NBER WORKING PAPER SERIES

THE MARKET FOR FINANCIAL ADVICE:AN AUDIT STUDY

Sendhil MullainathanMarkus Noeth

Antoinette Schoar

Working Paper 17929http://www.nber.org/papers/w17929

NATIONAL BUREAU OF ECONOMIC RESEARCH1050 Massachusetts Avenue

Cambridge, MA 02138March 2012

We thank Tim Adam, John Campbell, Matthew Gentzkow, Michael Haliassos, Steffen Meyer, JayRitter, and seminar participants at Berkeley, Columbia, MIT Sloan, NBER Behavioral EconomicsWorking Group, NBER Household Finance workshop, NetSpar Torino, SAVE Deidesheim, and ESSFMGerzensee as well as at the AEA and the AFA for their helpful comments. We thank Ximena Cadenaand Carter Powers for outstanding research assistance in implementing the project. We would alsolike to acknowledge financial support for the audit study from ideas42. All mistakes are of courseour own. Sendhil Mullainathan is the assistant director of research at the CFPB. The views expressedare those of the authors and do not necessarily represent those of the Director of the Consumer FinancialProtection Bureau nor those of the staff, nor the views of the National Bureau of Economic Research.

NBER working papers are circulated for discussion and comment purposes. They have not been peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies officialNBER publications.

© 2012 by Sendhil Mullainathan, Markus Noeth, and Antoinette Schoar. All rights reserved. Shortsections of text, not to exceed two paragraphs, may be quoted without explicit permission providedthat full credit, including © notice, is given to the source.

The Market for Financial Advice: An Audit StudySendhil Mullainathan, Markus Noeth, and Antoinette SchoarNBER Working Paper No. 17929March 2012JEL No. G02,G1,G11,G2,G23,G24

ABSTRACT

Do financial advisers undo or reinforce the behavioral biases and misconceptions of their clients? Weuse an audit methodology where trained auditors meet with financial advisers and present differenttypes of portfolios. These portfolios reflect either biases that are in line with the financial interestsof the advisers (e.g., returns-chasing portfolio) or run counter to their interests (e.g., a portfolio withcompany stock or very low-fee index funds). We document that advisers fail to de-bias their clientsand often reinforce biases that are in their interests. Advisers encourage returns-chasing behavior andpush for actively managed funds that have higher fees, even if the client starts with a well-diversified,low-fee portfolio.

Sendhil MullainathanDepartment of EconomicsLittauer M-18Harvard UniversityCambridge, MA 02138and Consumer Financial Protection Bureauand also NBERmullain@fas.harvard.edu

Markus NoethLS Banking & Behavioral FinanceVMP 5University of Hamburg20146 HamburgGermanymarkus.noeth@uni-hamburg.de

Antoinette SchoarMIT Sloan School of Management100 Main Street, E62-638Cambridge, MA 02142and NBERaschoar@mit.edu

1

1.IntroductionA large and still growing body of research demonstrates that individual investors oftenmake poor financial decisions if left to their own devices. Drawing on evidence frompsychology and behavioral economics, these studies suggest that investor beliefs anddecisionprocessesarepronetobiasesthatoftenresultinfinancialdecisionsatoddswithbasic portfolio theory. For example, retail investors are overconfident, engage in trend‐chasing,usenaïveheuristics,andaregenerallysusceptibletoanumberofdifferentbiases.BenartziandThaler(2007)documentsomeofthemostcommonbiases.1Usingtradedatafromretailcustomers,BarberandOdean(2000)documentthatexcesstradinginbrokerageaccountsandreturns‐chasingbehavior leads tosignificantly lowerreturnscomparedtoabuy‐and‐holdstrategy. Inshort, thepastdecadeofresearchhasproduceda largebodyofevidencesuggestingthathouseholdsmaybebadatchoosingportfoliosontheirown.2

Yet households do not make decisions in a vacuum. A variety of forces, from socialinteractionswithfriendsandfamilytoadvertisingandmedia,can influencetheirchoices.Oneparticularly importantsourceof inputscomesfromfinancialadvisers.3Inasurveyofretail investors, Hung et. al. (2008) found that 73%of all individuals surveyed consult afinancial adviser before purchasing shares or mutual funds. Given this central role ofadvisersintheinvestmentprocess,weaskwhetherornotthemarketforfinancialadviceservestode‐biasindividualinvestorsandthuscorrectpotentialmistakestheymightmakewithouttheseinputs.4Wedefine‘goodadvice’ as advice that moves the investor toward a low-cost, diversified, index-fund approach, which many textbook analyses on mutual fund investments suggest, see for example Carhart (1997). Alternatively, sincemany (though notall)advisersarepaidwithincentivesthatencouragethemtodirectmoneytospecificfundsand generatehigh fees,might advisers exploit thesebiases of retail investors in order tofurther their own interests? Additionally, if investors are unable to make portfoliodecisionsthatareinlinewithtextbookrecommendations,theymightbeequallyunabletodifferentiate between advisors that are either self‐interested or help investors to build awell diversified, low fee portfolio. In such a world, good advice, i.e. , to hold a well‐diversifiedportfoliocomposedof low‐fee funds,mightnotberewardedsufficiently.5Asa

1 Benartzi andThaler (2001) for example argue that employees follow a naïve diversification strategy of

mechanicallyspreading theirmoneyequallyacross the funds theyareoffered(what theycall1/nrule),generatingquiteperverseoutcomessincetheequitymixdependsontheinvestmentmenu.

2 It isnot inthescopeofthisarticletoreviewthisbodyof literature.SeeBarberisandThaler(2003)andCampbell(2006)foradditionalindepthoverviews.

3 Mostfinancialadvisersprovidepersonalizedadviceofwhatstocksorfundsto invest in.TheInvestmentAdvisers Act of 1940 defines (see section 202(11)): “`Investment adviser´ means any person who, forcompensation, engages in the business of advising others, either directly or through publications orwritings, as to the value of securities or as to the advisability of investing in, purchasing, or sellingsecurities,orwho,forcompensationandaspartofaregularbusiness,issuesorpromulgatesanalysesorreportsconcerningsecurities;(…).”Bolton,Freixas,andShapiro(2007)studycompetitionbetweenbanksanditsinfluenceonincentivesfortruthfulinformationrevelation.

4 Chen,De,Hu,andHwang (2011) focuson the impactofpeer‐based financialadvicevia socialmediaonaggregatestockmarketoutcomes.

5 Of course, understanding this effect requires a deeper understanding of a different source of cognitivebias‐people’sperceptionsofwhatconstitutesgoodadviceandwhatconstitutes independentadvice.SeeMoore,Cain,LoewensteinandBazerman(2005)forworkonthistopic.Inthiscontext, it isalsoanopenquestion why retail investors are mostly unaware of a conflict of interest between the provider ofinformationanditsrecipient.Malmendier and Shanthikumar (2007) show that retail investors are mostly naïve with respect to analysts’ recommendation incentives. In addition, Chater, Inderst, and Huck (2010) document in a

2

consequence, theseadvisersmightbedrivenoutof themarket inequilibrium, since theycannot attract a sufficient market share to account for their lower‐fee structure. TheexperienceofVanguard,whichofferedthefirstindexfunds,isacaseinpoint:thefirmhadto modify their sole reliance on low‐cost index funds since consumers seemed to besusceptibletohigh‐costadvertising(seeBogle,2000).Thus,themarketforfinancialadvicemaynotmitigatebehavioralbiasesandmayevenexacerbatethem.Despitethe(growing)importanceofadvisersintheinvestmentprocess,especiallyduetoanincreasingnumberofdefined contribution plans, very little is known about the market for financial advice.6Campbell(2006)highlightsourrelativeignoranceaboutthisimportantsector.

Understanding thismarket is also important from an aggregate level. Themarket forfinancial advice might influence how individual biases translate into aggregate marketoutcomes,e.g.,capitalflowsintodifferentinvestmentstrategiesorevenpricing.Howwellitde‐biases individuals is important for knowing how tomodel “representative” agents inmacro‐consumption models and how to model equity prices and for numerous policyapplications.

In this paperwe set up an audit study to test the quality of advice provided to retailinvestors in the market. The specific advisers we are looking at in this study are retailadviserswhom theaverage citizencanaccessvia theirbank, independentbrokerages,orinvestmentadvisoryfirms.Theseadvisersareusuallypaidbasedonthefeestheygeneratebutnotbasedontheassetsundermanagementor theperformanceof theportfolio.OnceclientshavemorethanUS$500kininvestiblecapital,theyhaveaccesstoabroadersetofadvisers with better compensation structures. We focus on pure investment advisers inordertofocusonanarrowsetofmeasureableoutcomes.Therefore,wedonotincludetaxadvisers,adviserswhoalsoprovideestateplanningservices,orprovidersofotherwealthmanagementservices.

Inparticular,wewanttounderstandwhetheradvisersactivelyde‐biastheirclientsorinsteadexaggerateexistingbiases,especiallybiasesthathelptheadviser’sowninterestsbyincreasing fees and turnover. For that purpose, auditorswere randomly assigned to fourdifferent treatments (portfolios) which represent different investment strategies andbiases.Wemakesurethatthelosstotheclientfromeachofthebiasesiscomparable.Therandomassignmentallowsustotesttheaverageresponseofatypicaladviserwithoutanyconcernforself‐selectionofclientstodifferenttypesofadvisers.Theauditorstrackedtheinformation requested from them, theadvicegiven, andother featuresof the interaction.Ourprotocolrecordstheadvicegivenviaauditorexitsurveys,aswellaswrittenmaterialswithportfoliosuggestionsbytheadviser.

Thefirst twoportfolios(“biasscenarios”)reflect twoof themostcommonlydescribedbiases in the literature. In the first scenario (“chasing fund returns”), the auditorholds aportfolioinwhich30%isinvestedinonesectorexchangetradedfundthatperformedwellin thepreviousyear,andheexpressesan interest in identifyingmore industries thathaddone well recently. In this case, the incentives of the adviser and of the client are notaligned: the adviser benefits from the bias of the client since it allows him to churn theportfoliomoreoftenandgeneratemorefees,whereastheclientwouldprofitfromabetter

European Survey with several thousand participants that retail investors ignore advisers’ potential conflict of interest. Inderst and Ottaviani (2011b) investigate theoretically the determinants of the compensation structure for brokers. Also see Schneider (2009) about trust and incentives structures.

6 ThemarketforfinancialadvicegeneratesbetweenUS$20bnandUS$50bnfeesperyeardependingonthedefinitionofadviceandcompensationmodelsforadvisers.

3

diversifiedportfolio.7Inthesecondscenario(“employerstocks”),anauditorholds30%ofhisportfoliointhecompanystockofhisassignedemployer.Thus,incentivesoftheadviserandoftheclientarealigned:itisinthebestinterestoftheadvisertoreduceoreliminatetheclient’sbiassinceholdingcompanystockalsoreducestheadviser’sabilitytogeneratefees.8 In the third scenario, theauditorholdsadiversified, low‐feeportfolio consistingofindexfundsandbonds‐‐ineffect,anefficientUSportfolio.Weintroducethisscenariototestif advisers arewilling tomove clients out of this portfoliowhichwould be closest to anallocationrecommendedinmost financetextbook.Finallywehaveacontrolgroup(“cashscenario”)inwhichtheadviseesimplyholdscertificatesofdepositsanddoesnotespouseaparticular view beyond a general willingness to increase risk for higher returns. Thisvariationintreatmentgroupswillallowustotesthowresponsiveadvisersaretotheneedsoftheirprospectiveclients.

Our audit produces three main sets of findings. First, advisers’ reactions to differentportfoliosorinvestmentscenariosvariedsignificantly.Adviserswerebroadlysupportiveofthetrend‐chasingportfoliobutmuchlesssupportiveofthecompanystockportfolio.Moststrikingly,theywereunsupportiveofthe(efficient)indexportfolioandsuggestedachangeto actively managed funds. Overall, advisers had a significant bias towards activemanagement. Innearly50%of thevisits, the adviserencouraged investing in anactivelymanaged fund; by contrast, in only 7.5% of the advice sessions (21 visits), advisersencouragedinvestinginanindexfund.Whenadvisersmentionedfees,theydidsoinawaythat downplayed themwithout lying. For example, they often used arguments like, “Thisfundhas2%feebut that isnotmuchabove industryaverage.”Theseresultssuggest thatthe market for financial advice does not serve to de‐bias clients but in fact exaggeratesbiases that are in the adviser’s financial interestwhile leaning against those that do notgenerate fees. In our index fund scenario, the advisers are even advocating a change instrategy(awayfromlowfeeindexfundsandtowardshighfeeactivelymanagedfunds)thatwouldmaketheclientworseoffthantheallocationwithwhichheorshestartedoff.

Second, consistent with portfolio theory, most advisers did ask clients about theirdemographic characteristics, which may have been used to determine risk preferences,time‐horizonandhumancapital risks, andcovariance.Overall,we find that inmore than75% of the visits, advisers asked for this kind of information, specifically income, othersavings (e.g.401(k)plan)besideswhat theywere investingwith theadviser,occupation,and marital and parental status. The recommended investment in stocks and domesticassets significantly increased with annual income, a fact that may be explained by anassumedhigherriskorlosstoleranceforthewell‐off.Marriedclientsweretoldtoholdlessinliquidassets.Thisisconsistentwithamodelofspousallaborsupplyprovidinginsurance,reducingtheneedforliquidity.However,inmanycases,theinformationdidnotgetusedinthewaypredictedby portfolio theory: the recommended exposure to equities decreasedwiththeamountinvested.Femaleclientswereaskedtoholdmoreliquidity,advisedtoholdlessinternationalexposure,andpushedlessfrequentlytoinvestinactivelymanagedfunds.Atthesametime,advisersdidnotseemtotailorportfolioadvicewiththeageoftheclientathand.Wefindnosignificantdifferencesinthemixofstocksandbondsforolderclients.By and large, though, this is the arenawhere adviserswere closest to traditional theory:

7 Inderst and Ottaviani (2009, 2011a) link advisers compensation and their advice quality theoretically

eitherbyfocusingontheagencyrelationshipbetweenthesellingfirmanditsemployedsalesforceorbyanalyzingcompetitionthroughcommissionsandkickbackspaidtoadvisersbythefundindustry.

8 The advice to sell the employer stocks and to invest themoney in a diversified portfolio enhances theclient’sportfolioandgeneratesfeesfortheadviser.

4

attempting tomatchportfolios to characteristics.The levelsofportfolio advicewerealsobroadly consistent with portfolio theory, with advisers suggesting a high equity mix(roughly 2/3) and thereby potentially reducing any bias that may generate an equitypremium.Interestingly,theywerealsomorelikelytomentionfeesspontaneouslywhentheauditor was older, possibly believing that older auditors would have asked themselves.9Theirresponsestothedifferentportfoliosreinforcethesefacts.

Third,wefindsomesuggestiveevidenceof‘catering’,i.e.advisersshowedsupportearlyonfortheclient’sexistingstrategies,mostlikelytoestablishcredibilityandnotalienateapotentialclient.The“initialreaction”toaclient’sstrategyvariedsignificantlyfromthelaterrecommendedcourseofaction. In fact, the initial reaction to index fund investmentswasverypositivewhileauditorswhowentinwithcompanystockorreturns‐chasingportfoliosfacedmoreambiguoussupport.These resultshighlight thatadvisershave tobeawareofthefactthattheyarefacingasalessituationandtheycannotbluntlycriticizewhatclientsmighthavedoneinthepast.However,itdoesnotappearthatadvisersareseverelylimitedin the investment recommendations they eventually give to their clients, since advisershave no problem discouraging clients from investing more in their current strategies(especiallyifitgoesagainsttheadviser’sinterests).Instead,theysuggestinvestmentsthatareorthogonaltotheclient’scurrentapproach.

Overall our findings suggest that the market for advice works very imperfectly. Theadvicebyandlargefailstode‐biasclientsandifanythingmayexaggerateexistingbiasesor,insomecases,evenmakestheclientsworseoff.Moreover,individualbiasescanhavefirst‐order implications for aggregate capital flows and pricing of risk, if there is not enoughinformedcapitaltoexploitarbitrageopportunitiesagainstcapitalflowsfrom“biased”retailinvestors.Itcanalsoshedlightonhowwemodelinformationaggregationinequilibriumifcompetitiveforcesinthemarketforfinancialadvicedonotleadtotheprovisionofthebestpossibleadvice.Competitionmightbelimitedbythefactthatfinancialadvisersexploitthebiases of naïve (or uninformed) retail investors. At the same time, advisers who areinterestedinprovidingbetteradvicemightbeunabletogainamarketshareifbiasedretailinvestorsareunable todifferentiategood frombadadvice.Whilewecannot ruleout thelatter force, our evidence suggests that advisers’ self‐interest plays an important role inprovidingadvicethatisnotinthebestinterestsoftheirclients.

It is important to note that our research design allows us to look at the quality ofinvestmentadviceprovidedtoclients.However,advisersmayprovidemanyotherbenefitsfor their clients, for example, by giving them the confidence and information to invest inriskyassetsinthefirstplace,byprotectingthemfromlosingmoneyinfraudulentfunds,orbyreducingtransactioncosts.Thesereasonsmightbeasimportantastheactualcontentoftheadvice.However,eveniftheseadditionalbenefitsareveryimportantforretailinvestors(perhaps even the primary reason that clients seek advisers), there is no reason whyadvisersshouldnotbeabletoalsoprovidehigh‐qualityadvicetotheircustomers.

Whileauditstudieshavebeenusedtomeasurediscriminationinthelaborandhousingmarkets (Fix and Turner (1998), Altonji and Blank (1999) and Heckman (1998) forreviews),theyhavenotbeenusedinthefinancialcontext,saveafewexceptions(seerecentworkbyIyerandSchoar,2009and2011).Importantly,theauditstudymethodologyallowsus tomeasure an adviser’s responsewhenwe exogenously vary the types of clients andbiases that the adviser is confrontedwith. This enables us to control for the selection of

9 Notethat“mentioningfees”mayincludestatements like“this isano‐loadfund”, i.e.notallrelevantfees

arementioned.

5

clients to advisers,which is oneof the centralproblemsplaguingnon‐experimentaldata.Thereasonisthatnon‐experimentaldata(e.g.,bankrecords)usuallydoesnotallowustodifferentiatehowmuchoftheobservedoutcomesaredrivenbytheadviserinfluencingtheclients’ decisions versus different clients selecting to certain types of advisers who willprovide them with the recommendations they want to hear. In addition, the auditmethodology enables us to vary the characteristics of the auditor exogenously either byselecting certain auditors or by assigning specific characteristics. In addition todemographiccharacteristicslikeage,gender,andnumberofchildren,wealsovariedtraitssuchasinvestiblewealthandhousingstatus.

Ourresultsarealsorelatedtoasmallbutgrowingliteratureonfinancialadvisers.OneofthemostnotedearlystudiesonfinancialadviceisCanner,Mankiw,andWeil(1997),whoexaminethegenericwrittenadvicegivenby investmentadvisersbasedonbroadrulesofthumb(seealsoBodieandCrane,1997).MorerecentworkbyBergstresser,Chalmers,andTufano (2009) orDel Guericio, Reuter, and Tkac (2011) show the role of incentives anddistributionchannelsintheprovisionoffinancialadvice.Bluethgenet.al.(2008),ChalmersandReuter (2011),Hackethaletal. (2011,2012),Kramer(2012),andBhattarchayaetal.(2012) use data on portfolio outcomes or trading volume to quantify the benefits offinancial advice. Georgarakos and Inderst (2011) show theoretically and provide someempirical evidence that trust in professional financial advice has a statistically andeconomically significanteffecton the stockmarketparticipation forhouseholdswith lowfinancial capability, i.e., most retail investors. Similarly, Inderst, and Ottaviani (2009,2011a) provide a theoretical framework that links adviser compensation with advicequality, focusingon theagencyrelationshipbetween theadvisory firmand its sales forceand competition via commissions and kickbacks which are paid to advisers by the fundindustry.

The rest of thepaperproceeds as follows: section2 provides a summary of the studydesignandsetupof theaudit. Section3shows thedescriptivestatisticsandconfirms therandomization.Wereportthemainsetofresultsinsection4andconcludeinsection5.2.StudyDesign2.1.OverviewofAuditInordertoinvestigatethequalityoffinancialadvicethatiscommonlygiventoclientsinthemarket, we set up an audit study in the greater Boston and Cambridge area. We senttrained,professionalauditorstoimpersonateregularcustomerswhoareseekingadviceonhow to invest their retirement savings outside of their 401k plan.10 Our auditors wererandomly assigned to four different treatment portfolios that reflect different types ofinvestmentstrategiesorbiases.Wewilldiscussthesestrategiesinmoredetailbelow.Wealsovary thewealth rangesof the clients, eitherbetweenUS$45,000andUS$55,000orbetweenUS$95,000andUS$105,000.Theserangeswerepickedtomimicthesavingsfor 10 If theshopperwasasked fora401(k)plan investment, thestandardizedanswerwasthata401(k)plan

existedbut that shewantedadviceonhow to invest theextramoney.Almostall advisersaccepted thisstatementeven though itmight lead toan inferiorhouseholdportfolio.However, ifadvisersasked fora401(k)portfolio,theymayhavefirstwantedtoenlisttheclientandwouldlaterincludethe401(k)planintheportfoliooptimization.Withrespecttoowningorrentingrealestate,wetoldourauditorstoalwayssaythattheyrentedtheirapartmenttoavoidsituationsinwhichthebestadvicewouldbetoreducethemortgagefirst.

6

averageUShouseholdsindifferentageranges.11Ourstudyfocusesonretailadvisersatthelowerendofthewealthspectrum,e.g.,wedonotincludeprivatewealthmanagersorhedgefunds.Themodaladviserinourstudyisworkingeitherforabank,retailinvestmentfirm,ortheirownindependentoperation,focusingonthelowerendoftheretailsegment.Mostofthemarepaidoncommissionbasedonthefeesandvolumesthattheygenerate,andonlyasmallsubsetof theadvisersare independentandwouldbepaidbasedoncapitalundermanagement. The fractionof this latter typeofadviser isverysmall inoursamplesincetheyusuallyonlydealwithwealthierclients.122.2.TreatmentsThis setup allows us to test how advisers react to pre‐existing investment strategies orclientbiases.For thatpurpose,wesetup fourdifferent treatments todifferentiatebiasesthatgoagainsttheadviser’sself‐interestversusthosewheretheadviser’sandtheclient’sinterestsarealigned.Wedesignedthebiasesinsuchawaythatthenetexpectedlosstotheinvestorissimilarinmagnitude.Asdiscussedintheintroduction,ifadvisersactpurelyoutofself‐interestedmotives,theyshouldcounteractclientbiasesthatleadtolow‐feeincome(e.g., excessive investment in company stock) but reinforce biases that increase theadviser’s ability to generate fees, such as trend‐chasing. However, if advisers areconstrained by having to cater to clients’ pre‐existing beliefs, we should expect thatadvisersaremorerestrictedintheadvicetheycangive,asincases,forexample,whentheclienthasstrongpriorbeliefsorisemotionallyattachedtothecurrentportfolio.Incontrast,an adviser is less restrictedwhen the client has no predetermined opinion, and thusweshouldnotseeadifferentialresponsetodifferenttypesofbiases.

To test the importance of these countervailing forces, we selected four differenttreatmentsthatarepresentedtotheadvisers(andimpersonatedbyourauditors).Asour“biasscenarios,”weselectedtwoofthemostcommonbiasesdocumentedintheliterature:chasingfundreturnsandinvesting inemployerstocks.Wecomplementedthesewithtwo“unbiasedscenarios”–adiversifiedlow‐feestock/bondportfolioandanall‐cashportfolio.

Inscenario1(“chasingfundreturns”)ourauditorsindicatedthattheyhadbeentryingtooutperformthemarketbyidentifyingindustriesthathadexcessreturnsintherecentpast.In the advice session, the auditor would present the adviser with a portfolio that isconcentrated ina few industrieswithhighreturns in the lastyearandask theadviser tohelp identify more stocks and industries of this type. Note that de‐biasing a client bydiversifyingthecurrentportfoliowould leadto(one‐time)returns fortheadviser,buthecouldprofitevenmorebycateringtothebiasandturningovertheportfolioatleastonceayear.We set up the portfolio such that 30% of the portfolio was invested in one sectorexchange traded fund that had performed well in the previous year (i.e. 2007).13 Thesesectors included telecommunication, oil & gas, metals & mining, and US aerospace &defense.Dependingontheagegroup(about30or45yearsold),20%or35%wasinvestedinintermediateUShigh‐creditqualitybondfunds.Therestoftheportfolio(50%or35%)

11 In addition, these amounts are varied around the average annual household income in theBoston area

(aboutUS$75,000).12 Financial advice by independent advisers,whoare compensatedby thehourorbasedon capital under

management,isoftennotavailableoritistooexpensiveatseveralhundreddollarsforafirstvisit.13 Wefixedtheproportionat30%fortwomainreasons.First,wewantedtogivetheadvisertheopportunity

toinvestmoreinthisstrategy,althoughthiswouldleadtolessdiversification.Second,webelievethatthemoreextremeaportfolioallocation(seealsoscenario2withemployerstock),thehighertheprobabilitythatanadviserwouldrememberaportfolios/hesawsomedaysagofromanotherpotentialclient.

7

wasinvestedinasingleS&P500indexfund.Wevariedtheselectionofindexfundsandtheexact amount invested to reduce the probability that an adviser would recognize theportfolio from a previous visit.14 The average under‐performance of the four selectedsectorscomparedtotheS&P500over1.5yearsaftertheendofourauditstudyhasbeenabout ‐6.5%p.a., i.e., the30% investment in this sector resulted inanunderperformancebetweenUS$1,000andUS$2,000peryeardependingon theportfolio size.Even thoughreturns‐chasing may be similar to a momentum strategy, we believe that this is notappropriateinthecontextofthismarketsinceinvestmenthorizonsofcustomersareratherlong and the frequencywithwhich people rebalance their portfolios is too low to allowthemeffectivelytotakeadvantageofmomentumstrategies.

In scenario 2 (“employer stock”), we assigned the auditor to one of the 50 largestemployers in the Boston area and assigned 30% of the person’s portfolio to companystock.15Dependingontheagegroup,20%or35%wasinvestedinbondsandtherestintheS&P 500 (as in scenario 1). In this scenario, it was in the interest of both the financialadviserandtheclienttorestructuretheexistingportfolio.Theadvisercanearnmoneyinthe portfolio rebalancing process, and the client will most likely end up with a betterdiversifiedportfolio.Toallocate30%oftheportfolio toonestock increasesportfolioriskeven ifwe ignorehumancapital risk. Let’s assumea standarddeviation in themarketof20% and a risk premium of 6% p.a., whereas the company stock has a 50% standarddeviation,unitbeta,and6%riskpremium.Then,a30%allocationtocompanystocklowerstheSharperatiofrom0.3to0.247,translatingforagivenrisktoareturnlossof1.29%orUS$1,032peryearforanUS$80,000portfolio.16AppendixAcontainsexamplesofscenario1and2portfoliosandadditionalclientbackgroundinformation.Again,wedidnotwanttoassignmorethan30%oftheportfoliotocompanystock,sinceonaveragepeopleusuallydonotholdmorethan50%ofassetsincompanystockandsothatweavoidedraisingadvisers’suspicion.

Inaddition to these twoscenarioswith inherentbiases, auditorswereassignedawelldiversified portfolio in scenario3, consisting of low‐fee US index stock funds and bonds,using thesameallocation tobondsdependingon theagegroup, as inallother scenarios.While the portfolio is the most efficient of all the treatments used in the study, thistreatment does have a (US) home bias, and thus a value‐enhancing adjustment to theportfolio might be to suggest more international diversification.17 Moving the low‐feeportfolio to an activelymanaged portfoliowith the same risk/return profile but averagemanagementfeeswouldresultinadditionalcostsofaboutonepercentagepointperyear,i.e.,betweenUS$500andUS$1,000inourscenario.

Scenario4isourcontroltreatment,sincetheavailablemoneyiscurrentlyinvestedinashort‐termcertificatesofdepositandtheauditordoesnotdisplayanypreconceivedbiases.In this scenario, only the investmentamount and thedemographics arevariedasbefore.Theadviserreceivesnohintshowtheclientwouldliketoinvestthemoneyexceptthattheclientwouldlikehelpwithabetterinvestmentstrategy. 14 Even though we used only ETFs or low‐cost index funds both for the diversified part of the portfolio

(treatments1‐3))andforthesectorsintreatment1,onlyoneadvisermentionedthisfact. 15 Auditors were assigned to one of these employers in all treatments but had employer stock in their portfolios only

in treatment 2. We provided them with some background information about their employer, including travel times to various points in Boston and Cambridge, such that auditors could answer basic questions about their working life.

16 WethankJohnCampbellforthisexample. 17 Given the high volatility of currency exchange rates over the last ten years, it is not obvious whether international

diversification helps to improve the portfolio characteristics.

8

Auditorswere randomly assigned to one of the four scenarios. They had to be college-educated and had to match our gender and age requirements.Onegroupwasintheirearly30s,andweassignedthema financialwealthofaboutUS$50,000.Theothergroup is in theirmid‐tolate40sandhadafinancialwealthofaboutUS$100,000.Allauditorswereassignedtooneofthe50largestemployersintheBostonarea.Inscenario2(employerstocks),theemployer had to be listed on an exchange. All other characteristics (like marital status,children, etc.) that auditors may have talked about with their advisers were their owncharacteristics,suchthat theycouldtalknaturallyabout them.The investmenthorizonofallauditorswasretirementage, i.e.,about30yearsforthefirstgroupandabout15yearsforthesecondgroup.2.3.LogisticsofAuditTo implement the actual logistics of the visits, we hired a financial audit firm thatspecializes in identifying and training auditors.Weworked very closelywith the firm toselect suitable people as auditors for the study, andwewere also intimately involved intrainingtheauditors.Wedesignedallthetrainingscriptsandsetupthescheduleofvisitswithpredefinedadviserstowhichauditorswererandomized.Toensurethatauditorswereabletounderstandtheadvicethatwasgiventothem,theyhadtoknowatleastsomebasicsof financial products and received some guidelines on how to ask for specific advice.Auditorswere trained first about basic financial literacy through our onlinemanuscript.Then,theyparticipatedinatrainingsessionviavideoconferencewiththesupervisorsandour staff. Finally, audit candidateshad to takea shortonline test toqualify for the study(about 10% of the pre‐selected auditors failed andwere excluded from this study. SeeAppendix B). Even after training, it is still possible that auditors do not retain all theinformationthatadvisersprovideinthemeeting.However,suchbehaviorshouldjustleadto more noise overall and not bias our results. Auditors were assigned only to onetreatment at a time to avoid confusion and retraining. After the first set of audits wasfinished,wereassignedauditorstoasecondtreatmenttocontrolforauditorfixedeffects.

Theauditfirmprovidedthelogisticsofmonitoringandimplementingtheschedulingofvisits, setting up online survey forms, and finding and compensating auditors. Auditorswerepaidon a per visit basis andwere told that theywouldnot be invited for a repeatassignment if we heard any complaints about their behavior.We also sent our researchassociate to do random spot checks in order to observewhether the clientwasmeetingwiththeadviser.Tominimizeanydemandeffects,wemadesurethatthestudywastripleblind:thefinancialadvisers,thefinancialauditfirm,andtheauditorsdidnotknowwhyandhowwechosespecificparameters.Thecompanyandtheauditorsweretoldthattheaimofourstudywastoconductanassessmentofthequalityofthemarketforfinancialadviceandthat any variation in the treatment arms was instituted to create variation in order tominimizedetectionandsuspicion.18

LogisticsoftheMeeting:Tosetupthein‐personmeeting,auditorscalledtheirassignedfinancialadviserandagreedonatimeconvenientforboth.19Asareasonforthevisit,theauditors stated that theywere seekingadviceonhow to invest privatelyheld retirementsavingstheyhadoutsideofemployer‐providedvehicles(401(k)anddefinedbenefitplans).

18 Note that we never mentioned different treatments, neither in our conversations with the audit firm nor with the

auditors. 19 A new adviser was only assigned after the previous visit had been completed. Advisers were at most

visitedonceaweekbydifferentauditors.

9

At theagreed‐upon time, theauditorswouldmeetwith theadvisers fora consultationofaboutonehour,usuallyintheadviser’soffice.Duringthemeeting,theauditorwouldfollowthe general script provided by us.Depending on their treatment assignment, theywouldexplain their existing investment strategy and ask for advice with their portfolios andinvestment strategies as described above. The auditors were asked to write down theirassignedportfolioonapieceofpaperorprintthemoutsothattheycouldshowthestatusquototheadviser.Therewasenoughvariationinthewaytheinformationwaspresentedthat adviserswould not be suspicious of any potential repetitions. Aside from the actualtreatmentassignment,auditorsweretoldtoanswertruthfullyanyinformationconcerningtheirname,socialsecuritynumber,oranyotherdemographicinformation,suchasnumberofchildrenormaritalstatus.

TrackingAdvice:We encouraged auditors to write down information that the adviserprovided in the courseof themeeting to increase accuracy.Takingnotes isnatural in anadvice situation and thus does not create any suspicion. Again, we made sure that theauditors had enough variation in how they would put down the notes to avoid anypotentially suspicious repetitions. One caveat about the scope of advice is that manyadvisersareunwillingtoprovidedetailed,personalizedadvice(e.g.,adviceontheallocationof assets to specific funds) unless the client hasmoved his or her funds to the adviser’sfirm.20 Since our auditors were not able to provide the adviser with those funds, someadviserswere reluctant toprovidevery specific adviceand rather commented ingeneraltermsaboutthequalityoftheclients’existingportfoliosandtherecommendedallocationgoingforward.Therefore,inmostofthestudywewillfocusonthetypeofadvicegivenandtheassociatedreasoning.

After the visit, auditorswere asked to fill out an online exit survey that hadmultiplechoicequestionswithfree‐textfields.Theyhad24hourstofilloutthisinformationaftertheconclusion of the visit to make sure that they had not forgotten the information theyobtained.Inaddition,eachauditorhadtosendintheadviser’sbusinesscardsuchthattheauditcompanycouldmakerandomcallstoverifythattheauditoractuallyhadshownuptothevisit.21Ifthequestionnairewasnotavailablewithin24hours,theauditorwascontactedby a supervisor and reminded toprovide the information.Thisprocedurehelped extracthigh‐quality and complete information after the visit.Moreover, auditorswere only paidafterfillingintheform.Formostquestions,a“Don’tknow/Don’tremember”optionexistedtoavoidrandomanswers.Ifauditorshadreceivedadditionalwritteninformationatorafterthe meeting with their adviser, they forwarded thesematerials to us and we coded thewrittenrecommendationsifanyweremade.Therewasonlyoneauditorwhodidnotfillouthissurveys inthenecessarytimeandwassubsequentlydroppedfromthestudy.Wealsoconductedanexit interviewwitheachauditorafter their firstvisit toverify thatauditorswerecomfortablewiththesetup.3.SummaryStatisticsandRandomizationTheauditdataof284clientvisitswascollectedbetweenAprilandAugust2008, i.e.afterthe problemsofBear Stearns surfacedbutbefore the bankruptcy of LehmanBrothers inmid‐September.Wehadinitiallyplannedforanauditof480observationsbutunfortunatelyhad to stop our audit study prematurely, since in the ensuing financial contraction the

20 This behavior is to some extent comparable to a car dealerwho asks first for a down payment before

agreeingtoatestdriveofacar.21 Informationontheadviser’sidentitywasnotpassedontous.

10

market for financialadvice in theBostonareawassignificantly restructured.22Moreover,the changing economic conditions were especially important for the chasing returnstreatmentsincetheoutperformingindustriesofthepreviousyearhadchanged.

Asaresult,ourfourscenariosarenotevenlycovered.Table1showsthedistributionofvisits across the four different scenarios: there are 103 visits in scenario 1, 62 visits inscenario 2, 49 visits in scenario 3, and 70 visits in the last scenario. However,Table1confirmsthat,despitethereducedsamplesize,therandomizationofvisitstoadvisersstillseemstobeintact.Theaverageageofauditorsdoesnotvaryacrossthetreatmentgroupsandiscenteredaround39‐40years.TheaverageassignedannualincomeisUS$80,000andagain there is no significant difference between the four cells. The same is true for theinvestment amounts; the average investment is between US$ 77,000 and US$ 80,000.Finally,onaverage,thefractionoffemaleauditorsisabout77%andtherearenosignificantdifferencesbetweenthedifferenttreatmentgroups.

Whilethepowerofthetestsis lowered,duetothesmallersamplesize,itisreassuringthattherandomizationlargelyholdsdespitethesmallersample.4.Results

4.1.DescriptiveStatisticsofaTypicalAudit

Table 2 provides an overview over the information the adviser collected and herrecommendation during the auditor’s visit. It is a prerequisite to understand the client’sfinancial situation, their ability to handle portfolio risk, and their current exposure tomarketriskthroughtheirotherinvestments.Someofthebasicinformationthatanadvisershouldaskaboutaretheincomeleveloftheclient,whethertheyhavesavingsina401(k)plan apart from the money they want to invest with the adviser, their occupation, andwhethertheyhavechildren.Weformindicatorvariablesequaltooneiftheadviseraskedfor the specific information at some point in the consultation and zero otherwise. TheresultsinthefirstfiverowsofTable2showthatinthevastmajorityofcasesadvisersstartoff by asking the auditor for basic personal characteristics such as age, income,whethertheyhavechildren,andwhethertheyhavea401(k)plan.23Onaveragethesequestionsareaskedinmorethan70%ofthevisits.24

InTable3,weregressadummyindicatingwhethertheadviseraskedabouttheclient’sageonthegenderandlogageoftheclient.Wefocusonthesetwocharacteristicssincetheyaremosteasilyobservablefromtheoutsetofthevisit.Wefindthatwomenwereaskedfortheiragelessoftenwhilethecoefficientonlogageisnotsignificantbutpositive.Similarly,incolumns(2)through(4),weseethatwomenauditorswereaskedabouttheirpersonalandfinancialsituationlessoftenthanmen.Columns(3)and(4)ofTable3showthatolderpeoplewereaskedabouttheirfinancialsituationconcerningincomeandwhethertheyhavea401(k)planmoreoften thanyoungerpeople.Theseresultscould indicate thatadvisersadjust theirapproach towardsapotential client to reflect their expectation about the future return from this client including the probability of recruiting her: older auditors have higher investment funds, and men are usually viewed as being more willing to move their account to

22 Financialadvicefirmsstartedtoconsolidatetheiradvisorybusinessbyreducingthenumberofadvisers.

Thus,arrangingvisitswithinourdesign,giventhepreviousvisits,becamealmostimpossible.23 Asmentionedabove,auditorssaidthattheyareinvestingina401(k)planbutdonotwanttodiscussthe

detailsoftheseinvestments.24 Advisersaskforthiscrucialinformationeventhoughitisnotalegalrequirementasinothercountries.

11

another adviser. In both cases, the expected return for the adviser is higher than with younger or female clients. As a consequence, advisersareless likelytoaskyoungerorfemaleauditorssomebasicquestionabouttheirfinancialsituation,anditalsoleadstoworseadvicesincetheadviserdoesnothavefullinformation.

Whenwelookattheadvisers’recommendationsinthesametable,weseethatadvisershadamuchhigherpropensitytosuggestactivelymanagedmutualfundsthanindexfunds.The advisers encouraged the client to invest in index funds in only 7.5% of the advicesessions (21 visits). In contrast, in 50% (or 142) of the visits, the adviser suggestedinvesting in actively managed funds. This stark discrepancy is a first indication thatadvisers might be trying to guide clients to high‐fee investments.25 In that context, it isinteresting to see that a majority of advisers mention (some) fees of the recommendedfunds spontaneously, without the client having to ask for it. But in many cases, the feediscussion is used to downplay the impact of fees. Given that clients usually come toadvisers to receive help with their investments, it might not be too surprising that onaverageadviserstrytochangetheclients’fundallocation.Whatisinteresting,however,isthattheytendedtomoveshoppersawayfromtheexistingstrategyregardlessoftheinitialportfolio,i.e.,evenwhentheylookedatalow‐feediversifiedportfolio.Sotheywerewillingtomaketheclienteffectivelyworseoff.

Another interesting finding, inferred from clients’ free‐form answers, is that someadvisers (84visits, or roughly30%) refused tooffer any specific adviceuntil the auditortransferred resources to the adviser. This is interesting because it illustrates a screeningproblem for customers: it is hard for them to judge the value of an adviser beforecommittingtotheadviser.

Finally,themixofassetallocationsisinteresting(seeTable6).Advisersrecommendedon average an investment of about 63.5% in equity, 23.8% in bonds, and12.7% in cash.However,themeanadvicesuggestsaninternationalequityallocationofabout27%oftheportfolio. While this is a smaller international allocation than optimal portfolio theorysuggests,thisisanaggressiveequityallocationandinthatsenseleansagainstanybiasthatmightgenerateanequitypremium.

4.2Advisers’reactionstothecurrentinvestmentstrategy

In a second step,we examine how treatment assignments affect the advice that auditorsreceivedfromtheadvisers.Asdiscussedabove,sincewerandomlyassignclientsandtheirportfolios to different advisers, we can test the average response of a typical adviserwithoutanyconcernofclients’self‐selectiontodifferenttypesofadvisers.First,inTable4wefocusontheadviser’soverallreactiontotheauditor’sassignedportfolioasafunctionofthe different scenarios presented by the auditors. We measure whether the adviser issupportive of the auditor’s current strategy and supports more investment in the sameportfolio or is critical of the current strategy and discourages further investment in thesame.26Thisanalysisexcludesthe“cashonly”treatment.Inthattreatment,itisnotpossible

25 Note thatcatering toclient’sbeliefsbasedontheircurrentportfoliowould leadadvisers torecommend

moreand/ordifferent indexfundssince70%(treatment1and2)or100%(treatment3)of thecurrentportfolioareinvestedinindexfunds.

26 Theencouragement/supportvariableisbasedonthequestion“Didtheadvisermakeanycommentsabouthow you should modify your existing portfolio?” Auditors could choose from three multiple choiceanswers:

12

fortheadvisertoexpresssupportforthe“strategy”sincetheclientexplicitlyasksforhelpinimprovingtheinvestmentstrategy.Ifadviserstrytode‐biastheirclients,theyshouldnotbe supportive of the trend‐chasing or employer stock strategy but supportive of thediversified portfolio strategy. At most, they should suggest some internationaldiversification, inthatcase. Incontrast, ifadvisersaimtomaximizetheir fee incometheyshouldbesupportiveoftrend‐chasingbutnotsupportiveoftheemployerstockorefficientportfoliostrategy,sincethesewillnotgenerateahighnumberoftransactionsandfees.

EachcellinTable4presentsthemean,thestandarddeviation,andthep‐valueofat‐testbetweenthesubsetpresentedinthatcellversusthesamplemean.Overall,advisersseemtosupportstrategiesthatresult inmoretransactionsandhighermanagementfees.Thefirsttwo rows show the mean responses for whether the adviser supported the currentinvestmentstrategyorsuggestedtochangeit.27Inthecaseofthereturns‐chasingportfolio,theadvisersweresignificantlymoresupportive than foreither thecompanystockor theindexportfolio.Thelikelihoodofasupportiveresponsewas19.4%forthereturns‐chasingportfolio,againstthesamplemeanof13.1%,butonly9.7%forthecompanystockportfolioandaremarkablylow2.4%inthecaseoftheindexportfolio.Whenwemeasurewhethertheadviserproactivelyencouragedtheclienttochangethecurrentinvestmentstrategy,weseeaparallelpattern.The incidentofanegativeresponse issignificantlybelowthemeanfor the returns‐chasing portfolio but significantly above the mean for the index fundportfolio: in 59.2% of cases, an adviser suggested a change in the current strategy. Incontrast, if the client had an index portfolio, the adviser suggested changing the currentinvestmentstrategyin85.4%ofthecases.

Theseresultsshowthateventhoughthemeetingbetweentheclientandadviserisalsoasalessituation,advisersarewillingtogoagainstthe(revealed)preferencesoftheclientandsuggest changes away from the current strategy in themajority of cases. Therefore, thecateringconcernsarelessstronginthiscontextthanonemighthaveconjectured.However,itispossiblethatthevisitsstartoffdifferentlyfromtheultimatelysuggestedstrategyandreflect more of the sales pitch: in the next two rows, we analyze the adviser’s initialresponsetotheauditor’sportfolioasopposedtothefinalrecommendationswesawbefore.This first reaction could be interpreted as a judgment of the client’s prior behavior.Weindeedsee in row(3) that theadviser’s initial reaction to thedifferentportfolios ismostlikely to be positive in the case of the index portfolio (24.4%), least positive for thecompany stock portfolio (12.9%), and about average for the returns‐chasing portfolio(16.5%).Explicitlynegative comments about theportfolio areobservedonly in23.3%ofthecaseswithreturns‐chasingportfoliobutinabout56%forboththeindexandcompanystockportfolios.

Theseresultsshowadviserstrytobemorepositiveinitially,eventhoughtheyultimatelywanttochangetheclients’strategies. Interestingly,advisersseemmorecompelledtosaysomethingexplicitlycomplementingaboutthe indexportfolioandare lesscomfortabletodoso for the returns‐chasingor thecompanystockportfolios. It ispossible thatadvisers

(a) “Adviserencouragedmetoinvestmoreintheexistingstrategy”,(b) “AdvisersaidthatIshouldnotchangetheallocationinmyexistingstrategybutnotinvestmore”or(c) “Adviserdiscouragedmetoinvestmoreintheexistingstrategy”.Theencouragementvariableiscodedas1iftheauditoranswered(a)and0otherwise.Negativecomments(discouragement) to the same question are coded as 1 if the auditor chose (c) and 0 otherwise. Thecorrelationbetweenthetwodummyvariablesis‐0.5374.

27 Wehavetoomanybanksandtoofewrepeatvisitsinoursamplesuchthatwecannotincludebankfixedeffects. Probit regressions or using random effects instead of clustering at the auditor level havequalitativelythesameresultsandareavailableonrequest.

13

feel constraint by what they know is the better advice, or that they do not want tounderminetheircredibilitybybeingoutrightdismissiveofdiversificationandindexfunds,assuming thatmostpeoplehaveat leastheard through themediaandothersources thatthesearegoodinvestmentguidelines.

Table5 containstheresultsofaparallelanalysisbutusingaregressionset‐up,whereweclusterobservationsattheindividualclientleveltocontrolforbase‐ratecharacteristics.Theregressionscontrolfortheageandincomeleveloftheauditors,sincethesewerethetwovariablesonwhichwestratifiedauditors.Wealsoincludemonthfixedeffectssincethestudywas implemented over a fivemonths period to reduce the likelihood of detection.Column(1)reportstheresultsfromaregressionofadummyindicatingencouragementonthe treatment dummies where the omitted category is the returns‐chasing portfolio(scenario1).Theresultsshowthatadviserswereleastsupportiveoftheefficientportfolio,followedby the company stock treatment,with coefficients of ‐.28 and ‐.17, respectively,i.e.,adviserssupportedthereturns‐chasingstrategysignificantlymoreoftenthantheothertwostrategies.28

In column (2),we replicate the regression setup of column (1) but include additionalcharacteristicsoftheclientssuchasgender,maritalstatus,andinvestmentamount.Noneofthe characteristics are significant and including these controls does not change thecoefficientson the treatmentdummies,as is tobeexpectedgivenrandomassignment. Inregressions(3)and(4),wenowbreakoutcustomerswithanassignedinvestmentamountof about US$ 100,000 and about US$ 50,000. The idea is to see if the advice across thescenariosdiffersforwealthierclients,andwefindnosupportforanysuchdifferences.

Incolumns(5)to(8),werepeattheregressionsetupofthefirstfourcolumnsbutuseasaleft‐handsidevariablewhethertheadviserstronglydiscouragedinvestingfurtherintheexistingstrategy.Weagainfindconsistentlywiththepriorresultsthatadvisersweremostnegativeontheefficientportfolio.Thecoefficientis.4andsignificantatthe1%level.Thereisnosignificantdifferenceforcompanystock,however.Theseresultsstronglysuggestthatadvisers try todissuadeclients frominvesting inanefficientportfolio, likelybecause thisminimizesthefeeincomefortheadviser.Interestingly,thisincentiveseemstobesostrongthat advisers are willing to push clients out of investments portfolios that are close toperfectly efficient (index scenario). However, it does seem that the investor is moreconstrained in the case of company stock, perhaps because of an attempt to cater to theclients’ bias. This produces a perverse situation where the adviser is actively leaningagainst an efficient portfolio but not willing to lean against what is actually a biasedstrategy.

In columns (9) and (10),we look at the initial reactionsof the adviser.Thedifferencebetweenthedependentvariablehereandincolumns(1)through(4)isthatwenowfocuson the adviser’s very first reaction to the client’s portfolio, i.e., looking backward. Incontrast, the dependent variable in columns (1) through (4) focuses on how the adviserarguedtheportfolioshouldbestructuredorrestructuredgoingforward.AsexpectedbasedonourresultsinTable4,wenowfindverydifferentadvisers’reactions.Initially,advisersreact significantly more positively to the index portfolio relative to the returns‐chasingportfolio. But there is no significant difference between company stock and returns ‐chasing. In addition, these differences are much less pronounced than in the first eightcolumns.Allthissuggeststhatadvisersaremoremoderatedintheir initialreactions.And

28 Intheseregressionsweusescenario2asthebenchmarkcasetodetectdifferencesbetweenscenarios2

and3.Theseresultsareavailableonrequest.

14

they seemmuch lesswilling tomake overtly negative comments about the clients’ priorchoices.

4.3AssetAllocation

In the next step, we evaluate the overall asset allocation (stocks, bonds, domestic orinternational investments)that isrecommendedbytheadvisers. Ina largefractionofthevisits,wedidnotgetverydetailedquantitativeadviceaboutwhichspecificfundstoinvestin,sincemanyadvisersinsistedthattheclientshouldfirstplacethefundswiththeadviser’scompany.Therefore,wecancomparedifferencesinthecapitalallocationacrossbroadassetclasses.IntheanalysisinTable6,wewillnowincludeallfourdifferentscenariosincludingthe cash treatment (scenario 4), sincewe look at the recommendations that the advisermakesfortheportfoliogoingforward.ThemeancomparisoninTable6 isparallel tothesetup in Table 4. The outcome variables here are measured as the recommendedpercentageof theportfolio that shouldbe invested inoneof the respectiveasset classes.Thefirstthreerowsreporttheresultsforthefractionofassetsinstockorbonds(weomitcashas theresidualcategory)andthe fractionofassets tobe invested internationally.Ascanbeseeninrow4,weloseabout45%ofthesampleinthisanalysissincemanyadvisersdidnotprovide specific enough recommendations.29 Second, themeans tests reported inTable6 indicatethatadvisersdidnotseemtotailor theassetallocationaccordingtothescenariothatauditorscomeinwith,sincenoneofthecomparisonsareevenclosetobeingsignificant.Theoneexceptionisthatauditorswhocameinwithanindexportfolioseemtohavereceivedahighersuggestedallocationtobondsonaverageacrossthesampleadvisers,suggest investing23.8%of theassets inbonds,but this fractiongoesto31.9%forclientsarrivingwithanindexportfolio.Thisissurprisingsinceourcontrolgroupmightbeseenasbeingevenmoreriskaversesincetheyhavehadnoexposuretoequities.Whilewecannotrule out that thismight be a spurious correlation, it couldpossibly suggest that advisersassessedpeoplewhoholdindexportfoliosasmoreriskaverseandthusfeltthatastrongerallocationtobondscouldbeinlinewiththeauditor’spreferences.Butatthesametimethisinterpretationisnotsoeasytosquarewiththefactthatadviserswerestronglypushingtheclientswhocomeinwithindexportfoliostomoveoutoftheseandintoactivelymanagedfunds.30

Advisers did notmention fees at a significantly different frequency depending on ourfour treatments (row 5). However, there are significant differences between treatmentswith respect to recommendingactivelymanagedor index funds (rows6and7).The twomost concrete dimensions of advice that we can measure are whether the adviserrecommendedactivelymanagedfundsand/or(passive)indexfundsasaninvestmenttotheclient.Thesetwodimensionsareofinterestsincealargebodyofliteratureonmutualfundreturnssuggeststhatactivelymanagedfundsonaveragehavelowernetreturnsbutallowfund companies and advisers to charge higher fees. In contrast, index funds have beenshown to be a better investment option for retail investors since they provide access to

29 We also used an indicator variable forwhether the adviser did at allmention investment in anyof the

abovementionedassetclassesandtheresultsarequalitativelyverysimilar.30 Anotherexplanationisrelatedtotheamountoffeesthatcanbeearnedwithstockfundscomparedtobond

funds.Asaconsequence, itdoesnotpayofftoknowmoreaboutbondfundsthanaboutstockfunds, i.e.adviserssimplydonotknowenoughtorecommendanotherbondfunds.Advisersmayalsobelievethatitiseasiertomoveclients fromindexbondfunds intoactivelymanagedstockfundsthanshiftingthemtoactivelymanagedbondfunds.

15

investing in a well diversified portfolio at a low fee structure (see for example Gruber1996).Rows6and7ofTable6reconfirmthatonaverageadvisersweremuchmorelikelytorecommendactivelymanagedfunds(49.7%ofthecases)versusindexfunds(only7.3%ofthecases).Interestingly,however,thelikelihoodsofwhenadviserssuggestoneofthesetwo typesof funds varieswith the scenario: activelymanaged fundswere recommendedespecially frequently toauditorswhocame inwithan index fundportfolio (61.0%of thetime)31 or just cash investments (75.0% of the time). In the case of returns‐chasingportfolios,activelymanagedfundsweresuggestedonlyin40.8%ofthevisitsandevenlessfrequentlyforcompanystockportfolios,whereitwasonly24.2%ofthetime.Incontrast,indexfundswerealmostnevermentionedbytheadvisers.Onlywhenpeoplecameinwithacashportfoliotheadviserwassignificantlymorelikelytorecommendindexfunds.

Asbefore,wenowusearegressionframeworktoconfirmtherobustnessoftheresults.Column(1)ofTable7followstheusualbaselinesetupwhereweregresstherecommendedfractionofbondson thedummies for the fourscenarios (again, scenario1 is theomittedone) and controls for the client characteristics. The fraction of bonds that wasrecommendedfortheclient’sportfoliodoesnotseemtovarywiththedifferentscenarios.Thesame is true incolumns (2) through (4)when lookingat theallocations to theotherassetclasses.Again,theadvicedoesnotseemtovarybyscenario.Theoneexceptionisthatclients in the all‐cash scenario 4 have significantly lower exposure to stocks (column 2).Thiscouldsuggestthatclientswhoareinscenario4areconsideredtobeveryriskaverse(orevenunsophisticated)bytheadvisersandthereforetheadvisersmightthinkthattheseclientswouldnotbeabletohandletheriskexposureofahighfractionofequities.Buttheresultsarequitenoisy,whichismostlikelyduetothemuchsmallersamplesize.

Therecommendedinvestmentinstocksanddomesticassetssignificantlyincreasedwithclients’annualincome,whichmaybeexplainedbyanassumedhigherriskorlosstolerance.However,therecommendedexposuretostocksdecreasedwithlargerinvestmentamounts.Marriedclientswereadvisedtohavesignificantlymorebondandstockinvestmentsattheexpenseofliquidity,whereasfemaleclientswereadvisedtoinvestsignificantlylessinbothasset classes based. Themonths following the collapse of Bear Stearns had an effect onrecommendeddomesticinvestmentsonly–clientsweretoldinMaythroughJulytoinvestdomesticallysignificantlyless.

Thesedifferent recommendationsbasedonpersonal characteristicsmaybe causedbythe adviser’s information collection process. Note that clients always disclosed theinvestmentamountat thebeginningof theconversationsince theyasked foradvicewithrespecttotheircurrentportfolio.Thelikelihoodofbeingaskedfortheircurrentoccupation(regression (2)) or annual income (regression (3)) significantly decreased with higherinvestmentamounts(seeTable3insection4.1).

4.4Investmentsuggestions:Adviserandclientincentives

Wenowalsoruntheaboveresultswithrespect toadvisers’recommendationsofactivelymanagedorindexfundsinregressionspecificationtomakesurethattheysurviveanumberofdifferentcontrols.Incolumn(1)ofTable8,weregressanindicatorvariableequaltooneif theadviserrecommended investing in index fundsandzerootherwiseondummies for

31 Thisresultmaynotbetoosurprisingatfirst.Butrecommendinganinternationallywell‐diversifiedlowfee

index fundswould have been an option fort he adviser andwould have counted as “recommending anindexfund”.

16

thefourscenarioswiththereturns‐chasingscenarioastheomittedcategoryandcontrolsfor theageand income levelof theclient.All the regressionsareclusteredat theauditorlevel. The results in column (1) show that advisers were significantly more likely torecommendindexfundstoclientswhocameinwithan“allcash”portfoliorelativetothereturns‐chasing treatment (the coefficient on the “all cash” scenario is 0.18 and issignificant at the 5% level). The coefficients on all other scenarios are close to zero andstatisticallyinsignificant.Incolumn(2),werepeatthesameregressionbutincludefurthercontrols for client characteristics such as gender,marital status, numberof children, andtheamounttheyclientwantstoinvest.Theresultsareunchangedcomparedtocolumn(1).In columns (3) and (4), we break out the sample into the visits where auditors wereassignedaninvestmentamountaroundUS$100,000versusUS$50,000.32

In contrast, in columns (5) to (8) ofTable8,we repeat the same regression setup asbefore but the dependent variable now is a dummy for whether the adviser suggestedactivelymanaged funds to theclients. It is important tonote thatadvisers recommendedactivelymanagedfundsinabouthalfofallvisitswhiletheyonlyrecommendedindexfundsinabout8%ofvisits.Thebaselineregressionincolumn(5)showsthatadvisersweremuchmorelikelytorecommendactivelymanagedfundstoclientswhocameinwitheithertheindexfundportfolio(scenario3)ortheall‐cashportfolioscenario4).Instead,adviserswerealmost20%lesslikelytomentionactivelymanagedfundstoclientsinthecompanystockportfolio(scenario2).Again, theresultsareunchangedwithadditionalcontrols forclientcharacteristics incolumn(6).Asbefore,wethenbreakout thesub‐sampleofclients thathave about US$ 100,000 to invest (column 7) and those that have about US$ 50,000 toinvest (column 8). Interestingly, clients with higher investment amounts tend to berecommended actively managed funds in scenarios 3 and 4 at a much higher rate thanclients that are less wealthy. Moreover, less wealthy clients received significantly fewerrecommendations in theemployerstockscenario,whichmayreflectadvisers’beliefs thattheprobabilityofde‐biasingtheseclientsisratherlow.

Overall,advisersseemedtomaximizefeesbyplacingmoreweightonactivelymanagedfundsthatcreatedahigherfeeincome.Moststrikingly,evenifaclienthadawell‐diversifiedindex fundsportfolio, the adviser encouraged investment in activelymanaged funds.Theobjectiveoftheadviserinthisbehaviormighthavebeentosignalthattheycouldaddvalueto the client by suggesting something different from the existing portfolio. This behaviorwasparticularlypronouncedforwealthierclientswherethefeeincomematteredmoretothe adviser. But advisers could also have achieved this goal by suggesting low‐feeinternationaldiversification.Ingeneral,advisersdidnotproactivelyreachouttoclientstorebalance theportfoliodue to changing circumstancesof the client,butonly to sell themnew funds and generate fees. The advice that we observe in our treatments are a goodproxy for the different situations that an adviser might encounter with their clientsthroughoutalongertermrelationship.Theevidencesuggeststhatmostoftheinteractionisdrivenbytheneedtogeneratefeesratherthantorespondtotheclientsrebalancingneeds.

But advisers also seem to attempt to cater to their clients’ perceived preferences. Forclients that came in with a company stock portfolio, advisers were much less likely tosuggestactivelymanagedfunds.Onecouldhypothesizethathighconcentrationincompanystockmightsuggesttotheadviserthatthisclient ismoreriskaverseandpassiveintheirinvestmentapproachandthusmightnotbecomfortableinvestinginanactivelymanaged

32Thepercentageimprovesto10.7%(from4.4%)intheindexfundsscenariowhenweeliminatethosevisits

inwhichadvisersarenotwillingtoprovideanyadvicebeforetheclienttransfersherportfolio.

17

fund.Giventhatactivelymanagedfundswererecommendedtohalfof theclients, it isnosurprise that advisers spontaneously mentioned fees just as often or as seldom in allscenarios and with respect to most control variables. However, advisers spontaneouslyaddressed fees with an increasing likelihood if clients were older and with a smallerinvestment amount (regression (10)). The first result can be explained by an assumedhigherexperiencelevel,i.e.,adviserspre‐emptedanoftenheardquestioninthisagegroup.The second result may seem surprising, given that less wealthy clients received morerecommendations for actively managed funds. But mentioning fees can involve talkingabout load fees (or discount of load fees) as well as management fees, which are moreimportantinthelongrun.Thus,advisersmaybemorelikelytotalkaboutthelessrelevantfeeswithlowerlevelsofwealthandlesslikelytotalkabout(all)feeswithhigherlevelsofwealth.

4.5PersonalCharacteristicsandCustomizationofAdvice

In Table 9, we first analyze the advice that was given as a function of clientcharacteristics.Characteristicssuchasgenderandagecanbe inferredrelativelyeasilybymeetingtheclient.Wealsoincludecharacteristicsoftheclient’spersonalsituationthattheadviser learns from the conversation, such as the client’s marital status, the number ofchildren,andincomelevel.Standardfinancetheorywouldsuggestthatadvisersshouldtakepersonalcircumstancesoftheclientintoaccountwhensettingtheinvestmentstrategy.Forexample, clientswitha shorter investmenthorizon, suchasolder clients,most textbookswouldsuggestthattheyshouldnotbeinvestinginriskylong‐termstrategies.Similarly,riskaverse clients should not be guided towards risky investments, such as having a highpercentageinoftheirportfolioofstocksoractivelymanagedfunds.

In column (1),we investigatewhether theadviser encouraged theauditor to invest inindexfundsasafunctionoftheauditor’scharacteristics.Standarderrorsinallregressionsareclusteredattheauditorlevel.Weseethatthecoefficientonlogofageispositivebutnotsignificant. Similarly none of the other coefficients on the client characteristics aresignificant.Thus,therearenodiscernibledifferencesbetweenclientsthatwereencouragedto invest in index funds.But it is important tokeep inmind that theoverall incidenceofsuggesting index fundswasvery low(8%). Incolumn(2),werepeat thesameregressionset‐up but use an indicator variable forwhether the adviser suggested activelymanagedfundsasthedependentvariable.Now,thecoefficientonlogofageislargeandstatisticallysignificant.Olderclientsweremoreoftenencouragedtoinvestinactivelymanagedfunds.Similarly, peoplewith childrenwere encouraged to invest in activelymanaged funds. Incontrast, the coefficient on the gender dummy is negative and highly significant. Thissuggests that advisers are less likely to recommend actively managed funds to femaleclients. This could be in linewith the belief thatwomen aremore risk averse and thuswould have a lesser tolerance for actively managed funds. Alternatively, women areperceivedasbeingmorefee‐sensitive.

Wefindthattheonlysignificantandeconomicallyverylargecoefficientofwhethertheadviser spontaneously mentioned the fees is log age (see column (3)). Advisers tend toexplainthefeestructureoffundsmuchmoretoolderclientsthantoyoungerones.Noneoftheothercoefficientsoncharacteristicsaresignificant.Thebeliefmustbethatolderpeopleare more cost‐conscious and potentially better informed about investment options.Therefore,advisersproactivelydiscussthefeesratherthantryingtoignorethetopic.

18

Finally,incolumn(4),weanalyzetheadviser’sattitudetowardstheclientrelationship.We learned through the free text answers of our auditors that some advisers refused tooffer any specific advice as long as the potential client had not transferred his or heraccounttothecompanyoftheadviser(n=84).Theintentionoftheadviserseemstobethatthey firstwanted tosignup theclientbeforepartingwithanyuseful information.On theonehand,itmakessensethatadviserswouldwanttoprotecttheirtimeandinsightssothatclientsdonotreplicatetheadviceforfree.However,thissituationispuzzling,sinceitforcesthe client to choose an adviser without being able to get any indication of the person’squalityupfront.The result in column (4) shows that thisbehaviorwasmostpronouncedtowardsfemaleauditors.Adviserswerealmost40%morelikelytotell femaleclientsthattheyfirsthadtotransferthebalancetothem.Onecouldimaginethatthisbehaviormightbebasedontheperceptionthatwomenaremoredocileorgullible.5.Conclusions

Overallourfindingssuggestthatthemarketforfinancialadvicedoesnotde‐biasretailinvestorsandifanythingmayexaggerateexistingbiases.Whileadvisorsseemtotakeintoaccount client characteristics such as age or family status when making portfoliorecommendations, they are unwilling to lean against biases that help them further theirown economic interest, e.g. maximize fees. We find that in some cases the advice evenpushes clients towards funds with higher expected fees with little change in portfoliodiversificationandthuswouldreducetheexpectedreturnsontheirportfolios(asdiscussedinthecaseoftheindexportfolio).Intermediationinthismarketonaveragedoesnotseemto correct individual biases. Therefore, these biases can have first order implications foraggregate capital flows and possibly pricing of riskmore broadly, if there is not enoughinformedcapitaltoexploitarbitrageopportunitiesagainstcapitalflowsfrom“biased”retailinvestors. While in times of normal economic activity these biases might be arbitragedaway, one could worry that in times of crisis they become particularly important whenarbitrage capital dries up. It can also shed light on how we should model informationaggregation in equilibrium if competitive forces in themarket for financial advicedonotleadtotheprovisionofthebestpossibleadvice.Competitionmightbe limitedbythefactthat financial advisers exploit thebiases of naïve (oruninformed) retail investors.At thesametime,adviserswhoareinterestedinprovidingbetteradvicemightbeunabletogainmarket share if biased retail investors are unable to differentiate advice that is in theinterestof theclient fromadvice that isonly in theself‐interestof theadvisor.Whilewecannotruleoutthatcateringtoclientbelievesplayssomerole,ourevidencesuggeststhatadviser self‐interest plays an important role in generating advice that is not in the bestinterestof theclients.Our findingsshowthatadvisersdonot feel constraint toonlygiveinvestment recommendations that affirm the clients’ prior beliefs for fear of losing theclient. Infact,advisorshavenoproblemdiscouragingclientsfrominvestingmore intheircurrentstrategiesifthisisnotintheinterestoftheadvisor.Butwedoseesomeevidencethattheadvisersarefullyawarethatthisisasalessituation,sinceintheinitialinteractionwiththeclienttheyalwayspraisetheclient’spriorinvestmentchoicesnomatterwhattheyinvestedin.

Theseresultsareintriguingbuttheyarealsoonlyafirststepinwhatisaveryimportantresearcharea.Theyopenthedoortoasetofotherquestionsaboutthemarketforadvicethat our current sample size or treatments do not allow us to answer. Three questionsstand out. First, does the nature and quality of the advice depend on the adviser’sincentives?Answering thisquestionwill requirea largersampleofadviserswithvarying

19

incentives, e.g., fee‐only advisers.Though suchadvisershavegrown in recent years, theyarestillaminorityinthemarket.Moreimportantly,auditingthemwouldraisethecostsofauditssignificantlysinceeachauditwouldnowentailafee.Whileourcurrenttestsreflecttheadvicethattherepresentativeadviserinthemarketprovides,itwouldbeinterestingtounderstand how variation in incentivesmight affect the interactionwith the clients. Forexample,itispossiblethatfee‐onlyadvisersprovidebetteradvicebuthavetochargesuchhighfeesthataverageretailinvestorsarebetteroffinthestatusquosituation.Second,doesthequalityofadvicechangeovertime?Inanongoingandevolvingadvicerelationship,anadvisermight either use the relationship to tailor the advicemore closely to the clients’needs or to be tempted to take advantage of the increasing trust Third, how does thedemandforfinancialadvicelooklikeandwhatroledoesitplayinshapingthetypeadvisersthatsurviveinthemarket?Specifically,howdoindividualsassessthequalityofadvice?Inthisauditstudy,auditorswerewillingtogobacktoabout70%oftheadviserstheyvisitedbutnowwiththeirownmoney(seeTable2).Inotherwords,mostadviserssucceededinconvincingtheirpotentialclientsandthustheyhavenoneedtochangetheiradvicegiving.To understand why clients like the currently available financial advice would require adifferentmethodologywheretheunitofobservationwouldneedtobepotentialinvestors.Thesetypesofquestionsareessentialgoingforwardtounderstandingtheforcesthatshapetheequilibriuminthemarketforadvice.

20

ReferencesAltonji, Joseph G. and RebeccaM. Blank, 1999, Race and gender in the labormarket, In Orley C.

Ashenfelter and David E. Card (eds.), Handbook of Labor Economics, edition 1, volume 3,chapter48,3143–3259,Amsterdam:ElsevierScience.

Barber,BradM. andTerranceOdean, 2000,Trading IsHazardousToYourWealth: TheCommonStockInvestmentPerformanceofIndividualInvestors,JournalofFinance55,773–806.

Barberis, Nicholas and Richard H. Thaler, 2003, A Survey of Behavioral Finance. In G.Constantinides,M.Harris, andR.Stulz (eds.),HandbookofEconomicsofFinance,1052–1121,Amsterdam:ElsevierScience.

Benartzi,ShlomoandRichardH.Thaler,2001,NaiveDiversificationStrategiesinRetirementSavingPlans,AmericanEconomicReview,91.1(March),79–98.

Benartzi, Shlomo and Richard H. Thaler, 2007, Heuristics and Biases in Retirement SavingsBehavior,JournalofEconomicPerspectives21.3(Summer),81–04.

Bergstresser,Daniel,JohnM.R.Chalmers,andPeterTufano,2009,AssessingtheCostsandBenefitsofBrokers:APreliminaryAnalysisoftheMutualFundIndustry,ReviewofFinancialStudies22,4129–4156.

Bhattacharya,Utpal,AndreasHackethal,SimonKaesler,BenjaminLoos,andSteffenMeyer,2012,IsUnbiased Financial Advice To Retail Investors Sufficient? Answers from a Large Field Study,ReviewofFinancialStudies(forthcoming).

Bluethgen,Ralph,AndreasGintschel,ArminMüller,andAndreasHackethal,2008,FinancialAdviceandIndividualInvestors’Portfolios,GoetheUniversitätFrankfurtamMain.

Bodie, Zvi and Dwight B. Crane, 1997, Personal Investing: Advice, Theory, and Evidence from aSurveyofTIAA‐CREFParticipants,http://papers.ssrn.com/sol3/papers.cfm?abstract_id=36158.

Bogle,JohnC.,2000,JohnBogleonInvesting:TheFirst50Years,NewYork:McGraw‐Hill.

Bolton,Patrick,XavierFreixas,andJoelShapiro,2007.ConflictsofInterest,InformationProvision,and Competition in the Financial Services Industry, Journal of Financial Economics 85.2(August),297‐330.

Carhart, Mark M., 1997, On persistence in mutual fund performance, Journal of finance, Vol. 52, No. 1, pp. 57-82.

Campbell,JohnY.,2006,HouseholdFinance,JournalofFinance61,1553–1604.

Canner,Niko,N.GregoryMankiw, andDavidN.Weil, 1997,AnAssetAllocationPuzzle,AmericanEconomicReview87.1(March),181–191.

Chalmers, Johnand JonathanReuter,2011,What is the ImpactofFinancialAdviceonRetirementPortfolioChoiceandOutcomes?,Workingpaper.

Chater, Nick, Roman Inderst, and Steffen Huck, 2010, Consumer decision‐making in retailinvestmentservices.ReporttotheEuropeanCommission(SANCO).

Chen, Hailiang, De, Prabuddha, Hu, Yu Jeffrey and Hwang, Byoung‐Hyoun, 2011, Customers asAdvisors:TheRoleofSocialMediainFinancialMarkets,doi:10.2139/ssrn.1807265.

Del Guercio, Diane, Jonathan Reuter and Paula Tkac, 2010, Demand for Financial Advice, BrokerIncentives,andMutualFundMarketSegmentation,NBERWorkingPaperNo.16312.

Fix,Michael andMargeryA.Turner, 1998,ANationalReport CardonDiscrimination inAmerica:TheRoleofTesting,Washington,DC:UrbanInstitutePress.

Georgarakos,DimitrisandRoman Inderst,2011,FinancialAdviceandStockMarketParticipation,

21

ECBWorkingPaperNo.1296.

Gruber,Martin, 1996,AnotherPuzzle: TheGrowth inActivelyManagedMutual Funds, Journal ofFinance51.3(July),783‐810.

Hackethal, Andreas, Roman Inderst, and Steffen Meyer, 2011, Trading on Advice, doi:10.2139/ssrn.1701777.

Hackethal, Andreas, Michael Haliassos, and Tullio Jappelli, 2012, Financial Advisors: A Case ofBabysitters?,JournalofBankingandFinance,36(2),509‐24.

Heckman,JamesJ.,1998,DetectingDiscrimination,JournalofEconomicPerspectives12.2(Spring),101‐16.

Hung,AngelaA.,NoreenClancy,JeffDominitz,EricTalley,ClaudeBerrebi,andFarrukhSuvankulov,2008,InvestorandIndustryPerspectivesonInvestmentAdvisersandBroker‐Dealers,TechnicalReport,RANDCenterforCorporateEthicsandGovernance.

Inderst,RomanandMarcoOttaviani,2009,MissellingthroughAgents,AmericanEconomicReview99.3(June),883‐908.

Inderst, Roman and Marco Ottaviani, 2011a, Competition through Commissions and Kickbacks,mimeo.

Inderst, Roman and Marco Ottaviani, 2011b, How (not) to pay for advice: A framework forconsumerfinancialprotection,mimeo.

Iyer,RajkamalandAntoinetteSchoar,2009,TheRoleofCultureinFinancialNegotiations:EvidencefromanAuditStudy, InternationalDifferences inEntrepreneurship,NBERConferenceVolume(forthcoming).

Iyer, Rajkamal and Antoinette Schoar, 2011, Ex Post (In)Efficient Negotiation and Breakdown ofTrade,mimeo.

Kramer, Marc, 2012, Financial Advice and Individual Investor Portfolio Performance, FinancialManagement(forthcoming),

Levitt, StevenD. andChadSyverson, 2008,MarketDistortionswhenAgents areBetter Informed:TheValueofInformationinRealEstateTransactions,ReviewofEconomicsandStatistics90.4(November),599‐611.

Malmendier, Ulrike and Devin Shanthikumar, 2007, Are small investors naive about incentives?,JournalofFinancialEconomics85.2(August),457‐489.

Moore,DonA.,GeorgeLoewenstein,DaylianCain, andMaxH.Bazerman (eds.),2005,ConflictsofInterest.CambridgeUniversityPress.

Schneider,HenryS.,2009,AgencyProblemsandReputationinExpertServices:EvidencefromAutoRepair,mimeo.

22

Table1:RandomizationacrossScenarios

Scenarios ChasingReturnsPF CompanyStockPF IndexFundsPF AllCash

(1) (2) (3) (4)

NumberofObservations 103 62 49 70 Age 39 40 40 40AnnualIncome US$81,000 US$80,000 US$82,000 US$81,000InvestmentAmount US$77,000 US$81,000 US$80,000 US$79,000%FemaleAuditors 77% 80% 78% 75% This table shows the distribution over the four different scenarios of the 284 audit visits to financial advisersbetweenApril andAugust2008. In scenario1, 30%of the client'sportfolio(PF)was investedin a sectorthat out‐performed the S&P 500 in the year prior to the audit study (Chasing Returns). In scenario 2, clients' currentportfolios contained30%companystock. Scenario3 refersto a diversifiedclientportfoliowithonly low‐feeindexfunds. These fundswere also used for the other 70% of the portfolioin scenario1 and scenario2. In scenario4,the client said that all the money was investedin a certificateof deposit at a local bank. The auditor'sactual agewasusedwhereastheannualincomeandtheinvestmentamountwereassignedbythestudy.

23

Table2:DescriptiveStatisticsabouttheAdviser‐ClientConversation

VARIABLES Yes No Total

Advisersencouragemoreofcurrentstrategy? 28 186 214Adviserssuggestchangeofcurrentstrategy? 144 70 214InitialReactionPositive 36 178 214InitialReactionNegative 85 129 214 Recommendindexfunds 21 263 284Recommendactivelymanagedfunds 142 142 284Spontaneouslymentionfees 160 121 281 Askaboutage 236 48 284Askaboutcurrentoccupation 217 67 284Askaboutannualincome 212 72 284Askabout401k 252 32 284Askaboutnumberofchildren 200 84 284

Auditorwouldgobacktothisadviserwithownmoney 200 84 284

This table contains descriptive statistics for the 284 audit visits at financial advisers. Auditors in thecontrol treatment did not answer the first four questions. As answers to the question "Did the advisermake any comments about how you should modify your existing portfolio?" auditors could choose frommultiple choice answers: (a) "Adviser encouraged me to invest more in the existing strategy", (b)"Advisersaid that I should not change the allocation in my existing strategy but not invest more" or (c)"Adviserdiscouraged me to invest more in the existing strategy". The answer is counted as an encouragement if theauditor answered (a). If the adviser picked answer (c), he suggested a change of the current strategy. Aninitial positive or negative reaction to the clients' current portfolio has been recorded, too. Auditors alsoentered in the online exit questionnaire whether index funds or actively managed funds wererecommended and whether any fees were mentioned by the adviser without being asked. In addition,auditorsrecordedwhethertheywereaskedatsomepointduringtheirvisitabouttheirage,currentoccu‐pation, annual income, the existence of a 401(k) plan or children. Finally, we asked auditors after eachvisitwhethertheywouldgobacktothisadviserwiththeirownmoney,ornot.

24

Table3:Personalcharacteristicsandadviser’sinformationgathering

(1) (2) (3) (4)VARIABLES

herage hercurrentoccupation

herannualincome

a401(k)plan

log(Auditor'sAge) 0.127 0.205 0.342** 0.321**(0.153) (0.169) (0.165) (0.141)

Gender ‐0.171* ‐0.106 ‐0.182* ‐0.107**(0.101) (0.0656) (0.0931) (0.0525)

Constant ‐5.234* ‐7.865*** ‐5.709** ‐1.243(2.737) (2.785) (2.201) (1.765)

Observations 283 283 283 283R‐squared 0.178 0.129 0.111 0.127

This table shows regression results for advisers' questions about the auditor's age, her current occupation, herannual income and the existence of a 401(k) plan (all variables yes: 1/no: 0) based on 284 audit visits atfinancial advisers between April (omitted) and August 2008. The additional explanatory variables are theauditor'sactualagehergender(female=1)asdummyvariables.

Adviseraskshisclientabout…

Robuststandarderrorsinparentheses;***p<0.01,**p<0.05,*p<0.1.

25

Table4: Advisers’ initial reaction and (non‐) supportive recommendations in differenttreatments

OverallSample ChasingReturns CompanyStock IndexFunds

13.11 19.41 9.68 2.44(39.74) (29.81) (15.62)0.007 0.0341 0.024

67.48 59.22 69.35 85.37(49.38) (46.48) (35.78)0.011 0.707 0.006

16.99 16.50 12.90 24.39(37.30) (33.79) (43.48)0.854 0.308 0.160

39.81 23.30 56.45 56.09(42.48) (49.98) (50.24)0.001 0.001 0.017

#observations 214 103 62 49

This table contains data collected in 214 visits (without the control treatment 4) at a financial adviser.Reported are the means, standard deviations and p‐values of tests between the sample mean and thetreatment mean. The two variables (supportive/not supportive) are coded based on the question "Didthe adviser make any comments about how you should modify your existing portfolio?" Auditors couldchoose from multiple choice answers: (a) "Adviser encouraged me to invest more in the existingstrategy", (b)"Adviser said that I should not change the allocation in my existing strategy but not investmore" or (c)"Adviser discouraged me to invest more in the existing strategy". The answer is counted asan encouragement (=1) if the auditor answered (a) and 0 otherwise. The answer is seen as discouraging(=1) if the auditor chose (c) and 0 otherwise. An initial positive (yes: 1/no: 0) or negative (yes: 1/no: 0)reactiontotheclients'currentportfolioareusedforrows(3)and(4),respectively.

Istheadvisersupportive ofthecurrentstrategy?

Initalreactionpositive

Initalreactionnegative

Istheadviseragainst thecurrentstrategy?

26

Table5:Advisers’initialreactionand(non‐)supportiverecommendations

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)VARIABLES positive

initialreaction

negativeinitialreaction

InvestmentAmount ≈$100k ≈$50k ≈$100k ≈$50k

CompanyStockPF ‐0.165** ‐0.185** ‐0.216 ‐0.205* 0.110 0.162 0.0550 0.297** 0.0137 0.252*(0.0807) (0.0780) (0.154) (0.102) (0.0908) (0.111) (0.179) (0.117) (0.0766) (0.125)

IndexFundsPF ‐0.284** ‐0.304** ‐0.418* ‐0.300** 0.397*** 0.438*** 0.476** 0.521*** 0.199** 0.195(0.111) (0.116) (0.210) (0.108) (0.0989) (0.109) (0.176) (0.147) (0.0884) (0.167)

log(Auditor'sAge) 0.0965 0.0121 0.991 ‐0.681 ‐0.0175 0.380 0.394 1.888** 0.327* 0.265(0.152) (0.396) (0.732) (0.489) (0.181) (0.485) (0.481) (0.775) (0.181) (0.275)

log(AnnualIncome) ‐0.266 ‐0.249 ‐0.124 ‐0.801*** 0.358 0.381 0.123 0.751** 0.194 ‐0.0901(0.228) (0.260) (0.349) (0.166) (0.252) (0.262) (0.390) (0.275) (0.357) (0.353)

log(InvestmentAmount) 0.0190 ‐0.186(0.258) (0.293)

MaritalStatus ‐0.0354 0.0866(0.0355) (0.0924)

Children ‐0.125* 0.135(0.0682) (0.119)

Gender ‐0.0119 ‐0.0871(0.108) (0.127)

Constant 2.612 2.561 ‐2.026 11.24*** ‐3.911 ‐3.539 ‐2.442 ‐14.86*** ‐3.287 0.0809(2.273) (2.185) (3.556) (2.984) (2.781) (3.039) (3.308) (4.765) (4.006) (3.494)

Observations 214 214 105 109 214 214 105 109 214 214R‐squared 0.093 0.116 0.137 0.158 0.107 0.127 0.148 0.165 0.075 0.172

The explanatory variables are dummy variables for the first three scenarios: In scenario 1 (omitted), clients were chasing fund returns. In scenario 2,clients' current portfolios (PF) contained 30% company stock. Scenario 3 refers to a diversified client portfolio with only low‐fee index funds. Theauditor's actual age, her annual income assigned by the study, and the investment amount assigned by the study are used, as well. Finally, the client'sactualmaritalstatus(married=1),theexistenceofatleastonechildinthehousehold,andtheclient'sgender(female=1)enterasdummyvariables.

Robuststandarderrorsinparentheses;***p<0.01,**p<0.05,*p<0.1.

This table shows regression results for advisers' reactions to the clients' current portfolios based on 284 audit visits at financial advisers between April(omitted) and August 2008 ‐‐ did they encourage the client to pursue the current strategy or did they discourage the client? The two variables arecoded based on the question "Did the adviser make any comments about how you should modify your existing portfolio?" Auditors could choose frommultiple choice answers (a) "Adviser encouraged me to invest more in the existing strategy", (b)"Adviser said that I should not change the allocation inmy existing strategy but not invest more" or (c)"Adviser discouraged me to invest more in the existing strategy". The answer is counted as anencouragement (=1) if the auditor answered (a) and 0 otherwise. The answer is discouraging (=1) if the auditor chose (c) and 0 otherwise. An initialpositive(yes:1/no:0)ornegative(yes:1/no:0)reactiontotheclients'currentportfolioareusedasdependentvariablesforregressions(9)and(10).

adviserdiscouragedclienttoinvestmoreintheexistingstrategy

adviserencouragedclienttoinvestmoreintheexistingstrategy

27

Table6:Averageadvisers’assetallocationrecommendationsandfees

OverallSample ChasingReturns CompanyStock IndexFunds AllCash

allocationbonds% 23.79 23.71 21.84 31.94 21.65(16.16) (13.07) (15.64) (19.42) (16.43)

0.964 0.369 0.007 0.258

allocationstock% 63.48 67.13 63.65 67.27 57.89(25.97) (22.85) (26.21) (19.91) (30.59)

0.220 0.964 0.411 0.056

allocationinternational% 26.81 29.46 24.67 23.23 27.78(17.21) (14.91) (19.49) (11.06) (20.45)

0.199 0.414 0.234 0.662

observations(allocation)

56.94 59.41 48.39 60.53 58.75(49.61) (49.35) (50.38) (49.53) (49.53)

0.533 0.124 0.638 0.701

activelymanagedfunds 49,65 40.78 24.19 60.98 75.00(50.09) (49.38) (43.18) (49.38) (43.57)

0.024 0.000 0.117 0.000

indexfunds 7.34 2.91 4.84 0.00 18.75(26.13) (16.89) (21.63) (0.00) (39.27)

0.031 0.395 0.051 0.000

observations(all) 284 103 62 49 70

advisermentionsfeesspontaneously

This table contains data collected in 284 visits at financial advisers in the Boston/Cambridge area in 2008.Reported are the means, standard deviations and p‐values of tests between the sample mean and thetreatment mean. There are fewer observations of the allocation data since not all advisers provided allocationadvice. Mentioning fees (yes: 1/no: 0), actively managed funds (yes: 1/no: 0) and index funds (yes: 1/no: 0)areusedforrows5,6and7,respectively.

28

Table7:Advisers’assetallocationrecommendations

(1) (2) (3) (4)VARIABLES

inbonds instocks internationally domestically

CompanyStockPF ‐0.351 0.0278 ‐4.347 ‐0.306(4.569) (5.735) (6.797) (7.785)

IndexFundsPF 7.108 3.776 ‐3.595 10.24(4.963) (6.792) (6.094) (6.699)

AllCashPF ‐2.435 ‐9.851* ‐3.595 ‐13.06(6.046) (5.143) (8.691) (8.925)

log(Auditor'sAge) 7.866 ‐0.812 5.149 ‐2.739(8.620) (12.66) (10.34) (14.09)

log(AnnualIncome) 12.57 27.52** ‐19.63 33.42*(11.96) (12.72) (15.05) (19.23)

log(InvestmentAmount) ‐2.444 ‐3.982*** ‐0.127 ‐2.437*(1.526) (0.905) (0.766) (1.293)

MartialStatus 0.908 20.44*** 10.92** 14.26*(3.890) (3.804) (4.676) (7.364)

Children ‐6.962 6.187 4.131 ‐15.49*(4.221) (4.496) (6.536) (7.871)

Gender ‐11.57** ‐9.621* ‐11.31** ‐10.76(4.314) (5.252) (4.970) (8.484)

Constant ‐116.7 ‐205.4 237.6 ‐251.3(150.3) (156.4) (177.0) (255.5)

Observations 167 172 152 152R‐squared 0.174 0.211 0.128 0.334

This table shows regression results for advisers' portfolio allocation recommendations for bonds, stocks,internationally and domestically (all in %) based on 284 audit visits at financial advisers between April (omitted)and August 2008. The explanatory variables are dummy variables for all four scenarios: In scenario 1 (omitted),clients were chasing fund returns. In scenario 2, clients' current portfolios (PF) contained 30% company stock.Scenario 3 refers to a diversified client portfolio with only low‐fee index funds. In scenario 4, the client claimed thatall the money was invested in a certificate of deposit at a local bank. The auditor's actual age, her annual incomeassignedbythestudy,investmentamountassignedbythestudy.Inaddition,weusetheclient'sactualmartialstatus(married=1), the existence of at least one child in the household, and the client's gender (female=1) as dummyvariables.

adviserrecommendstoinvest%ofportfolio

Robuststandarderrorsinparentheses;***p<0.01,**p<0.05,*p<0.1.

29

Table8:Advisers’recommendationsandmentioningfees

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)VARIABLES

InvestmentAmount ≈$100k ≈$50k ≈$100k ≈$50k ≈$100k ≈$50k

CompanyStockPF 0.0166 0.0157 0.0222 0.102 ‐0.198** ‐0.149* ‐0.00673 ‐0.455** ‐0.108 ‐0.0967 ‐0.0133 ‐0.0774(0.0540) (0.0567) (0.0246) (0.179) (0.0906) (0.0737) (0.0796) (0.180) (0.0966) (0.103) (0.152) (0.157)

IndexFundsPF ‐0.0191 ‐0.0218 0.0442 0.105 0.255* 0.308** 0.663*** ‐0.101 ‐0.0149 ‐0.0444 0.372* ‐0.248(0.0354) (0.0415) (0.0429) (0.100) (0.138) (0.119) (0.0978) (0.167) (0.142) (0.133) (0.215) (0.146)

AllCashPF 0.183** 0.168** 0.256*** 0.181 0.318*** 0.292*** 0.517*** 0.0750 ‐0.0106 ‐0.0274 0.0860 0.0421(0.0827) (0.0697) (0.0658) (0.137) (0.104) (0.105) (0.0704) (0.144) (0.114) (0.107) (0.167) (0.118)

log(Auditor'sAge) ‐0.0847 0.129 ‐0.541 1.559 0.287 0.504** ‐0.231 1.600 0.325 0.494** ‐0.563 1.190(0.120) (0.110) (0.410) (0.969) (0.172) (0.193) (0.560) (1.050) (0.196) (0.205) (0.911) (0.782)

log(AnnualIncome) 0.0313 0.0292 0.142 ‐0.165 ‐0.170 ‐0.267 ‐0.0744 0.186 ‐0.134 ‐0.131 0.223 ‐0.526(0.141) (0.126) (0.119) (0.183) (0.291) (0.273) (0.272) (0.458) (0.240) (0.220) (0.401) (0.312)

log(InvestmentAmount) ‐0.109*** ‐0.0783*** ‐0.0620***(0.0110) (0.0220) (0.0175)

MaritalStatus 0.0271 0.00512 0.106(0.0365) (0.0623) (0.0889)

Children ‐0.0521 0.272*** ‐0.0638(0.0518) (0.0913) (0.0997)

Gender 0.0553 ‐0.134* 0.0409(0.0490) (0.0744) (0.101)

Constant ‐0.0528 0.321 0.217 ‐3.570 0.884 2.188 2.067 ‐7.605* 0.849 0.934 0.0365 2.371(1.349) (1.285) (1.847) (3.211) (3.311) (2.968) (3.105) (4.085) (2.862) (2.673) (5.195) (4.307)

Observations 284 284 148 136 284 284 148 136 281 281 148 133R‐squared 0.101 0.177 0.185 0.144 0.180 0.226 0.301 0.211 0.028 0.049 0.070 0.094

This table shows regression results for advisers' investment recommendations of index funds or actively managed funds (yes: 1/no: 0) based on 284 audit visits atfinancial advisors between April (omitted) and August 2008. In addition, regression results are shown for advisers' mentioning of fees (yes: 1/no: 0) without beingprompted by the client. The explanatory variables are dummy variables for all four scenarios: In scenario 1 (omitted), clients were chasing fund returns. In scenario 2,clients' current portfolios (PF) contained 30% company stock. Scenario 3 refers to a diversified client portfolio with only low‐fee index funds. In scenario 4, the clientclaimed that all the money was invested in a certificate of deposit at a local bank. The auditor's actual age, her annual income assigned by the study, investment amountassigned by the study. In addition, we use the client's actual marital status (married=1), the existence of at least one child in the household, and the client's gender(female=1)asdummyvariables.

adviserrecommendsindexfunds adviserrecommendsactivelymanagedfunds

advisertalksspontaneouslyabout(any)fees

Robuststandarderrorsinparentheses;***p<0.01,**p<0.05,*p<0.1.

30

Table9:Adviceasafunctionofauditorcharacteristics

(1) (2) (3) (4)VARIABLES adviser advice

indexfunds activelymanagedfunds

mentionsfees aftermoneytransferonly

log(Auditor'sAge) 0.190 0.643*** 0.527** ‐1.283(0.118) (0.217) (0.197) (0.811)

log(AnnualIncome) ‐0.0683 ‐0.361 ‐0.106 0.488(0.120) (0.272) (0.219) (0.350)

MaritalStatus 0.0557 0.107 0.130 ‐0.124(0.0366) (0.0876) (0.0804) (0.105)

Children ‐0.0203 0.320*** ‐0.0471 0.138(0.0438) (0.0813) (0.101) (0.151)

Gender 0.0194 ‐0.209** 0.0146 0.383***(0.0417) (0.0979) (0.101) (0.116)

Constant 1.366 2.884 0.574 ‐11.20**(1.248) (3.068) (2.688) (4.488)

Observations 284 284 280 128R‐squared 0.113 0.124 0.045 0.218

adviserrecommends

This table shows regression results for advisers' recommendations of index funds and actively managedfunds based on all 284 audit visits at financial advisers between April (omitted) and August 2008. The otherregressions evaluate when advisers spontaneously mention (any) fees and why advisers offer advice onlyafter the portfolio or money is transfered to the new account. The additional explanatory variables are theauditor's actual age, her assigned annual income, her actual marital status (married=1) and the client'sgender(female=1)asdummyvariables.

Robuststandarderrorsinparentheses;***p<0.01,**p<0.05,*p<0.1.

31

Appendices–Auditstudydesignexamplesandtestquestionsforpotentialauditors

AppendixA–Portfolioandbackgroundinformationexamples

Scenario1(chasingfundreturns)

Scenario2(employerstocks)

Shopper ID 3

Scenario 1 Company Haemonetics

Gender Female Occupation Product Development

Age 30‐35 Annual Income $65.000

$ to Invest $45.000

Portfolo ID 1T1

Portfolio $45.000,00

SPDR S&P Metals & Mining (XME) $13.500,00

Vanguard 500 (VFINX) $22.500,00

Vanguard Interm-Term Bond Index (VBIIX) $9.000,00

On‐line account on E‐Trade (mention only if asked)

Shopper ID 52

Scenario 2 Company Analog Devices

Gender Male Occupation Intellectual Property

Age 30‐35 Annual Income $75.000

$ to Invest $49.000

Portfolo ID 1T2

Portfolio $49.000,00

Analog Devices Stocks (ADI)* $14.700,00

Vanguard 500 (VFINX) $24.500,00

Vanguard Interm-Term Bond Index (VBIIX) $9.800,00

On‐line account on E‐Trade (mention only if asked)

32

AppendixB–Testforpotentialauditors

1.Pleasegiveanexampleofafundfamily?a. GeneralMillsb. Vanguardc. USGovernmentd. SovereignBank

2.Whatdoadvisersmeanwhentheytalkofactivelymanagedfunds?a. Fundswherethemanagerpickssecuritiesinwhichtoinvestthemoneyb. Anyfundthathasamanagerc. Fundsthattrackthemarketportfolio

3.Usuallyindexfundshavelowerfeesthanactivelymanagedfunds?True/false

4.Thedefinitionofdividendsisa. theprofitsthatacompanyretainseachyeartoinvestinnewprojectsb. anamefortheincreaseinannualstockpriceofacompanyc. thepartofprofitthatacompanypaystoitsinvestorseachyear

5.Whichofthefollowingisnota“security”?a. ProcterandGamblestockb. DividendfromCoca‐Colac. USGovernmentbondd. OppenheimerCapitalAppreciationA(OPTFX)

6.Adviserswillusuallyrecommendthatyourportfolioshouldbediversifiedbetweenbonds,stocksandmutualfunds.ImagineyouhadUS$20,000toinvestandwantedtoinvest60%inequityand40%inbonds,whichofthefollowingportfoliosshouldyoupick?a. US$6,000inVanguardS&P500index

US$4,000inMFShighIncomeMunicipal US$10,000inCertificateofDeposit

b. US$8,000USGovernmentbonds US$1,000FordMotorstock

US$6,000inVanguardS&P500index US$5,000inAllianceBernsteinIntlGrowthA

c. US$8,000MFShighIncomeMunicipal US$10,000IBMstock

d. US$8,000Vanguard500Index US$12,000USGovernmentBonds

7.Whatistherelationbetweenriskofdefaultandreturnofacorporatebond?a. Thehighertheriskofdefaultthelowershouldbetheinterestrateb. Thelowertheriskofdefaultthelowertheinterestratec. Moretrustworthydebtorsofferhigherinterestratesd. Returnisindependentofrisk

8.Afrontloadisa. asalescommissiontheinvestorhastopayperiodicallyforholdingafundb. afundwithaheavyconcentrationinshorttermsecuritiesc. ameasureoftheoperationalcostsofthefundd. asaleschargetheinvestorpayswhenbuyingamutualfund