama, E. F. (1998) ‘Asset management:

Engineering portfolios for betterreturns’, Senior Consultant, PCTPublishing, Richmond, VA, May, pp.1–6.

42 DiBartolomeo, D. and Witkowski, E.(1997) ‘Mutual fund misclassification:Evidence based on style analysis’,Financial Analysts Journal, Vol. 53, No.5, pp. 32-43.

43 Sharpe, W. F. (1992) ‘Asset allocation:Management style and performancemeasurement’, Journal of PortfolioManagement, Winter, pp. 7-19.

44 Kim, M., Shukla, R. and Tomas, M.(1999) ‘Mutual fund objectivemisclassification’, Working paper, July.

45 Sirri, E. and Tufano, P. (1993) ‘Buyingand selling mutual funds: Flows,performance, fees, and services’,working paper, Harvard BusinessSchool, Cambridge, MA.

46 Bodie et al., ref. 11 above.47 Grinblatt, M. and Titman, S. (1989)

‘Adverse risk incentives and the design

of performance-based contracts’,Management Science, Vol. 35, No. 7,pp. 807–822.

48 Falkenstein, E. G. (1996) ‘Preferencesfor stock characteristics as revealed bymutual fund portfolio holdings’, Journalof Finance, Vol. 51, No. 1, pp. 111–135.

49 Payne, B. (2000) ‘BGI accepts tracking-error fee’, Pensions and Investments, Vol.28, No. 8, 17th April, p. 18.

50 Risk Magazine (2003), Vol. 16, No. 1,January.

51 Chase Fleming Asset Management(2000) ‘Global Equity Management: APresentation to Consultants’, GlobalPortfolios Group, October, London.

52 Investment Company Institute (2001)Factbook.

53 Flick, U. (1998) ‘An Introduction toQualitative Research’, Sage, London.

54 Ibid.55 Patton, ref. 7 above.56 Patton, M. Q. (1999) ‘Enhancing the

quality and credibility of qualitativeanalysis’, Health Services Research,December, Vol. 34, pp. 1189-1208.

57 Webb, E. J., Campbell, D. T.,Schwartz, R. D. and Sechrest, L. (1966)‘Unobtrusive Measures’, RandMcNally, Chicago, IL.

58 Denzin, N. K. (1978) ‘The ResearchAct: A Theoretical Introduction toSociological Methods’, McGraw-Hill,New York.

59 Ibid.60 Ibid.61 Ibid.62 Ibid.63 Silverman, H. I. (2005) ‘An

Investigation into the Dynamics of theUS Stock Market Boom (1995–2000)’,Doctoral Thesis, University of London.

64 Hammersley, M. and Atkinson, P. (1983)‘Ethnography: Principles in Practice’,Tavistock Publications, London.

65 Gamst, F. (1980) ‘The Hoghead: AnIndustrial Ethnology of the LocomotiveEngineer’, Holt Rinehart and Winston,New York.

Silverman


66 Patton, ref. 7 above.67 Silverman, D. (1993) ‘Interpreting

Qualitative Data: Methods forAnalysing Talk, Text and Interaction’,Sage, London.

68 Hammersley and Atkinson, ref. 64above.

69 Altheide, D. L. (1987) ‘Ethnographiccontent analysis’, Qualitative Sociology,Vol. 10, No. 1, pp. 65–77.

70 Plummer, K. (1983) ‘Documents ofLife: An Introduction to the Problemsand Literature of a HumanisticMethod’, George Allen & Unwin,London.

71 Altheide, ref. 69 above.72 Altheide, D. L. (1996) ‘QualitativeMedia

Analysis’, Sage, Newbury Park, CA.73 Altheide, ref. 69 above.74 Smith, T., Sells, S. and Clevenger, T.

(1994) ‘Ethnographic content analysis ofcouple and therapist perceptions in areflective team setting’, Journal ofMarital and Family Therapy, Vol. 20,No. 3, pp. 267–286.

75 Altheide, ref. 69 above.76 Plummer, ref. 70 above.77 Hodder, I. (2000) ‘The interpretation of

documents and material culture’, inDenzin, N. and Lincoln, Y. (eds)‘Handbook of Qualitative Research’(2nd edn), Sage, Thousand Oaks, CA.

78 Chambers, E. (2000) ‘Appliedethnography’, in Denzin, N. andLincoln, Y. (eds) ‘Handbook ofQualitative Research’ (2nd edn), Sage,Thousand Oaks, CA.

79 Flick, ref. 53 above.80 United States General Accounting

Office, ‘Content Analysis: AMethodology for Structuring andAnalyzing Written Materials’, USGAOPublication No. 10.1.1, USGovernment Printing Office,Gaithersburg, MD.

81 Altheide, ref. 72 above.82 Glaser, B. G. and Strauss, A. L. (1967)

‘The Discovery of Grounded Theory:Strategies for Qualitative Research’,

Aldine, New York.83 Hammersley and Atkinson, ref. 64

above.84 Altheide, ref. 69 above.85 Guba, E. G. (1978) ‘Toward a

Methodology of Naturalistic Inquiry inEducational Evaluation’, Monograph 8,UCLA Center for the Study ofEvaluation, Los Angeles, CA.

86 Altheide, ref. 69 above.87 Ibid.88 Smith et al., ref. 74 above.89 Altheide, D. L. (1985) ‘Format and

ideology in TV news coverage of Iran’,Journalism Quarterly, Vol. 62, pp. 346–351.

90 Smith et al., ref. 74 above.91 Denzin, ref. 58 above.92 When changes are made to the original

fund prospectus, these are filed asamendments with the SEC under Rule485 of the Securities Act 1933. Theprospectuses employed in this studyappear on the SEC website as‘Conformed Submission Type:485BPOS’ and reflect filing monthswhich may vary from year to year.Thus, in some cases, the length of timebetween prospectus update and annualreport may be more or less than 12months.

93 Guba, ref. 85 above.94 Weber, R.P. (1990) Basic Content

Analysis’, 2nd edn, Sage, NewburyPark, CA.

95 Scott, W. A. (1955) ‘Reliability ofcontent analysis: The case of nominalscale coding’, Public Opinion Quarterly,Vol. 19, pp. 321–325.

96 Krippendorff, K. (1980) ‘ContentAnalysis: An Introduction to ItsMethodology’, Sage, Newbury Park,CA.

97 Ibid.98 Grinold and Kahn, ref. 21 above.99 Field, A. (2000) ‘Discovering Statistics

Using SPSS for Windows’, SagePublications, Inc, London.

100 Markowitz, ref. 9 above.



101 Jacobs and Levy, ref. 13 above.102 Ibid.103 Grinold and Kahn, ref. 21 above.104 Space limitations precluded inclusion of

the authors’ discussion on VanguardAsset Allocation Fund (VAAPX) andBrandywine Fund (BRWIX), copies ofwhich are available on request.

105 Fabozzi, ref. 8 above.106 Farrell, J. (1974) ‘Analyzing covariation

of returns to determine homogenousstock groupings’, Journal of Business,Vol. 47, No. 2, pp. 186–207.

107 Farrell, J. (1976) ‘The multi-indexmodel and practical portfolio analysis’,The Financial Analysts ResearchFoundation Occasional Paper No. 4.

108 French, K. R. and Fama, E. (1992) ‘Thecross-section of expected stock returns’,Journal of Finance, Vol. 47, No. 2,pp. 427–466.

109 Basu, S. (1983) ‘The relationship

between earnings’ yield, market valueand return for NYSE common stocks:Further evidence’, Journal of FinancialEconomics, Vol. 12, No. 1, pp. 129–146.

110 Webb et al., ref. 57 above.111 This large holding in one security may

have been a violation of anothercovenant in the LMVTX 1999prospectus which states that the Fundmay not: ‘With respect to 75% of totalassets, invest more than 5% of its totalassets (taken at market value) insecurities of any one issuer. . .’.

112 Altheide, ref. 69 above.113 Plummer, ref. 70 above.114 Altheide, ref. 72 above.115 Smith et al., ref. 74 above.116 Altheide, ref. 89 above.117 Hodder, ref. 77 above.118 Jacob, ref. 34 above.119 DiBartolomeo and Witkowski, ref. 42

above.

Silverman
