Successful Online Crowdfunding
Campaigns:
An Application of Parametric and Semiparametric
Two-step Selection Models
Annemieke Denters
Student number: 5804205
Supervised by: Dr. J.C.M. van Ophem Second reader: Dr. K.J. van Garderen Date: 13-08-2015
Abstract
Onl i ne c r ow df undi n g i s a no vel al t ernat i ve wa y t o f un d vent ures . Howe ver , l i t erat ur e on crowdf undi n g i s st i l l r el ati vel y sparse. In t hi s t hesi s we research t o what ext ent f undi n g t ype and p l at f or m choi ce are predi ct ors for successful crowdfundi n g ca mpai gns on pl at f or ms K i ckst ar t er and In di ego go when al l owi n g f or sel ect i on and endo genei t y. T o corr e ct for sa mpl e sel ect i on, we est i mat e t wo -st ep sel ect i on model s using Heckman’s two -step estimator, Newey’s series estimator and Coslett’s selection model f or fundi n g t ype choi ce and pl at f orm choi ce. T o correct for endo genei t y, t he second st ep of t he t wo -st ep sel ect i on mo del s i s est i mat ed wi t h a t wo -st a ged l east squar es speci fi cat i on. In addi t i on, w e est i mat e t hese model s wi t h t he l ess rest ri ct i ve swi t chi ng r e gressi o n approach, al l owi n g t h e coeff i ci ent s and err ors t o be di f fer ent bet ween t he t r eat men t s. Fi nal l y, we est i mat e t he avera ge t reat ment effect for t he t r eat ed ( AT ET ) . We find no si gni f i cant t reat ment eff ect f or t he di f ferent fundi ng t ypes or for pl at f or ms K i ckst art er and Indi e go go . T her efore, we concl ude t hat i t i s not evi dent t hat ent repr en eurs’ deci si ons on f un di ng t ype a nd pl at for m have a di rect ef f ect on t he success of onl i ne crowdfun di n g ca mp ai gns .
1
1 Contents
2 Introduction ... 4
2.1 Theoretical framework ... 4
2.2 Online crowdfunding pl atforms ... 6
2.2.1 Kickstarter ... 7
2.2.2 Indiegogo ... 8
2.2.3 Research question: fixed versus flexible funding ... 8
3 Data and Variables ... 8
3.1 Data collection process ... 9
3.2 Dependent variables ... 9
3.3 Funding type choice and platform choice ... 11
3.3.1 Funding types on Indiegogo and Kickstarter ... 11
3.3.2 Success rates ... 11
3.4 Descriptive statistics ... 13
3.4.1 Correlation matrix of explanatory variables ... 14
3.5 Dummy variables ... 15
3.5.1 Categories ... 16
3.5.2 Verified non-profit organizations ... 18
3.5.3 Starting years ... 18 3.5.4 Countries ... 18 3.6 Additional considerations ... 19 3.7 Empirical challenges ... 19 3.7.1 Endogeneity ... 19 3.7.2 Funding goal ... 21
3.7.3 Truncated data for Indiego go and Kickstarter ... 22
3.7.4 Extremely successful projects ... 24
3.7.5 Right-skewed variables ... 24
4 Binary choice models for sample selection ... 26
4.1 Bivariate probit with sample selection ... 26
4.2 Variable specification for funding type choice and platform choice ... 28
2
5.1 Variable specification ... 31
5.2 Sample selection correction ... 31
5.3 Parametric estimation : Heckman’s Two -Step-Selection model ... 32
5.4 Semiparametric estimation ... 33
5.4.1 Coslett’s selection model ... 33
5.4.2 Newey series estimator ... 34
6 Switching regression for funding type and platform ... 36
6.1 Switching regression ... 36
6.2 Fixed funding as a treatment ... 37
6.3 Platform Indiegogo as a treatment ... 38
7 Endogeneity: Hausman’s specification test ... 38
7.1 Endogeneity of amount of backers ... 39
7.2 Hausman specification test for amount of backers ... 39
7.3 Instruments ... 40
7.3.1 Test on the amount of Facebook friends ... 41
8 Average treatment effects ... 41
9 Results ... 44
9.1 Bivariate probit with sample selection : Heckprobit ... 44
9.2 Probit models for funding type choice and platform choice ... 45
9.3 Hausman test for endogeneity ... 47
9.4 Exogeneity of instruments ... 48
9.5 Two-step-selection models funding type choice ... 50
9.5.1 Explanatory variables ... 50
9.5.2 Selection correction terms ... 51
9.5.3 Partial conclusion two-step selection models ... 52
9.6 Switching regression funding type choice ... 53
9.6.1 Switching regression for fixed funding ... 53
9.6.2 Switching regression for flexible funding ... 56
9.6.3 Partial conclusion switching regression for funding type ... 57
9.7 Switching regression for platform choice ... 59
3
9.7.2 Switching regression for Indiegogo and fixed funding ... 62
9.7.3 Switching regression for Kickstarter ... 64
9.7.4 Partial conclusion switching regression for platform choice ... 66
9.8 ATET results ... 67
10 Conclusions ... 69
11 References ... 72
4
2 Introduction
New ventures require capital at some point in order to sustain existence and growth. But without proven customers and/or profits a new venture might find it hard to acquire a bank loan. Traditionally, such ventures would turn to ‘angel investors’ that grant a large sum of money often in return for shares in the venture. Angel investors can be professional investors or friends and family that believe in the entrepreneur and the business. An alternative wa y to raise capital is crowdfunding where a net work of peopl e fi nances the capital . An int uitive exampl e is a music art ist that wants to record a new al bum and as ks fans to donat e mone y or pre -purchas e the album . In this exampl e the investor and the consum er are the s ame. Crowdfundi ng can thus allow consumers t o pla y a m ore active rol e in devel opm ent - instead of choosi ng what to buy aft er producti on consum ers can activel y influence what wil l be produced.
Websit es l ike Ki ckst art er.com and Indiegogo.com hav e revolut ioniz ed crowdfundi ng in recent years . In 2015, it is expecte d that the indus tr y will rais e twi ce the amount of funds it did i n 2014 to 34.4 billi on USD (M as solot ion , 2015). An advant age for the ent repreneur is t hat crowdfundi ng websit es lower the t hreshold for entrepreneurs t o reach potenti al inves tors outsi de t hei r own soci al network. The new plat form com es with new nam es as wel l: ent repreneurs are referred to as project founders and investors are called project backers. In contrast to the angel investors in traditional financing, these backers contribute relatively small amounts of money thus greatly reducing the barrier for people to support project they find attractive. As a relatively new phenomenon, there are still m an y i mportant questi ons reg arding crowdfunding websites. When a venture considers crowdfunding t hrough a websit e it is faced with a coupl e of decisi ons . There are s everal online crowdfunding pl atforms that provide ent repreneurs wit h wa ys to raise funding. Thes e pl at forms offer di ffe rent funding goal t ypes, such as fixed and fl exible fundi ng. Wit h these opti ons , th e ent repreneurs will have to decide on whi ch of the platform s avail abl e to l aunch its fundi ng campai gn. Also t he y have t o decide whether to s et a fixed funding goal, or whet her t o rais e t hrough so call ed fl exible fundi ng. This thesi s will anal yz e pl at form choice and funding goal t ype choi ce on crowdfunding websit e dat a and t akes sampl e s el ection and endogeneit y into account.
2.1 Theoretical framework
Online crowdfunding is a rel ati vel y new res earch topic wi th a s mal l number of papers publi shed. The foll owi ng paragraph wi ll dis cuss thes e. Fi rst we dis cuss profit bas ed crowdfunding vers us product bas ed crowdfunding. S econdl y we discuss the import ance of the soci al network of proj ect founders and t he ir
5 spat ial dist ance t o the backers. Thi rdl y we di scuss the import ance of updates to proj ects .
Before starting a crowdfunding campaign, the project founder has to decide on the form of funding she will try to raise. Bellaflame et al. (2012) compared the two most common forms of crowdfunding: profit based and product based. Profit based crowdfunding is similar to the classical entr epreneur-investor relationship: project founders share their profit or equity securities with their backers in exchange for money. The backers can purchase the product after production, but are not necessarily its consumers. Profit based funding will not be discussed further in this thesis as t he crowdfunding websites and dat a analyzed do not include profit based f unding campaigns. Instead, this thesis focuses on product based crowdfunding.
Product based crowdfunding enables the backers to pre-purchase the product, enabling the project founder to finance the production of the product. The backers are also the consum ers of this product. Product based crowdfunding has three advantages. Firstly, founders retain full ownership of their business and do not have to cede profits in the form of dividends. Secondly, founders receive an estimate for product demand enabling bet ter alignment of investments. Thirdly, working capital requirements are greatly reduced as customers pay before production or even development. As a result, risks are greatly reduced for the project founder. Moreover, product based crowdfunding can allow f ounders to price discriminate very specifically: founders can price discriminate between consumers who pre-purchase the product and re gular consumers in the market. Under some conditions price discrimination enables the entrepreneur to gain a larger profit (Varian, 1985). Bellaflame et al. (2012) find that founders prefer product based funding to profit-sharing, unless a large capital requirement per product is needed.
As with many other internet related revolutions, social information is a relevant aspect to crowdfunding websites. In the early stages of the online crowdfunding campai gn founders will want t o increase t he probabilit y of s uccess of the campai gn. Thes e earl y s tages of a funding campai gn will res olve around informi ng pot enti al backers. Traditio nall y invest ors prefer a short ph ys i cal dist ance to t he proj ect , as this is essential in order t o acquire informat ion, monitor progress and provide input (Agrawal et al ., 2011). However, Agrawal et al. find t hat dis tance does not necess aril y hold t he s am e rel evance for crowdfundi ng: on m usic crowdfunding websit e S ell aband t he average dist ance bet ween artist and backers is 3000 mil es. Inst ead, geographical proximit y is onl y i mport ant i n t he fi rst stages of invest ment to gain t he support of earl y
6 backers (Cont i et al., 2010). These initi al backers provide a si gnal of ent repreneuri al com mitment. Agr awal et al. (2011) found t hat i niti al backers consist mainly of the project founder’s social network: friends, family and acquaint ances . A st rong social net work s ug gests trust worthi ness of the founder (J ohnst on et al ., 2013). Thi s si gnal has a pos itive effect on the decis ions of pot enti al backers that have a l arger dis tance to the proj ect founder. P roj ects requi re 'soci al proof': t he concept that backers are more lik el y to support proj ects t hat are alread y backed b y m an y others (Ci aldini, 2009). To concl ude, a crowdfunding campaign’s success does not depend on the physical distance of the founders to the backers , but does depend on the siz e and scope of the s ocial net work of the founders .
Frequent updates help to increase chance of a successful funding campaign, especially in the final days before the end of a campaign (Kuppuswamy and Bayus, 2013). An update is a message from the founder that is posted on the campaign page about the progress of the campaign. An update is visible for everyone and is also delivered to backers that are supporting the campaign via email. Kuppuswamy and Bayus (2013) find that the success of a project is mostly determined in the final stages of the online crowdfunding campaign for two reasons. A late peak of backer support increases the probability of a project succeeding. The researchers hypothesize that backers, noticing the imminent deadline, gather support for the project in their own soc ial network thus boosting the success chances. When the project deadline is approaching, updates tend to increase for all projects. However, projects that update relatively more frequent are more successful.
The dis cus sed arti cl es have thus anal yz ed predi ctors for a s uccess ful crowdfundi ng campaign in t hree st ages of the proj ect. Fi rstl y, b efore t he proj ect starts t he founder choos es profit based or product based crowdfunding. S econd y, in the earl y st ages of the crowdfundi ng campai gn the social net work of the founder i s import ant . Fi nall y, at the end of a funding campaign the am ount of updat es and the i ncrease of additional backers are predi ct ors for a s uccess ful proj ect. This thesis will focus on deci si ons founders will face at the st art of a crowdfundi ng campaign. Speci fi call y, t hi s thesis will anal yz e fixed and flexibl e funding as predictors for success on online crowdfunding platforms
2.2 Online crowdfunding platforms
In this paragraph we discuss online crowdfunding platforms. Firstly, we discuss online crowdfunding platforms in general. Secondly, we discuss online
crowdfunding platform Kickstarter in paragraph 2.2.1 and Indiegogo in
paragraph 2.2.2. Finally, we discuss funding type choice on these platforms and formulate the research question of this thesis in paragraph 2.2.3.
7 Crowdfunding pl at forms have grown strongl y over the past 5 years. In 2012 to 2013 fundi ng rais ed b y crowdfunding websi tes have i ncreas ed b y 81% t o 2.7 billion USD. Furt hermore, in 20 14 platforms expanded b y 167 percent to rais e 16.2 billi on USD on campai gns, up from 6.1 billion US D in 2013. In 2015, it is expect ed that t he indust r y will doubl e t o rais e 34.4 billi on USD on campai gns (Massolotion ,2015). At this mom ent , Kickstart er and Indiegogo are the most popul ar crowdfundi ng comm uniti e s accordi ng t o Google P ageR ank shown in tabl e 1 (bl og. al exa.com in J anuar y 2015). This thesi s therefore focus es on Kickst arter and Indi egogo.
T a b l e 1 G o o g l e P a g e R a n k W e b s i t e W o r l d w i d e G o o g l e P a g e R a n k K i c k s t a r t e r . c o m 4 1 4 I nd i e g o go . c o m 1 1 8 0 G o f u nd me . c o m 1 6 9 2 T i l t . c o m 2 1 7 9 3 U l e l e . c o m 2 6 2 1 9
Online crowdfundi ng platform s Ki ckstart er and Indiegogo have m an y similariti es . Both pl atforms clai m as t heir miss ion t o offer a plat form to enabl e creative cam pai gns t o connect with backer s. Pot enti al backers can s ee t he l evel of support from other backers before they can choose to invest. On both crowdfunding platforms campaigns are open to investments for a limited amount of time up to 90 or 120 days. They both disallow starting a crowdfu nding campai gn for a profi t -bas ed proj ect on t hei r platform . Inst ead, backers usuall y receive a product or an experi ence in exchange for thei r investm ent . A product can ent ail what the founders aim to m ake, merchandis e wi th the l ogo of t he proj ect founder or a pos tcard. An experi ence can entail a lunch with t he founders to nami ng a charact er in the comput er gam e or the com ic book aft er you. Finall y, proj ect founders own the intell ectual propert y on both Ki ckst art er and Indi egogo: the crowdfunding webs i tes offer m erel y a pl atform and are not invol ved in the devel opm ent of the projects.
2.2.1 Kickstarter
Kickst arter was founded in 2009 b y Yance y St rickl er, Charl es Adl er and P err y Chen. Kickst arter has s ince raised 1.2 billion US D t o over 77,000 creative proj ects b y 7 milli on backers (Kickst art er.com/hel p/st ats ). Kicks tart er does not allow an y proj ect. P roj ects have t o go through an appl ication process . From 2009 until 2012, Ki ckst art er onl y al lowed proj ect founders locat ed in the US from 2009 to 2012 . Since then, the all owed locati ons have expanded to include the UK (oct 2012), Canada (s ep 2013), New Zeal and and Australi a (nov, 2013), Net herl ands (april , 2014), S weden, Denmark, Norwa y and Irel and (oct, 2014).
8 Kickst arter m entions that projects cannot rais e funds for charit y or involve prohibit ed i tem s. P rohibit ed i tem s include among ot hers: political fundraisi ng, medi cal drugs, weapons and t obacco. Campai gns can run up to 90 da ys . (Ki ckstart er.com )
2.2.2 Indiegogo
Indi egogo was founded in 2008 b y Sl ava R ubi n, Danae Ri n gelmann and Eri c Schell. Indiegogo does not publ ish st atis t ics online. Indi egogo menti ons on their websit e t hat there is no applicati on proces s. The pl atform puts fewer requi rements on project founders and proj ects t han Ki ckst arter. For i nst ance, Indi egogo does allow for politi cal fundrai sing and proj ects founders to com e from ever y countr y. In addition to entrepreneurial and creati ve proj ects that can also be found on Ki ckst art er, Indi egogo also allows personal cam pai gns (rai se mone y for you, a loved one i n need or t o celebrat e a li fe event ) and non -profi t campai gns. C am pai gns can run up to 120 da ys . ( Indi egogo.com)
2.2.3 Research question: fixed versus flexible funding
One of the ke y di fferences between Ki ckst art er and Indi egogo is the t ype of funding goal. Ki c kst art er onl y uses a fixed funding goal, whereas Indi egogo al so allows for a fl exibl e fundi ng goal . Through fl exible funding project founders alwa ys receive pl edged mone y regardl ess of m eeting the preset funding goal. Fixed fundi ng ret urns pledged m one y ba ck to backers if t he funding goal i s not met. It is not cl ear, whether proj ects raise more funds t hrough fixed or fl exible funding. This rai ses the question: To what ext ent can fundi ng t ype choice and platform choice predict differences in crowd funding campaign success of Kickstarter and Indiegogo when allowing for selectivity and endogeneity?
This thesis researches fixed and flexible funding as predictors for success on online crowdfunding platforms and proceeds as follows. In section 3 we describe the data and variables, in section 4 we describe the methodology and model specifications for funding type choice and platform choice, in section 5 we describe parametric and semi -parametric two-step-selection models, in section 6 we describe switching regression models for funding type choice and platform choice, in section 7 we discuss endogeneity and introduce 2SLS instrumental variables estimation methods , in section 8 we discuss suitable measures for the treatment effects of platform choice and funding type c hoice, in section 9 we present the results and finally, we conclude in section 10.
3 Data and Variables
In this section we describe the data, the relevant independent and dependent variables and the empirical challenges concerning the data and data collecti on
9 process. Firstly, we discuss the data collection process in paragraph 3.1.
Secondly, in paragraph 3.2 we discuss and choose the dependent variable. Thirdly, in paragraph 3.3 we discuss the independent variables of interest, namely funding type choice an d platform choice. Fourthly, in paragraph 3.4 we discuss the remaining continuous independent variables and their descriptive statistics. Fifthly, in paragraph 3.5 we discuss dummy variables. Sixthly, in paragraph 3.6 we discuss our additional consideratio ns and finally in paragraph 3.7 we discuss empirical challenges regarding the data and the data collection process.
3.1 Data collection process
In this paragraph we discuss the data collection process. The collected data consists of 210,406 observations f rom crowdfunding websites Kickstarter and Indi egogo. The dat a is coll ected using a dat a scraper t o cop y t he url s and corresponding information from Kickstarters website. The urls from Indiegogo could not be scraped directly from their website so a database of urls was first used, following Lau (2014). The urls in this database were collected in 0 1-2014 and were used to scrape the corresponding information from Indiegogo. The final Kickstrater and Indiegogo data were scraped between 10-09-2014 and 23-09-20141. From the scarped observations a total of 7, 755 observations were removed: 417 were removed because the url was either removed from the
internet or broken. 7,339 cam pai gns were live at t he tim e of dat a coll ection and had remaining time to raise funding, for the analysis they were filtered out resulting in a total of 202,650 fi nished campaigns: 160,396 observations for Kickstarter and 42,254 for Indiegogo.
3.2 Dependent variables
In this paragraph we consider possible dependent variables and their properties. There are three ways to measure the success of an online crowdfunding project.
1. The rai sed funding i n US D
Continuous.
2. Is t he funding goal reached?
Binary - Yes or No.
3. The funding rati o: ratio of rais ed m one y over funding goal .
Continuous. 1 I n c o l l a b o r a t i o n wi t h J o r g v a n d e r H a m , M S c E n t r e p r e n e ur s h i p s t ud e nt o f C o p e n ha ge n B u s i n e s s S c ho o l 2 F o r m o a nd a ( h t t p : / / w w w. o a nd a . c o m/ c ur r e n c y/ a ve r a ge )
10 From the graph below the distribution of funding ratio is displayed. 43% of the campaigns reach their funding goal, thus have a funding ratio larger than 1. Only a few projects fail their funding target by a slim margin: Campaign with funding ratios between 0.6 and 0.8 account for only 1% of the sample. Whereas the category is composed of campaigns that just reach their target funding , campaigns with a funding ratio between 1.0 and 1.2 , account for 24% of the sample. Campaigns that were not able to raise more than 20% of their funding goal account for 42% of the sample . There is a number of campaigns with a funding ratio of 5 and up, this could be due to projects choosing a (too) small funding goal.
F i g u r e 1 D i s t r i b u t i o n o f t h e f u n d i n g r a t i o ( X - a x i s i s t h e f u n d i n g r a t i o , Y - a x i s i s t h e p e r c e n t a g e o f c a mp a i g n s )
For an entrepreneur the goal is to raise funds and other parameters are a means to achieve that end. There are several parameters that an entrepreneur can
influence in the process of launching a crowdfunding campaign, such as type of funding goal, platform, and funding goal. On platforms Kickstarter and
Indiegogo the ‘is successful’ and the funding ratio are used as measures for success. But, as funding ratio and ‘is successful’ are affected by the funding goal, these measures for successfulness contain less information – e.g. changing the funding goal with the aim of raising additional funding would affect funding ratio and ‘is successful’ as they are derived from the funding goal . This is
further supported by the observation that the correlation between funding goal and raised funding is negligible, which we will show later in this thesis.
Therefore, we argue that raised funding (USD) is the best measure for the success of an online crowdfunding campaign.
00% 05% 10% 15% 20% 25% 30% 35% 40% 45% 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.5 5 50 % fixed funding % flexible funding
11 In this paragraph we considered different dependent variables and argued that, based on economic theory, raised funding in USD is the best dependent variable for our model.
3.3 Funding type choice and platform choice
In this paragraph we discuss th e independent variables of interest funding type choice and platform choice. We are interested in the effect of choosing fixed versus flexible funding on raised funding. To research the effect of funding type choice, we also research the effect of using platform Indiegogo versus platform Kickstarter on raised funding. Firstly, we discuss differences between fixed and flexible funding on platforms Indiegogo and Kickstarter in paragraph 3.3.1 . Secondly, we discuss the general differences between fixed and fle xible funding and their success rates in section 3.3.2.
3.3.1 Funding types on Indiegogo and Kickstarter
Firstly, we discuss the differences between fixed and flexible funding goals for platforms Indiegogo and Kickstarter. In table 2 the platform fees for fixed and flexible funding are shown.
T a b l e 2 P l a t f o r m F e e s P l a t f o r m f e e s G o a l i s me t G o a l i s no t me t ( F i x e d F u nd i n g) G o a l i s no t me t ( F l e xi b l e F u nd i n g) K i c k s t a r t e r 5 % 0 % - I nd i e go go 4 % 0 % 9 %
Indiegogo facilitates both fixed and flexible goals. Indiegogo charges a fee of 4% if a funding goal is met, for both fixed and flexible goals. For flexible goals the raised funds are also paid out if the goal is not met. In that case Indiegogo charges 9% of the raised funds. If a fixed goal is not met Indiegogo charges a fee of 0% (Indiegogo’s website).
All goals on Kickstarter are fixed goals. Kickstarter charges a fee of 5% of the raised funds if the goal is met and 0% if the goal is not met. If the goal is not met, the funds are returned to the backers (Kickstarter ’s website).
Indiegogo offers a choice between fixed and flexible funding. As a result, 95.6% of the Indiegogo campaigns use flexible funding and 4.6% of the campaigns use fixed funding.
3.3.2 Success rates
In paragraph 3.3.1 we describ e that Indiegogo offers the option to choose flexible funding. As a result a large majority ( 95.6%) of the campaigns founders choose to use flexible funding.
12 In this section we first argue that campaign founders are more likely to choose flexible funding because they are naturally risk -averse. In prospect theory (Tversky and Kahneman , 1981) it is assumed that people in the domain of gain are naturally risk-averse. People prefer a certain (no risk) gain in money above uncertain (risk) gain with equal expect ed outcome. However, if people are in the losing domain, they naturally act risk -seeking to avoid losing more money. We apply this principle to the funding types on crowdfunding platforms. The gaining domain is comparable to flexible funding, as money is g ained with every additional backer. The losing domain is comparable to fixed funding, because money with every additional backer is ‘lost’ until the funding goal is met. As people prefer the gaining domain to the losing domain, we argue that people prefer flexible funding when offered the choice.
Secondly, we argue that fixed funding campaigns are more likely to meet their funding goal than flexible funding campaigns. Although people prefer the winning domain of flexible funding, we argue that campaign foun ders are more likely to succeed in the losing domain of fixed funding. As suggested by prospect theory, in the losing domain of fixed funding, the founders act risk -seeking and with flexible funding act risk averse. We suggest that with fixed funding the founders are more motivated not to lose additional money, than the flexible funding founders are to gain additional money. To avoid ‘losing’ money, founders may prefer flexible funding: they receive the pledged money, even if their projects funding goal is not met. Therefore, we argue that flexible funding campaigns are less likely to meet their funding goals than fixed funding campaigns.
The success rates for fixed and flexible funding are reported in table 3. Fixed funding campaigns have a higher probabi lity to meet their funding goal (43.4%) than flexible funding campaigns (31.4%). Indiegogo campaigns that use fixed funding are reported to have the highest probability (52.7%) to meet their funding goal. T a b l e 3 S u c c e s s r a t e s f o r f i x e d a n d f l e x i b l e f u n d i n g c a mp a i g n s S u c c e s s p e r F u n d i n g t y p e F l e x i b l e ( I n d i e g o g o ) F i xe d ( T o t a l ) F i xe d ( K i c k s t a r t e r ) F i xe d ( I n d i e g o g o ) T o t a l S uc c e s s f u l 1 3 2 8 6 6 9 6 1 8 6 8 5 9 2 1 0 2 6 8 2 9 0 4 U ns u c c e s s f ul 2 8 9 6 8 9 0 7 7 8 8 9 8 5 8 9 2 0 1 1 9 7 4 6 T o t a l 4 2 2 5 4 1 6 0 3 9 6 1 5 8 4 5 0 1 9 4 6 2 0 2 6 5 0 S uc c e s s f u l % 3 1 . 4 % 4 3 . 4 % 4 3 . 3 % 5 2 . 7 % 4 0 . 9 %
In this paragraph we discussed the differences between fixed and flexible funding. On platform Indiegogo 95.6% of the campaigns use flexible funding
13 and 4.4% use fixed funding. On Kickstarter all ca mpaigns use fixed funding. Fixed funding campaigns have a higher probability to meet their funding goal than flexible funding campaigns. Fixed funding campaigns on Indiegogo are reported to have the highest probability to meet their funding goal.
3.4 Descriptive statistics
In this paragraph we present and discuss the descriptive statistics and the correlations between the continuous variables.
The descriptive statistics of the relevant variables are reported in table 4. From table 4 we observe that the average funding goal is double for flexible funding (68175 USD), compared to fixed funding (31 347 USD). However, the funding result is half for flexible funding (4119 USD), compared to fixed funding (8 927 USD). The data suggests that the difference in funding result is mostly due to the number of backers as flexible funding campaigns have only half the number of backers on average compared to fixed funding (54 and 103 respectively). However, on average, backers of flexible funding campaigns invest more ( 90.31 USD) than backers of fixed funding campaigns ( 66.47 USD).
Based on the differences between fixed and flexible funding a hypothesis can be formed that fixed funding projects convince a broader audience to invest. From the data it can be observed that fixed funding projects have twice the amount of backers, but raised funds per backer is lower. It is possible that fixed funding projects have a higher conversion because backers are ensu red that they won’t lose their invested money thus lowering the ‘hurdle’ fo r backing a fixed funding campaign. Theoretically, one would expect those potential backers that are less enthusiastic or with lower budgets to be more concerned about investing in a campaign with the risk of no return, which is the case for flexible fundi ng.
14 T a b l e 4 D e s c r i p t i v e S t a t i s t i c s f i xe d f u n d i n g f l e xi b l e f u n d i n g V a r i a b l e M e a n s t d . d e v . M i n ma x M e a n s t d . d e v . mi n M a x f u nd i n g go a l 3 1 3 4 7 . 3 1 , 9 6 e + 0 6 0 . 0 1 6 . 6 0 e + 0 8 6 8 1 7 4 . 7 9 , 7 3 e + 0 6 4 8 0 2 . 0 0 e + 0 9 f u nd i n g r e s 8 0 5 2 . 4 8 1 3 2 3 . 8 0 1 . 3 3 e + 0 7 4 , 0 7 5 . 4 2 2 2 9 1 . 5 0 1 . 9 6 e + 0 6 s uc c e s s f u l ? 0 . 4 3 0 . 5 0 0 1 0 . 3 1 0 . 4 6 0 1 f u nd i n g r a t i o 2 . 0 1 1 4 3 . 3 0 4 1 5 3 5 0 . 6 4 1 . 4 2 0 1 4 2 # b a c ke r s 1 0 3 . 3 8 2 8 . 2 0 1 . 0 6 e + 0 5 5 4 . 3 3 2 6 . 8 0 3 3 2 5 3 # up d a t e s 5 . 0 6 8 . 9 6 0 3 0 1 4 . 3 3 8 . 3 2 0 2 4 7 M e a n i n v e s t me n t 6 6 . 4 7 1 2 1 . 3 0 1 0 , 0 0 0 9 0 . 3 1 1 2 8 . 4 0 1 0 , 0 0 0 # c o m me n t s 3 5 . 4 1 2 0 6 . 9 0 3 2 5 , 5 3 5 2 3 . 8 2 0 0 . 6 0 2 8 , 2 7 2 # fo u nd e r F a c e b o o k fr i e nd s 4 6 6 . 6 7 9 7 . 5 0 5 , 8 9 1 4 5 0 . 9 7 1 6 . 7 0 4 , 9 6 6 # d a ys r u n ni n g 3 5 . 1 1 4 . 1 1 3 7 8 4 7 . 4 2 7 . 2 0 3 6 3 # i ma ge s 3 . 8 0 7 . 8 0 0 3 7 8 1 . 5 3 4 . 6 4 0 1 2 7 vi d e o ? 0 . 7 8 0 . 4 1 0 1 0 . 6 9 0 . 4 6 0 1 N u mb e r o f o b s e r v a t i o n s 1 6 0 3 9 6 4 2 2 5 4 1 ) o n l y t h e f i x ed f u n d i n g c a m p a i g n s f r o m K i c k s t a r t er
To conclude, the composition of raised funding is different for fixed and flexible funding. Raised funding is the product of average funding and mea n backer investment. Fixed funding campaigns have a larger number of backers that invest a smaller amount compared to flexible funding campaigns that have a smaller amount of backers that invest a larger amount of money. In section 3.4.1 we discuss the correlations between these variables to get more insight.
3.4.1 Correlation matrix of explanatory variables
In this section we firstly discuss the correlations between the different
variables. The correlations between the different variables are presented in table 5. Correlations with a value larger than 0.10 are displayed in bold. The funding goal is not correlated (0.01) with the funding result. This indicates that funding goal is not a good predictor for raised funding. The number of images and the presence of a video are positively correlated with fixed funding (0.13 and 0.09, respectively). Indiegogo allows projects to run for a larger number of days (up to 120 days, in contrast to a maximum of 90 days on Kickstarter) , which
explains the negative correlation ( -0.27) with fixed funding. The amount of backers is strongly positively correlated with funding result (0.76), as additional backers directly raise the funding result. The number of updates and amount of images are also positively correlated with the fundi ng result (respectively 0.20 and 0.16), this could be explained as both are indications for the online effort of the founders: the correlation between the amount of updates and amount of images is also positive (0.32). The amount of comments is also strong ly
15 positively correlated with funding result (0.41), but this could be explained by the strong correlation between the amount of comments and the amount of backers (0.45). The comments are an indication of the backer involvement. The backers are also positively correlated with the amount of images ( 0.17).
T a b l e 5 C o r r e l a t i o n M a t r i x Correlations fix ed ? fu n d in g g o al fu n d in g r es Is s u cc ess fu l? fu n d in g r atio # b ac k er s # u p d ates Me an in v estme n t # co m m en ts # Face b o o k # d ay s ru n n in g # i m ag es h as v id eo ? fixed? 1.00 funding goal 0.00 1.00 funding res 0.02 0.01 1.00 is successful? 0.10 -0.01 0.09 1.00 funding ratio 0.00 0.00 0.01 0.01 1.00 # backers 0.03 0.01 0.76 0.11 0.02 1.00 # updates 0.03 0.00 0.20 0.37 0.01 0.23 1.00 mean investment -0.08 0.00 0.01 0.10 0.00 0.01 0.06 1.00 # comments 0.00 0.00 0.41 0.03 0.00 0.45 0.15 0.01 1.00 # facebook 0.01 0.00 0.01 0.10 0.00 0.01 0.08 0.00 -0.02 1.00 # days running -0.27 0.00 0.00 -0.14 0.00 -0.01 0.03 0.05 0.00 -0.02 1.00 # images 0.13 0.00 0.16 0.07 0.01 0.17 0.32 0.06 0.11 0.00 -0.07 1.00 has video? 0.09 0.00 0.04 0.13 0.00 0.05 0.16 0.07 0.01 0.08 -0.02 0.15 1.00
Indi egogo m akes a distinction for the am ount of da ys a proj ect can run. Fixed funding cam pai gns can run up t o 60 da ys and fl exibl e funding campai gns can run up to 120 da ys ( Indi egogo.com ). Ki ckstart er al lows thei r fi xed funding campai gns to run up to 90 da ys .
In t his secti on we di scuss ed t he correl ati ons bet ween t he expl anator y vari ables. To s um up t he mos t rel evant fi ndings: we find a hi gh correl ati on (0.76) bet ween rai sed funding and amount of backers, but a fai rl y low correl a tion (0.01)
bet ween rai sed fundi ng and funding goal. This indicates that funding goal isn’t a good predictor for raised funding. Next to the variables discussed here, w e
introduce rel evant dummies in s ection 3.5. 3.5 Dummy variables
In this section we describ e the explanatory dummy variables. Firstly, we discuss the categories in paragraph 3.5.1. Secondly, we discuss verified non -profits, starting years and countries in paragraphs 3.5.2, 3.5.3 and 3.5.4 respectively .
16
3.5.1 Categories
In this paragraph we discuss t he comparability of the categories on Kickstarter and Indiegogo. Firstly, we discuss the different categories the online
crowdfunding platforms offer. Secondly, we describe how we covert different categories to make them comparable over platforms. Finally, we discuss the properties, descriptive statists and differences in raised funding for these created categories.
Both Kickstarter and Indiegogo divide their projects in different categories. The most popular categories (largest amount of projects) on both platforms are film, design, technology and music. The categories differ per platform and to compare the platforms we look at similar categories: Kickstarter has 15 categories, while Indiegogo has 25. Based on similarity we use 14 categories for further analysis. These categories are listed below with the number of campaigns between parentheses. They cover art (13,241), comics (4,397), dance (2,690), design (8,745), fashion (6,437), film & video (45,116), food (7,948), games (10,768), journalism (1,277), music (32,169), photography (4,852), publishing (16,436), technology (6,601) and theater (9,225).
The following considerations were made while comparing categories:
Kickstarters publishing category is treated as equivalent to Indiegogo’s writing category.
Kickstarters film&video category is treated as equivalent to truncated Indiegogo’s categories film and video/web.
Indiegogo’s transmedia category is treated as equivalent to Kickstarters journalism category.
Kickstarter has 1 category not covered by Indiego go: Crafts, which entails 1,812 projects.
Indiegogo has 10 categories not covered by Kickstarter. The categories are listed below with the number of projects between parentheses. They cover animals (952), community (4,910), education (2,995), environment (561), health (2,897), no category (5), politics (27 5), religion (335), small business (1,602) and sports (756). These categories are summarized in category other.
Table 6 displays the raised funding per category per funding type. For both fixed and flexible funding, the mean of raised funding per category is shown as well as the number of observations.
17 T a b l e 6 R a i s e d f u n d i n g p e r c a t e g o r y - d e s c r i p t i v e s t a t i s t i c s C a t e g o r y F i xe d f u n d i n g F l e x i b l e f u n d i n g T o t a l M e a n S t d . d e v . O b s . M e a n S t d . d e v . O b s . O b s . A r t 3 1 6 6 1 4 7 3 6 1 2 9 2 9 3 0 1 8 5 0 1 1 1 5 9 1 1 4 5 2 0 C o mi c s 7 2 1 4 2 9 7 3 6 4 2 4 7 3 8 4 0 1 1 8 7 8 3 2 3 4 5 7 0 D a nc e 3 5 3 9 4 7 2 5 2 0 5 4 2 4 6 8 3 5 7 0 7 2 0 2 7 7 4 D e s i g n 2 2 6 7 7 1 9 6 9 3 4 8 6 9 7 1 1 8 1 8 7 0 5 8 2 4 0 1 9 0 9 8 F a s hi o n 5 6 9 4 2 6 5 2 6 6 9 5 0 5 6 2 7 2 2 5 7 9 4 3 7 7 3 8 7 F i l m & V i d e o 6 3 3 1 4 1 3 1 5 3 6 6 7 2 3 7 8 0 1 0 9 8 7 1 2 6 1 2 4 9 2 8 4 F o o d 6 1 2 5 2 7 4 7 3 8 1 9 7 4 4 1 8 7 6 4 3 6 3 2 8 8 2 9 G a me s 2 4 4 1 7 1 6 4 7 3 7 1 0 8 4 2 6 1 3 0 3 3 9 5 2 4 4 7 1 1 2 8 9 J o ur na l i s m 3 5 2 0 1 0 8 5 3 1 2 4 8 4 2 2 1 7 5 2 8 2 3 6 1 4 8 4 M u s i c 3 9 3 6 1 1 3 6 1 3 0 7 1 6 3 3 6 7 7 5 2 3 4 4 3 8 3 5 1 5 4 P ho t o gr a p h y 3 0 0 8 9 4 9 2 5 0 5 0 3 7 2 6 1 5 0 1 5 5 0 8 5 5 5 8 P ub l i s h i n g 3 0 2 3 1 2 3 2 0 1 7 8 8 6 2 8 2 9 3 8 9 3 1 0 4 6 1 8 9 3 2 T e c h no l o g y 3 3 0 7 7 2 3 3 4 7 6 6 0 5 9 1 8 1 9 1 1 1 2 8 0 2 1 0 0 4 7 0 6 3 T he a t e r 4 1 3 9 9 8 3 3 6 5 1 8 2 8 9 6 4 3 5 1 3 0 9 4 9 6 1 2 O t he r1 2 9 5 6 1 4 9 9 0 2 3 3 1 3 7 9 1 1 4 6 2 2 1 4 7 6 5 1 7 0 9 6 T o t a l 1 6 0 3 9 6 4 2 2 5 4 2 0 2 6 5 0 1 ) In c lu d e s 11 c a t eg o r i es t h a t a r e n o t s h a r ed b y t h e p la t f o r m s In d i e g o g o a n d Ki c k s t a r t e r
Table 7 shows the share of total funding raised per category. For instance the category technology has raised 15.5% of the funds in USD using fixed funding with only 3.9% of the total campaigns. For flexible funding the category ‘Other ’ is responsible for 32.6% of the campaigns raising 34.8% of the total funds.
T a b l e 7 R a i s e d f u n d i n g p e r c a t e g o r y - r e l a t i v e s t a t i s t i c s F i xe d f u n d i n g F l e x i b l e f u n d i n g C a t e g o r y t o t a l ( ml n s ) % U S D % o b s t o t a l ( ml n s ) % U S D % o b s A r t 4 0 . 9 3 . 2 % 8 . 1 % 4 . 8 2 . 8 % 3 . 8 % C o mi c s 3 0 . 6 2 . 4 % 2 . 8 % 1 . 2 0 . 7 % 0 . 8 % D a nc e 7 . 3 0 . 6 % 1 3 . 7 % 1 . 8 1 . 0 % 1 . 7 % D e s i g n 1 9 7 . 1 1 5 . 3 % 5 . 8 % 4 . 7 2 . 7 % 0 . 9 % F a s hi o n 3 9 . 6 3 . 1 % 4 . 2 % 2 . 5 1 . 5 % 1 . 0 % F i l m& V i d e o 2 3 2 . 0 1 8 . 0 % 2 2 . 5 % 4 7 . 7 2 7 . 7 % 3 0 . 1 % F o o d 5 0 . 2 3 . 9 % 5 . 1 % 2 . 8 1 . 6 % 1 . 5 % G a me s 2 6 5 . 0 2 0 . 5 % 7 . 2 % 2 . 7 1 . 6 % 1 . 0 % J o ur na l i s m 4 . 4 0 . 3 % 0 . 7 % 1 . 0 0 . 6 % 0 . 6 % M u s i c 1 2 1 9 . 4 % 1 9 . 2 % 1 4 . 9 8 . 7 % 1 0 . 5 % P ho t o gr a p h y 1 5 . 2 1 . 2 % 3 . 0 % 1 . 9 1 . 1 % 1 . 2 % P ub l i s h i n g 5 4 . 1 4 . 2 % 1 0 . 6 % 3 . 0 1 . 7 % 2 . 5 % T e c h no l o g y 2 0 0 . 0 1 5 . 5 % 3 . 9 % 1 8 . 3 1 0 . 6 % 2 . 3 % T he a t e r 2 7 . 0 2 . 1 % 4 . 2 % 9 . 0 5 . 2 % 7 . 4 % O t he r1 6 . 9 0 . 5 % 1 . 4 % 5 6 . 0 3 2 . 6 % 3 4 . 8 % T o t a l 1 2 9 1 1 0 0 % 1 0 0 % 1 7 2 1 0 0 % 1 0 0 % 1 ) In c lu d e s t h e 11 c a t eg o r i es t h a t a r e n o t s h a r ed b y t h e p la t f o r m s In d i e g o g o a n d Ki c k s t a r t er
In this paragraph we discussed the different categories that are comparable for Kickstarter and Indiegogo. The categories Design, Film&Video, Games and Technology are categories that have the highest raised funding per campaign for
18 fixed funding. For flexible funding campaigns the categories Design, Games and Technology raise the highest funding mean.
3.5.2 Verified non-profit organizations
Indiegogo allows campaigns to raise funds on behalf of a nonprofit organization . These campaigns will have a "Verified Nonprofit Campaign" badge placed on the campaign page to certify to contributors that funds will go directly to a verified nonprofit organization. (Indiegogo.com)
3.5.3 Starting years
The amount of campaigns for Indiegogo and Kickstarter are shown in table 8 for the year the campaign started. The data for Kickstarter are recorded until 09-2014 and for Indiegogo until 01-2014. This explains why there is only one observation in 2014 for Indiegogo and a smaller number of observations for Kickstarter compared to Kickstarte rs earlier years. Growth has continued for both parties since the founding in 2008 for Indiegogo and 2009 for Kickstarter. As shown in the table, Indiegogo’s growth was largest in 2012 (12.110
additional campaigns) and Kickstarter ’s growth was largest (15. 290 additional campaigns) in 2011. T a b l e 8 C a mp a i g n s s t a r t e d p e r y e a r Y e a r ( s t a r t ) I n d i e g o g o K i c k s t a r t e r 2 0 0 9 4 9 1 2 8 1 2 0 1 0 1 1 9 3 1 0 2 6 0 2 0 1 1 5 8 4 1 2 5 5 5 0 2 0 1 2 1 7 9 5 1 4 0 0 1 9 2 0 1 3 1 9 1 6 5 4 4 0 7 9 2 0 1 4 1 3 7 2 6 1 T o t a l 4 4 2 0 0 1 5 8 4 5 0 3.5.4 Countries
The four countries with the largest amount of raised funds are the large native English speaking nations: United States, United Kingdom, Canada and Australia. The United States is home to most projects for both fixed and flexible funding.
T a b l e 9 R a i s e d f u n d i n g p e r c o u n t r y C o u n t r y F i xe d F u n d i n g F l e x i b l e F u n d i n g U ni t e d S t a t e s 8 6 . 5 % ( 1 3 8 6 8 7 ) 7 3 . 3 % ( 3 0 9 0 5 ) U ni t e d K i n gd o m 1 0 . 0 % ( 9 9 7 9 ) 3 . 9 % ( 1 6 3 9 ) C a n a d a 3 . 6 % ( 3 6 2 0 ) 9 . 6 % ( 4 0 2 8 ) A u s t r a l i a 1 . 5 % ( 1 4 6 4 ) 1 . 7 % ( 7 1 3 ) O t he r 4 . 4 % ( 6 6 4 6 ) 1 1 . 6 % ( 4 9 6 9 ) T o t a l 1 6 0 3 9 6 4 2 2 5 4
19 There are more Canadian projects on Indiegogo than on Kickstarter. The ‘big four ’ countries have 191,035 campaigns and campaigns from the remaining countries total 11,615 campaigns.
3.6 Additional considerations
In this paragraph we discuss additional considerations for variable comparison over Kickstarter and Indiegogo. Firstly, we discuss additional information on Indiegogo. Secondly, discuss the comparability of the amount of Facebook friends. And finally we discuss the conversion of the currencies used on both platforms.
Not every category of data is available on both platforms. Information about the project initiator is only available on Indiegogo. This data includes the amount of projects the initiator starte d and the amount of projects the initiator backed publicly.
Another consideration is the Amount of Facebook friends of the founder. Indiegogo allows for multiple founders to be shown whereas Kickstarter allows for only a single person to be the founder. F or the purpose of recording the number of Facebook friends of the founder, we have therefore only used the number of friends of the founder with the most Facebook friends.
The currencies (AUD, CAD, NZD, GBP, EUR, USD) are converted to US dollars using the monthly exchange rate at the end date of the project. Oanda average monthly exchange rates are used. The average of the bid and the ask rate is us ed with an interbank rate of 0% 2.
3.7 Empirical challenges
In this section we discuss empirical challenges to s tudying online crowdfunding decisions. Firstly, in paragraph 3.7.1, we discuss endogeneity. Secondly, we discuss challenges regarding the funding goal. Thirdly, we discuss truncation, extremely successful campaigns and right skewedness of the data.
3.7.1 Endogeneity
In this section we discuss whether variables in this study are considered to be endogenous.
There are empirical challenges to studying online crowdfunding decisions. A problem that occurs is endogeneity. A regressor is defined as endogenous if it is correlated with the error term. In general, if any of the regressors is endogenous all estimates become inconsistent (Cameron & Trivedi, 2005, p.92). Table 10 shows whether variables in this study are considered to be endogenous.
2
20 T a b l e 1 0 E n d o g e n o u s a n d e x o g e n o u s v a r i a b l e s D e p e n d e n t v a r i a b l e C o m me n t R a i s e d f u nd i n g C o nt i n uo u s F u nd i n g r a t i o C o nt i n uo u s I s s uc c e s s f u l D i s c r e t e S e l e c t i o n D V P l a t fo r m c ho i c e E nd o g e no us f u nd i n g t yp e c ho i c e E nd o g e no us E n d o g e n o u s r e g r e s s o r s # B a c ke r s E v e r yt h i n g t h a t e xp l a i n s # b a c ke r s a l s o e xp l a i ns r a i s e d f u nd i n g F u nd i n g go a l E xo g e n o u s r e g r e s s o r s # U p d a t e s # I ma ge s # F a c e b o o k fr i e nd s fo u n d e r s # U s e r c o m me n t s i n f l ue n c e s me a n i n v e s t m e n t S t a r t ye a r 6 d u m mi e s C a t e go r y 1 5 d u m mi e s , i n fl u e nc e s me a n i n v e s t me n t C o u nt r i e s 5 d u m mi e s V e r f i e d no n -p r o f i t d u m m y I n s t r u me n t s f o r # b a c k e r s # U p d a t e s E xo g e no us ( i n f l ue nc e s a mo u n t o f b a c ke r s ) # I ma ge s E xo g e no us ( i n f l ue nc e s a mo u n t o f b a c ke r s ) # F a c e b o o k fr i e nd s fo u n d e r s E xo g e no us ( i n f l ue nc e s a mo u n t o f b a c ke r s )
Firstly, a founder decision to utilize fixed or flexible funding depends on their platform choice. Founders are only able to utilize flexible funding by choosing platform Indiegogo. Therefor e funding type is considered to be endogenous. Secondly, a founder decision to utilize platform Indiegogo or Kickstarter is also considered to be endogenous, as Kickstarter allows campaigns from only a few countries. Suppose the independent variable fixed funding attracts ‘better ’ campaigns and raises hi gher funding for campaigns than flexible funding, ceteris paribus. This correlation between good campaigns and fixed funding gives a biased upward effect for fixed funding on funding.
Thirdly, amount of backers is considered to be endogenous. This will be further discusses in section 7.1.
Fourthly, the amount of updates, the amount of images and the amount of Facebook friends of the founders are considered to be exogenous as well as suitable instruments for the amount of backers. As is shown in previous r esearch in section 2.1, the amount of updates and the amount of Facebook friends have a positive effect on amount of backers (Agrawal et al., 2011) This is also the case for amount of images.
21 Fifthly, funding goal is considered to be endogenous. Utilizing fixed funding is a way of signaling to the backers that the project founders are more committed to make the project successful, as they commit to not start the project if they do not raise the funding goal. However, the funding goal is not a predictor fo r a successful campaign, which is also shown in other crowdfunding literature (Belleflamme et al., 2013, 2014; Mollick and Kuppuswamy, 2014).
Furthermore, the category dummies are also considered to be exogenous, although Kickstarter does not allow all th e categories that Indiegogo offers. Finally, we consider properties of the campaign founders, e.g. their country, their amount of Facebook friends, raising for a verified non -profit and the start year of the campaign to be exogenous.
To conclude, we consider the amount of backers and the funding goal to be endogenous. The funding goal also brings other empirical challenges that are discussed in section 3.7.2.
3.7.2 Funding goal
In this section we discuss additional challenges to studying the funding goal. The distribution of funding goal is different for fixed and flexible funding. Indiegogo offers fixed and flexible funding goals and requires the funding goal to be at least 500 money units (USD, Euro, Australian dollar, GBP or New -Zealand dollar). Opposed to Indiegogo, Kickstarter offers fixed funding exclusively and has no minimum funding goal.
From table 11 we observe that 1,519 Indiegogo funding goals are set equal to 500 money units, making up 3.43 % of the Indiegogo sample. On Kickstarter 11,298 Kickstarter goals are set lower or equal to 500 USD, 1,968 goals are set equal to or lower than 100 USD and 80 campaigns have goals equal or lower than 1 USD3. 3 I t i s a r g ua b l e t ha t K i c k s t a r t e r f i xe d f u nd i n g c a mp a i g n s wi t h f u nd i n g g o a l s b e l o w 5 0 0 m o n e y u ni t s a r e s i mi l a r t o fl e xi b l e f u nd i n g c a mp a i g n s , a s t he y i n s ur e a r e a c he d f u nd i n g go a l . A n o p t i o n i s t o c o n s i d e r f i xe d f u nd i n g c a mp a i g n s o n K i c ks t a r t e r wi t h f u n d i n g go a l s b e l o w 5 0 0 USD to utilize flexib le fund ing. Ho wever , o nly 58 % o f these ‘flexib le’ fund ing campaign pled ge their fund ing goal. T herefor e we chose no t to take this ‘flexib le funding’ at K i c k s t a r t e r i nt o a c c o u n t.
22 T a b l e 1 1 D i f f e r e n c e s i n f u n d i n g g o a l a t K i c k s t a r t e r a n d I n d i e g o g o F u n d i n g g o a l # K i c k s t a r t e r C a mp a i g n s # I n d i e g o g o C a mp a i g n s ≤ max 1 5 8 4 5 0 ( 1 0 0 % ) 4 4 3 2 9 ( 1 0 0 % ) ≤500 1 1 2 9 8 ( 7 . 1 3 % ) 1 5 1 9 ( 3 . 4 3 % ) ≤100 1 9 6 8 ( 1 . 2 4 % ) - ≤1 8 0 ( 0 . 0 5 % ) -
Besides founders that set (too) low funding goals for their campaigns, there also appears to be a challenge on the platforms side with campaigns th at raise a (too) low amount of funding. This is discussed in section 3.7. 3 below.
3.7.3 Truncated data for Indiegogo and Kickstarter
In this section we discuss truncation at Kickstarter and Indiegogo. Although both platforms do not report removing campaigns tha t raise no funding, our data suggests otherwise.
Our data suggests that Indiegogo removes projects that raise less than 500 money units. With the scraper the urls of all projects were copied in A ugust 2014, whether they were finished or still live. When w e scraped additional information in September 2014 using these urls, we found that some urls did not seem to be working. After doing some research we found that these urls did not raise the minimum of 500 USD. This data is truncated rather than censored, a s there is no information in the dataset. In total there are 16,155 campaigns with no backers, making up 8% of the our data. Besides having no backers, these campaigns have not raised funding and have a funding ratio equal to 0. The mean investment per backer is not interpretable.
T a b l e 1 2 N o n - f u n d e d a n d l o w f u n d e d c a mp a i g n s Y e a r K i c k s t a r t e r , f u n d i n g = 0 K i c k s t a r t e r , funding≤500 O b s I n d i e g o g o , f u n d i n g = 0 I n d i e g o g o , funding≤500 O b s 2 0 0 9 1 5 . 2 % 4 6 . 4 % 1 2 8 1 0 % 6 . 1 % 4 9 2 0 1 0 1 4 . 2 % 4 2 . 3 % 1 0 2 6 0 0 % 2 . 7 % 1 , 1 9 3 2 0 1 1 1 0 . 9 % 3 8 . 3 % 2 5 5 5 0 0 % 1 . 4 % 5 8 4 1 2 0 1 2 9 . 0 % 3 7 . 6 % 4 0 0 1 9 0 % 1 . 1 % 1 7 9 5 1 2 0 1 3 6 . 5 % 3 3 . 8 % 4 4 0 7 9 2 . 5 % 7 . 2 % 1 9 1 6 5 2 0 1 4 1 2 . 9 % 4 7 . 9 % 3 7 2 6 1 0 % 0 % 1
From table 12 it appears that both Kickstarter and Indiegogo remove campaigns that raise no funding. Firstly, looking at Kickstarter ’s percentages we find that the percentages of campaigns that raise no funding are between 6.5% and 15.2%. These percentages decrease from 2009 to 2013. However, the number of campaigns are increasing in number over these years. It appears that Kickstarter
23 removes a share of campaigns that raise no funding, and has increasingly done so over the years. We base this argument on the ‘rise’ of non -funded campaigns in 2014. In 2014 the percentage of campaigns with no funding is equal to 12.9%. This could be an indication that non -funded campaigns in 2014 are not yet removed. Comparing Kickstarter ’s percentages to Indiegogo’s, we argue that it is very likely that Indiegogo removes campaigns that raise no funding. On Kickstarter there are still campaigns that appear to raise no funding, however between 2009 and 2012, there appears not one campaign in our dataset that that raised no funding on Indiegogo. In 2013, 2.5% of campaigns raised no funding, however when we checked 50 of these urls randomly, we found that all of them were removed. As we only recorded one Indiegogo campaign in 2014, we do not take this year into account for Indiegogo. On Indiegogo it also appears that campaigns that raise below 500 money units are removed. An explanation could be that Indiegogo does not allow a funding goal below 500 money units, thus choosing to remove campaigns that raise less. When we compare Indiegogo’s percentages to those of Kickstarter we find that Kickstarte r has percentages between 33.8% and 47.9% for campaigns that raise less than 500 USD and Indiegogo has percentages between 7.2% and 1.1%.
To conclude, it appears that a number of campaigns that raise less than 500 money units to no funding are missing fro m the data. However, there are still campaigns in our dataset that have not raised funding. These campaigns and their properties are discussed in section 3.7.3.1 below.
3.7.3.1 Unfunded projects
The amount of backers and therefore also the funding result, are zero for 8% of the sample. Campaigns that raise no funding, naturally have no backers. The amount of updates, the amount of images and the amount of user comments are significantly lower for non -funded campaigns than for funded campaigns as is shown in table 13. T a b l e 1 3 U n f u n d e d c a mp a i g n p r o p e r t i e s V a r i a b l e u n f u n d e d c a mp a i g n s F u n d e d c a mp a i g n s # B a c ke r s 0 . 0 0 ( 0 . 0 0 ) 1 0 1 . 1 1 ( 7 8 3 . 4 4 ) # U p d a t e s 0 . 2 1 ( 3 . 5 3 ) 5 . 3 2 ( 9 . 1 0 ) # I ma ge s 0 . 9 2 ( 3 . 0 9 ) 3 . 5 3 ( 7 . 5 3 ) # U s e r c o m me n t s 0 . 0 5 ( 0 . 3 1 ) 3 5 . 8 3 ( 1 1 2 3 . 3 3 ) T he me a n i s r e p o r t e d wi t h t he s t a nd a r d d e v i a t i o n i n p a r e nt he s i s
24 As we are interested in how to crowdfund successfully, we argue that these observations are not relevant for our analysis. Therefore , we remove unfunded campaigns, resulting in 186492 observations: 142764 Kickstarter campaigns, 41810 Indiegogo campaigns that use flexible funding and 1918 Indiegogo campaigns that use fixed funding. Removing the unfunded campaigns is also a way to deal with the truncation of non -funded campaigns.4
Besides campaigns that raise no funding, there are also campaigns that raise a very high amount of funding. These campaigns are discussed below in section 3.7.4
3.7.4 Extremely successful projects
In this section we discuss extremely successful c ampaigns that raise more than a million USD. In total 91 campaigns raised more than a million USD, 82 on
Kickstarter and 9 on Indiegogo. 7 campaigns raised more than 5 million USD and 3 campaigns raised more than 10 million USD.
RyanGrepper, coolest cooler: 21s t century cooler that’s actually cooler – raised over 13.2 million USD Kickstarter
Ubuntu edge
– raised over 12.8 million USD on Indiegogo, this project is still live and therefore open for new backer contributions.
Pebble e-paper watch for iPhone and Android – raised over 10.3 million USD on Kickstarter
These extremely successful campaigns could be indication for a right -skewed dependent variable. In section we 3.7.5 we research if the amount of funding is right-skewed.
3.7.5 Right-skewed variables
In this section we di scuss if amount of funding and amount of backers are right -skewed.
The variables raised funding and the amount of backers are right skewed as is displayed in the table below. In our crowdfunding data, there is a lot of variance in raised funding. The amount of USD a campaign pledges is the product of the
4 A s ui t a b l e ma n n e r t o ha nd l e t r u nc a t i o n i s a T o b i t mo d e l t h a t t a ke s t r u nc a t i o n i nt o a c c o u nt .
An exa mple is Heckman ’s two -step estimator, wher e the first step co nsists o f a prob it r e g r e s s i o n o f b i na r y v a r i a b l e d wh i c h i s o ne i f for campaigns that are censored if no f u nd i n g i s r a i s e d ( o r 5 for campaigns that are censored if raised funding is below 500 mo ne y u n i t s ) . T he s e c o n d s t e p c o n s i s t s o f a n O L S e s t i ma t i o n o f t h e t r u n c a t e d d a t a wh e r e a n e s t i ma t i o n o f t h e p r o b i t mo d e l i s s ub s t i t ut e d ( C a me r o n & T r i ve d i , 2 0 0 5 , p . 5 4 3 ) . H o we v e r , fo r t hi s t he s i s we l i mi t t he s c o p e o f r e s e a r c h t o p r o b l e ms c o n c e r n i n g s a mp l e s e l e c t i o n a nd e n d o g e ne i t y.
25 amount of backers that invest in the campaign and the average amount of USD they invest. As in crowdfunding there is a large amount of backers investing a small amount, we argue that this large variance is due to a large variance in the amount of backers. Therefore, we assume that the right -skewedness of raised funding can be explained by the right -skewedness of the amount of backers.
T a b l e 1 4 D e s c r i p t i v e s t a t i s t i c s o f f u n d i n g a n d l o g ( f u n d i n g ) V a r i a b l e F u n d i n g L o g ( F u n d i n g ) # b a c k e r s L o g ( # b a c k e r s ) M e a n 7 2 2 3 6 . 5 9 3 2 . 9 M e d i a n 1 2 9 5 7 . 2 2 2 3 . 1 S ke wn e s s 9 6 . 2 -1 . 1 6 6 . 6 -0 . 1 8 K ur t o s i s 1 3 8 8 8 3 . 8 2 6 5 8 7 2 . 9 4
For the variables funding result and the amount of backers, we take two complications into account. Firstly, funding result is right-skewed with a mean of 7223 USD that is much larger than the median of 1 295 USD. Applying a logarithmic scale to funding result reduces the skewness statistic from 96.2 to -0.7 and transforms the kurtosis from 13888 to 3.86, close to the normal value of 3. The mean is transformed to 7.1 which is closer to the median of 7.4. Secondly, the amount of backers is also right -skewed with a mean of 93 that is larger than the median of 22. Applying a logarithmic scale to the amount of backers reduces the skewness statistic from 66.6 to -0.18 and transforms the kurtosis from 6587 to 2.94, closer to the normal value of 3. The mean is transformed to 6.5 which is closer to the median of 3.1.
We argue that using a lo garithmic transformation on raised funding and amount of backers is preferable. Firstly, using a logarithmic scale reduces heterogeneit y of the error terms. Secondly, compressing the means and the standard deviations fits our crowdfunding data well. We argue that the differences between a campaign that raises 10. 000 USD and a campaign that raises 30. 000 USD, are similar to the differences between a campaign that raises 100.000 USD and a campaign that raises 300.000 USD.
To conclude, the raised funding and the amount of backers are considered to be right-skewed. We argue that using a logarithmic transformation is suitable.
26
4 Binary choice models for sample selection
In this section we discuss binary choice models that take sample selection into account. Firstly, we discuss the sample selection bias in our data and the
different subgroups. Secondly, we specify a bivariate probit model that takes sample selection into account. Thirdly, we discuss the exclusion restriction. Fourthly, we specify separate probit models for funding type choice and platform choice.
The crowdfunding campaigns in our data are divided into two subgroups using either fixed or flexible funding. These subgroups do not emerge randomly, but emerge as the result of an underlying selection process. There are two selection problems that occur. Firstly, sample selection : The funding result for fixed
funding is only observed if fixed funding was used, Kickstarter only offers fixed funding and Kickstarter did not allow countries from outside the US from 2009 until 2012. Secondl y, s elf-sel ection : Founders choose the plat form Indiegogo or Kickstarter and on Indiegogo choose fixed or flexible funding. Sample selection and self-selection have similar estimation problems (Cameron & Trivedi, 2005, p. 546) and therefore we focus on sample selection
For sample selection models evaluation problems occur as counterfactual situation is unobserved. For instance it is unobserved how much a flexible funding campaign would pledge if it used flexible funding. Th e analysis of only one subgroup is not sufficient and different econometric methods should be used. An appropriate evaluation method is selection models, as they take the subgroups structure of population sensitivity into account.
For sample selection mod els consistent estimation relies on relatively strong distributional assumptions. This even holds for semi parametric estimation. Cameron and Trivedi (2005) mention experimental data studies as an alternative. Experimental data studies use random assignmen t, avoiding selection problems. However, these models are difficult to implement.
Another alternative are bivariate selection models . In section 4.1 we discuss a bivariate probit model to research the relation between funding type choice and platform type choice. Firstly, we describe a bivariate probit model that takes sample selection into account. Secondly, we discuss the exclusion restriction and specify relevant variables for funding type choice and platform choice in section 4.2.
4.1 Bivariate probit with sample selection
In this section we describe a bivariate probit model for funding type choice and platform choice. To test if a two separate probit models or a bivariate probit model is a better fit, we consider the joint model with two dependent variable s
27 platform choice and funding type ( , ). Campaign i is = 1 for Indiegogo and = for Kickstarter, and = 1 for flexible funding and = for fixed funding. Than for campaign i we define
𝑝 = Pr[ = 𝑗, = 𝑘] , 𝑗 = ,1, 𝑘 = ,1 (1)
In theory 𝑝 defines four mutually exclusive events, but as Kickstarter does not allow for flexible funding, 𝑝 = . Therefore, we have three mutually exclusive events with probabilities 𝑝 , 𝑝 and 𝑝 and ∑ ∑ 𝑝 = 1. We define four binary indicator variables = 1 if ( = 𝑗, = 𝑘 ) and = otherwise. Then the joint density for campaign i is
𝑓( , ) = ∏ ∏ 𝑝
(2)
The log-likelihood of this equation is
∑ ∑ ∑
ln (
𝑝 ) (3)
To take sample selection into account we describe a bivariate probit model also known as Heckprobit, from Cameron and Trivedi (2005, p. 547). This model consists of 2 equations: the participation equation and the resultant outcome equation. In our case the participation equation describes platform type choice and the outcome equation describes the fu nding type choice.
Participation equation
= {1 𝑖𝑓 𝑖𝑓 ≤ (4)
And resultant outcome equation (funding type choice)
= {− 𝑖𝑓 𝑖𝑓 ≤ (5)
indicates platform choice with = 1 for Indiegogo and = for Kickstarter. indicates the funding type choice which is equal to for Indiegogo. As Kickstarter offers no choice , does not take any meaningful value for Kickstarter. The underlying model is linear with additive errors.
