Shares in Equity Crowdfunding - Uni Trier · Pricing shares in equity crowdfunding Abstract We...
Transcript of Shares in Equity Crowdfunding - Uni Trier · Pricing shares in equity crowdfunding Abstract We...
Research Papers in Economics
No. 7/15
Pricing Shares in Equity Crowdfunding
Lars Hornuf
Matthias Neuenkirch
Pricingsharesinequitycrowdfunding
LarsHornufa,bandMatthiasNeuenkircha
aUniversityofTrier
bMaxPlanckInstituteforInnovationandCompetition
Firstdraft:30June2015
Thisversion:28July2016
Correspondingauthor:
MatthiasNeuenkirchDepartmentofEconomicsUniversityofTrierD‐54286TrierGermanyTel.:+49–651–2012629Email:neuenkirch@uni‐trier.de
* The authors are highly indebted to themanagers and owners of Innovestmentwhoprovided the investor data as well as Matthias Schmitt who calculated the data ongeographicdistances.TheyalsothankGregorDorfleitner,LarsKlöhn,AlexanderRajko,ElisabethSchulte, andparticipantsof the3.CrowdinvestingSymposium (UniversityofMunich) and Munich Summer Institute 2016 (Bavarian Academy of Sciences andHumanities),whoprovidedvaluablecommentsandsuggestionsonpreviousversionsofthepaper.Anearlierversionofthepaperwascirculatedunderthetitle“PricingtheValueofCashFlowRightsinCrowdinvesting:AnAnalysisofInnovestmentBackers.”
2
Pricingsharesinequitycrowdfunding
Abstract
Weanalyzethepricingofcashflowrightsinstart‐upcompaniesusingauniquedataset
of44equitycrowdfundingcampaigns.Oursampleconsistsof499backerswhoinvested
during the period fromNovember 6, 2011, toMarch 25, 2014, on the German equity
crowdfunding portal Innovestment. In contrast with all other European equity
crowdfunding portals, Innovestment runs a multi‐unit second‐price auction in which
backersthemselvescanspecifythepriceofaninvestmentticket.Weexploitthisunique
auctionmechanismtoanalyzebackers’willingnesstopayforcashflowrights.Wefind
thatcampaigncharacteristics,investorsophistication,progressinfunding,herding,and
stock market volatility influence backers’ willingness to pay in an economically
meaningfulmannerwhilegeographicdistance,learningeffects,andsnipingattheendof
anauctionhavenoeffect.
Keywords:Auctions,Equitycrowdfunding,Valuationofshares
JELClassifications:D44,G11,M13
3
1. Introduction
Aroundtheglobe, lawmakersaretakingactionstobringequitycrowdfundingundera
specific legal umbrella. Equity crowdfunding (also referred to as investment‐based
crowdfunding, securities crowdfunding, or crowdinvesting1) constitutes a financial
innovation in securities issuance that gives small entrepreneurs access to the general
public.Regulatoryeffortsoftenpursuetheobjectivetofacilitateentrepreneurialactivities
while also putting a minimum level of investor protection in place. To balance this
tradeoff,regulatorsmustconsidertheactualbehaviorofinvestorsinthesemarkets.In
thispaper,weinvestigatehowbackerspricethevalueofcashflowrightsinastart‐up
companywhenengaginginanequitycrowdfundingcampaign,usingauniquedatasetof
Innovestmentbackers.
Prior studies on Internet‐based entrepreneurial finance have mainly focused on
donation‐based crowdfunding (Bøg et al. 2012; Burtch et al. 2013;Koning andModel
2013;Meer2014;SaxtonandWang2014),reward‐basedcrowdfunding(Agrawaletal.
2013; Belleflamme et al. 2014; Colombo et al. 2015; Kuppuswamy and Bayus 2014;
MaromandSade2013;Mollick2013,2014;YounkinandKashkooli2013;Zvilichovskyet
al. 2013), andcrowdlending (Burtchet al. 2014;Linet al. 2012;LinandViswanathan
2013).Inoneofthefirststudiesonequitycrowdfunding,Agrawaletal.(2013)analyze
therevenue‐sharingmodelofSellaband.UndertheSellabandmodel,backersreceivea
portionof the futurereturns thatanartistgeneratesbyproducingmusic.Ahlersetal.
(2015)investigateinvestorsontheAustralianequityportalASSOB.Theyfindthatstart‐
upslistedontheportalusesignalswithregardtofinancialroadmaps,riskfactors,andthe
internalgovernanceofthefirmthatencouragecrowdinvestorstoparticipate.Blocketal.
(2016),HornufandSchwienbacher(2015),andVismara(2015)investigatethefunding
dynamics inequitycrowdfunding.They find that investorsbase theirdecisionson the
1Inthispaper,werefertothenewassetclassasequitycrowdfunding,asthisisthetermmostfrequently
used in the literature. See also the JOBS Act, which includes the term ‘crowdfunding’ referring totransactionsinvolvingtheofferorsaleofasecurity,andAhlersetal.(2015,p.955),whodefinetheterm‘equitycrowdfunding’asa‘formoffinancinginwhichentrepreneursmakeanopencalltosellaspecifiedamountofequityorbond‐likesharesinacompanyontheInternet’.TheFCAConsultationPaperCP13/13‘TheFCA’sregulatoryapproachtocrowdfunding(andsimilaractivities)’andtheEuropeanSecuritiesandMarkets Authority ‘Opinion Investment‐based crowdfunding’ use the term ‘investment‐basedcrowdfunding’.Knightetal.(2012)andtheUSSecuritiesandExchangeCommission(17CFRParts200,227,232etal.Crowdfunding,ProposedRule)refertothisnewactivityas‘securitiescrowdfunding’.Theterm‘crowdinvesting’isprobablythemostuseful,asitencompassesallfinancialinstrumentsfoundinpractice,regardlessofwhethertheyareclassifiedassecuritiesorinvestmentsorlackalegaldefinitionaltogether(KlöhnandHornuf2012).
4
informationofferedbytheentrepreneurintheformofupdatesandbypeerinvestments
and comments of other crowd investors. Moreover, there is evidence for a collective
attentioneffectandherdingbehavior.
Inwhat follows,weanalyze thepricingof cash flowrights in a start‐upcompanyby
equitycrowdfundingbackers.IncontrastwithallotherEuropeanequitycrowdfunding
portals,Innovestmentdeviatesfrombrokeringfixed‐priceinvestmentticketsonafirst‐
come, first‐served basis. Instead, the portal implemented a multi‐unit second‐price
auctioninwhichbackerscanthemselvesspecifythepricetheyarewillingtopayforeach
ticket,with a lower threshold being specified by Innovestment and the start‐up to be
listed.Asaconsequence,backerscanoutbideachotherwhenacquiringcashflowrights
inastart‐upcompany.
Ourkeycontributiontotheliteratureistoexploitthisuniqueauctionmechanismand
present an analysis of backers’ willingness to pay. We test whether (1) campaign
characteristics,(2)investorsophistication,(3)theprogressinthefundingcampaign,(4)
herdingbehavior,(5)stockmarketvolatility,(6)thedistancebetweenthebackerandthe
start‐up,and(7)snipingattheendofanauctionplayarolewhenbackersdecidehow
muchmoneytheyarewillingtopayforaticket.Oursampleconsistsof44campaignsthat
Innovestmentacceptedtobelistedonitswebsite.Ourresultsarebasedon1,450bids
madeby499backersduringtheperiodfromNovember6,2011,toMarch25,2014.
Ourkeyfindingsarethatcampaigncharacteristics,investorsophistication,progressin
thefundingcampaign,herding,andstockmarketvolatilityinfluencebackers’willingness
to pay in an economically meaningful manner. We find no evidence that geographic
distance, learning effects, or snipingbehavior at the endof the auction influences the
pricingofcashflowrightsinastart‐upcompany.Theresultssuggestthatself‐imposed
portaldesignsandtheorganizationofequitycrowdfundingcampaignscanexertastrong
impact on backers’willingness to pay for cash flow rights and company sharesmore
generally.
Theremainderofthispaperproceedsasfollows.Section2providessomebackground
onequitycrowdfundingingeneralandadetailedexplanationoftheauctionmechanism
oftheequitycrowdfundingportalInnovestment.Section3introducesthedatasetand
derives the paper’s hypotheses. Section 4 presents the empirical results. Section 5
concludesandprovidespolicyimplications.
5
2. Theoreticalandinstitutionalbackground
2.1Definingequitycrowdfunding
Crowdfundingcombinestheideaofmicro‐financewithcrowdsourcing(Mollick2013).In
the United States, crowdfunding campaigns are run under either the donation or the
reward model. Under the former, backers donate money to support a philanthropic
project without expecting any compensation. Under the latter, backers are promised
tangibleorintangibleperks,suchasasupportercoffeemugorbeingmentionedonthe
campaign website. For some of the most popular projects, rewards resemble a pre‐
purchaseof theproduct or service tobedevelopedby the founder. In the caseof the
Pebblesmartwatch,forexample,68,929backersspentmorethan10millionUSDintotal
toobtainawatchthatconnectswiththesmartphone.Thefirst200backerspre‐purchased
ablackwatchfor99USD.Another40,799backersthenpre‐paid115USDforthevery
samewatch.Theremainingbackerspre‐paidaslightlyhigheramounttoobtainafancier
versionofthewatch.
The crowdfunding businessmodel is different from crowdlending, inwhich backers
invest in consumer or business loans to receive a pre‐determined periodic interest
payment from debtors. Equity crowdfunding is a combination of crowdfunding and
crowdlending. Backers spendmoney in equity crowdfunding campaigns to support a
founder, who is working to develop a sustainable product or service, and expect a
monetaryreturnaftertheinvestmentcontractexpiresorthestart‐upcompanyisbought
byaventurecapitalist.Inthemajorityoftheequitycrowdfundingcampaigns,however,
backersdonotpre‐purchasetheproductorservicetobedeveloped.IntheUnitedStates,
equitycrowdfundingwasrestrictedforalongtimetoaccreditedinvestorsanddidnot
takeplace inanysignificantmanner.Although in2012 theUnitedStateswas the first
jurisdiction to pass a law specifically regulating equity crowdfunding activities, the
Securities and Exchange Commission implemented specific rules on Title III of the
Jumpstart Our Business Start‐ups (JOBS) Act only in May 2016. At that time, equity
crowdfundingbysolicitingthegeneralpublicbecamelegal.
Under German securities law, equity crowdfunding by non‐accredited investors has
always been possible. Since 2011, more than 30 equity crowdfunding portals began
operating. The crowd participates in the future cash flows of a firm by investing in
mezzaninefinancialinstruments.Mostfoundersdonotoffercommonsharesinaprivate
limitedliabilitycompany(LLC),asanotaryneedstobeinvolvedtoallowforthetransfer
6
ofsuchshares(Braunetal.2013).Moreover,theminimumcapitalrequirementaswellas
theoperatingcostsofapublicLLC(whichdoesnotrequiretheinvolvementofanotaryto
transfershares)oftenoverburdensthefoundersofastart‐upcompany.Commonshares
ofapublicLLCarethereforerarelyusedinequitycrowdfundingcampaigns.Asaresult,
Germanstart‐upsmostoftenuseprofit‐participatingloans,cooperativecertificates,and
silentpartnershipswhenrunninganequitycrowdfundingcampaign,whichthenreplicate
thefuturecashflowsofthefirm.
2.2Innovestment
One of the oldest German equity crowdfunding portals is Innovestment. The start‐up
Particularcompleted its firstsuccessfulcampaignthroughtheportalonDecember25,
2011,thesameyearmarket leaderandfirst‐moverSeedmatchappearedontheequity
crowdfundingmarket.Inmanyrespects,InnovestmentissimilartoSeedmatchandmany
otherequitycrowdfundingportalsinEurope(HornufandSchwienbacher2014).Before
acampaigngoesonline,Innovestmentandthefoundersmustagreeonavaluationofthe
start‐up,andevenbeforethat,thefoundersofthestart‐upmustdecidehowmuchcapital
theywanttoraise.Afterconsideringthefinancialneedsofthefirmandthevalueofthe
firmthatwasnegotiated,Innovestmentadaptsastandardizedfinancialcontract(asilent
partnership agreement) replicating an equity share in the start‐up. Becoming a silent
partner allows investors toparticipate in the future cash flowsof the firmduring the
lifespanofthecontractandagainwhenthesilentpartnershipagreementexpires.
Manystart‐upsrunningcampaignsonInnovestmentintendedtoraiseEUR100,000and
offeredEUR1,000 investment tickets tobackers. If the initialvaluationof thestart‐up
was, for example, negotiated to be EUR 1,000,000 and the firm raised EUR 100,000,
backersbuyingasingleinvestmentticketobtainedarighton0.091%ofthecashflow,
provided that the price of the investment ticket did not rise during the auction. It is
importanttonotethatbackerswhoultimatelybecomesilentpartnersofastart‐updonot
receive any of the rights attached to a common equity share, such as voting rights;
however,theyalsodonotparticipateinthelossesofthestart‐up.Furthermore,thesilent
partnership agreements Innovestment uses are senior to ordinary shares and
shareholderloansbutrankafterallordinaryliabilities.Theseusuallyexpireafterthree
to sevenyearsandcannotbe tradedona secondarymarketafter the initialallotment
takesplace.
7
While in many respects Innovestment is similar to all other European equity
crowdfunding portals, it also differs in one important respect and therefore is worth
analyzing in further detail. European equity crowdfunding portals uniformly allocate
equitysharesoroneoftheaforementionedfinancialinstrumentsthroughafixed‐price
first‐come,first‐servedallocationmechanism.Thatis,theportalstipulatesafixedprice
perinvestmentticketthatusuallyappliesforallitsinvestorsandcampaigns.Thenumber
ofticketsbeingofferedduringacampaignisthendeterminedbytheoverallfundinglimit
asdefinedbythefoundersandthefixedpriceperticket.Thelowerthepriceperticket,
themoreticketscanbesoldgiventheparticularfundinglimit.Asaresult,theportalstops
sellingsilentpartnershipagreementstothecrowdwhenthefundinglimitand,thus,the
pre‐determinednumberofticketsarereached.
Innovestment has deviated from stipulating a fixed price per investment ticket and
instead has adapted a multi‐unit second‐price auction. In theory, under a sealed‐bid
second‐priceauctionadominantstrategyforbackersistorevealtheirtruewillingnessto
pay for the cash flow rights in a start‐up company (Kagel and Levin 2001). The
Innovestment auction is particular as it involves three stages.Beforedescribing these
threestagesinmoredetail,wenotethatitisonlyattheendofapre‐determinedfunding
period(usually30days2)thatunitsareallottedtotheinvestorsandalegaltransferof
moneyaswell as silentpartnershipagreements takesplace.Before that, backersonly
committobuyingcashflowrightsaccordingtotheirbids,andfundsarefrozenonatrust
account.Moreover, the portal reveals only three types of information to backers: the
current price per ticket, the overall funding amount reached, and, thus, whether the
funding goalwas reached or not. Nevertheless, individual bids by other investors are
sealed(seeFigureA1intheAppendixfortheentrymask).
Duringthefirstphaseoftheauction,backerscanmakepledgesbyspecifyingthenumber
of tickets they want to buy and the price they are willing to pay for each ticket.
Innovestment and the start‐up determine a lower threshold for the price of a single
investment ticket,which is oftendetermined to beEUR1,000. Everyonewhopledges
moneyisallottedthedesirednumberofticketsduringthefirstphaseoftheauction,and
thelowestbidappliestoeveryone.Inprinciple,thereisnoreasonforinvestorstooutbid
thelowerthresholdatthisphase,asthereisyetnoscarcityinticketsandindicatingtheir
2 ChemlaandTinn (2016)show theoretically thata limitedcampaign length isessential toovercome
moralhazard.
8
truewillingnesstopaywouldonlydriveupthepriceperticket.However,backersmay
anticipatethattheauctionwillruninthesecondphaseandindicatetheirtruewillingness
topayforcashflowrightsfromtheoutsettoavoidthepotentialtransactionscostofbeing
outbid and bidding again later.3 Importantly, the Innovestment auction also operates
under an all‐or‐nothing funding model (Cumming et al. 2015). Under this model,
Innovestment and the start‐up determine a minimum funding goal that needs to be
reachedwithin a pre‐determined funding period. If theminimum funding goal is not
reachedwithinthistimeframe,thecapitalpledgedbythebackersisreturnedtothem.
Thesecondphaseoftheauctionbeginswhenapre‐determinednumberofinvestment
ticketsaresoldtothecrowd.Thenumberoftickets,andthusthebeginningofthesecond
stageoftheauction,isnotknowntotheInnovestmentbackersuntilthesecondstageis
finallyreached.Thenumberofinvestmentticketssoldbytheendofthefirstauctionphase
alsodeterminesthenumberthat isavailable throughoutthesecondphaseand is then
kept constant. From now on, investors can only outbid each other by posting higher
prices.Backersanticipatingthatthesecondstageoftheauctionwillbereachedshould
now rationally reveal their true willingness to pay, given that this phase of the
InnovestmentauctionisequivalenttoaVickrey(1961)auction.Importantly,thesecond
phaseoftheauctionisnotrestrictedtoinvestorsfromthefirstphase.Everyinvestorwho
isregisteredontheportalcanstilljointhebiddingprocess.Thesecondphasecontinues
until the funding limit is reached.After that, theauctionenters the thirdstage,during
whichallregistereduserscanstilloutbidinvestors.Atthispoint,however,itisnolonger
possible to increase the overall sum of funds received by the start‐up. Higher bids
consequentlyresultintheoverallnumberofinvestmentticketsbeingreduced.Because
theoverallsumoffundsstaysconstant,whilethenumberofticketsisreduced,thecash
flowrightsthestart‐upmustsellforagivenamountofcapitalaredecreased.4
Whatshouldbecleartothecrowdisthatthedifferentphasesoftheauctionmechanism
havenohard‐endingrule;thatis,silentpartnershipagreementscannotselloutastheir
availability only depends on backers’willingness to pay. Everyone can invest at each
phaseoftheauctionuntilthepre‐determinedfundingperiodends.Thus,unlikeunderthe
3Indeed,theCEOofInnovestmentmadethisargumentwhenshewasaskedwhyinvestorsoverbidthe
lowerpricethresholdduringthefirstphaseoftheauction.4ThesecondphaseoftheauctionwasabolishedfromNovember1,2012,onward.Consequently,thefirst
phasecontinueduntilthefundinglimitwasreached.Thereafter,thethirdphasestartedimmediately.Intheempiricalanalysis,wetakethischangeinportaldesignintoaccount.
9
fixed‐pricefirst‐come,first‐servedallocationmechanism,inwhichitmightberiskyfor
thecrowdtopostponeaninvestmentdecision,investorshaveanincentivetorevealtheir
truewillingness topayandmay theoretically investatany timeof the fundingperiod
underthemulti‐unitsecond‐priceauctionmechanism.
3. Empiricalmethodologyanddata
Our data set consists of 42 start‐ups that used the equity crowdfunding portal
InnovestmentfortheirfundingcampaignsduringtheperiodfromNovember6,2011,to
March25,2014.Intotal,weobserve1,627bidsfor44fundingcampaigns,5withatotal
volumeofEUR4,525,062pledged.Totalbidsbyindividualsoverthe2.5‐yearperiodvary
fromEUR500toEUR149,839.Duetodataavailabilityissuesforsomeoftheexplanatory
variables (average income according to postal code; see subsequently), our sample
contains1,450bidsmadeby499backers.
3.1Dependentvariable:premiumoverticketprice
Asthedependentvariable,wemeasurebackers’willingnesstopayforcashflowrightsby
calculatingtherelative‘premium’overtheinitialticketpriceinpercentage:
1 100
.
Table 1 shows descriptive statistics for the observed premia. Figure 1 shows the
distributionofthepremia,splitforseveralsub‐groups.Thefirstsub‐groupconsistsofall
bidsbeforethefundinggoalwasreached,thesecondsub‐groupconsistsofallbidsafter
thefundinggoalwasreachedbutbeforethefundinglimitwasreached,andthethirdsub‐
groupconsistsofallbidsafterthefundinglimitwasreached.
Overall,457investmentbids(31.5%ofallbidsinthesample)aremadewithoutany
premium.Mostofthesebidsweremadebeforethefundinggoalwasreached(370;
48.9%ofallbidsinphase1).6However,thefactthatmorethan50%ofallbidsinthat
sub‐samplearemadewithapositivepremiumconfirmsthatsomebackersavoidthe
5Twostart‐upsinoursampleranmultiplefundingcampaigns.6Thestartofthesecondorthirdstageoftheauctiondoesnotnecessarilycoincidewiththefundinggoal
orfundinglimitbeingreached.
10
transactioncostsofbiddingagainlater,eventhoughpostingapremiumcandriveupthe
secondpriceinthefirstroundoftheauction.
Table1:Descriptivestatistics:premiumoverticketprice
Mean SD Median Min. Max. Obs. 0All 18.32 25.87 8.00 0 203 1450 457Goalnotreached 10.67 20.81 1.00 0 203 757 370Goalreached 13.70 19.68 9.13 0 150 281 86Limitreached 35.53 29.67 29.80 0 203 412 1
Note:Column‘0’indicatesthefrequencyofbidswithoutanypremium.
Fig.1:Distributionofpremiaoverticketprice
Note:y‐axisshowstherelativefrequencyofpremiainthethreephasesoftheauction.
Theaveragepremiumovertheticketpriceis18.3%andisincreasingoverthethreesub‐
samples:Itwas10.7%beforethefundinggoalwasreached,13.7%afterthefundinggoal
wasreached,and35.3%after the funding limitwasreached,andthesedifferencesare
statisticallysignificant.7 Inaddition, thestandarddeviationdiffersconsiderablyacross
sub‐groups.Itis1.5timesaslargeinthethirdsub‐groupasinthefirstandsecondsub‐
groups,andthisdifferenceisalsostatisticallysignificant.8Thisisalsoreflectedintheright
panelofFigure1whichismoreuniformlydistributedoverthedifferentlevelsofpremia
thantheleftandmiddlepanels.
7Theresultsoft‐testsfordifferencesinmeansacrosssub‐groupsareasfollows:goalnotreachedvs.goal
reached:t=–2.17,p‐value=0.03;goalnotreachedvs.limitreached:t=–15.11,p‐value=0.00;goalreachedvs.limitreached:–11.65,p‐value=0.00.8Theresultsofvariance‐comparison testsacross sub‐groupsareas follows:goalnot reachedvs. goal
reached:f=1.12,p‐value=0.27;goalnotreachedvs.limitreached:f=0.49,p‐value=0.00;goalreachedvs.limitreached:f=0.44,p‐value=0.00.
11
Ofthe499backersinoursample,255(51.1%)madeasinglepledgeduringthewhole
sampleperiod,another107(21.4%)pledgedtwice,andonly24(4.8%)made11pledges
ormore.In527(36.3%)ofthe1,450totalbids,backersmadearepeatedpledgeinone
campaign. The maximum number of bids by one backer in one campaign is 11. The
averagestartingbidoffirst‐timebiddersbeforethefundinggoalwasreachedis10.8%(n
=310),whichisalmostthesameastheoverallaveragebidduringthatphase(10.7%,n=
757,seealsoTable1),implyingthattherearenodifferencesbetweenfirst‐timeandmore
experiencedbidders.
ToaccountfortheabolitionofthesecondphaseoftheauctiononNovember1,2012,
andtoinvestigatewhetherinvestorsadapttheirwillingnesstopayovertime,weplotthe
premiaagainstthenumberofpledgesaninvestorhadalreadymadeatthetimeofthe
focalpledgeforthetwosub‐samplesfromNovember6,2011,toNovember1,2012,and
fromNovember2,2012,toMarch25,2014.Figure2showsthatduringthesecondsub‐
sample, investors’ mean unconditional willingness to pay for an investment ticket
remainedlargelyconstantwhenbidding.Inthefirstsub‐sample,however,weobservea
hump‐shaped patternwhen the premia are plotted against the number of pledges. In
general,wefindthatinthefirstperiod,investors’willingnesstopayismuchlarger,with
ameanpremiumpaidof22.7%(n=710),thanthatinthesecondperiod,withamean
premiumpaid of only 14.1% (n= 740), and this difference is statistically significant.9
Whiletheobserveddifferencecouldbeattributedtotheabolitionofthesecondstageof
the auction after November 1, 2012, it might also be due to changes in the market
environment, inwhichovertimemoreportalsprovidedfundingopportunitiesandthe
additionalcompetitiondrovedowninvestors’willingnesstopay.
In theempiricalanalysis that follows,weruna regressionon the full sampleof1,450
observationsandfurtherfocusontheperiodfromNovember2,2012,toMarch25,2014,
toaccountforapotentialstructuralbreakduetothechangeintheplatformdesign.In
addition,wetruncateboththefullsampleandthesecondsub‐samplebyleavingout164
and 50 observations, respectively, where we observe a premium larger than 50% to
exploretherobustnessofourresults.10
9Theresultsofat‐testfordifferencesinmeansacrosssub‐groupsisasfollows:premiumNovember6,
2011,toNovember1,2012vs.November1,2012,toMarch25,2014:t=6.39,p‐value=0.00.10Thisthresholdcorrespondstoroughlytwostandarddeviationsintheobservedpremia.Anotherreason
forleavingoutrelativelylargepremiaistoavoidtypingerrorsbytheinvestors.Forexample,in25casesweobserveapremiumof100%,anditmightbethecasethatinvestorswantedtobuytwoticketswithoutanypremiuminsteadofoneticketwithapremiumof100%.
12
Fig.2:Averagepremiumandnumberofpledges(byinvestor)
Note:Theleft(right)panelshowstheaveragepremiumfordifferentnumbersofpledgesatthetimeofthepledge(byinvestor),alongwith95%confidencebandsduringthefirst(second)sub‐sample.Thedashedlinesrepresentthemeansfromtherespectiveothersub‐sample.
3.2Explanatoryvariables11andhypotheses
3.2.1Campaigncharacteristics
Ourfirstsetofexplanatoryvariablesreflectscampaigncharacteristicsthatareobservable
to all backers on the portal website. For each start‐up, Innovestment reports an
assessmentofthefirm’svalue,whichvariesfromEUR420,000toEUR10,000,000inour
sampleof44 fundingcampaigns. Inaddition,each firmmustannouncea fundinggoal,
which varies from EUR 36,000 to EUR 150,000. We conjecture that the backers can
interpret both the firm value and the funding goal as effective signals in the spirit of
Spence (1973) forpotentially lucrative investments.This isbecause thevaluationand
fundinggoalarebotheasilyobservable,andifchosensuchthattheyaretoohigh,theyare
costlyforthefounderbecausethecampaignmightreceivenotenoughornofundingat
all.Forthefundinggoal,ahigherfundinggoalsignalstothecrowdthattheentrepreneur
isconfidentthatheorshewillatleastcollectthepre‐determinedamountofmoney.Ifthe
thresholdisnotmet,themoneypledgedisgivenbacktothefundersandthecampaign
fails. However, in case of the pre‐valuation there is also a channel that works in the
oppositedirection.Ahigherpre‐valuationimplies,forasingleinvestmentticket,alower
11TableA1intheAppendixreportsdescriptivestatisticsfortheexplanatoryvariables.
13
shareoffuturecashflowsand,consequently,makessuchaninvestmentlessattractive.
Accordingly,ourfirsthypothesisisasfollows:
H1:Theeffectofthefirm’spre‐valuationonthepremiumisambiguous.Thepremiumis
increasinginthefundinggoal.
3.2.2Backersophistication
We conjecture that more sophisticated backers understand the underlying auction
mechanismbetterthantheirlesssophisticatedpeers.Asmentionedpreviously,weexpect
no extensive investment premia in the first phase of the auction, though backers
anticipatingthesecondstageoftheauctionmightrationallyposttheirreservationprice,
whichmayliewellabovetheminimumticketprice.Inaddition,weexpectsophisticated
backersto indicatetheirtruewillingnesstopayforcashflowrights inthesecondand
third stages. The differences across different types of backers might even be more
relevantundertransactioncosts,asmoresophisticatedbackerstypicallyfacerelatively
low costs when investing because they are more specialized in evaluating start‐up
companies.Aswecannotmakeanyconjecturesabouthowthewillingnesstopaydiffers
amongsophisticatedandunsophisticatedinvestors,wedonotspecifyafirmpriorabout
conditionaldifferencesinthepremiumacrossthesesub‐groups.
Weincludeasecondsetofexplanatoryvariablesthatproxybackersophistication.First,
more sophisticated investors typically undertake relatively large investments.
Consequently,weusethenumberofticketsasingleinvestorbidsforinasinglepledge,
whichvariesbetween1and40,asanexplanatoryvariable.Similarly,ahigherminimum
priceperticketasdefinedbyInnovestmentcanserveasanentrancebarrier forsmall
investors.Thus,itismorelikelythatmoresophisticatedinvestorsundertakebidsifthe
minimumticketprice,whichvariesbetweenEUR500andEUR25,000,isrelativelyhigh.
Moreover, backers might better understand how the auction mechanism works after
pledginginmultiplecampaignsandbecomemoresophisticatedbyinvestingmoreoften
on the portal. To capture possible learning effects, we consider how often a backer
pledged on the portal before the current investment (see also Figure 2). Next,
Innovestmentrequireseverybackertocompleteashortquestionnaireabouthisorher
investmentexperienceinthefollowingsevencategorieswhenregisteringwiththeportal:
bonds,commodities,fundsandcertificates,realestate,stocks,termdeposits,andother
14
equity.Backerswhoclaimtohaveexperienceinatleastoneofthesecategoriesconducted
52.3%ofthebids.Intheempiricalanalysis,weincludeasetofdummyvariablesforall
sevencategories,whichtakethevalueof1ifabackerhasexperienceinthatparticular
categoryand0otherwise.Finally,Innovestmentrecordsthepostalcodeofeachbacker.
Thus,weareable to include theaverage income in thebacker’shomeregion in2011,
which varies between EUR 16,239 and EUR 28,900 in our sample, as a proxy for the
backer’sincomeandsophistication.12Therefore,oursecondhypothesisisasfollows:
H2:Thepremiumwilldifferdependingonthenumberofticketsbought, thepriceper
ticket, the number of pledges the backer previously made, the backer’s general
investmentexperience,andtheaverageincomeinthebacker’shomeregion.
3.2.3Progressinthefundingcampaign
Athirdhypothesistakesintoaccounttheprogressinthefundingcampaign.Backersare
wellawareoftheoverallpercentageoftargetedfundingaccomplishedatthetimeoftheir
decision.BecausetheauctionmechanismofInnovestmentallowsforbidsevenafterthe
fundinggoalorlimithasbeenreached,13theaccomplishedfundingshareatthetimeofa
bidvariesbetween0%and100%.Consequently,weincludeanotherexplanatoryvariable
thatmeasuresthefundingshareinpercentage.Inaddition,backersknowwhetherornot
thefundinggoalorthefundinglimithasbeenreached.Thus,wealsoconsidertwonon‐
disjunctivedummyvariables,whichmeasure(1)whetherthefundinggoalwasreached
butthefundinglimithasnotyetbeenreachedand(2)whetherthefundinglimithasbeen
reached.Asreachingthefundinggoalremovestheuncertainty inwhetherthefunding
actuallytakesplace,backerswithstrongliquiditypreferencesnolongerneedtofearthat
they are simply putting theirmoney on hold because the campaign in the end fails.14
Furthermore,reachingthefundinggoalandfundinglimitmightbeasignalofdemandfor
theparticularinvestmentopportunityandthepotentialqualityofthestart‐up.Thus,we
expectedastrongpositiveinfluenceofthesetwodummyvariablesonthepremiumand,
inparticular,forthefundinglimit.Accordingly,ourthirdhypothesisisasfollows:15
12Wecannotretrievethisinformationforsomeoftheforeigninvestorsand,therefore,loseapartofthe
1,627observationsowingtotheinclusionofthisvariable.1347.8%(28.4%)ofallbidswererecordedwhenthefundinggoal(limit)wasreached.14However,backersstillcanbeoutbidatthisstage.15Wedonotdifferentiatebetween thedifferentstagesof theauction in theempiricalmodel,because
includingadummyvariableforthethirdstageoftheauctionalongsideinteractiontermsofthisdummy
15
H3: The premium is increasing in the share of targeted funding, which has been
accomplishedandishigherifthefundinggoalorfundinglimithasbeenreached.
3.2.4Herding
Herdingisawell‐documentedphenomenoninfinancialmarkets(ScharfsteinandStein
1990),andithasalsobeenobservedincrowdlending(Herzensteinetal.2011;Leeand
Lee2012)andequitycrowdfunding(HornufandSchwienbacher2015;Vismara2015).
To test whether herding affects the pricing of cash flow rights on Innovestment, we
includethesumofinvestmentbidsinastart‐upthatweremadeearlieronthesameday,
as additional explanatory variable. The variation in this variable is surprising, as it is
betweenEUR0andEUR217,000.Thus,ournexthypothesisaimstotestwhetherherding
behaviorinequitycrowdfundingaffectsthepremiumofferedbybackers:
H4: Thepremium is increasing in the sumof bidsmade earlier on the sameday in a
particularstart‐up.
3.2.5Stockmarketvolatility
Oursampleperiodconsistsofepisodesoffinancialmarketstress,inparticularduringthe
euroandsovereigndebtcrisis.Consequently,stockmarketvolatilityasmeasuredbythe
German VDAX varies considerably over this period (between 11.47and 37.28%).
Moreover, portfolio diversification of equity investors largely increased during the
financial crisis as investors had a higher demand for similar but uncorrelated assets
(Vermeulen2013).Thus,ifbackersconsiderstocksandcrowdinvestmentssubstitutes,
higherstockmarketvolatilitymightleadtohigherdemandforthisassetclassandalarger
premiabeingpaidforcrowdinvestments.16Thus,ournexthypothesisaimstodetectsuch
asubstitutioneffect:
H5:Thepremiumisincreasinginstockmarketvolatility.
withthefundingshare,thefundinggoal,andthefundinglimitneithergeneratessignificantestimatesnorchangestheresultsoftheotherexplanatoryvariables.16Dornetal.(2015)documentthatinvestorsconsiderinvestmentandgamblingproductssubstitutes.
16
3.2.6Distancebacker/start‐up
Weusethedistancebetweenthebackerandthestart‐upasanadditionalexplanatory
variable. This variable takes values between 0 and 644 km in our sample. A greater
distance to a specific investment might imply higher search costs to obtain accurate
information about a start‐up and, as a consequence, a lowerwillingness to pay and a
reducedpremium.Inaddition,weobservealocalbiasinfinancialmarkets(Baltzeretal.
2015;CummingandDai2010).HornufandSchmitt(2016)provideevidencethatbackers
onInnovestmentalsoexhibitalocalbias.Ifbackershaveahigherdemandformorelocal
start‐upsthanfordistantstart‐ups,ahigherpremiumforgeographicallyclosefirmscould
result. Both the aforementioned channels indicate a negative relationship between
distanceandpremium,whichleadstooursixthhypothesis:
H6:Thepremiumisdecreasinginthedistancebetweenthebackerandthestart‐up.
3.2.7Sniping
Awell‐knownphenomenon inauctions is sniping—that is, theauctionprice increases
drasticallytowardtheendoftheauctionprocess(Arielyetal.2005;RothandOckenfels
2002).AsInnovestmentpoststhecurrentsecondpricethatappliestoeveryone(seealso
FigureA1intheAppendix),backersmightbidlatetoavoidrevealinginformationabout
theirwillingnesstopaytootherbackers,whichcouldultimatelydriveupthepriceper
ticket. Indeed, roughly25%of thebidsaremadeon the lastdayof theauction,which
providessomedescriptiveevidenceinfavorofsniping.Totestwhethersnipingisalso
relevantinamultivariateanalysis,weincludetheremainingtimemeasuredindaysasan
additionalexplanatoryvariable.Totestforpotentialnon‐linearitiesandtocapturethe
often‐documentedmassive increase toward the end of the auction, we also include a
quadratictermthatmeasuressquaredremainingtimeindays.Ifsnipingisprevalent,we
wouldobserveanegativesign;thatis,thepremiumislowerthemoretimeisremaining
intheauctionprocess.Thus,ourlasthypothesisisasfollows:
H7:Thepremiumisdecreasingintheremainingtime.
Table 2 provides an overview of all seven hypotheses and explanatory variables
employedintheempiricalanalysis.
17
Table2:Summaryofhypotheses
H1:Campaigncharacteristics H2:Backersophistication(+/–)Pre‐valuation(+/–) Numberoftickets(+/–)Fundinggoal(+) Priceperticket(+/–) Investmentexperience(+/–)H3:Progressinthefundingcampaign Averageincome/region(+/–)Fundingshare(+) Numberofpledges(+/–)Fundinggoalreached(+) Fundinglimitreached(+) H4:herding(+) H5:Stockmarketvolatility(+) H6:Distancebacker/start‐up(–) H7:Remainingtime(–)
3.3Econometricmodel
We explain the relative premium over the ticket price with all explanatory variables
describedintheprevioussub‐section.Econometrically,weuseordinaryleastsquaresand
standarderrorsclusteredatthebackerlevel.17InSection4,wepresentfourdifferentsets
of results. First, we show estimates that are based on the full sample of all 1,450
observations.Second,weestimatethesamemodelfortheperiodfromNovember2,2012,
toMarch25,2014,usingonly740observationstoaccountforapotentialstructuralbreak
duetothechangeintheplatformdesign.Inaddition,wefurthertruncatethesampleand
thesecondsub‐sample,respectively,andexplainthe1,286and690investmentsinwhich
the premium is lower than or equal to 50%of the ticket price. In all regressions,we
controlforday‐of‐the‐weekeffects,withMondayasthereferencecategory.
4. Empiricalresults
Table3reports theresults for the full sampleperiodandallbids(column(1)). Italso
presents theresults for the full sampleperiodandbidswithapremiumofup to50%
(column(2)),forthesecondsub‐samplefromNovember2,2012,toMarch25,2014,and
allbids(column(3)),andforthesecondsub‐sampleandbidswithapremiumofupto
50%(column(4)).
17 Our empiricalmodel does not contain campaign‐fixed effects. Otherwise, wewould not be able to
identify the effect of campaign characteristics (H1) and the price per ticket (H2) on the premium. Inaddition,ourmodeldoesnotcontainbacker‐fixedeffects.Otherwise,wewouldnotbeabletoidentifytheeffectofexperience(H2),theaverage incomeinthe investor’sregion(H3),andthegeographicdistance(H6)onthepremium.Finally,ourmodeldoesnotcontaintime‐fixedeffects.Otherwise,itwouldbedifficulttoidentifytheeffectofstockmarketvolatility(H5)onthepremium.
18
Table3:Explainingequitycrowdfundingpremia
FullSample Nov2,2012–Mar25,2014 AllBids Prem.≤50% AllBids Prem.≤50% (1) (2) (3) (4) Pre‐valuation 0.001 0.001 0.005 * 0.003 * (0.001) (0.001) (0.002) (0.001) Fundinggoal 0.270 ** 0.067 ** –0.049 –0.042 (0.053) (0.024) (0.076) (0.045) Numberoftickets 0.309 0.146 0.226 0.193 * (0.197) (0.089) (0.201) (0.097) Initialprice/ticket –0.928 –0.971 ** –1.613 –1.145 * (0.633) (0.319) (0.970) (0.465) Bonds –1.831 0.542 –2.593 –1.345 (2.941) (1.351) (3.469) (1.509) Commodities –3.320 –0.634 –0.312 0.478 (2.705) (1.397) (3.373) (1.639) Funds/certificates 3.974 –0.004 5.702 –1.216 (3.308) (1.805) (4.281) (2.085) Realestate –5.317 –2.842 * –9.809 * –5.375 ** (2.903) (1.279) (3.789) (1.536) Stocks 3.787 3.145 9.367 * 5.729 ** (2.779) (1.777) (4.088) (2.106) Termdeposits –2.039 –1.777 –3.736 –0.326 (3.165) (1.515) (4.670) (1.528) Otherequity 2.936 0.877 –0.321 0.364 (2.290) (1.094) (2.637) (1.242) Disposableincome 0.112 0.119 0.584 0.178 (0.292) (0.146) (0.423) (0.187) Numberofpreviouspledges 0.059 0.168 ** –0.190 0.132 (0.099) (0.037) (0.168) (0.071) Fundingshare 0.030 0.025 0.041 0.034 (0.033) (0.017) (0.043) (0.021) Fundinggoalreached 3.729 3.513 ** –3.443 0.022 (2.334) (1.168) (3.240) (1.452) Fundinglimitreached 17.748 ** 6.640 ** 9.021 * 3.938 ** (2.369) (1.109) (3.707) (1.410) Bidsearlierthatday 0.079 ** 0.111 ** 0.128 ** 0.125 ** (0.020) (0.009) (0.022) (0.009) VDAX 0.789 ** 0.521 ** –1.130 * 0.088 (0.152) (0.063) (0.502) (0.243) Distancebacker/start‐up 0.225 –0.205 –0.308 –0.278 (0.381) (0.182) (0.502) (0.222)
19
Table3:Explainingequitycrowdfundingpremia(continued)
Daysremaining –0.013 0.510 ** –0.207 0.422 ** (0.261) (0.098) (0.236) (0.126) Daysremaining2 0.004 –0.012 ** 0.014 * –0.011 ** (0.007) (0.003) (0.006) (0.003) Tuesday –0.987 1.126 2.348 1.918 (2.331) (1.305) (2.918) (1.582) Wednesday –1.354 0.367 –0.270 –1.088 (2.392) (1.338) (2.175) (1.331) Thursday 0.258 –0.677 4.734 –0.303 (2.917) (1.254) (3.476) (1.454) Friday 2.588 0.158 8.900 * 1.017 (2.728) (1.340) (4.073) (1.506) Saturday 0.034 –1.083 5.493 0.012 (2.906) (1.410) (3.022) (1.523) Sunday 1.375 1.402 8.018 * 2.881 (2.722) (1.278) (3.367) (1.720) Constant –29.614 ** –15.751 ** 2.997 –4.675 (6.989) (3.694) (11.729) (6.181) AdjustedR2 0.254 0.357 0.220 0.480 ExclusiontestDotW 0.74 1.23 2.18 * 1.61 Observations 1450 1286 740 690 Note:Dependentvariable:Premium(inpercentage).Standarderrors(clusteredatthebackerlevel)areinparentheses.**/*indicatessignificanceatthe1%/5%level.
4.1Campaigncharacteristics
Intheregressionsforthefullsampleperiod,wefindthatthepremiumisincreasinginthe
sizeofthefundinggoal,whichconfirmsH1.Backersarewillingtoofferapremiumof27.0
basispoints(bps)(column(1))and6.7bps(column(2))foreachEUR1,000increasein
the funding goal, which evidences that the funding goal indeed serves as a signal to
potential investors. To put this figure into perspective, we compare two groups of
campaignsandusethemoreconservativeestimateincolumn(2).Inourdataset,wehave
15campaignswithafundinggoalofEUR50,000andanother14campaignswithafunding
goal of EUR 70,000. This difference of EUR 20,000 corresponds to a ceteris paribus
differenceof1.34percentagepoints (pp) in thepremium.These findingsarearguably
driven by the first sub‐sample, as the coefficients on the funding goal are no longer
significantwhenweconsideronlytheperiodafterNovember2,2012.Incontrast,firms’
pre‐valuationinfluencesthesizeofthepremiumonlyinthesecondsub‐sample.Foreach
EUR1,000increaseinpre‐valuation,thepremiumincreasesby0.5bps(column(3))and
20
0.3bps(column(4)),respectively,whichprovesthatahigherpre‐valuationalsoserves
as a signal for a potentially lucrative investment. The ceteris paribus difference for
campaignswithpre‐valuationsofEUR800,000andEUR1,000,000is0.6pp(basedonthe
estimatesincolumn(4)).18
4.2Backersophistication
The estimates for our proxies of backer sophistication yield conflicting results. The
premiumincreasesinthenumberofticketsboughtby19.3bps(column(4))andinthe
numberofpriorpledgesbyaninvestorby16.8bps(column(2)).Inaddition,investors
withexperienceinthestockmarketarewillingtopayhigherpremiaduringthesecond
sub‐sample(9.8ppincolumn(3)and5.7ppincolumn(4)).Incontrast,eachEUR1,000
increase in theminimumpriceper ticket leads toa0.97pp (column(2))and1.15pp
(column (4)) decrease in the premium. This implies that the total difference between
campaignswithEUR500tickets(6campaigns)andEUR1,000tickets(29campaigns)is
48.6bps(basedontheestimatesincolumn(2)).Furthermore,backerswithexperience
inrealestateinvestmentsofferasignificantlylowerpremiumthanbackerswithoutany
experienceinthatcategory.Onepotentialdriverofthisdifferencebetween–2.84ppand
–9.81ppcouldbetheexperienceinassessingafinancingplan.Finally,theaverageincome
in the backer’s region is insignificant in all estimations. In summary, similar to the
descriptiveanalysisintheprevioussection,wefindnoconclusivedifferencesbetween
sophisticatedandunsophisticatedinvestors.
4.3Progressinthefundingcampaign
ConfirmingH3,theprogressinthefundingcampaignpositivelyinfluencesthepremium
backers offer. The premium of bids made after the funding goal was reached is, on
average, 3.51 pp larger than bidsmade before the goal was reached, but only in the
completesampleperiodandforbidswithapremiumupto50%(column(2)).Reaching
the funding limit is significant in all four specifications as we observe an additional
increaseinthepremiawhenthefundinglimitwasreached.Thisincreasevariesbetween
18Notethatthedifferences intermsof(non‐)significanceofbothvariablesacrossthedifferentsetsof
resultsmightbeduetocollinearity,asthebivariatecorrelationbetweenpre‐valuationandthefundinggoalis=0.76.
21
3.98ppand17.75pp.Finally,theaccomplishedfundingshareitselfdoesnotsignificantly
influencethepremium.
4.4Herding
Forthesumofinvestmentbidsinastart‐up,whichweremadeearlieronthesameday,
weagainobservepositiveandsignificantcoefficients inall fourmodels.Thepremium
increasesbetween7.9bpsand12.8bpsforeachEUR1,000,whichisaclearindicationof
herdingbehaviorandconfirmsH4.Multiplyingthepointestimateof11.1bps(column
(2))bythestandarddeviationofthisvariable(EUR38.023)indicatesthatthevariation
inthepremiacausedbyherdingbehaviorisalsoeconomicallyrelevant(4.22pp).
4.5Stockmarketvolatility
Forthefullsampleperiod,backerstendtobidhigherpremiaduringepisodesoffinancial
marketstressandconsiderstocksandcrowdinvestmentssubstitutes.Aone‐unitincrease
intheVDAXleadstoa78.9bps(column(1))and52.1bps(column(2))largerpremium.
Toputthelatterpointestimateintoperspective,weconsidertheeffectofaonestandard
deviation change in theVDAX (6.752%).Thisback‐of‐the‐envelope calculation reveals
thattheVDAXaccountsforavariationof3.52ppinthepremium.Similartothefindings
for the fundinggoal, theresults forstockmarketvolatilityaredrivenby the first sub‐
sample,asweevenobserveanegativeandsignificantcoefficientforthisvariableinthe
secondsub‐sample(column(3)).
4.6Distancebetweenbackerandstart‐up
Thedistancebetweenabackerandastart‐upis insignificant inall foursetsofresults.
Consequently, we find no evidence of distance influencing the premium paid, which
rejectsH6.
4.7Sniping
For thecomplete sampleperiodandallobservations, thepointestimates for thedays
remainingand(daysremaining)2are individuallyand jointly insignificant(F(2,498)=
1.43).Whenconsideringonlypremiaupto50%,wefindahump‐shapedinfluenceofthe
remainingtimeonthepremiuminthecompletesampleperiodandinthesecondsub‐
samplestartingonNovember2,2012.From0to21(19)daysremaininginthefullsample
22
period(secondsub‐sample),thepremiumisincreasing;thereafter,itisdecreasing.19We
interpretthisasbackerspostingtheirreservationpriceatsomepointduringtheauction,
whichiswellinlinewiththedominantstrategyinVickreyauctions.Anotherexplanation
forthefindingisthatInnovestmentextendsthefundingperiodconsecutivelyforanother
15minutes if additionalbidsaremade toward theendof the campaign,whichmakes
sniping literally impossible. Consequently, we find no evidence for sniping behavior
towardtheendoftheauction,whichrejectsH7.
Finally,wecanexcludeday‐of‐the‐weekeffectsinthreeofthefourmodels(columns(1),
(2),and(4)).Forthesecondsub‐sampleandallbids(column(3)),bidsonFridaysand
Sundaysare8.90ppand8.02pphigher,respectively,thanthereferenceday(Monday).
5.Conclusion
In thispaper,weanalyze thepricingofcash flowrights instart‐upcompaniesusinga
uniquedatasetofequitycrowdfundingbackers.Oursampleconsistsof44campaignsand
includes1,450bidsmadeby499backersduringtheperiodfromNovember6,2011,to
March25,2014,on theGermanequitycrowdfundingportal Innovestment. Incontrast
with all other European equity crowdfunding portals, Innovestment runs amulti‐unit
second‐priceauctioninwhichbackerscanspecifythepricetheyarewillingtopayforan
investmentticket,withtheportalandstart‐upspecifyingalowerthreshold.Weexploit
thisuniqueauctionmechanismtoanalyzebackers’willingnesstopayforcashflowrights
inastart‐upcompany.
First,campaigncharacteristicsplayameaningfulroleinthedeterminationofbackers’
willingnesstopay.Boththefundinggoalandpre‐valuationserveassignalsforpotentially
lucrative investments, as an increase in these variables is associated with a higher
premium.Second, theestimates forbackersophisticationyieldconflicting results.The
premiumincreasesinthenumberofticketsboughtandinthenumberofpriorpledgesby
aninvestorbutdecreases intheminimumpriceperticket.Backerswithexperiencein
realestateinvestments(thestockmarket)offer,onaverage,alower(higher)premium
thantheircounterpartswithoutanyexperienceintherespectiveclassofassets.Third,
marketforcesarealsoofparticularrelevance,asreachingthefundinggoal leadstoan
increaseinthepremium(comparedwithinvestmentsinwhichthefundinggoalhasnot
19NotethattheU‐shapedpatternincolumn(3)isnotsignificantfor0to28daysremaining.
23
beenreached),asdoesreachingthefundinglimitwithanadditionalsignificantincrease.
Fourth,backersrespondtothesumofinvestmentbidsinastart‐up,whichweremade
earlieronthesameday,byincreasingthepremiumintheirbids.Fifth,backerstendtobid
higherpremiaduringepisodesoffinancialmarketstressandconsiderstocksandcrowd
investmentssubstitutes.Wefindthiseffect,however,onlyforthefirstsub‐samplefrom
November6,2011,toNovember1,2012.
Incontrast,wefindthatgeographicdistance,learningeffects,andsnipingdonotaffect
thepremiumpaid.Ifbackersweretoindicatetheirtruewillingnesstopayonlyattheend
oftheauction,asisregularlythecaseoneBay(RothandOckenfels2002),theremightbe
a risk that some bids are not successfully transmitted and investors with a higher
willingnesstopayarelockedout.Ourresultsdonotindicatethatthisisthecaseinequity
crowdfundingthattakesplaceunderamulti‐unitsecond‐priceauction.Conversely,under
afirst‐come,first‐servedmechanismwithahard‐endingrule, investmentticketsmight
quicklysellout(HornufandSchwienbacher2016)andinvestorswithahigherwillingness
topaycouldbeinefficientlydebarred.Whetherequitycrowdfundingportalsshouldadopt
anauctionmechanism,however, alsodependson the returns that inventorsearnand
whethertheirbidsexceedthevalueoftheauctionedasset.Apromisingavenueforfuture
researchwouldbetotesttheauctionmechanismafterdataoninsolvenciesandactual
payoutsofthefundedfirmsbecomeavailable.
Ourresultscontributetotheliteratureonportaldesignandcampaigncharacteristicsin
equitycrowdfunding.Theysuggestthatportaldesignandthespecificfeaturesofhowan
equitycrowdfundingcampaignisrunsignificantlyinfluencebackers’willingnesstopay
forfuturecashflowrightsinastart‐up.Thisisalsoinlinewithpriorstudiesthatfindthat
campaigncharacteristicssuchastheamountofequityofferedandfinancialprojections
matter for funding success (Ahlers et al. 2015). Furthermore, in linewithHornuf and
Schwienbacher (2015) and Vismara (2015), who find that information cascades
determine funding success in equity crowdfunding, our results provide evidence that
herdingalsoaffectsbackers’willingnesstopayforsharesinastart‐up.Asmoredataon
theultimatesuccessandfailureofstart‐upsbecomeavailable,itwouldbepromisingto
investigate whether herding in the context of equity crowdfunding is rational or
irrational.Ontheonehand,someinvestorsmightrationallyrelyonthebehaviorofothers
becauseinformationcostsarehigh.Ontheotherhand,thecrowdcouldalsomakefaulty
decisions by engaging in what has been termed ‘groupthink’ (Janis 1972). If it were
24
necessary to guard investors from herding, equity crowdfunding portals that run an
auctionmechanismcould implement someof the rules that are common to electronic
tradingsystemsonregularstockmarkets.
25
References
Agrawal,A.,Catalini,C.,&Goldfarb,A.(2013).Crowdfunding:socialfrictionsintheflatworld.NBERWorkingPaper16820.http://www.nber.org/papers/w16820.
Ahlers, G., Cumming, D., Guenther, C., & Schweizer, D. (2015). Equity crowdfunding.Entrepreneurship Theory and Practice, 39(4), 955–80.http://dx.doi.org/10.1111/etap.12157
Ariely,D.,Ockenfels,A.,&Roth,A.E.(2005).AnexperimentalanalysisofendingrulesinInternet auctions. RAND Journal of Economics, 36(4), 891–908.http://dx.doi.org/10.2139/ssrn.429963
Baltzer,M.,Stolper,O.,&Walter,A.(2015).Home‐fieldadvantageoramatterofambiguityaversion?LocalbiasamongGermanindividualinvestors.EuropeanJournalofFinance,21(6),734‐754.http://dx.doi.org/10.1080/1351847X.2013.877514
Belleflamme,P.,Lambert,T.,&Schwienbacher,A.(2014).Crowdfunding:tappingtherightcrowd. Journal of Business Venturing, 29(5), 585–609.http://dx.doi.org/10.1016/j.jbusvent.2013.07.003
Block,J.,Hornuf,L.,&Moritz,A.(2016).Whichupdatesduringanequitycrowdfundingcampaignincreasecrowdparticipation?”TrierResearchPapersinEconomics,No.7/15.http://ssrn.com/abstract=2612998.
Bøg,M.,Harmgart,H.,Huck,S.,&Jeffers,A.M.(2012).FundraisingontheInternet.Kyklos,65(1),18–30.http://dx.doi.org/10.1111/j.1467‐6435.2011.00525.x
Braun,R.,Eidenmüller,H.,Engert,A.,&Hornuf,L.(2013).Doeschartercompetitionfosterentrepreneurship? A difference‐in‐difference approach to European company lawreforms. Journal of Common Market Studies, 51(3), 399–415.http://dx.doi.org/10.1111/jcms.12003
Burtch,G.,Ghose,A.,&Wattal,S.(2013).Anempiricalexaminationoftheantecedentsandconsequencesofinvestmentpatternsincrowd‐fundedmarkets.InformationSystemsResearch,24(3),499–519.http://dx.doi.org/10.2139/ssrn.1928168
Burtch, G., Ghose, A., & Wattal, S. (2014). Cultural differences and geography asdeterminants of online pro‐social lending. MIS Quarterly, 38(3), 773–794.http://dx.doi.org/10.2139/ssrn.2271298
Chemla, G., & Tinn, K. (2016). Learning through crowdfunding. SSRNWorking Paper.http://ssrn.com/abstract=2796435.
Colombo,M.G.,Franzoni,C.,&Rossi‐Lamastra,C.(2015).Internalsocialcapitalandtheattractionof early contributions in crowdfundingprojects.EntrepreneurshipTheoryandPractice,39(1),75–100.http://dx.doi.org/10.1111/etap.12118
Cumming, D., & Dai, N. (2010). Local bias in venture capital investments. Journal ofEmpiricalFinance,17,362–380.http://dx.doi.org/10.1016/j.jempfin.2009.11.001
26
Cumming,D.,Leboeuf,G.,&Schwienbacher,A.(2015).Crowdfundingmodels:keep‐it‐allvs. all‐or‐nothing. Working Paper, http://leeds‐faculty.colorado.edu/bhagat/CrowdfundingModels‐KeppItAll‐AllorNothing.pdf.
Dorn,A.,Dorn,D.,&Sengmueller,P.(2015).Tradingasgambling.ManagementScience,61(10),2376–2393.http://dx.doi.org/10.1287/mnsc.2014.1979
Herzenstein,M.,Dholakia,U.,&Andrews,R.(2011).Strategicherdingbehaviorsinpeer‐to‐peer loan auctions. Journal of Interactive Marketing, 25(1), 27–36.http://dx.doi.org/10.1016/j.intmar.2010.07.001
Hornuf, L., & Schmitt,M. (2016). Does a local bias exist in equity crowdfunding? Theimpact of investor types and portal design.Max Planck Institute for Innovation &CompetitionResearchPaper,No.16‐07.http://ssrn.com/abstract=2801170.
Hornuf, L., & Schwienbacher, A. (2014). The emergence of crowdinvesting in Europe.MunichDiscussionPaperNo.2014‐43.http://ssrn.com/abstract=2481994.
Hornuf, L., & Schwienbacher, A. (2015). Funding dynamics in crowdinvesting. SSRNWorkingPaper.http://ssrn.com/abstract=2612998.
Hornuf, L., & Schwienbacher, A. (2016). Should securities regulation promotecrowdinvesting?SmallBusinessEconomics,forthcoming.
Janis,I.(1972).Victimsofgroupthink:psychologicalstudyofforeign‐policydecisionsandfiascos,2nded.,Boston,MA:HoughtonMifflin.
Kagel,J.,&Levin,D.(2001).Behaviorinmulti‐unitdemandauctions:experimentswithuniform price and dynamic Vickrey auctions. Econometrica, 69(2), 413–454.http://dx.doi.org/10.1111/1468‐0262.00197
Klöhn, L., &Hornuf, L. (2012). Crowdinvesting inDeutschland:Markt, Rechtslage undRegulierungsperspektiven.JournalofBankingLawandBanking,24(4),237–266.
Knight,T.B.,Leo,H.,&Ohmer,A.(2012).Averyquietrevolution:aprimeronsecuritiescrowdfundingandtitleoftheJOBSAct.MichiganJournalofPrivateEquity&VentureCapitalLaw,2,135–153.
Koning, R., & Model, J. (2013). Experimental study of crowdfunding cascades: whennothing is better than something. Academy of Management Proceedings, 2014(1),16683.http://dx.doi.org/10.5465/AMBPP.2014.16683abstract
Kuppuswamy,V.,&Bayus,B.L. (2014).Crowdfunding creative ideas: thedynamicsofproject backers in Kickstarter. SSRN Working Paper.http://ssrn.com/abstract=2234765.
Lee, E., & Lee, B. (2012). Herding behavior in online P2P lending: an empiricalinvestigation. Electronic Commerce Research and Applications, 11(5), 495–503.http://dx.doi.org/10.1016/j.elerap.2012.02.001
Lin,M.,Prabhala,N.,&Viswanathan,S.(2012).Judgingborrowersbythecompanytheykeep:friendshipnetworksandinformationasymmetryinonlinepeertopeerlending.ManagementScience,59(1),17–35.http://dx.doi.org/10.1287/mnsc.1120.1560
27
Lin,M.,&Viswanathan,S.(2013).Homebiasinonlineinvestments:anempiricalstudyofan online crowdfunding market. Management Science, 62(5), 1393–1414.http://dx.doi.org/10.1287/mnsc.2015.2206
Marom,D.,&Sade,O.(2013).Arethelifeanddeathofanearlystageventureindeedinthepowerofthetongue?Lessonsfromonlinecrowdfundingpitches.SSRNWorkingPaper.http://ssrn.com/abstract=2255707.
Meer, J. (2014). Effects of the price of charitable giving: evidence from an onlinecrowdfundingplatform.JournalofEconomicBehaviorandOrganization,103,113–124.http://dx.doi.org/10.3386/w19082
Mollick, E. (2013). Swept awayby the crowd?Crowdfunding, venture capital, and theselectionofentrepreneurs.SSRNWorkingPaper.http://ssrn.com/abstract=2239204.
Mollick, E. (2014). The dynamics of crowdfunding: an exploratory study. Journal ofBusinessVenturing,29(1),1–16.http://dx.doi.org/10.1016/j.jbusvent.2013.06.005
Roth,A.,&Ockenfels,A.(2002).Last‐minutebiddingandtherulesforendingsecond‐priceauctions: evidence from eBay and Amazon auctions on the Internet. AmericanEconomicReview,92(4),1093–1103.http://dx.doi.org/10.3386/w7729
Saxton, G., & Wang, L. (2014). The social network effect: the determinants of givingthrough social media. Nonprofit and Voluntary Sector Quarterly, 43(5), 1–19.http://dx.doi.org/10.1177/0899764013485159
Scharfstein, D., & Stein, J. (1990). Herd behavior and investment.American EconomicReview,80(3),465–479.http://dx.doi.org/10.1257/aer.90.3.705
Spence,M.(1973).Jobmarketsignaling.QuarterlyJournalofEconomics,87(3),355–374.http://dx.doi.org/10.2307/1882010
Vermeulen,R.(2013).Internationaldiversificationduringthefinancialcrisis:ablessingfor equity investors? Journal of International Money and Finance, 35, 104–123.http://dx.doi.org/10.1016/j.jimonfin.2013.01.003
Vickrey,W.(1961).Counterspeculation,auctions,andcompetitivesealedtenders.JournalofFinance,16,8–37.http://dx.doi.org/10.1111/j.1540‐6261.1961.tb02789.x
Vismara,S.(2015).Informationcascadesamonginvestorsinequitycrowdfunding.SSRNWorkingPaper.http://ssrn.com/abstract=2589619.
Younkin, P., & Kashkooli, K. (2013). A crowd or a community? Comparing threeexplanations for the decision to donate to a crowdfunding project.https://www.funginstitute.berkeley.edu/sites/default/files/A_Crowd_Or_Community.pdf.
Zvilichovsky,D.,Inbar,Y.,&Barzilay,O.(2013).Playingbothsidesofthemarket:successand reciprocity on crowdfunding platforms. SSRN Working Paper.http://ssrn.com/abstract=2304101.
28
Appendix
TableA1:Descriptivestatisticsofexplanatoryvariables
Mean SD Min. Max. Yes CorrPre‐valuation(inEUR1,000) 1,051.350 770.547 420 10,000 –0.047 Fundinggoal(inEUR1,000) 60.325 14.904 36 150 0.050 Numberoftickets 2.270 2.821 1 40 0.019 Initialprice/ticket(inEUR1,000) 1.136 1.577 0.5 25 –0.019 Bonds 0.303 0.460 439 0.005 Commodities 0.228 0.420 331 –0.030 Funds/certificates 0.416 0.493 603 0.047 Realestate 0.306 0.461 443 –0.013 Stocks 0.460 0.499 667 0.049 Termdeposits 0.434 0.496 629 0.025 Otherequity 0.309 0.462 448 0.036 Disposableincome2011(inEUR1,000) 21.219 2.569 16.239 28.900 0.025 Numberofpreviouspledges 4.998 6.979 1 55 0.030 Fundingshare(in%) 69.469 36.353 0 100 0.240 **Fundinggoalreached 0.478 0.500 693 0.309 **Fundinglimitreached 0.284 0.451 412 0.419 **Bidsearlierthatday(inEUR1,000) 15.660 38.023 0 217 0.227 **VDAX(in%) 19.568 6.752 11.47 37.28 0.271 **Distancebacker/start‐up(in100km) 2.942 1.758 0 6.440 0.033 Daysremaining 14.911 14.101 –0.047 76.058 –0.061 *Tuesday 0.112 0.316 163 –0.072 **Wednesday 0.157 0.364 228 –0.055 *Thursday 0.114 0.318 165 –0.045 Friday 0.140 0.347 203 –0.007 Saturday 0.105 0.306 152 –0.036 Sunday 0.292 0.455 423 0.191 **
Note:Column‘Yes’indicatesifadummyvariabletakesthevalueof1.Column‘Corr’showsbivariatecorrelationswiththepremium.**/*indicatessignificanceatthe1%/5%level.
29
Fig.A1:Auctionentrymask
numberofticket
priceperticket
totalprice
fundinggoal fundinglimit
fundingreached
currentprice
valuationdaysremaining