Master Thesis
Your name: Pauline Dekker
Your student number: 6285724/10001974
Specialization: Organizational Economics
Field: Crowdfunding and the beauty premium
Number of credits thesis: 15 ECT
Title of your research: The possible effect of appearance on success rates in funding a project through Kickstarter.
The thesis coordinator will assign a teacher to supervise your thesis. Assigned supervisor (to be filled in by thesis coordinator):
Jeroen van de Ven
If a teacher has already accepted to supervise your thesis, please provide the name. Name of supervisor:
2 Summary
Online crowdfunding has recently become a popular way of financing your innovative ideas. Although some articles have reviewed several aspects that might influence your chance of getting funded, hardly any literature exists on the influence of gender and appearance on your funding success. Therefore, surveys were used in this research to gain data on
appearance, and for this purpose 100 Kickstarter projects were used to review the effect of gender and appearance differences on the performance of idea founders. It appeared that men were not more likely to get their ideas funded then women; however some interesting differences occurred between men and women. And though the appearance categories had their predicted signs, attractiveness only had a slight negative influence if the founder was in the average category.
Statement own work
I hereby declare, Pauline Dekker, I wrote this thesis myself and I take full responsibility for its contents.
I confirm that the text and the work presented in this thesis are original and that I did not make use of sources other than those mentioned in the text and in the reference.
The Faculty of Economics and Business is solely responsible for the guidance to submitting the thesis, not the content.
3 Table of contents Summary ……….……… 2 1. Introduction………. 4 2. Theoretical framework ……… 5 2.1 Beauty premium ……… 5 2.2 Gender differences ………. 9 2.3 Crowdfunding ……….. 12 3. Methodology ………..… 15 3.1 Kickstarter ………..….. 15 3.2 Hypotheses ……… 17 3.3 Project data ………..………… 18
3.4 Survey and treatments ………..…..… 18
3.5 Raters ………..……. 20
4. Results ……….….… 20
4.1 Descriptive statistics projects ………..… 21
4.2 Descriptive statistics survey ………..…… 22
4.3 Appearance ………... 24
4.4 Gender differences ………..….. 28
4.5 Regression analysis projects ……….… 31
4.6 Regression analysis survey ……….… 33
4.7 Possible limitations ………...…. 35
5. Conclusion ……….…… 36
6. References ……….……… 37
4 1. Introduction & Motivation
Recently it has become increasingly popular that entrepreneurs turn to crowdfunding as a way to finance their creative ideas. Crowdfunding involves relatively small contributions from a relatively large number of customer-investors using the internet over a fixed time period, usually a few weeks, and generally no longer than a month (Kuppuswamy & Bayus, 2013). Since crowdfunding is a relative new phenomenon, literature on this specific topic is still very incomplete. Previous articles on online crowdfunding particularly focused on which platforms are most popular and what characteristics are important to become successful (Antonenko, Lee & Kleinheksel, 2014); what role social information plays (Kuppuswamy & Bayus, 2013), and the underlying dynamics of success and failure regarding product features and founders’ social network (Mollick, 2014).
Something that is quite striking is that none of these previous articles took into consideration the physical characteristics of the founders. While in previous literature it has been found that appearance influences earnings in the labor market (Biddle and
Hamermesh, 1994) and can benefit the company one works for (Biddle, Bosman,
Hamermesh & Pfann, 2000). It therefore seems obvious there is room for research that takes into consideration this positive effect your appearance can have, called ‘the
beauty-premium’, when analyzing factors that influence online funding success.
Whereas appearance is related to gender, this general characteristic also deserves some attention. Although there is a lot of literature on gender differences, there is only one article (Marom, Robb & Sade, 2014) that reviews the possible effect of gender on funding a project through online crowdfunding as far as this research is concerned. The authors examined gender dynamics and biases in the process of raising funds for new projects via crowdfunding on Kickstarter, and find that women enjoy higher success rates.
In this research the focus will therefore be on the effects gender and appearance can have on others’ decision to fund your project on Kickstarter. More specifically, under some circumstances it is conjectured that people will let their investment decision depend on gender and appearance of the presenter, and choose the ideas of the prettier
entrepreneurs. For this purpose 100 Kickstarter projects are analyzed and rated through surveys. Subjects were asked to rate both the project and the founder, and two different treatments were used to explore the effect of appearance on the raters opinion; one
5 showing the picture of the founder before asking any question and one treatment with pictures after the questions. This produces a rich data set with both characteristics of the projects together with gender and appearance ratings. Based on the above, the research question is formulated as follows: “Does your appearance and gender significantly increase your chance of being funded on Kickstarter?”
The possible existence of the beauty premium could be very important to aspiring entrepreneurs for the reason that if the effect exists, one can increase their chance of success by spending more time on their appearance rather than on their business idea . Besides implications for the use of Kickstarter, these findings can be applied to a broader area, your appearance might for example also influence other job related opportunities. At first sight there also appeared to be a beauty premium in this thesis, but after further analysis the premium was not found to exist and thus the null hypothesis that founders in all appearance categories are as likely to get their ideas funded cannot be rejected. Though men did perform slightly better, there was also no proof for men to be more likely to get funded on Kickstarter than women.
The remainder of this thesis is organized as follows. Section 2 reviews the existing related literature on gender differences, the beauty premium and crowdfunding; section 3 describes the subjects, procedure, treatments and hypotheses. In section 4 the results of the data analysis are presented and discussed. Section 5 will conclude and in the appendix an example survey and some tables with additional information can be found.
2. Theoretical framework
In this section all available relevant information on the beauty premium, gender differences, and crowdfunding will be reviewed.
2.1 The beauty premium
The beauty premium has been demonstrated in many articles and is already reviewed in the eighties by Hatfield and Sprecher (1986) in their book on physical attractiveness. Thereafter the existence of the ‘beauty premium’ is still widely discussed in previous literature and has proven to exist many times.
6 favoritism for the beautiful and the possible discrimination in the labor market against the ugly. The authors used ratings of physical appearance in combination with three data sets. This gives the authors the great advantage of combining three sets of data into a dataset with sufficient sample size to make clear inferences on the effect of beauty. The findings suggested that more attractive people are indeed paid more; however the penalty for looking bad is larger than the premium for looking good. The bad look penalty is 9% in earnings among men and the good look premium is estimated to be 5% in earnings. And for women this premium was found to be 4% on average, compared to a 5% bad look penalty, although both percentages were not significant for women. Besides these effects, there was slight evidence that the labor market sorts the best-looking people into jobs where their looks are most productive. Though this evidence is fairly weak, it is slightly disadvantageous for this research that the alternative sources of earning differentials are hard to disentangle. The authors concluded that better-looking people earn more than bad-looking people and that these penalties and premiums reflect beauty effects in all its aspects.
Belot, Bhaskar and Ven (2012) also addressed that appearance can be in your favor. The authors analyzed the beauty premium in their article on a TV game show with high stakes. In the show contestants had to answer questions and after each round the best player could eliminate another player. While the authors used a group of participants with a variety of occupations who were not recruited by any specific criteria, the participants may not be representative of the population, causing possible external validity issues. The
authors found that less attractive people are substantially more likely to be eliminated, even though they do not perform any worse or do even better. Only 27% of them made it to the final round, while the most attractive players made it to the final round in 49% of the cases. This difference cannot be attributed to a difference in performance or cooperation. It was also found that the elimination of less attractive players is costly, on average 25% of the median stake, while there is no financial benefit. Compared to the less attractive people, attractive players earned a premium because they were not eliminated by their fellow players. The authors therefore concluded that the beauty premium is a form of taste based discrimination.
Besides the fact that your beauty can be in your favor, it has also been proven that your appearance can be beneficial to the company you work for. In their research among Dutch advertising firms, Biddle et al. (2000) found that firms with better-looking executives
7 have higher revenues. With this research, the authors are one of the few to focus on to what extent the associated differences in beauty are due to differences in ability or to
discrimination. This is particularly interesting for this research because it shows company related benefits of appearance which could be of importance for start-ups. The executives’ beauty was measured by rating black and white yearbook pictures. The use of these black and white pictures can be seen as a disadvantage since hair color and eye color are not visible in this case, but are proven to affect appearance ratings (Belot et al., 2012). Besides that, the use of estimates from other studies on earnings instead of the real earnings of the subjects is decreasing the validity of this research. The effect found exceeded the effect of the executives’ beauty on their own possible earnings increase, 63% at most is reaped by the executives in the form of higher wages. This implies that beauty creates firm-specific
investments where the returns are shared by the executive and the firm.
Likewise in daily life, the premium has also been found in experimental settings several times. These laboratory experiments are particularly useful because they can be designed to disentangle the different sources of the beauty premium. In experimental settings one can actually rule out other sources besides your appearance in influencing your topic of research, therefore the main experimental findings on the beauty premium are presented hereafter.
Mulford, Orbell, Shatto and Stockard (1998) used the prisoner’s dilemma and the role of perceived physical attractiveness to find out whether the beauty premium had an effect in everyday exchange. This article was the first to also measure perceptions of subjects’ own physical attractiveness besides the perception of others. One drawback of their study implies that their results are based on only six interactions, which is not really representative for the real world. Also, in real life the settings are much more complicated than the dilemma used for this process. Mulford et al. (1998) results showed that subjects cooperated and played more often if they found the other attractive. Only 28% cooperated when the other was perceived unattractive compared to 48% when the other was perceived attractive. It also appeared that subject’s self-assessment played a role; subjects who rated themselves low also had a low expectation of cooperation from others in 60% of the cases, while this was only 24% if the subject saw himself as attractive. The authors concluded by mentioning that it is more profitable for those who are seen as attractive to involve in ‘everyday exchange’. However, Shinada and Yamagishi (2014) found that the negative relationship
8 between cooperation and physical attractiveness in a prisoners dilemma only showed up with young men, but not once they are older or among women.
Another experimental study was presented by Mobius and Rosenblat (2006), who used students in an experimental labor market as ‘employers’ and ‘workers’. Workers had to perform a simple task that required true skills, which were shown to be independent of physical attractiveness. Employers on their turn had to estimate productivity and set wages. Even though it was about true skills, beauty raised both the workers and the employers’ productivity estimates. In their research the authors used the interview process, while the beauty premium might have different implications and causes on the long term after interacting repeatedly. As the authors mentioned, only the interview process was used for this research while in the long term tasted-based discrimination might again influence the beauty premium, as other contributors might decrease. The results indicated a beauty premium identified through three transmission channels; prettier people were more confident and their confidence increased wages; given the level of confidence, employers viewed attractive workers as more able; and for a given confidence level the more beautiful people had oral skills that increased their wages when interacting with the employer. Solnick and Schweitzer (1999) on the other hand examined the effect of physical attractiveness in an ultimatum game experiment but found no significant influence of appearance and gender on the game decisions. Even though the more attractive persons and men were offered more in the game, also more was demanded from them.
In an article by Andreoni and Petrie (2008) it was examined whether beauty and gender mattered in a public goods experiment. Players were shown digital photos of all other members of their group each round, one with and one without individual
contributions. Based on the game, the authors found a beauty premium; however this premium was found to disappear once information on individual contributions was provided. In case the individual contributions were not observable, attractive people made more money than unattractive people, even though there was no reason for, based on their willingness to cooperate. Therefore the authors were not completely able to draw
conclusions and the implication of their findings stays vague. Besides that, a gender effect was observed, however not always favoring men. Women earned more money when only group contributions were known, but in case individual contributions were known, men earned 15% more compared to women. The differences between men and women will be
9 further discussed in the next paragraph.
Though there exists substantial literature on the beauty premium, it is still a new subject where many aspects yet have to be explored. Most previous literature is either experimental or work atmosphere related, while your appearance might have a much broader range in influencing your daily life.
2.2 Gender Differences
Dreber and Johannesson (2008) researched gender differences in a deception game where subjects were either the receiver or the sender; therewith they were the first to investigate this in an economic setting. Receivers had to choose between two actions, one that yielded them more money and one that yielded the other party more money without knowing the payoffs; senders knew these payoffs and had to send a true or a false message containing which option, A or B, earned the receiver more. Women appeared to lie significantly less than men to secure a monetary benefit, 55% of the male senders lied compared to 38% of the female senders. The authors also tested for the extent receivers trusted the message of the sender, perceived trustworthiness and whether it mattered if the message was sent by a male or a female, but found no significant differences here. Though the subjects were students and the study was experimental, the study found that men were more likely to lie, something that might have a big effect on their performance in daily life. Men might also lie to make their ideas on Kickstarter more appealing and thereby influence project supporters’ donations. From another study it also appeared that men are more selfish (Andreoni & Vesterlund, 2001), and less generous than females (Eckel & Grossman, 1998) which could also lead to performance differences between males and females.
In their experiment, Gneezy, Niederle and Rustichini (2003) examined whether gender differences existed in several different controlled experimental settings. This allowed them to exclude discrimination and expected discrimination. A small minus however was that the authors were not fully able to measure possible psychological effects, as they mentioned themselves. Subjects were given the task to resolve a maze and three payment methods were used; piece wages, a competitive mixed tournament payment and a random payment. There was no gender gap when participants were paid a piece wage, however the results of Gneezy, Niederle and Rustinchini (2003) did indicate that women were less
10 performance of the male participants significantly, but not the performance of females. The effect was found to be weaker in single-sex competition compared to the situation where women had to compete against men. The authors concluded this meant there is a
substantial impact on performance if tournament incentives are used. These results are particularly interesting since Kickstarter can be seen as a mixed-sex tournament competitive environment, meaning men should also be outperforming females in this research. Though the difference was not significant in this research, men did slightly outperform women.
Besides in strategic games, similar results were found in social dilemma games. For example Dodge, Van de Kragt and Stockard (1988) used experimental social dilemmas to examine sex differences in cooperation. Subjects were recruited by advertisements and undergraduate classes and had to choose between cooperating with the group or defecting and thus keeping all the benefits to themselves. In the first experiment participants were allowed to discuss the decision and more cooperation was observed, once discussion was no longer allowed the rate of defection increased. In both experiments females were slightly more cooperative than males. However women were more likely to justify their behavior as altruistic and more socially-oriented, whether or not they cooperated.
The results indicated that the experimental settings had much more influence on behavior than the gender of the participant. Therefore the authors concluded that sex differences in cooperation might be overstated and the conditions of the experiment have far more influence on behavior than the sex of the participant. Despite the fact that the authors used a relative old data set and only focused on cooperative behavior while the results might differ for other kinds of behavior, the results were quite interesting. This research is one of the few that took the effect of the experimental design into consideration and concluded that gender effects are overstated since the experimental design had much more influence. Likewise in this research it appeared that gender differences didn’t play as much of a role as other influences.
The results of Dodge et al. (1988) were later confirmed by Eckel and Grossman (2000), who reviewed the results found in literature on ultimatum, dictator and public goods experiments. The results presented by different authors on the differences between men and women in these games are not consistent, though some patterns emerged. In case risk is involved there were no significant differences between men and women. However once the participants were no longer exposed to risk, the systematic differences appeared.
11 Women’s choices appeared to be more socially-oriented compared to the more individually-oriented choices of men. All the results were conditioned by the level of risk and therefore the authors could conclude that differences in results between men and women were dependent on the payoff structure and the experimental procedure. With this unclear conclusion this article shows there is still room for a clarifying article on gender differences, especially in an environment like Kickstarter where people take the risk to set up a business and try to find financial support for it.
An attempt to summarize all findings on experimental games considering gender differences was presented by Croson and Gneezy (2009), who reviewed the experimental literature on gender differences in their article. Their main insight was that men are less risk averse than women, with exceptions for managers. Their explanations for the difference included overconfidence, framing and emotions. Second, the authors found that most literature agrees men and women differ in their social preferences but the literature on this topic varies. And third, a noteworthy emergence in literature found, was that men are more competitive in bargaining and competitive situations. The fact that men were found more competitive than women could indicate that men are more efficient in a competitive environment like Kickstarter.
A field experiment was conducted by Marom et al. (2014) who used Kickstarter to investigate gender dynamics in the funding process. The authors used a very rich data set with a total of about 25.000 projects, about the same amount of entrepreneurs, about a million investors and over 120 million dollars of investments. The main focus was on whether there exists a barrier for female entrepreneurs to raise capital. Results indicated that men quested for higher amounts of capital and raise more funds compared to women. However, even after controlling for goal amount and category, females experienced higher rates of success in funding their projects (69.5% versus 61.4%). Most projects that were led by females were also mainly financed by women and women make up 45% of the investors, a larger percentage than the amount of female project leaders. To investigate the underlying reasons why, the authors conducted a survey among Kickstarter investors and found some evidence for taste-based behavior. The authors concluded by mentioning there is some evidence towards increased participation of women in crowdfunding platforms but note that further research is necessary. This article is very relevant since it uses the Kickstarter website in combination with gender differences as one of the firsts. However the article leaves room
12 for research that does not focus on the female’s participation numbers in online
crowdfunding but more on the effect gender could have on their funding success. The topic of crowdfunding will be discussed in detail in the following paragraph.
All in all there is an incredible amount of literature on gender related issues available. Even though most of the previous presented literature is experimental, it is relevant for this research since it indicates gender differences in a laboratory environment where the
influences of other factors determining your outcome variable are minimized. Previous authors used substantial data sets and gender differences are measured in several different areas. However, not much has been published about gender in combination with
appearance yet. It may very well be that the appearance of females has different implications than the appearance of males.
2.3 Crowdfunding
Crowdfunding is a modern way of raising capital for a new venture idea and it is usually about small contributions by a lot of backers. As stated by Mollick (2014), considering its fast rise in popularity, the dynamics of crowdfunding have been unstudied for a large part. There are some articles before the year 2000 that mention the word crowdfunding but the recent trend of online crowdfunding platforms is largely undiscovered. Since there are hardly any articles on online crowdfunding platforms like Kickstarter in combination with the influence of certain characteristics of the founder, first some more general articles on the dynamics of successful crowdfunding are reviewed. This lack of literature already shows there is
considerable scope for a study like this.
One of the earlier works on using alternative sources to fund your venture idea is presented by Macmillan, Siegel and Subba Narasimha (1985) who reviewed criteria that are used by venture capitalists to judge a new venture idea. Experience, characteristics of the product/service, characteristics of the market, financial consideration and personality were measured using questionnaires filled out by venture capitalists form New York. Of course around the time this article was published, the internet did not play a substantial role in crowdfunding yet, so this research could complement in also involving the internet. The authors also only focused on venture capitalists and ignore other potential investors. The main finding from the article was that the entrepreneurs’ experience and personality are the most important criteria. Thus, the business plan is only to indicate the entrepreneur’s ability
13 to implement the idea but personality is the deciding factor for venture capitalist to fund a project.
These findings are more or less confirmed 24 years later by Gera, Goldfarb and Kirsch (2009) who explored venture capital opportunity screening. The authors analyzed the form and content of business plans in their sample of 722 applications by project founders. Their main finding stated that these documents and the content are only associated with venture capitalists funding decisions for a small part. The most critical evaluation used for
information was found to be obtained through alternative channels by the venture
capitalists. Although this paper only focused on venture capitalists and also not involved the internet, the authors did use a very large dataset and related actual decisions to
characteristics.
However, contradicting results were found by Kaplan, Sensoy and Stromberg (2009), who studied firm characteristics to review ventures from their early business plan till their initial public offering. The sample consisted of 50 firms, a rather small sample, that went public in an IPO and had an early business plan. Despite of their small data set, the authors believe their results support the non-human capital asset theory. The main result suggested that the business is more important than the management team. In the business plan, IPO and annual report the most important factor cited, by respectively 100%, 98% and 91% of the companies, was the belief that the product/service offered is unique. The authors
argued that the core business activities, customers and competitors tend to stay the same or broaden over time while the management team often changes in time. Even though an initial strong business might not be sufficient, an initial strong business plan is necessary for a company to succeed.
Although previous literature is not fully relevant on online crowdfunding, it
represents an interesting debate on whether the idea or the founder of the idea is the most important aspect in successful crowdfunding. This is also a very important question
considering online crowdfunding on Kickstarter, since the main aim of this research is to find out whether gender and appearance are significant determinants of your potential success. Since this was not proven to significantly be the case, there is a little more evidence towards the conclusion that it is the product that mainly determines the success in getting funded. The following literature is more focused on online crowdfunding like Kickstarter.
14 campaigns on educational technology startups in their article and thereby shed some light on some possible beneficial characteristics one could use for your funding idea to succeed. Even though the authors used educational startups, their result are still interesting since they used Kickstarter in their analysis. The authors used content and basic frequency analysis to find support for their hypotheses which platforms are most popular and what successful characteristics are used to become successful. The results indicated that requesting a reasonable amount for your project, communicating, and informing supporters, increases the chance to become a top performing educational project and most startups used RocktHub, Kickstarter or Indiegogo.
One of the few articles that focus mainly on Kickstarter is written by Kuppuswamy and Bayus (2013), who analyzed two years of data on projects available at the Kickstarter website to study the role of social information in backer funding. The authors showed that past backer support negatively affected new backers, because potential backers suppose that others will finance these projects since they already have a lot of support. Besides that, reduced diffusion of responsibility and an increase in project updates towards the deadline lead to an augmentation in backer support. Project creators thus benefit from posting public and private updates towards the end of their funding cycle. This study used a large dataset on Kickstarter projects and actually gave clear implications for project creators; however the authors paid little attention to the characteristics of the founder, like gender and
appearance.
In the most recent article on Kickstarter, Mollick (2014) tried to find the underlying dynamics of success and failure in crowdfunding. The author used the online crowdfunding platform Kickstarter and the dataset consisted of over 48.500 projects. The results indicated that the project quality and the personal network of the entrepreneur are associated with success, which was also emphasized by Belleflamme et al. (2014). An interesting observation was that geography also seemed to influence success rates but also the types of projects proposed. Results also indicated that entrepreneurs could improve their chance of success by showing preparedness and by using their social network. Furthermore setting a
reasonable goal prevents the founder from late delivery. Therefore, the main conclusion was that careful planning will increase your chance of being funded and the chance of a
successful execution of the project. Although Mollick (2014) used a substantial data set and took into account a lot of possible dynamics of success, also in this article the effect of your
15 gender and appearance is omitted while previous literature has proven it to be very
influential in other job and business related subjects.
Summarizing the subject of crowdfunding it is a relatively new subject in literature, though some articles have made fierce attempts to capture the ins and outs of
crowdfunding. Especially the effect of founder characteristics in crowdfunding through websites like Kickstarter is not studied sufficiently yet while these results could have great impact on the results you can achieve by using an online crowdfunding platform.
3. Methodology
In this section the data gathering process is described and explained. Some essential background information on Kickstarter is provided, the project data is described and the surveys used for this research are clarified. Besides that, the different treatments are illustrated and the hypotheses are appointed.
3.1 Kickstarter
Kickstarter is a relatively new online crowdfunding platform and an alternative source in funding project ideas. Since its launch in 2009, 8.6 million people have pledged more than $1.7 billion, funding 84.000 projects (Kickstarter, 2015). Kickstarter divides projects in fifteen different categories; art, comics, crafts, dance, design, fashion, film & video, food, games, journalism, music, photography, publishing, technology and theater. In order to fully
understand a typical Kickstarter page, the definitions of the key variables are summarized in table 3.1.
Table 3.1
Key variable Explanation
Project goal The amount founders aim to raise using
crowdfunding.
Funding level The percentage of the project’s goal that has already been raised.
Backers The number of people supporting the
16
Pledge The promise of backers to support the
project by a specific amount, only the total pledge is visible on Kickstarter, no individual pledges.
Campaign The campaign explains the project and
optionally shows pictures, videos and project progress.
Updates Founders can use the update page to update
their backers on their progress and for example new features.
Comments Here all project backers can comment on the
project and potential backers can ask the project creator questions.
Overfunding If projects raise more money than their
original goal specified.
Days to go The remaining time the project will be online and funders still have the possibility to support the project
From Kickstarter (2015) one can learn that anyone can launch a project, as long as the rules are followed. Founders can place their projects online in order to raise money for their project. Kickstarter uses an all-or-nothing system, meaning that the project will only be funded if its project goal is reached. In that case, the founder has to implement his idea and meet the agreements with the backers. If not, backers will get their money back from the creator and the project will not be launched. Usually, only projects that offer their product, service or something else in return for backer support get (over)funded. This just means that the project creator has to keep more promises to more backers, not that there is any money left for personal use. A 5% fee is collected by Kickstarter from the projects total funding, if successfully funded; there are no fees if the project was unsuccessfully funded. Creators stay the owners and Kickstarter or the backers never control the creators’ work (Kickstarter, 2015). There can be many reasons to invest in projects on Kickstarter; most people are supporting friends or projects they already admired for a while. Other people just think the
17 idea is good and support it because they want to see it exist in the real world. Another
simple motivation is the projects reward; some just want the product, an experience or something else promised in return for their pledge. These rewards can range from a thank you card to a private cruise party with a barbeque (Kickstarter, 2015A).
3.2 Hypotheses
Prior literature has demonstrated that men and women differ from each other in a lot of ways. Men are for example more efficient in competitive environments (Gneezy et al., 2003). Besides, men are more likely to lie (Dreber and Johannesson, 2008), more selfish (Andreoni & Vesterlund, 2001) and less generous (Eckel & Grossman, 1998). In this research we expect that men might perform better in the Kickstarter environment, since this a highly
competitive environment, and will therefore be more likely to get funded. Consequently, based on the previous literature and to support the research question, the following hypothesis is formulated:
Hypothesis 1: men are more likely to get their ideas funded on Kickstarter.
Besides gender, appearance has been proven to affect ones surroundings in previous literature. For example there is evidence that attractive people get paid more (Biddle & Hamermesh, 1994), that firms with better-looking executives have higher revenues (Biddle et al., 2000) and that prettier people are more confident which increases their wages (Mobius & Rosenblat, 2006). Therefore in this research prettier people are expected to be more likely to get funded through Kickstarter than average looking people and below average looking people. This leads to the following hypothesis:
Hypothesis 2: Prettier people have a bigger chance in getting funded through crowdfunding.
In this research, the main focus will therefore be on gender and beauty. However, if these variables have an effect on funding this might have different causes, for example because attractive people are seen as more creative. Hence, this research will take other possible explanations in consideration when designing the research and analyzing the data.
18 3.3 Project data
For this thesis 100 projects were selected from the Kickstarter website and four categories were used; technology, games, food and fashion. Two categories that might found to be more interesting by men, and two categories that are expected to be more feminine. The category selection was made in order to be able to make the data comparable between men and women, if fifteen categories were to be used, the projects might differ too much to draw conclusions. Kickstarter has a discover option on its website where all the life projects can be viewed, either by popularity, magic, ending date, newest or most funded. Considering the time limit to write about this topic, projects were sorted by ending date, because all projects have a given ending date this sorting is completely random. There is no reason to believe that certain projects would appear on top of the page for other reasons than their upcoming ending date. Projects were recorded in excel, writing down their category, gender of the founder, city of the founder, goal in dollars, whether the goal was reached, how many days this took, what the final percentage funded was and whether the project was picked by the Kickstarter staff. This last item was recorded because many of Kickstarter staff picked projects rise in their popularity and get seen more often, which might lead to an increase in successful funding. Due to the fact that some project goals were stated in a currency
different from the US dollar, all other curries were converted to dollars on 19-4-2015. The exchange rates at that date are presented in table 3.3, included in the appendix.
3.4 Survey & treatments
Besides the data from Kickstarter, data on appearance was necessary for this research. Since no such data is available yet and the purpose is to measure beauty as objective as possible, several opinions are needed to construct data on the appearance of the project founders. In order to obtain this data on the founder’s appearance, surveys were used to collect opinions on appearance, creativity, productivity and the founder’s idea1.
All surveys started with a short introduction explaining the questionnaire and why the help of the respondents would be appreciated, after that a short summary on Kickstarter was included to make sure everyone understood the concept of online crowdfunding.
Because possibly not everyone is familiar with the projects on Kickstarter, first an example
1
19 project was provided. After the short introduction and explanation the survey started with three general questions considering age, gender and whether the subject is a student or not. The answers to these questions will be used as control variables in the analysis. Perhaps older people are more generous in judging the project ideas or females might have different opinions on how creative projects are. Including these variables gives the option to correct for these differences.
After the general questions, four questions about the project followed. The first three questions were about whether the rater thought the project was a good idea and how long they thought it would take for the project to reach its goal. These three questions were designed to get an opinion of the rater on the quality of the project. Without opinions about the project one cannot find out the underlying reason why people rate some projects better than others. To find the underlying source the fourth question asked the raters to what extend they thought the idea was creative. This question in combination with the three questions on the founder of the idea can give an idea of the underlying reason why the rater did (not) like the project. The three questions about the founder measured attractiveness, productivity and ability. This gives the opportunity to check whether judgments on project quality were affected by its founders’ appearance. In order not to confuse the participants and to keep the data comparable, choice options were given for each of the questions about the projects and its founder.
For the survey, two treatments were used. In the first treatment subjects were shown projects with the accompanying founder and questions were asked on both the project and the founder of the project. In the second treatment subjects were first shown the projects and were asked to fill out some questions about them, there after they were shown the founders of the projects and were asked to answer some questions about them. These different treatments were used to measure the underlying reasons of the possible differences between the appearance and gender categories. If for example women are found to receive less funding, the survey allows one to check the reason why. Perhaps women are seen as less creative or productive or perhaps it is just discrimination. By
designing these two treatments, it becomes possible to also see whether the same projects receive different ratings depending on whether or not the rater knows the gender of the entrepreneur.
20 survey is included from treatment 1 in the appendix, because all surveys are similar though they are on different projects.
3.5 Raters
Every rater received a questionnaire containing some general questions and some project and founder related questions. They were all asked to rate 10 projects each. It would be perfect if all raters rated all the 100 projects because this would give more data points, but also more consistent data. However this was not an option considering the length of the questionnaire. The questionnaire already consisted of around 15 pages each when 10 projects were included. Considering that the subjects were all volunteers and had no financial incentives the questionnaire could not be too long. Besides that it would make the data less valuable if subjects get tired from answering questions and do not dedicate enough attention to each project. Rating the ten projects took on average around 10 and 15
minutes. To make sure the subjects would still take the time to finish the survey, 10 projects was perceived as the maximum amount of projects for one survey, since the only incentives the subjects had were personal.
Raters were recruited through a personal network and Facebook where no specific criteria were used. Raters differed in age from 16 till 55 and most raters were around the age of 20. Most of them are college, bachelor and master students from all different
specializations and thirty percent of the raters were male. Unfortunately the response rates were a bit low and people took long time to return the surveys, therefore eventually 6 ratings per project are available. Three of them saw the picture of the founder immediately after the project was presented and three of them were shown the pictures of the founders after they answered questions about the project.
4. Results
In this paragraph the data on the 100 Kickstarter projects and the answers the raters gave in the 60 surveys will be analyzed. The results are presented and will be used to test the hypotheses.
21 4.1 Descriptive statistics: projects
From table 4.1 it can be seen that the average goal the founders set for their project was $17.687,95, however this average amount varies from $5.838,84 (Fashion) to $44.092,32 (Technology) in the different project categories. Within the categories the goals specified range between $200 and $289.666. The average percentage of the goal reached was 143,20% taken over all projects, also the non-funded projects, with the highest average of 257,59% in the technology category and the lowest average of 81,33% in the fashion category.
The average percentage of projects that reached their goal was 52%, varying per project category. This is slightly higher than the average success rate of 44% reported on Kickstarter (2015), a reason for this could be that projects were sorted to make sure data was available on both successful and unsuccessful ideas. Approximately half of all the founders were male, with higher male ratios in the technology and game categories. This was expected, because technology and games are more popular amongst men. In the fashion and food category, more founders were women as expected. Since Kickstarter is an American website it seems no more than logical that the largest percentage of its users are American citizens, as we can see from the table the average non-American users is only 17%.
Table 4.1
Survey statistics by project category Average goal Average % goal reached % of projects that reached their goal Percent -age of males Percentage of team Percentage of non-US founders Technology $ 44.092,32 257,59% 48% 60% 20% 28% Games $ 10.972,20 131,91% 56% 60% 16% 24% Fashion $ 5.838,84 81,33% 44% 28% 16% 32% Food $ 9.848,84 101,63% 60% 32% 16% 12% Total $ 17.687,95 143,20% 52% 45% 17% 24%
In table 4.2 the summary statistics of the projects from Kickstarter are presented. From the table it can be seen that 44 of the 83 gender observations were male and that 74 of all the projects were American. Since most projects were created by Americans and the other projects came from all different countries, the country variable only measured whether the
22 project was created by an American or not. As seen in table 4.2, taking all projects into account, the average percentage of the goal reached is 143.2% ranging from 0% (the projects that did not get any funding) to projects that raised 2386.51% of their specified goal. The average number of days it took successful founders to reach their goal was 18.44 days, ranging from projects that took only 1 day to succeed till projects that needed 45 days. Besides that, 12% of all the projects were selected by Kickstarter as the ‘Kickstarter Staff pick’, which might increase their chance of being funded.
Table 4.2
Summary project statistics
Obs. Mean Std. Dev. Min Max
Gender 83 0.530 0.502 0 1 Country 100 0.740 0.441 0 1 Goal 100 17687.95 34147.39 200 289666 Reach 100 0.520 0.502 0 1 Percentage 100 1.431 2.932 0 2386.51 Days 52 18.442 12.161 1 45 Staff 100 0.120 0.327 0 1
Note: the 83 observations on gender result from 17 projects being team projects. The Days variable only has 52 observations because 52 projects succeeded and 48 did not so there is no data on how long it took them to succeed.
4.2 Descriptive statistics: surveys
In table 4.3 the summary statistics on the survey data are presented. As can be seen from the table, 30% of the respondents were male; the average age of the raters was about 26 years, ranging from 16 to 55 years, 65% being students. In the survey, raters were asked whether they thought they would invest in the project and could answer on a scale from 1 to 5, with 1 meaning totally agree and 5 totally don’t agree. On average the raters gave a 3.07, close to the middle ‘don’t know’ option.
Raters were slightly less optimistic about whether the project would reach its goal or not, with an average of 2.93 on a scale of 1 to 5. Survey takers estimated the average
23 number of days the project would take to reach its goal close to actual average day it took, respectively 23.34 compared to 18.44. To the question whether the rater thought if the founder would raise more money than their goal specified, the average answer was 0.34 on a scale from 0 to 1. The lower amount of observations can be explained by survey takers indicating not to have an idea about the question asked.
Creativity, productivity and attractiveness were rated on a ten point scale. The raters were not too optimistic about the average attractiveness, which was found to be 5.17 with a standard deviation of 2.03. The average productivity was rated more positively with a 6.35 and a standard deviation of 1.71, the average creativity was 5.50 with a standard deviation of 2.12. The survey ended with the question whether raters considered the projects founders as suitable to execute the project and their answers indicated a 2.37 on a 1 to 5 scale. Data on these three variables were not available for all projects because some of the projects were created by teams and did not contain a picture to show the raters to evaluate.
Table 4.3
Summary survey statistics
Obs. Mean Std. Dev. Min Max
Male 600 0.300 0.459 0 1 Age 600 26.133 10.857 16 55 Student 600 0.650 0.477 0 1 Idea 600 3.068 1.183 1 5 Goal 600 2.925 1.129 1 5 Days 600 23.340 8.682 1 31 Money 528 0.349 0.477 0 1 Creativity 600 5.498 2.126 1 10 Productivity 498 6.349 1.713 1 10 Attractiveness 498 5.170 2.031 1 10 Suitability 498 2.376 0.880 1 5
Note: Money represents the question whether the rater thought the project would fund more money than needed. The 528 observations for this variable are explained by some raters answering ‘I don’t know’.
24 4.3 Appearance
In order to define the attractiveness of each project founder, the average of the 6 raters per project were added together and the average of these six ratings determined the overall attractiveness rating. Subsequently these ratings were divided in three categories; below, average and above. The below category consisted of all ratings below half a standard deviation under the mean (< 4.1545), the average category consisted of all ratings between half a standard deviation below the mean and half a standard deviation above the mean (4.1545 < X < 6.1855), and the above category consisted of all ratings above half a standard deviation on top of the mean (> 6.1855). This resulted in 16 observations in the below category, 50 observations in the average category and 17 observations in the above category.
To analyze the effect of appearance, first the percentage of projects that reached their goal per appearance category were calculated and presented in the graph below. From the graph it appeared that the above category performed best with 64.17% of the project reaching its goal. Surprisingly, people in the below attractive category reached their goal in more cases than people who were seen as average, an appearance phenomenon that was not found earlier in literature.
Note: The x-axis presents the three classes of attractiveness with 1 representing the below average looking people. The y-axis presents the percentage of projects that reached their goal in each appearance group.
Besides the percentage of the projects that reached their goal in each appearance category, it would also be interesting to see the average amount of dollars funded per category.
0 20 40 60 80 100 1 2 3
Percentage of projects that
reached their goal
25 However, this graph would give a distorted picture because there are more observations in the average group compared to the below and above group, and besides that the average goals set in each appearance category differed, ranging from 6731.88 for the above category to 24319.84 in the average attractiveness category. Therefore the average percentage of the goal reached was calculated and presented below to sketch a more representative picture. This outlines an interesting picture since it appears that the below category outperforms both the average as the above category with a 136.44% reached on average. One would have expected the above category to outperform the other categories; especially since the average goal set in the above category was the lowest.
Note: The x-axis presents the three classes of attractiveness with 1 representing the below average looking people. The y-axis presents the average percentage of the goal reached.
To further analyze the effect of the founders’ appearance, the appearance data was added to the project data and some Kruskal-Wallis tests were performed. These tests determined whether there is a difference between the distributions between the three appearance categories for several other variables measured in the project data. This non-parametric test was chosen because the data is not symmetric, more than two groups are used, and this test does not require the data to be normal but uses the rank of the data values instead. As can be seen from table 4.4 the gender of the founder differed significantly at the 1% level for the different categories of attractiveness. Meaning the mean ranks were not the same in the
0 20 40 60 80 100 120 140 1 2 3
Average Percentage of the goal
reached
26 three appearance categories. The goal set and whether the project was picked by the
Kickstarter staff were the only two other variables that differed significantly for the three categories of attractiveness of the founder, however only at the 10% level. From the table it also becomes apparent that there is no indication that above average looking people are more likely to reach their goal. With a p-value of 0.228, the null hypotheses, that the means of the three appearance categories are equal, cannot be rejected. Subjects in different appearance groups neither differed in their means for the percentage of their goal reached, or the amount of days it took to achieve the goal.
Table 4.4
Project characteristics and Attractiveness
Chi-squared P-value Mean Below Mean Average Mean Above
Male 14.437 0.001*** 0.089* 0.300 0.228 0.340 0.056* 0.625 9003.50 136.44 0.563 17.78 0.000 0.640 24319.84 105.23 0.396 17.14 0.143 0.118 6731.88 86.17 0.684 23.82 0.056 Goal 4.839 Percentage 2.406 Reach 2.953 Days 2.160 Staff 5.773
Note: the table presents 6 different Kruskal-Wallis tests and all tests are by attractiveness categories and corrected for ties. Significant values are indicated by: *p<0.10,**p<0.05 and ***p<0.01.
Perhaps attractive people are seen as more creative, productive or suitable. As Mobius and Rosenblat (2006) already found, that attractive workers are seen as more able. Therefore it would be interesting to see whether the raters differed in their mean answers on other survey questions for different categories of attractiveness. To see the possible effect of appearance on the survey answers the raters gave, several Kruskal-Wallis tests were performed on the survey data by appearance category.
In table 4.5 the means are not presented because it would make it confusing to add these for all 10 categories of attractiveness. Hence appearance cannot be divided in three categories in this case because all raters gave different answers on the question whether they thought the founder was attractive, and since all projects were rated by six raters the founders would fall in several categories at the same time. Even though there is no
27 extinction made between the levels of attractiveness, it is interesting to see that a lot of answers differed significantly in their means for the different ratings of appearance. From the table it appears that only for the question whether the raters thought the project would get funded more money than needed, the mean answers of the raters per appearance category did not significantly differ from each other. On both the questions about whether the raters thought it was a good idea and whether they thought the project would reach its goal, raters answered significantly different for the appearance categories. Also the question about how many days the rater thought it would take the founder to reach its goal was influenced by the appearance of the founder. This also held for the question on creativity, productivity and suitability, where raters also were influenced by the appearance of the founder while answering the questions. Therefore one can conclude that for all but the money variable it does not hold that the mean answers of the raters were equal in the ten appearance categories. This is a very interesting finding because it appears that your looks influence other people’s opinion about you and your project significantly.
Table 4.5
Survey characteristics and Attractiveness Chi-squared P-value Idea 22.851 0.007*** Goal 23.057 0.006*** Days 17.112 0.047** Money 10.527 0.310 Creativity 37.620 0.000*** Productivity 77.767 0.000*** Suitability 43.694 0.000*** Note: the table presents 7 different Kruskal-Wallis tests and all tests are by attractiveness categories and corrected for ties. Significant values are indicated by: *p<0.10, **p<0.05 and ***p<0.01.
28 4.4 Gender differences
From table 4.4 one could already see that the gender of the founder differed significantly at the 1% level for the different categories of attractiveness, only not whether men and women were equally likely to be in either of the categories. Therefore it would be interesting to see whether these differences in mean ranks for men and women differ significantly from each other in each of the three categories. Therefore in table 4.6 several Mann-Whitney U tests were performed to review the gender difference per appearance category. This test was used because two independent groups are compared that are not normally distributed and the independent variable is ordinal. From table 4.6 one can see that men and women are not equally likely to be found either average or above average looking, there was no difference in means observed for the below category. It appeared that men are more often seen as below average than women with a p-value of 0.014. However, in the above category women were significantly seen as more attractive than men at the 1% level. This also followed from the survey data, where 62.50% of the below average looking people were male, 64% was male in the average category and only 11.80% was male in the above average category.
Table 4.6
Attractiveness characteristics and Gender Z-value P-value Mean
male Mean female Attractiveness Below -0.841 0.400 0.227 0.154 Attractiveness Average -2.454 0.014** 0.727 0.462 Attractiveness Above 3.798 0.000*** 0.045 0.385 Note: the table presents 3 different Wilcoxon Mann-Whitney tests and all tests are by the gender of the founder. Significant values are indicated by: *p<0.10,**p<0.05 and ***p<0.01.
Also the project characteristics might differ for the gender of the founder. From table 4.7 we can see that the goal the founders set is significantly different for men and women since the means cannot be assumed to be equal with a p-value of 0.04. Although the mean for men (140.13%) was higher than the mean for women (70.35%), the percentage of the goal reached cannot be proven to be different between men and women. Also for the reach
29 variable, which indicates whether the founder has reached its goal or not, it applies that the means between men and women cannot be proven to be non-equal. This means that men are thus not more likely to reach their goal then females, even though the mean was higher for females. Herewith the null hypothesis that men and women are equally likely to get funded through Kickstarter cannot be rejected. One of the few authors that also did not find a significant gender influence were Solnick and Schweitzer (1999).
On the other hand, men did differ significantly in their mean from women on the days variable, meaning men needed significantly less days to reach their goal (in case they actually reached their goal). The staff variable, indicating whether the project was picked by the Kickstarter staff as recommended, also differed significantly at the 5% between men and women. From the mean values we can see that men (0.159) are more likely to get picked by the Kickstarter staff than women (0.026).
Table 4.7
Project characteristics and Gender
Z-value P-value Mean males Mean females
Goal -2.049 0.040** 19417.50 15900.21
Percentage -1.547 0.122 140.13 70.35
Reach 0.321 0.748 0.477 0.526
Days 2.866 0.004*** 13.238 25.200
Staff -2.044 0.041** 0.159 0.026
Note: the table presents 5 different Wilcoxon Mann-Whitney tests and all tests are by gender of the founder. Significant values are indicated by: *p<0.10,**p<0.05 and ***p<0.01.
The gender of the founder might also have an effect on the answers the raters gave in the survey. From table 4.8 one can see that both the idea and goal variable did not differ in their means between men and women. Surprisingly, raters did not think males or females would either have better ideas or be more likely to reach their goal. Though this was perhaps not expected, the raters’ opinions did match the finding that men were indeed not more likely to reach their goal. On a 10% level the raters did think that men would need fewer days to get
30 funded then women would. To the question whether the raters thought the project would get funded more money than needed, the mean answers did not differ significantly for the gender of the founder. It seemed that both attractiveness and productivity were not rated the same for male and female founders. Men were seen as more productive whereas women were seen as more attractive. This does not only mean that your appearance influences a lot of the answers to the survey questions as seen before, but also that
appearance significantly differs between men and women. However, there is no significant difference in the mean answers on how suitable or creative the raters consider the founder based on their gender.
Table 4.8
Survey characteristics and Gender
Z-value P-value Mean males Mean females
Idea 0.686 0.493 3.043 3.113 Goal 0.779 0.436 2.919 2.996 Days -1.873 0.061* 23.422 23.633 Money -0.599 0.550 0.682 0.625 Creativity 0.181 0.856 5.601 5.233 Attractiveness 4.862 0.000*** 4.775 5.604 Productivity -2.062 0.039** 6.500 6.196 Suitability 1.557 0.119 2.314 2.442
Note: the table presents 8 different Wilcoxon Mann-Whitney tests and all tests are by the gender of the founder. Significant values are indicated by: *p<0.10,**p<0.05 and ***p<0.01.
31 4.5 Regression analysis on the projects
To further determine the effect of appearance and to avoid possible reversed causality, the appearance ratings were added to the Kickstarter project data and several regressions were conducted, presented in table 4.9. These regressions are presented because gender and attractiveness are correlated as seen before, and adding them both to the regression gives the opportunity to see the effect on the project features when controlling for both variables and some control variables.
In the first column a regression with only the gender variable and some control variables is presented to further analyze the hypothesis that men are more likely to reach their goal than women. The coefficient shows that the gender variable is not significant, meaning that the gender of the founder was not of significant influence on the project reaching its goal. Also in the other specifications the gender coefficient stays non-significant. Herewith it also seems that the null hypothesis that men and women are equally likely to get funded through Kickstarter cannot be rejected. The country variable however is significant and stays significant in all regressions, meaning that US citizens were significantly more likely to reach their goal than non-US founders. This could be caused by the over representation of US founders compared to founders from other countries.
It seems that the goal founders set barely has any influence on whether the project reaches its goal or not, which is quite strange actually. One would expect that for example a $50.000 goal is harder to reach than a $5.000 goal, though in all columns it appeared that the goal set had no significant impact in determining whether the project would reach its goal. This could perhaps be explained by people that target higher goals are more motivated and prepared than people who set lower goals. Also the staff variable appeared not to be significant in either of the regressions. This can be interpreted as Kickstarter staff picked projects not having a bigger chance in reaching its goal than projects that were not picked. In the second column the attractiveness variable was used in a regression together with some control variables. Here appearance is not yet divided in categories to see the overall effect of appearance on whether the project reaches its goal. As we can see from the table, attractiveness is significant at the 10% level. With a positive coefficient, this means that your appearance positively influences the chance of the project reaching its goal. In the third column three category dummies for the technology, game, and fashion
32 category were added to check whether being in one of the categories made a significant difference for reaching the projects goal. The food category was omitted to avoid perfect multi-collinearity because all projects fell in one of these four categories. It appears tha t being in either of the categories is not of significant impact for your performance. The appearance coefficient stays significant at the 10% level and the control variables are, all but the country variable, not of significant impact.
In the fourth column the appearance variable was divided in three categories: below, average and above and added to the model. Since each project that was rated on
appearance is in one of these categories, these variables are dependent on each other and therefore the above dummy was omitted. It can be seen that the below average category had its predicted negative sign, however being in this category had no significant negative impact on whether the goal was reached or not compared to being in the above category. Remarkably, the average category did have a small significant negative effect on whether a project would reach its goal compared to the above category. It was checked whether the coefficients and standard errors differed once another appearance category was left out and it appeared that the above category did have a positive coefficient when one of the other categories was omitted, however this difference was not significant.
These category effects stayed the same once the dummy variables for the different categories were added in the fifth column. As in the third column, being in either of the four categories did not make a significant difference for the likeliness of reaching the goal of the project. At last the 83 observations can be explained by the fact that 17 of the 100 projects were performed by teams and therefore there are no gender or appearance observations for these 17 projects.
33 Table 4.9
Determinants of a project reaching its goal
(1) (2) (3) (4) (5) Male 0.026 (0.116) 0.098 (0.120) 0.067 (0.125) 0.095 (0.122) 0.086 (0.130) Country 0.375*** (0.130) 0.391*** (0.129) 0.380*** (0.133) 0.371*** (0.130) 0.375*** (0.134) Goal -0.000 (0.000) -0.000 (0.000) -0.000 (0.000) -0.000 (0.000) -0.000 (0.000) Staff 0.238 0.198 (0.185) 0.183 (0.189) 0.294 (0.189) 0.297 (0.195) Attractiveness 0.746* (0.045) 0.087* (0.048) Attr Below -0.127 (0.175) -0.125 (0.186) Attr Average -0.264* (0.148) -0.265* (0.152) Technology 0.160 (0.168) 0.107 (0.169) Game 0.100 (0.161) -0.009 (0.161) Fashion 0.006 (0.150) 0.024 (0.152) N 83 83 83 83 83
Note: Dependent variable: whether the founder of the project reached its goal (yes=1, no=0). The table presents coefficients from OLS regressions. The Attr variables represent the categories of appearance. Standard errors are in parentheses: *p<0.10,**p<0.05 and ***p<0.01.
4.6 Regression analysis on the survey
Possible effects might also occur because attractive people are seen as more creative, productive or suitable as we saw in table 4.5. Therefore in table 4.10 several regressions are presented on the dependent variable idea, a variable describing the question whether the rater thought the project was a good idea, to see if these effects dominate each other. In the first regression the attractiveness variable was regressed only with some control variables and one can see that the overall appearance ratings had a significant negative effect on whether the rater though the project was a good idea. In the second column, only the control variables were used in a regression to see their influence on the idea variable before adding attractiveness. In the third column one can see that the
34 suitability. This is probably because these variables are correlated with appearance and vice versa. Therefore in table 4.10B in the appendix, the correlations between these four
variables are presented and it appeared that all the variables are significantly correlated with each other.
From the table we can also see that creativity is significant at the 1% level in all the regressions, raters thus consider creativity as a crucial determinant of an idea being good, which makes perfect sense. However against expectations, the negative sign of both the creativity and the productivity variable indicate that these variables had a negative impact on whether the raters thought the project was a good idea. Perhaps raters considered the more creative ideas as more risky and therefore as ideas they would not invest in as much as other ideas.
In the fourth column a dummy for treatment is included; this variable indicates whether the raters saw the picture of the founder before or after having answered questions about the project, with 1 meaning the raters saw the picture of the founder after the project questions. This variable negatively influenced the raters’ opinion on the project, however not significantly. This means raters were less optimistic about projects when they did not see the picture of the founder before answering questions about the project. In the fifth column this treatment effect was added to the attractiveness variable and some control variables and it appeared that also here the effect of attractiveness was no longer significant, the coefficient was even almost negligible.
In the last column, dummies for the categories were added to see if the opinions of the raters were influenced by the categories. The food category was omitted to avoid perfect multi-collinearity. It appeared that the technology category had a significantly worse impact on whether the rater thought the idea was good than the food category. Both the game and fashion category had a slightly more positive influence then the food category, though not significant. To see the impact of the categories also the other categories were left out and it appeared that the technology category had a significant negative effect compared to all the other categories. The attractiveness coefficient remained practically the same compared to the previous column and thus had no significant impact.
The same regressions as in table 4.10 were ran with goal as the dependent variable to see if the effects were the same on whether the raters thought the project would reach its goal, this robustness check is presented in table 4.10C, included in the appendix.
