Gender bias in crowdfunding

(1)

Faculty Economie en Bedrijfskunde

Gender bias in crowdfunding

Bachelor Thesis

Ason Berhe

Specialization: Finance and Organization

Supervisor:

Dhr. A.R.S. Woerner January 2017

(2)

Statement of Originality

This document is written by Student Ason Berhe who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

Gender bias on crowdfunding platforms is researched. Kickstarter’s reward-based projects in the Netherlands are used as the data sample. 310 projects were used. Regressions were run on it and the results proved that women benefitted on crowdfunding. 30.3% of the sample got success and when this is separated for both genders 26.9% of men get success while women had a success rate of 45.6%. This difference might be explained due to that women are better storytellers than men and therefore can interact better with the crowd. Or that feminist backers try to democratize entrepreneurship by actively supporting female-led businesses. Further, the variables amount of money set as goal, having a video in your campaign and campaigning in the music category are all significant.

(4)

1. Introduction

Raising large amounts of money without financial intermediaries is becoming more and more practicable with time.

Using social media to raise money for a project was not popular ten years

ago. In a short amount of time a new digital medium for raising finances for your project without financial intermediaries has arisen and the new phenomenon is getting more and more attention from entrepreneurs. The new phenomenon is called crowdfunding and is changing the game in many aspects. Instead of going to a bank you can now campaign a project online on different platforms and raise money from different investors who appeal to your idea. A medium where having a large social media network has a positive correlation with funds raising (Colombo et al. 2014). This shifts the power away from the bank to businesses and private investors. Slowly on this will give consumers a bigger leverage in deciding how the market will get shaped.

Over the years crowdfunding has been growing enormously. According to Douw & Koren (2016), 2016 has been a record year for the Netherlands. The total sum of crowdfunded capital in the Netherlands was 170 million euro, which is 52 million euro more than 2015. The threshold is becoming lower to be an investor or an entrepreneur. Traditionally men have always been the funder and the executor in entrepreneurship. So do women also benefit from the lowered threshold? Sharkey and Thébaud (2014) encountered that women still have more difficulty in acquiring funding for their businesses in the small business lending markets. So what interesting to know is whether gender bias also plays a role in crowdfunding. Do women also encounter disproportionate effects in acquiring funds due to gender bias? Is crowdfunding reflecting this gender inequality gap or does it reduce the gap. So the research question for this thesis is as follows: will women encounter a disadvantage in acquiring funds in crowdfunding relative to men due to gender bias? This

research uses the data from reward-based Kickstarter projects campaigned in the Netherlands of the year 2016 to research the gender bias. All the data is processed manually. The reason why reward-based projects are chosen to investigate gender bias in crowdfunding is because of the expectation that it will give stronger results for gender bias than any other funding types. In this funding type backers don’t gain financially but still have personal interests in getting the project funded. So instead of getting ensured to get something tangible back, backers provide to projects they trust to succeed more likely. It is expected to reflect the gender bias clearer. Now it will show more clearly if backers think that a male or female is more capable to execute a project. Grobben (2016) says that women are better storytellers and that because of that they can resonate better

(6)

with the wishes of the potential investors. So this is the hypothesis for the thesis.

There are three regressions done to estimate the gender variable as clear as possible. The most significant model is used to estimate the gender variable. The first model is the standard model consisting of the female variable, duration of the project, whether the campaign has a video or not, the logarithm of the goals set and all category dummies. To the second model there is an interaction term added between the female variable and the video variable. To the third model there is an interaction term added between the female variable and the female categories variable. There is also a robustness check done to ensure the results.

This paper is structured as follows. First a little background about crowdfunding and Kickstarter is discussed. Second there is an overview of relevant academic papers concerning crowdfunding success. After that a chapter that discusses the independent variables will follow. Then the empirical approach and results follow. Subsequently the discussion and the conclusion will end the paper.

(7)

2. About Crowdfunding and Kickstarter

Crowdfunding

(Online) crowdfunding started as means to raise donations for potential start-ups through various online platforms. The first crowdfunding campaign was done by the British band Marillion in 1997. In order to be able to pay for their tickets for their U.S. tour they asked their fans for contributions. Their original goal was 30.000 dollar. They eventually ended up with 47.000 dollar. In 2003 ArtistShare was founded by Brian Camelio (Chaney, 2010, pp. 44-48). This is the first online crowdfunding platform for musicians. This platform was donation based. Nowadays the goal of crowdfunding did not change. However there are several means of generating funding from the crowd, also called backers. The four funding types of crowdfunding are discussed below.

Donations

As mentioned above the first form of crowdfunding is through donations. This form of crowdfunding distinguishes itself by funders who donate without wanting a return (Ahlers et al, 2013, p. 8). Nowadays mainly charitable organizations and individuals use this form of crowdfunding. The biggest donation based crowdfunding platform is GoFundMe. It was launched in 2010 and has more than 25 million donors. GoFundMe has raised over 3 billion dollar so far. Everyone can start their personal campaign.

Equity

In this form of crowdfunding start-ups offer shares. This form distinguishes itself by the funder who will receive a payoff in the form of equity or equity-like arrangements (Bradford, 2012). Until 2011 Profounder was the leading equity based crowdfunding platform. They were not allowed to sell securities unless they registered themselves as broke dealers. There are strict regulations on this form of crowdfunding. In 2012 Obama signed the JOBS act which also contained the CROWDFUND ACT (Stemler, 2013). Under this act the regulations for crowdfunding platforms became less harsh. If the sites called themselves “funding portals”, they were allowed to raise capital. However there are still requirements in order to prevent illegal activities such as fraud. Important to note is that entrepreneurs have share all the legal information and numbers of the company in their campaign.

(8)

Lending model

Since the financial crisis it got harder to get a bank loan. In order to make it possible for entrepreneurs to get a loan, a new model came up. This model is called peer-to-peer lending (P2P) (Lin, Prabhala & Viswanathan, 2013). In this model entrepreneurs will get a loan from the crowd. Usually they have to pay it back with either a fixed interest or a percentage of the revenue. The business model of P2P platforms is that they receive a service fee. The lending club is the biggest p2p platform.

The reward-based model

Ever since ArtistShare the amount of reward based platforms are growing. The biggest reward based platforms are Indiegogo and Kickstarter. Besides art, the latter two platforms also host campaigns for businesses and social causes. In this model entrepreneurs presell their products and services in order to launch their business or social cause. Examples of these social causes are educational and environmental campaigns. Important to note is that the campaign holder does not lose any form of equity. In this form the investor is an early consumer.

Kickstarter

In the research I used the numbers and information of Kickstarter. Therefore I will elaborate on the Kickstarter platform. Kickstarter is a reward based platform. Their mission is “to help bring creative projects to life” (Kickstarter, 2017). Ever since their launch in 2009, they have had over twelve million people supporting a project and raised almost three billion dollar. At the moment of writing this 118501 projects are funded. Kickstarter distinguishes itself by the use of a so-called all-or-nothing-funding. This means that when a campaign does not reach its goal within the given timeframe, it will not receive the funding. This is a harsh rule compared to other crowdfunding platforms that give the option to let campaigners take the money that they accomplish to raise. Other rules are that Kickstarter does not fundraise for charity and cannot offer equity. It really focusses on start-ups in its early phases and new products. It has helped many launches. The Pebble Smart watch is the highest funded campaign in the Kickstart history with over 10 million dollar.

(9)

3. Literature review

Mollick (2014) researched the underlying dynamics of crowdfunding to see which factors are crucial for getting funds. This research is one of the biggest that tries to gain an analytical understanding of crowdfunding. Mollick did that by analyzing 48,500 projects on Kickstarter with combined funding over 237 million dollars and pointed out important factors that have an effect in getting projects succeeded or not. The research shows that successful fundraising a project is associated with the personal networks of the campaigner and the underlying quality of the project. So having large numbers of friends on social media while signaling high quality level, gives the biggest chance to success. It also suggested that crowdfunding projects tend to have a bigger chance of failing when a large amount of money is set. Furthermore, the paper proves that the geographic component can’t be underestimated; projects in certain categories have a bigger chance in a certain geographic location.

Consistent with the previous research of Mollick. Mollick & Kuppuswamy (2014) conducted a survey of projects that examined the long-term implications of crowdfunding. They found that traditional entrepreneurship appears to be supported and led by reward-based found crowdfunding. They saw that 90% of successful projects remained ongoing after 1 to 4 years being campaigned. Successful projects on Kickstarter employed new employees by an average of 2.2 employees. 32% even has yearly revenue over a $100,000. This paper also suggests that crowdfunding benefitted multiple parties, beyond the creators themselves. Crowdfunding helps provide an access to customers, employees, press and outside funders.

Etter et al. (2013) encountered that only one of two Kickstarter projects got success. Therefore they investigated what kinds of campaigns are more likely to succeed. So they proposed a model that could predict funding success in the early stage of a campaign with high accuracy. They accomplished this by using direct information and the social features. They used time series of money pledges to estimate the probability of success and the information they gathered from observing tweets of the Kickstarter’s project and the backers. The combination of this data ensures that they can predict whether projects get funded or not with an accuracy of more than 76% within 4 hours.

Colombo et al. (2014) conducted an equivalent research that investigated whether there is an empirical relationship between contributions received early in a campaign and succeeding. They

(10)

also tried to know what the role of social capital is in succeeding. Social capital is the potential funds that might be raised through the social contacts of the campaigner. _{An econometric analysis} was done on 669 Kickstarter projects. They found that due to the contribution of social contacts in the early stage of the campaign that the project was more likely to succeed later on. So having a big social network can cause a snowball effect that eventually can get a campaign succeeded.

Kuppuswamy also conducted a research with Bayus (2015) so they could add an empirical understanding to the project funding cycle of crowdfunding campaigns. Two years of publicly available online data is used to search a pattern in Kickstarter’s projects. They found that succeeded projects typically were funded with relatively small contributions of many Individuals in a U-shaped pattern. Meaning that backers in most cases contribute more in the first and the last week of a project compared to the middle period of the funding cycle. They also confirmed the positive correlation that social contacts have in obtaining success.

Burcht et al. (2013) analyzed the decision-making process of backers in crowdfunding. They found out that popularity influenced the decision making process. Backers would reflect the preferences of influential marketplace members when deciding their contribution decision.

(11)

4. Data

For gathering the data to answer the research the question data from Crowdsurfer was used. Crowdsurfer is a database that keeps up with projects that are being campaigned in over 200 crowdfunding platforms. Including the big ones as Kickstarter, Lending club, Kiva, Indiegogo, etcetera. Crowdsurfer gives a panel available to cleanly search and analyze the crowdfunding industries. With this transparent information, the opportunity is given to navigate through its big data set and analyze the relation between certain variables.

From the data on Crowdsurfer we used Kickstarter for the sample. Kickstarter is used because it’s one of the most dominating reward-based crowdfunding platforms (Agrawal, Catalini & Goldfarb, 2013, p. 4). Also Kickstarter has a relatively simple pay model were campaigners have to pay a constant fee of 5% (Kickstarter, 2017). Because of the limitation that all the information has to be processed manually the data is limited to only Kickstarter’s projects.

The data contained projects campaigned in the Netherlands, from in the year 2016. There were 512 projects reward-based projects campaigned in the Netherlands on Kickstarter that year. From the 512 projects, 310 were used for this research. 202 projects were eliminated because the project was either campaigned by a business or by a team with mixed genders. Of those 310 projects 57 were classified as female. A project is classified as female when a woman has campaigned the project or a team of all women without a man did.

Furthermore, different independent variables were used to estimate the gender variable as precise as possible. Any variable that is likely to have a reasonable correlation with the outcome is added to the standard model so the coefficient is estimated the best possible.

The goal of the campaign is set as a variable because of the assumption that how larger the goal is, how more difficult it is to get the goal as a whole raised. The logarithm of the goal is used to make the variable more manageable because of the skewness of the variable. Having a minimum of 100 euro’s and a maximum of 1.1 million euro’s, with a mean of 33.296.

The variable for how long a project is campaigned is added to the regression as duration in months. Because there is an expectation that when a project is campaigned for a longer amount of time that it’s more likely to get funded. So to neutralize this fact, the variable duration is added to the standard model.

The variable whether a campaign has a video or not is also added to the model. Whether a campaign has a video or not indicates the effort put in the campaign by the campaigner. It is logic to say that how more effort is put in the campaign how bigger the chances are that the campaign

(12)

will succeed and the other way around. Also, the promotion of the campaign is more complete with a video then without, giving the backer a better insight in the project. So this variable shouldn’t be omitted.

The categories of the campaigns are also used as variables. The variables are added to estimate the success chance with respect to the category where is being campaigned. Kickstarter projects are classified in thirteen categories, from: Art, Comics, Dance, Fashion, Film & Video, Food, Music, Photography, Publishing, Technology and Theater. Comics, dance and theater are omitted in the regression because they weren’t observed in the data sample. Design is used as the reference group since it has been observed the most. Having a reference group prevents perfect multicollinearity in the model.

Independent variable

Description

Female 1 if campaigned by a female; 0 if campaigned by a male

Loggoal The log of the amount that is set as goal in euros

Duration The duration of the campaign in months

Has_video 1 if the campaign has a video; 0 if the campaign hasn’t one

Art 1 if it is campaigned in the art category; 0 if otherwise

Fashion 1 if it is campaigned in the fashion category; 0 if otherwise

FilmVideo 1 if it is campaigned in the film/video category; 0 if otherwise

Food 1 if it is campaigned in the food category; 0 if otherwise

Games 1 if it is campaigned in the games category; 0 if otherwise

Music 1 if it is campaigned in the music category; 0 if otherwise

Photography 1 if it is campaigned in the photography category; 0 if otherwise

Publishing 1 if it is campaigned in the publishing category; 0 if otherwise

Technology 1 if it is campaigned in the technology category; 0 if otherwise

Table 1 The independent variables

(13)

5. Empirical approach

In this section the multiple models that are regressed will be discussed. From those multiple models the most explanatory one is used to estimate the female variable . The regression is a linear approximation, so a linear probability model is the outcome. A linear probability model is an OLS regression. The regression’s assumptions are that it is unbiased and consistent. The dependent variable is binary. So the estimated probabilities lie between 0 and 1. The coefficients give marginal effects. The standard model regressed is:

FUND =

β β β β β β β 0 + 1Female + 2Loggoal + 3Duration + 4Has_video + 5Art + 6Fashion +

β β β β β β ₇FilmVideo + ₈Food + ₉Games + ₁₀Music + ₁₁Photography + ₁₂Publishing +

β ₁₃Technology

The first alternative model will have an interaction term added. The interaction term added will consist of the female variable and the variable whether somebody got a video. This variable is added because there may be an additional effect when a woman is shown in the video. Since the looks of a woman are perceived as more appealing in society. So this correlation may be shown in the big data sample and shouldn’t be overlooked. The regression of the model with the additional variable is now:

FUND =

β β ₁₃Technology + ₁₄Female*Has_Video

The second alternative model has an interaction term added consisting of the female variable and the so-called female categories variable. The female categories variable consist of categories where women campaigned relatively more than men. In table 2 a table is presented where the proportions of where each gender campaigns is displayed. In the table we see that the categories art, fashion, food, music and publishing women are relatively more attractive for female campaigners. All these categories are included in the female categories variable. Due to the limited observations there is only one big female categories variable interacting instead of multiple interaction terms of the female variable with all the categories individually. The reason why this interaction term is added is because categories were women tend to campaign relatively more than men may be seen as more natural for women to operate in. So if the interaction variable is added to the model and the extra estimate indicate significance with respect to the

(14)

standard model. This may point out that there may be a decrease or increase in gender bias for certain categories. The regression for this model is:

FUND =

β β ₁₃Technology + ₁₄Female*FemaleCategory Art Design Fashio n Film/Vide o Food Games Music Photograp hy Female 12.28% 22.81% 12.28% 10.53% 8.77 % 5.26% 10.53 % 1.75% Male 5.53% 21.34% 8.30% 5.93% 3.16 % 17.39 % 5.93% 2.77% Publishin g Technol ogy 10.53% 5.26% 5.93% 23.72%

Table 2 The proportions of campaigning of each gender in the different categories

A robustness check with a logit regression is also done. The logit regression is an alternative way to obtain a binary choice model. So to verify the OLS regression’s outcome a logit regression is done. If the logit regression gives the same sign, that is being either negative or positive, for the significant coefficients, it will invigorate the OLS regression.

(15)

6. Results

All the regressions are run and the regression results can be observed below. The regression results display that the standard model (1) has the highest adjusted R-square compared to the first and second alternative model, (2) and (2). The adjusted R-square’s respectively are 0.1472, 0.1454 and 0.1470. So the standard model decreased with the extra interaction terms.

Model Variables (1) (2) (3) (4) Female 0.138** 0.236 0.215** 0.705** (0.065) (0.168) (0.103) (0.353) Log(Goal) -0.086*** -0.084*** -0.087*** -0.498*** (0.016) (0.016) (0.016) (0.102) Duration -0.119 -0.120 -0.124* -0.646 (0.074) (0.074) (0.074) (0.443) Has_video 0.202** 0.219** 0.207** 1.364** (0.067) (0.072) (0.067) (0.454) Art -0.088 -0.078 -0.056 -0.596 (.110) (0.111) (0.115) (0.628) Fashion 0.069 0.063 0.098 0.436 (0.097) (0.097) (0.101) (0.519) FilmVideo -0.014 -0.013 0.015 -0.148 (0.109) (0.109) (0.113) (0.593) Food -0.143 -0.013 -0.107 -1.167 (0.132 (0.132) (0.137) (0.944) Games -0.026 -0.025 -0.016 -0.187 (0.082) (0.082) (0.083) (0.463) Music 0.227** 0.227** 0.257** 0.963* (0.108) (0.108) (0.112) (0.570) Photography 0.153 0.144 0.159 0.729 (0.163) (0.164) (0.163) (0.895) Publishing 0.054 0.059 0.084 0.263 (0.107) (0.107) (0.111) (0.559) Technology -0.041 -0.040 -0.028 -0.394 (0.076) (0.076) (0.077) (0.472) Female *Has_video -0.115 (0.182) Female* femalecat -0.129 (0.133) Constant 1.032 1.010 1.026 (0.175) (0.179) (0.175) Observations 310 310 310 310 15

(16)

Adjusted R-square

0.1472 0.1454 0.1470

F-test 5.09 4.74 4.79

*** p<0.01, ** p<0.05, * p<0.1

Table 3_{The regression results}

The standard model gives significance for the female variable with a significance level of 5%. The coefficient estimates that women have 13.8% bigger chance of getting their project funded than men, holding all other variables constant. The logarithm of the goal comes out as the most significant. It is significant with a level of 1% and has a negative relation with getting funded. Having a video and a project in the music category are also significant with a level of 5% and they are both positive related with the dependent variable.

The descriptive statistics in the table below shows that 30.3% of the data sample got funded. In appendix 1A&B the descriptive statistics of the model is shown when gender is separated, one model with only males and the other model with only females. Here it’s displayed that there is a big difference in the expectations between men and women in the chance of being funded or not. Respectively, 26.9% and 45.6%. This also shows a big difference, in benefit of women.

Variable Obs Mean Std. Dev Min Max

Fund 310 0.3032 0.4604 0 1 Female 310 0.1839 0.3880 0 1 Loggoal 310 9.2508 1.6478 4.6052 13.9108 Duration 310 1.1206 0.3367 0.25 2 Has_video 310 0.8252 0.3804 0 1 Art 310 0.0667 0.2517 0 1 Fashion 310 0.0903 0.2871 0 1 FilmVideo 310 0.0677 0.2517 0 1 Food 310 0.0419 0.2008 0 1 Games 310 0.1516 0.3592 0 1 Music 310 0.0677 0.2517 0 1 Photography 310 0.0258 0.1588 0 1 Publishing 310 0.0677 0.2517 0 1 Technology 310 0.2032 0.4030 0 1

Table 4 Descriptive statistics

The robustness check (4) confirms the legitimacy of the significant coefficients of the standard model. The gender, goal, having a video and the music category variable came out with the same significant level and sign.

(17)

7. Discussion

There are two big limitations in this research. The first one is that the data couldn’t be processed automatically and so that it had to been done manually. Because of that the choice had to be made to stick only to one crowdfunding platform. This may give a distorted reflection of the overall gender bias in all crowdfunding platforms. If the dataset was bigger, there would be more female campaign-led projects in the set. Now it was only limited to 57. The alternative models would maybe have been more significant if there were more observations added to the research. In particular the second alternative model. So it may be too soon to conclude that gender-bias doesn’t differ per category.

The second and the biggest limitation is that only a correlation effect is estimated and not a causal one. So from this research we can’t make sure that whether gender bias plays a role or that women are just better funds raisers on crowdfunding. Grobben (2016) beliefs it is because of that women are better storytellers and that because of that they interact easier with the crowd, receiving more funds. While Spors (2014) explains the different by activism. She advocates that female feminist backers actively support women-led projects to democratize entrepreneurship.

Considering those two arguments it seems that maybe the advantage for women in crowdfunding is explained by the fact that crowdfunding comes more naturally for women entrepreneurship or that they encounter their positive gender bias by their own gender who is trying to fight sexism. Perhaps a combination of both. A sequel research where not only the gender of campaigners is researched, but also the gender of the backers might be the outcome to further investigation in gender bias on crowdfunding platforms.

(18)

8. Conclusion

This paper asked the question whether there is a gender bias on crowdfunding towards women who try to acquire funding for their projects. This question was asked because women still encounter more difficulties in acquiring funding for their businesses in the small business lending markets then man did. So the question that arose was if whether if crowdfunding would reflect this or would make the gap smaller and be solution for female entrepreneurs. The hypothesis stated that women would have a benefit over men based on the statement of Grobben.

Different regressions were run to estimate the gender variable as precise as possible. The standard model included the independent variables female, duration of the campaign in months, the logarithm of the goal, whether a video was included and the categories of the different industries. The first alternative model included an extra estimate that consisted of an interaction term between female and having a video. The second alternative model included an extra interaction variables consisting of the female variable and the so-called female categories variable_.

The standard model gave the largest adjusted R-square, 0.1472 respectively to 0.1454 and 0.1470. Making this model the most accurate to estimate the significance of the female variable. This model gave a p-value of 0.036 for the gender variable, what makes the variable significant with a significance level of 5%. So this point out that gender, at least statistically, matters in the chance of getting funded in this research. Contrary to the small business lending markets women are in an advantage relative to men. The coefficient estimates that women have 13.8% bigger chance of getting their project funded than men. And when the data is separated into two separate groups of only females and males and the regression is run again, we observe that 26.9% of male projects and 45.6% of female projects got funded. This also reflects a large difference.

Due to the limitation that it we merely can estimate a correlation effect and not a causal effect it is unknown whether gender bias really plays a role. The benefit that women encounter could be explained by female-led activism. While Grobben advocates it is because of that women can interact with the crowd than men. So the hypothesis can’t be accepted or rejected yet.

Further, the logarithm of the goal comes out as the most significant with a significant level 1%. It is negative related with the dependent variable. So how larger the goal is set, how smaller the chance is in getting funded. Having a video and campaigning in the music category are also significant with a level of 5%. These variables are positive correlated to the dependent variable.

(19)

Meaning that campaigning in the music projects are in favor compared to other projects in other industries and having a video will enlarge your chance on success.

Overall what this paper proves for certain is that female reward-based campaigns significantly got funded more in the Netherlands in the year of 2016 on the crowdfunding platform Kickstarter than their male counterpart. And that it is statistically proven that there is positive correlation between being a female campaigner and the chance of getting your project funded. What causes this can’t be really made sure of from this point. And it may still be too premature to conclude that this reflects it all for all the crowdfunding platforms out there. But the results of this research shouldn’t be overlooked or be neglected. Crowdfunding may really lead to increased participation of women in entrepreneurship and change the status quo regarding gender bias in the long run.

(20)

References

Agrawal, A., Catalini, C., Goldfarb, A. & 2010. The geography of crowdfunding. _{SSRN Electronic}

Journal.

Ahlers, G. K., Cumming, D. J., Günther, C. & Schweizer, D. (2013). Equity crowdfunding. Available

at SSRN 2362340

.

Bradford. (2012). Crowdfunding and the federal securities laws.Columbia Business Law Review

,

2012(1), 1–150.

Burtch, G., Ghose, A., & Wattal, S. (2013). An empirical examination of the antecedents and consequences of contribution patterns in crowd-funded markets. _{Information Systems}

Research

, 24(3), pp.499-519.

Chaney, D. (2010). International Journal of Arts Management Vol. 12, No. 2 (WINTER 2010), pp. 44-48

Colombo, M.G., Franzoni, C. & Rossi-Lamastra, C. (2014). Internal social capital and the attraction of early contributions in crowdfunding. Entrepreneurship: Theory and Practice

, pp.75–100.

Douw & Koren. (2017) Retrieved from _{http://www.douwenkoren.nl/}

Etter, V., Grossglauser, M., & Thiran, P. (2013). Launch hard or go home!: predicting the success of kickstarter campaigns. In Proceedings of the first ACM conference on Online social networks

,

pp.177-182. ACM.

Grobben, W. (2016). Retrieved from

http://www.crowdfunding.nl/study-indicates-women-are-winners-when-it-comes-to-crowdfu nding/

Kickstarter. (2017). Retrieved from https://www.kickstarter.com/

Kuppuswamy, V. & Bayus, B. (2013). Crowdfunding creative ideas: the dynamics of project backers in Kickstarter. Working Paper.

Lin, M., Prabhala, N. R. & Viswanathan S. (2013). Judging Borrowers by the Company They Keep:

(21)

Friendship Networks and Information Asymmetry in Online Peer-to-Peer Lending. Management

Science

, 59(1), pp. 17-35.

Mollick, E. (2014). The dynamics of crowdfunding: An exploratory study. Journal of Business Venturing

, 29(1), pp.1–16.

Mollick, E. & Kuppuswamy, V. (2014). After the Campaign: Outcomes of Crowdfunding. Available

at SSRN.

Sharkey, A. J. & Thébaud, S. (2014). Unequal hart times: The influence on gender bias in entrepreneurial investment. Center for the study of social organization.

Spors, K. (2014). Retrieved from

https://www.americanexpress.com/us/small-business/openforum/articles/why-crowdfundin g-favors-women-owned-businesses/

Stemler, A.R. (2013) ‘The JOBS act and crowdfunding: Harnessing the power and money of the masses’, _{Business Horizons}

, 56(3), pp. 271–275. doi: 10.1016/j.bushor.2013.01.007.

(22)

Appendix

Appendix 1A&B

Appendix 1A: Summary statistics for females

Variable Observations Mean Std. Dev Min Max

Fund 57 0.4561 0.5025 0 1 Loggoal 57 8.8771 1.7622 4.6052 12.6112 Duration 57 1.1600 0.3774 0.25 2 Has_video 57 0.8596 0.3505 0 1 Art 57 0.1228 0.3311 0 1 Fashion 57 0.1228 0.3311 0 1 FilmVideo 57 0.1053 0.3096 0 1 Food 57 0.0877 0.2854 0 1 Games 57 0.0526 0.2253 0 1 Music 57 0.1053 0.3096 0 1 Photography 57 0.0175 0.1325 0 1 Publishing 57 0.1053 0.3096 0 1 Technology 57 0.0526 0.2253 0 1

Appendix 1B: Summary statistics for males

Variable Observations Mean Std. Dev Min Max

Fund 253 0.2688 0.4442 0 1 Loggoal 253 9.3350 1.6126 4.6540 13.9108 Duration 253 1.1118 0.3270 0.25 2 Has_video 253 0.8175 0.3871 0 1 Art 253 0.0553 0.2291 0 1 Fashion 253 0.0830 0.2764 0 1 FilmVideo 253 0.0593 0.2366 0 1 Food 253 0.0316 0.1753 0 1 Games 253 0.1739 0.3798 0 1 Music 253 0.0593 0.2366 0 1 Photography 253 0.0277 0.1643 0 1 Publishing 253 0.0593 0.2366 0 1 Technology 253 0.2372 0.4262 0 1 22

Gender bias in crowdfunding

Faculty Economie en Bedrijfskunde