Mobilizing the social network in crowdfunding

(1)

Master thesis:

Mobilizing the social network in

crowdfunding

University of Amsterdam

Faculty of Business and Economics

MSc. In Business Administration - Entrepreneurship and Innovation track

Final version, 23 June 2017 Submitted by Kimman, Dicky (11431881)

(2)

Statement of originality

This document is written by Student Dicky Kimman 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

Purpose – The purpose of this paper is to provide insights in how entrepreneurs can mobilize their social network in order to increase their chances of success in crowdfunding.

Design / methodology / approach – This paper reflects on the current state of the literature about crowdfunding in relation to social network theory. Additionally, it collected quantitative data from the crowdfunding platform Kickstarter and social network sites Facebook and Twitter. Consequently, the data is analyzed to reflect on the literature and to offer new insights in how an entrepreneur can tap in and mobilize the social network around him.

Findings – The word-of-mouth on social networks sites is an important indicator for the success rate in crowdfunding. An entrepreneur can trigger this word-of-mouth, also social media diffusion, by mobilizing his social network to spread the word by means of updates. Having more Facebook friends plus creating and linking a Facebook page to the crowdfunding campaign positively affect the social media diffusion. The reputation within the crowdfunding community is also an important indicator for the success rate.

Research limitations / implications – As the sample size is limited (N=83 and N=146) the findings in this paper should be interpreted with caution and future research is needed to check the validity in larger data sets. Nonetheless provide interesting insights for entrepreneurs searching for funding on crowdfunding platforms.

(5)

1. Introduction

To attract financial resources, nowadays entrepreneurs can tap directly in the crowd rather than following the traditional route to funding where a single or small number of investors (e.g. banks, angel investors or venture capital funds) fund projects. It is now possible to access a huge pool of potential investors on crowdfunding platforms (Zheng et al., 2014). On these platforms, entrepreneurs can create an account, post their ideas and include a text and video to present their plans in order to collect funds.

Crowdfunding entrepreneurs tend to be extremely innovative. Many important projects in consumer electronics as of 2013, such as 3D printers, electronic watches and video game consoles, were funded through crowdfunding campaigns (Mollick and Kuppuswamy, 2014).

Thanks to the advancement of web 2.0 technology, crowdfunding has developed rapidly, and over 450 platforms have emerged worldwide (Cordova et al., 2015).

In the era of the Social Web, crowdfunding has become an increasingly more important approach for entrepreneurs or small enterprises to raise the essential capitals from the crowd to support or ‘kick start’ their projects or businesses. Crowdfunding websites such as Kickstarter and IndieGoGo behave as online intermediary agents that allow project founders to quickly reach a large number of individual investors with minimal costs.

The role of social capital is stressed as important in crowdfunding (Mollick, 2014; Colombo et al., 2014; Zheng et al., 2014). As the social capital is embedded within the social network (Coleman et al., 1988) and in line with Mollick (2014) this paper will further elaborate on the social network. In addition, the role of social network is underlined as important in funding new ventures (Hsu, 2007; Shane and Cable, 2002). Friends, family and social contacts within the community are important to spread the word of the campaign (Colombo et al., 2014; Ordanini et al., 2011).

However, the question remains how can a founder tap into his social network? How can he mobilize his social network? Therefore, the main research question this paper will address is:

“How does the social network influences the success of a crowdfunding campaign and how can this social network be mobilized?”

(6)

The paper is structured as follows. It starts with a literature review defining the concept of crowdfunding. Subsequently it will elaborate on the factors social network, reputation in the social network and social media. Concluding the literature review with a conceptual model.

Next, the research method and data collection method are discussed. The relationships between the constructs as pointed out in the conceptual model are tested in the following chapter by means of quantitative research on social network data. What follows is a discussion where the findings about each construct are related to theory and logic. Besides, the discussion provides implications for literature and practice, addresses the limitations of this study and points out directions for future research. To end the paper, the research question will be answered in the conclusion.

(7)

2. Literature review

First this chapter will introduce the phenomenon crowdfunding. Second, the role of social network in crowdfunding is described. Thirdly, the importance of reputation within the social network is emphasized. At last, how social network sites like social media platforms influence crowdfunding success are discussed.

2.1. Crowdfunding

Crowdfunding is a relatively new way of financing novel ventures. Crowdfunding comes forward from concepts like microfinance and crowdsourcing, but it represents its own unique category of fundraising (Mollick & Kuppuswamy, 2014). Projects vary greatly both in objective and significance, from small projects to founders searching big amounts of financial resources in seed capital (Schwienbacher and Larralde, 2010).

Building on the definition of Schwienbacher and Larralde (2010), Mollick (2014) defines crowdfunding as follows: “Crowdfunding refers to the efforts by entrepreneurial individuals and groups - cultural, social, and for-profit to fund their ventures by drawing on relatively small contributions from a relatively large number of individuals using the internet, without standard financial intermediaries”.

In crowdfunding projects the products offered are often bought in advance when regular sale has not started yet. Based on the amount funded, backers will receive a monetary or nonmonetary reward. Belleflamme et al. (2013) argue that the involving the crowd in the production process is a major advantage of crowdfunding over traditional ways of funding projects. This involvement enhances the overall experience of the consumer.

Mollick (2014) demonstrates that there are four categories of crowdfunding campaigns, namely donation-based, loan-based, reward-based or equity-based projects. For the scope of this research, only reward-based crowdfunding will be discussed in more detail.

In reward-based crowdfunding funders receive a reward for backing a project. Backers are treated as early customers, given an earlier buying date or better price. Pre-selling of products to backers is another regular feature for crowdfund projects that produce more traditionally resemble entrepreneurial ventures (projects producing novel software, hardware or consumer products).

(8)

Also, participation from the backer in the product is common, such as being credited into a movie and invitations on opening nights of restaurants or even the creation of personal characters in video games.

2.2. Role of social network

Literature around crowdfunding stresses the role of social capital (Mollick, 2014; Colombo et al., 2014; Zheng et al., 2014). Social capital is defined as “the sum of the actual and potential resources embedded within, available through, and derived from the social contacts of an individual or an organization” (Nahapiet and Ghoshal 1998, as cited in Colombo et al., 2014). These social contacts together are called the social network. Or in other words, the social network is the source in which social capital is embedded (Coleman et al., 1988). The success of funding entrepreneurial financing is influenced by the social network of the individuals who are looking for funding also called founders in crowdfunding. The social network provides the founder with connections to potential backers as well as endorsements of crowdfunding project quality for the potential backers (Shane and Cable, 2002; Sorensen and Fassiotoo, 2011). Kickstarter gives the opportunity to founders to link their Facebook account to their Kickstarter campaign. If founders do so, Kickstarter shows the number of Facebook friends publicly. Mollick (2014) shows that social network size, as measured in number of Facebook friends, predicts the success of the crowdfunding campaign. Contradictory, founders having only a few online connections are better off not linking their Facebook account to the Kickstarter campaign. Other research from Lu et al. (2014) state that the social network size of the founder, as measured in number of followers on Twitter, has low correlations with the crowdfunding success as it only indicates how popular a person is, not how many people are reading or acting on the posts.

The social capital of a founder on crowdfunding platforms exists out of two types of social capital according to Colombo et al. (2014). The research makes a clear distinction between external social capital and internal capital. Founders can rely on social contacts outside the crowdfunding platform, such as family, friends and social media contacts. These contacts are referred to as the external social network. Founders can also benefit from the social capital within the crowdfunding platform by establishing relationships with other founders and backers. These contacts are called the internal social network; contacts

(9)

developed within a collective or community. Internal social capital is proven to influence performance of individuals and organizations, increasing the ability to complete complex projects and increase their innovative capabilities (Colombo et al., 2014).

Research by Lin, Prabhala and Viswanathan (2013) on a platform for peer-to-peer lending found that borrowers who have more online friends within the platform, so a larger internal social network, are more likely to be funded, get lower interest rates and have a lower probability of ex-post default. In sum, social contacts developed within an online community increases the success getting a loan.

The social contacts of a founder within and outside the Kickstarter platform increases the success of the campaign based on three mechanisms according to Colombo et al. (2014). At first, these contacts operate as promoters of the campaign, spreading information around the campaign beyond the founders own social circle. This happens by word-of-mouth, which is defined as the communication between consumers about a product, service, or a company in which the sources are considered independent of commercial influence (Arndt, 1967; Litvin et al., 2008). Word-of-mouth communication can either occur face-to-face or online.

There is also a second mechanism going on within the crowdfunding platform which triggers reciprocity through a feeling of perceived obligation (Coleman, 1990 as cited in Colombo et al., 2015). This entails that social contacts that have received funding from the founder feel obliged to help the founder. This mechanism is called specific reciprocity. Faraj and Johnson (2011) add, that a founder benefits from the online community when he backed many projects in the past. Moreover, people feel obliged to support other projects as they have received funding in the past or expect to need funding in the future. This mechanism is called the norm of generalized reciprocity, which is the third mechanism (Colombo et al., 2014). Zheng et al. (2014) also state that a founder can develop internal social capital, so within the Kickstarter community, by investing in other entrepreneurs’ crowdfunding projects. This may trigger the willingness among other entrepreneurs to fund the founders’ own project (Staber, 2006).

As Kickstarter publicly shows the number of projects that a member has supported in the past, this makes the reciprocating behavior visible to others in the community. This visibility is important for generalized reciprocity, where the member should be seen as a giver (Bolino, Turnley and BLooddgood, 2002).

(10)

At last, Zvilichovsky, Inbar and Barzilay (2015) researched owners of crowdfund projects, also called founders, that back other projects also called backer-owners. They found that backer-owners, are more successful in financing their campaigns compared to owners who did not back others. They also found that backer-owners are more active on the platform than other user types: they back and create more projects than other backers and non-backers. In other words, they have more social contacts within the community. This is in line with aforementioned argument about the impact of obligation. Backing others has a cumulative effect: the more a founder backs his colleagues, the higher the number of overall backers he secures and thus the probability of financing success (Zvillochovsky et al., 2015). Accordingly, Zheng et al. (2014) found as well that backing others was a significant predictor of crowdfunding success. In line with abovementioned reasoning and research, this research expects that backing others has a positive influence on how likely a project is to succeed. Resulting in the hypothesis which captures the internal social network of a founder:

H1. The more a founder backed other projects, the more likely his own crowdfunding campaign is to succeed

Building on Mollick’s (2014) findings on the number of Facebook friends of the founder, this research also expects that having a larger external social network has a positive effect on crowdfunding success. This results in the first two hypotheses:

H2. The larger the external social network of a founder, the more likely his crowdfunding campaign is to succeed

(11)

2.3. Reputation in social network

Backing others, possibly develops the founder’s reputation within the crowdfunding community which could create more trustworthiness and in increase in social capital (Zheng et al. 2014). As the reputation within social networks is discussed next, Shane and Cable (2002) state that besides the private information advantage that social contacts of the founder have, there is also a ‘social obligation’ rationale. It entails that investment decisions depend, in part, on the relationships themselves, rather than competence-based criteria. Therefore, Shane and Calbe (2002) suggest that investors are more likely to fund when the founder has a positive reputation from the perspective of the potential backer.

Within social networks, reputation is the extent to which users can identify the standing of others in the social media setting. Reputation is a matter of trust which in social media settings is based on aggregate user-generated information. A measurement for reputation within a community can be the activity or number of posts within this community (Kietzmann et al. 2011). Translated to Kickstarter, created projects and updates are two possibilities to post within the Kickstarter community (Kickstarter, 2017).

Accordingly, previous created projects can lead to an increase in social capital as Kickstarter allows founders to bulk e-mail supporters of (previous) projects (Greenberger and Gerber, 2014). In this way, the social network of the founder increases and is easy to reach when launching a second or third campaign for example.

Moreover, the reputation also serves as a signal of future performance based on perceptions of past performance (Dimov and Shepherd 2005) and on the other hand the visibility of past performance or activities is an important aspect of reputation (Lang and Lang, 1988). Kickstarter does show the number of projects a founder created previously, accordingly potential backers have access to that information. Side note, it is not directly showed whether the projects were indeed successful (Kickstarter ,2017). However, according to Cope et al. (2011) the success of previous attempts is less important, as Cope et al. describe that failed entrepreneurial ventures are often better prepared to proceed forwards in their next attempt due to learning experiences. Moreover, a qualitative research from Greenberger and Gerber (2014) indicated that indeed founders of failed projects felt more experienced and better prepared to crowdfund again.

(12)

In sum, this research will examine the number of projects created regardless of their success since the aim is to determine the effect of the founders’ reputation on the success rate of the crowdfunding campaign.

Moreover, founders can uphold a reputation of a responsible and accessible project creator by addressing questions and posting regular updates (Hui et al. ,2012). Frequent project updates on the project page is already empirically detected as a success predictor in prior research. As updates within the project can motivate potential backers to actual back the project (Kuppuswamy and Bayes, 2013). Also, updates within the first days of the project are positively correlated to fundraising success (Mollick, 2014). According to Kickstarter, the option “updates” is designed to “keep backers informed of a projects’ progress” (Kickstarter, 2017). Kickstarter further states “backers appreciate regular, insightful, and honest updates. Do not be hesitant to communicate delays or changes to your original plans – or to just check in”.

Additionally, posts within the Kickstarter community are a form of communication and frequent communication increases benevolence within the community (Cohen et al., 2010). At last, projects without an update had a significantly lower success rate (32,6%) compared to projects with updates (58,7%) in a research conducted on 8,529 campaigns by Xu et al. (2014).

Combining the views of Shane and Calbe (2002) about the positive effect of reputation on investment with the views of Kietzmann et al. (2011) about reputation measured in number of posts and with the views of Lang and Lang (1988) about the visibility of past activity plus the positive relation between updates and reputation from Hui et al. (2012), this research expects that the reputation within the Kickstarter community will positively affects the success the campaign. Hence, the following hypothesis is proposed:

H3. The higher the reputation of a founder within the Kickstarter community the more likely his campaign is to succeed.

(13)

2.4. Social media

Founders encourage their potential backers and actual backers to share the project on their social media by means of updates. It was found that most updates (23%) contained a plea to share the project on the social media of the backers (Xu et al., 2014). Research of Ordanini et al. (2011) state that activating the backers or the social network to share the campaign in the last phase indeed increases the chances of success. Kickstarter also supports the sharing by providing the button “share” on the project page, by clicking on this button the project can directly be shared on the different platforms: Facebook, Twitter, Tumblr and Pinterest (Kickstarter, 2017). The word-of-mouth as aforementioned can be done effortlessly by using this “share” button (Colombo et al., 2014). Instead of spreading the word to a few friends in traditional forms of word-of-mouth, consumers now are able to tell thousands of other people with a simple click (Mangold and Faulds, 2009). Social media platforms such as Facebook and Twitter are often used to execute word-of-mouth marketing strategies in online environments (Groeger and Buttle, 2014).

Using one’s own or business social media accounts to promote the crowdfunding campaign are highly recommended by successful crowdfund founders. Moreover, also asking your online friends to post it on their social media accounts is recommended (Xu et al., 2014). Social network sites, Facebook and Twitter, are found to be important platforms for founders to connect to potential backers and friends who are willing to provide financial and informational support (Bechter et al., 2011). As potential backers, do not have the opportunity to experience the quality of the products and services before consumption it is necessary to inform and convince these potential backers, which can occur on social network sites (Zheng et al., 2014). This communication with potential investors, also backers, is an important element of leveraging social capital (Nahapiet et al., 1998; Hazleton and Kennan, 2000). Communication with potential backers happens easily on social network sites (Lambert and Schwienbacher, 2010) and according to Ordanini et al. (2011) it is a necessary to post crowdfunding projects on Facebook and Twitter to gain visibility. Earlier research already found that linking a Twitter account to the project to spread the word of one’s crowdfunding campaign positively relates to the success of a crowdfunding project (Beier and Wagner, 2014).

(14)

In sum, in order to connect and communicate with potential backers and friends who are willing to offer financial or informative resources combined with the results of Beier and Wagner (2014) this research proposes the following hypothesis:

H4. Having a social media page on Facebook or Twitter linked to the crowdfunding campaign the more likely the campaign is to succeed

As it is not only the founder who is posting this research also analyses the effect of the posts in total as explained below. Visitors, or potential backers, can become promoters of the campaign when they promote the project among their friends through their online social networks. The research from Lu et al. (2014), which researches the diffusion of crowdfunding campaigns on Twitter, state that once a project is promoted by someone, that one is more likely to back the project in the future. Which is common-sense, as sharing a campaign comes with a certain sign of interest in the campaign. Also, in the research of Lu et al. (2014) the number of backers is highly correlated to the volume of the promotional activities, while crowdfunding success is more correlated to the design of the promotional campaign. The research concludes that reaching a lot of potential backers is done by massive promotion and crowdfunding success is established by intensive interactions with those potential backers.

The question remains how to get to get these consumers to share the campaign with their social network. Mangold and Fauls (2009) conclude in their research that consumers like to network with others who share interests and desires with them. Moreover, consumers are more likely to communicate through social media and traditional word-of-mouth when they are engaged with the product, service, or idea. If a project is more persuasive to intrigue consumers to discuss it in their social network, either online or traditional, the project is also more attractive to investors. Therefore, to enhance success projects should target specific groups that have interest in the campaign. Then when a group shares this interests this group is more correlated to the funding ratio (Lu et al., 2014). The research also states that is not simply the size of the online social network of the founder but the diffusion of information through social media. This makes sense, the more people get to know a crowdfunding project, the higher the chance that a group or individual with interest in the campaign encounters the project. Accordingly, the research of Thies et al. (2014), argues that the word-of-mouth

(15)

generate traffic to the project page compared to using the Facebook account linked to the project page.

This research expects that sharing or posting the link of a campaign, also called the social media diffusion, increase the likelihood of a successful crowdfunding campaign resulting in the hypothesis:

H5. The higher the social media diffusion the more likely a campaign is to succeed As mentioned earlier founders encourage the visitors of their projects page and their social network to share the project link on social media (Xu et al., 2014, Ordanini et al., 2011). Zooming in on this effect, this research expects that the bigger the social network of the founder, internally and externally, will possibly influence the social media diffusion positively. As founders encourage their network and leverage their reputation to share the project on social media by means of updates or messaging their existing contacts. Therefore, this research expects that the social media diffusion acts as a mediator between reputation and social network on the one hand and success rate of the crowdfunding project on the other hand. Elaborating on that, the greater the size of the social network or reputation the greater the social media diffusion.

In contrast, with the use of logic reasoning this research also expects that there is a certain trade-off or moderating effect between the social network and reputation of a founder and the social media diffusion of the campaign. Since, the more ‘viral’ a campaign will go on social media, the less important the social network of the founder will become to predict the success rate of the campaign. It is expected that social media diffusion lowers the effect of the internal reputation and social network on the success rate of a crowdfunding project. As these effects are merely gut feelings, this research examines the effect of social media diffusion on the relation between the constructs reputation and social network on the success rate of the crowdfunding campaign. Hence, this research will check for moderating effects as for mediating effects.

(16)

In sum, this research researches the following hypothesis:

H1. The more a founder backed other projects, the more likely his own crowdfunding campaign is to succeed.

H2. The larger the external social network of a founder, the more likely his crowdfunding campaign is to succeed.

H3. The higher the reputation of a founder within the Kickstarter community the more likely his campaign is to succeed.

H4. Having a social media page on Facebook or Twitter linked to the crowdfunding campaign the more likely the campaign is to succeed.

H5. The higher the social media diffusion the more likely a campaign is to succeed. The visualization of these hypotheses on the outcome success rate are captured in a conceptual model which is presented below (figure 1).

(17)

3. Methodology Part

This research used primary data about crowdfunding extracted from Kickstarter, the most dominant reward-based crowdfunding site. Data from Kickstarter is also used in previous studies (Kuppuswamy and Bayus, 2013; Mollick, 2014; Colombo et al. 2014). To measure the social media activities around the crowdfunding projects, social media posts and pages are collected from social network sites: Facebook and Twitter. At first this chapter will briefly explain how the data is collected. Secondly, how the different constructs are measured. Lastly, validity and reliability are discussed.

3.1. Data collection

The data used in this research is gathered from publicly available information on the Kickstarter, Facebook and Twitter websites. The dataset obtains projects launched after the 1st_{of April 2017 and completed before 1}st_{of June 2017. This time-period captures the} campaign lengths, ranging from 1 day to 60 days, as allowed by Kickstarter (Kickstarter, 2017). For each project, this paper recorded different variables as further elaborated on in the construct measurement section. The method used to gather how often a certain Kickstarter project is shared on Facebook is web scraping. A freely accessible program called “shared count” is used to automatically scrape data from Facebook. Other variables from the platforms Kickstarter, Twitter and Facebook are subtracted by hand.

3.2. Construct measurements

To facilitate the understanding of the measurement of the constructs used in this paper the constructs and how they are measured will be briefly explained below.

Success rate is the dependent construct used in this research, this entails the success percentage of a crowdfunding campaign. To measure the success rate, this paper measures the: Funding ratio this is the ratio between the goal as set at the beginning of the campaign and the pledged money at the end of the campaign. A funding ratio of 1 or higher indicates a successful project. Projects can also be overfunded, then the projects raise more than their goal and therefore the ratio will be higher than 1. Pledged, is the amount of money that is raised by backers for a project at the end of the campaign. Goal, is the set goal in terms of money that the creator seeks for his project. Kickstarter follows an ‘all or nothing’ model where the founder can only collect the money from the backers if the funding goal is reached

(18)

(Kickstarter, 2017). Therefore, Success is a dummy variable of the success rate and it is defined that it is 1 when the amount of money raised exceeds the funding goal of the project within the set campaign duration in other words when the success rate is higher or equal to 1. For example, if a project raised 80% of the goal at the end of the campaign, this research talks about a Success rate of 0.80 and Success will be 0.

Internal Social Capital, Colombo et al. (2014) argued that the internal social capital within crowdfunding sites can be measured by the number of projects that the creator had backed at the time of launching her own campaign (backed_others). The authors state that this variable shows to which degree the creator has been supportive of other projects and therefor has established social contacts within the Kickstarter community

External Social Capital, is measured in Facebook friends of founder (FBF_founder). Creators can link their Facebook page with their project, if they do so, then Kickstarter shows the amount of Facebook friends that a creator has. This measurement for external social capital is conducted from Mollick (2014), his research records FBF_founder by the time of data collection rather than at the time of project initiation.

Reputation, to quantify and measure the reputation of the founder within the Kickstarter community, the number of updates of the measured Kickstarter project are counted resulting in the measurement: Updates. Also, this research collected the number of crowdfunding projects that a founder previously created on Kickstarter. This resulted in the measurement: CFP_created.

Social media page, to quantify the efforts from the founder to reach out to his potential (online) external social network, having a Facebook page or Twitter page linked to the campaign at the start of the campaign are measured. Having a Facebook or Twitter page offers the opportunity for potential backers outside the Kickstarter platform to be directed to the campaign and the pages also foster communication between founder and potential backer. This research measured this construct when the social media page link was either directly shown on the Kickstarter project page. Or either when typing the crowdfunding project name into the search option of Facebook or Twitter, a page with the same name showed up. For example, when a project was called “Pebble watch 2” and one searched Facebook for “Pebble watch 2” and the Facebook page “Pebble watch” or “Pebble watch 2” popped up, this was counted as having a Facebook page linked.

(19)

Social media diffusion, is measured as the number of times the Kickstarter project link is shared on Facebook resulting in the measurement: KS_shares.

Additional this paper will control for one control variable which is commonly used in crowdfunding research (Kuppuswamy and Bayus, 2013; Mollick, 2014; Zheng, 2014; Colombo et al., 2015) which is the duration of a project (Duration), this variable is the length of the campaign for a project measured in days.

3.3. Validity and Reliability

To keep the measurement error to a minimum, validity and reliability are considered. Validity entails that a test measures what it set out to measure conceptually (Field, 2009). All measurements are determined from previous academic researches and therefore are considered valid content and criterion wise. Social media diffusion is applied to another social network site, and should not impact the validity of the results. Reliability is the ability of the measure to produce the same results under the same conditions (Field, 2009). Since, the constructs are the same as in previous studies, it is expected that applying these measurements and methods on a future research, the same results would be produced.

(20)

4. Data research

This chapter report on the results derived from the empirical analysis. First this research describes how missing data is handled followed by descriptive statistics, normality checks and correlations for all variables. Second, the transformation of variables is described to fit normality followed by the descriptive statistics and correlations for the transformed variables. This chapter concludes with the hierarchical regression analysis and the exploration of the moderating and/or mediating effect of social media diffusion is described.

4.1. Missing data-points and descriptive analysis

While collecting the data, the missing data points where indicated by ‘99999’. The missing data were only related to the variable FBF_founder. If a founder of a crowdfunding project did not link his personal Facebook account to the project, this was indicated as a missing variable and “99999” was collected as number of Facebook friends from the founder. As this research examines the relationship between the external social capital of the founder and the project success rate, these missing variables might carry important information. As Mollick (2014) argues that it is better to not link the number of Facebook friends to the project if you have little Facebook friends, it might be that founders comply with these findings. To act upon this information, this research created a new variable FBF_connected, this variable is 0 when a founder did not connect his Facebook account to the project and 1 if he did. Creating this new variable possibly tackles another problem, as the sample size would be considerably reduced as there are 63 missing cases on the variable FBF_founder. The descriptive statistics can be found in table 1 below.

(21)

Table 1 Descriptive statistics

Variables N M SD Min. Max. skewness kurtosis

Funding_ratio 146 1.52 3.42 0 33.33 6.68 55.17 Success 146 0.54 0.50 0 1 -0.17 -2.00 Pledged 146 11660.40 23390.02 0 152413 3.29 12.68 Goal 146 27138.44 166936.09 19 2000000 11.57 137.20 Duration 146 30.57 9.85 7 61 1.13 3.06 Updates 146 3.68 6.09 0 38 3.03 11.11 CFP_created 146 2.15 4.43 1 37 5.98 38.94 backed_others 146 12.37 50.62 0 387 5.88 36.74 FB_page 146 0.61 0.49 0 1 -.47 -1.80 FBF_connected 146 0.57 0.50 0 1 -.28 -1.95 FBF_founder 83 880.84 1044.60 0 5000 2.29 5.91 TW_page 146 0.39 0.49 0 1 .454 -1.82 KS_shares 146 454.48 2049.11 0 24320 11.08 129.24

To check if the data is normally distributed, the skewness values from table 1 and the histograms of these variables are checked. Skewness values above ± 2 are considered acceptable to prove normal distributed data (Fields, 2009). Based on the skewness, there is a lack of symmetry in the variables Funding_ratio, Pledged, Goal, Updates, CFP_created, backed_others, and FBF_founder. Indicating that these variables are not normally distributed. The histogram plots, who visualize the data conform the same thing, the data is skewed to the left. As only continuous variables can be normal distributed there is no need to check the dichotomous variables Success, FB_page, TW_page and FBF_connected

To check the correlations of the variables the Spearman’s rank correlation efficient is used. As this method provides a non-parametric measure of rank correlation, it can be used when the data violates the parametric assumptions for normally distributed data (Field, 2009).

(22)

Correlations tell something about the relationship between the variables, which can be used to get some more understanding about the relationships and how variables interact with each other. As can be seen in table 2 some of the independent variables are related to the dependent variable. The Spearman’s correlation test shows that there is a positive relationship between the dependent variable funding ratio (Funding_ratio) and the following variables: the dummy variable success (Success) (r = .863; p<.01) and the amount of money pledged (Pledged) (r = .775; p<.01) also the funding ratio is negatively related to the goal of the campaign (Goal) with r = -.163 and p<.05.

The correlations between these variables are not surprising as Success is 1 for a Funding_ratio above or equal to 1 and Funding_ratio is derived from Goal and Pledged money. More interesting, is the negative correlation between Goal and Funding_ratio which could possibly indicate the more money a founder askes, the less likely his crowdfunding campaign is to succeed.

As this research works with the dependent variable funding ratio, only correlations with that variable are further elaborated on. The funding ratio, is also positively related with the independent variables: number of updates (Updates) (r= .649; p<.01) , number of created projects by the founder (CFP_created) (r=.221; p<.01) , number of backed others projects by the founder (backed_others) (r=.343; p<.01) , number of times the Kickstarter project link was shared on Facebook (KS_shares) (r=.547; p<.01) and the number of Facebook friends of the founder (FBF_founder) (r=.268; p<.05). These are promising results, as it entails that most the independent variables are related to the dependent variable.

Further there are strong positive correlations between the number of updates within a project (Updates) the number of backed other projects (backed_others_ (r = .394; p<.01) , whether there is a Facebook page linked to the Kickstarter campaign (FB_page) (r=.1.84; p<.01) , whether there is Twitter page linked the Kickstarter campaign (TW_page) (r=.222; p<.01) and the number of times the Kickstarter project link was shared on Facebook (KS_shares)(r=.583; p<.01) .

(23)

Other strong correlations are between the number of projects created by a founder (CFP_created) and the number of other projects backed (backed_others) (r=.358; p<.01). The correlation between backed other projects (backed_others) and number of times the Kickstarter project link was shared on Facebook (KS_shares) is r =.202 with p<.05. Having a Facebook page linked to the Kickstarter project (FB_page) is strongly correlated with having a Twitter page linked to the project (TW_page) (r=.377; p<.01) and to the number of Kickstarter project link shares on Facebook (KS_shares)(r=.334; p<.01). As these variables, all refer to the social media activity of the founder, these correlations are not that surprising.

Lastly, having a Twitter page linked to the project (TW_page) is correlated with the number of Kickstarter project link shares on Facebook (KS_shares) with r = .210 and p<.05.

If a founder has connected his personal Facebook page to his Kickstarter project, so that potential backers can see the number of Facebook friends of the founder (FBF_connected) is positively correlated with the number of times a Kickstarter project link is shared on Facebook (KS_shares) (r=.247; p<.05). The relation between FBF_connected and FBF_founder is missing as FBF_connected is a dichotomous dummy variable for FBF_founder.

(24)

Table 2 Spearman's rho correlations Variables Mean SD 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 1. Funding_ratio 1.52 3.42 - 2. Success .54 .50 .863** - 3. Pledged 11660.40 23390.02 .775** .689** - 4. Goal 27138.44 166936.09 -.165* -.120 .352** - 5. Duration 30.57 9.85 -.056 -.078 -.004 .110 - 6. Updates 3.68 6.09 .649** .534** .662** .105 -.050 - 7. CFP_created 2.15 4.43 .221** .141 .087 -.157 -.049 .154 - 8. backed_others 12.37 50.62 .343** .241** .292** -.019 -.114 .394** .358** - 9. FB_page .61 .49 .087 .052 .277** .197* -.031 .175* .007 .065 - 10. FBF_connected .57 .50 -.012 .086 -.004 .031 .103 .024 .010 .115 .011 - 11. FBF_founder 880.84 1044.60 .268* .292** .272* -.060 -.060 .082 .111 .112 .142 . - 12. TW_page .39 .49 .150 .117 .264** .178* -.013 .222** .099 .160 .381** .017 .164 - 13. KS_shares 454.48 2049.11 .547** .506** .808** .330** .008 .583** -.067 .202* .337** .057 .247* .210* -

(25)

4.2. Transformation and descriptive statistics

To address the issue of non-normally distributed data, the variables have been transformed by using the natural logarithm. As the descriptive table showed, there are variables with values of zero. Since the natural logarithm of zero is undefined, there is a 1 added to these variables. The transformation then becomes Ln(1+X), where X stands for the old variable. The measurements Goal and CFP_created are transformed using the Ln (x) as they do not have values of zero. The variables Funding_ratio, Pledged, Updates, backed_others, KS_shares and FBF_founder are transformed using the Ln (x + 1) as they possibly have values of zero. Table 3 contains the new measurements and how they are computed based on the old measurements.

Table 3 Computed measurements

New measurement Computed as

LN_Funding_ratio Ln ( 1 + Funding_ratio ) LN_Pledged Ln ( 1 + Pledged ) LN_ Goal Ln ( Goal ) LN_Updates Ln ( 1 + Updates ) LN_CFP_created Ln ( CFP_created ) LN_backed_others Ln ( 1 + backed_others ) LN_KS_shares Ln ( 1 + KS_shares ) LN_FBF_founder Ln ( 1 + FBF_founder )

After the transformation of the variables the same descriptive analyses are conducted on the data; table 4. It can be stated that the distribution improved significantly. Although still some variables appear to have relatively non-normal distributions based upon the skewness, LN_CFP_created (skewness = 2.80) and LN_backed_others (skewness = 1.98). Although this research adheres the variables as normal distributions.

(26)

Table 4 Descriptive statistics transformed variables

Variables N M SD Min. Max. skewness kurtosis

LN_Funding_ratio 146 0.64 0.63 0 3.54 1.49 3.59 Success 146 0.54 0.50 0 1 -0.17 -2.00 LN_Pledged 146 7.01 3.07 0 11.93 -0.84 0.13 LN_Goal 146 8.51 1.65 2.94 14.51 -0.14 1.33 Duration 146 30.57 9.85 7 61 1.13 3.06 LN_Updates 146 1.00 0.99 0 3.66 0.59 -0.63 LN_CFP_created 146 0.31 0.68 0 3.61 2.80 8.42 LN_backed_others 146 0.82 1.37 0 5.96 1.98 3.44 FB_page 146 0.61 0.49 0 1 -0.47 -1.80 FBF_connected 146 0.57 0.50 0 1 -0.28 -1.95 LN_FBF_founder 83 6.10 1.40 0 8.52 -1.30 3.98 TW_page 146 0.39 0.49 0 1 0.45 -1.82 LN_ KS_shares 146 4.30 2.13 0 10.10 -0.41 -0.31

To check the correlations of the transformed variables the Pearson’s correlation efficient is used. As this method provides the correlation between normally distributed variables (Fields, 2009). As can be seen in table 5 the correlations do not change a lot. However, the correlation between the number of Facebook friends of the founder (LN_FBF_founder) is not statistically significant related with LN_Funding_ratio anymore compared to the Spearman’s correlation. Also, the correlation between number of created projects (LN_CFP_created) is not correlated significantly anymore with the number of updates (LN_Updates).

(27)

4.3. Outliers

Observations in a data set are considered outliers when they are distinctively different in from main trend of the other observations or data (Field, 2009). Outliers can be data points which are type errors and then should be excluded from the regression analyses as they would bias the results. However, outliers can also be unique cases which carry unique characteristics. If these unique cases are excluded from the analyses these unique characteristics would get lost or will not be discovered (Field, 2009). Moreover, by transforming the data with the natural logarithm significant different observations which can cause skewness will be reduced (Field, 2009). Therefore, this research decides to keep all the data points as they did not carry any typing errors.

(28)

Table 5 Pearson's associations Variables Mean SD 1 2 3 4 5 6 7 8 9 10 11 12 13 1. LN_Funding_ratio .64 .63 - 2. Success .54 .50 .758** - 3. LN_Pledged 7.01 3.07 .663** .658** - 4. LN_Goal 8.51 1.65 -.174* -.157 .184* - 5. Duration 30.57 9.85 -.096 -.131 -.071 .116 - 6. LN_Updates 1.00 .99 .627** .523** .619** .072 -.121 - 7. LN_CFP_created .31 .68 .270** .162 .113 -.138 -.082 .206* - 8. LN_backed_others .82 1.37 .258** .254** .249** -.052 -.122 .378** .577** - 9. FB_page .61 .49 .073 .060 .273** .193* -.029 .175* .042 .089 - 10. FBF_connected .57 .50 -.076 .086 .014 .028 .053 .017 -.030 .119 .019 - 11. LN_FBF_founder 6.10 1.40 .165 .201 .176 -.042 -.096 .032 .088 .100 .139 .c - 12. TW_page .39 .49 .104 .117 .230** .119 -.036 .212* .110 .156 .377** .017 .180 - 13. LN_KS_shares 4.30 2.13 .470** .508** .792** .246** -.046 .554** .028 .167* .312** .107 .200 .187* - ** Correlation is significant at the 0.01 level (2-tailed).

* Correlation is significant at the 0.05 level (2-tailed).

(29)

4.4. Assumptions check

To draw conclusions bout data based on a regression analysis done in a sample several assumptions must be true (Berry ,1993; Field, 2009). The variable types, multicollinearity and normality of residuals are discussed in this chapter.

Variable types

All variables included in the regression analyses should be quantitative or categorical (with two categories), and the outcome variable must be quantitative, continuous and unbounded. In this research, all variables used meet these requirements. Accordingly, the predictor variables are either dichotomous variables or ratio (discrete) variables, the dependent variable is a ratio variable.

Multicollinearity

As there should not be a perfect linear relationship between two or more predictors. So, the predictor variables should not correlate too highly (Field, 2009). Therefore, the variance inflation factors (VIF) of the variables are checked. Variables should have VIF’s below 10 and correlations below .90 (Field, 2009), which were met by all variables; table 5 above and table 16 and 17 in Appendix A. The VIF’s were checked after this research ran the regressions as reported on in chapter 4.5. In sum, the assumption of no multicollinearity is met for all variables.

Normality of residuals

The assumption to be met is that the residuals in the model are random normally distributed variables with a mean of 0. In order to test this assumption, the histogram and P-P plot of the standardized residuals are examined; which can be found in Appendix A. The histogram indicates that the standardized residuals of LN_Funding_ratio are normal distributed. Also, the P-P plots roughly indicates that the standardized residuals are normally distributed. Lastly a scatterplot of the standardized residuals and standardized predicted value is presented in Appendix A. The scatterplot also meets the assumptions of randomly and evenly dispersed throughout the plot (Field, 2009). Hence, the assumptions of the normality of the residuals is met.

(30)

4.5. Hierarchical regressions

This research will examine the relation between the independent variables and the dependent variable Funding_ratio. To check the direct effects of the independent variables on Funding_ratio. A hierarchical regression analysis is used to examine the relation between the independent variable and Funding_ratio. This type of regression was conducted to research the ability of the independent variables to predict the crowdfunding Funding_ratio, after controlling for campaign Duration. As a first step of the regression, this control variable Duration is entered as predictor. This was done so that a shared variability of this variable with the independent variable Funding_ratio can be controlled. Thus, the observed effect of the independent variables on Funding_ratio is independent of the effect of this control variable. As the sample size will differ when using either the dichotomous predictor FBF_connected or the continuous predictor LN_FBF_founder, this research will analyze two hierarchical regressions. One with FBF_connected used as predictor and one with LN_FBF_founder used as predictor.

First hierarchical regression

The first hierarchical regression was conducted to predict project Funding_ratio for 82 projects using Duration, LN_Updates, LN_CFP_created, LN_backed_others, FB_page, LN_FBF_founder, TW_page, LN_KS_shares as predictors.

In the first step of hierarchical multiple regression, one predictor was entered: Duration. This model was not statistically significant F (1, 82) = 0.75; p = .39 thus p > 0.05 this model explained 0,9% of variance in Funding_ratio.

After entry of the other independent variables at Step 2 the total variance explained by the model as a whole was 46,6% F (8, 82) = 8.07; p < .001. The introduction of the independent variables explained additional 45,7% in variance in Funding_ratio, after controlling for Duration (R2 Change = .457; F (7, 81) = 9.04; p < .001). In the final model two out of eight predictor variables were statistically significant, with LN_Updates recording a higher Beta value (ß = .34, p < .001) than LN_CFP_created (ß = .19, p < .05). In other words, if LN_Updates increases for one, the LN_Funding_ratio will increase for 0.34 units. On the other hand, if LN_CFP_created increases for one, then LN_Funding_ratio will increase for 0.19 units.

(31)

The results derived from step 2 in the model indicate that the relationship between the more other project a founder backed and the success rate of the project is statistically insignificant and has a negative relationship. This result implies that the number of projects backed by the founder does not impact the success rate of his own project. Hence, hypotheses one (H1)) is not supported. Also, the number of Facebook friends of the founder are not statistically significant related to the success rate of the crowdfunding project. Indicating that the external social network of a founder does not impact the success rate of the project. Hence, the second hypothesis (H2) is not supported.

This research did found support for hypothesis 3 (H3). As the number of updates has a strong and positive relation with the success rate of the campaign. This entails that when increasing the number of updates, the success rate is increased as well. Also, when the number of previous projects created by the founder are increased the success rate will increase as well, since they are strongly and positive related. However, hypothesis 4 (H4) and hypothesis 5 (H5) are not supported by this regression as having a Facebook page or Twitter page connected to the Kickstarter project do not have a statically significant relation with the success rate. Moreover, the number of times a Kickstarter project link is shared on Facebook is also statically insignificant related to the success rate of a crowdfunding project. The first hierarchical regression model can be found in table 6 below.

(32)

Table 6 First hierarchical Regression LN_Funding_ratio

Predictor variable R R2 _R2_{Change B} _S.E. _ß _t

Step 1 .10 .01 Duration -.01 -.01 -.10 -.86 Step 2 .68 .47*** .46*** Duration -.00 .01 -0.01 -.12 LN_Updates .34*** .07 .53 4.76 LN_CFP_created .19* .10 .21 1.97 LN_backed_others -.04 .05 -.10 -.86 FB_page -.11 .12 -.08 -.87 LN_FBF_founder .05 .04 .12 1.29 TW_page -.05 .12 -.04 -.45 LN_ KS_shares .06 .03 .20 1.81

(33)

Second hierarchical regression

Next a second hierarchical regression is conducted were the independent variable LN_FBF_founder is replaced with the dichotomous variable FBF_connected. The useful sample size increases to 146 by doing so. So, LN_Funding_ratio is now predicted based on the independent variables LN_Updates, LN_CFP_created, LN_backed_others, FB_page, FBF_connected, TW_page and LN_KS_shares while controlling for Duration.

In the first step of hierarchical multiple regression, one predictor was entered: Duration. This model was not statistically significant F (1, 145) = 1.33; p = 0.25 thus p > 0.05 this model explained 0,9% of variance in Funding_ratio.

At step two the other independent variables were added to the model. The total variance explained by the model as a whole was 46,2% F (8, 145) = 14.72; p < 0.001. The introduction of the independent variables explained additional 45,3% in variance in Funding_ratio, after controlling for Duration (R2 Change = .453; F (7, 144) = 16.48; p < <.001). In the final model three out of eight predictor variables were statistically significant, with LN_Updates recording a higher Beta value (ß = .32, p < 0.001) than LN_CFP_created (ß = .18, p < .01) and LN_KS_shares (ß = .07, p < .01) .In other words, if LN_Updates increases for one, the LN_Funding_ratio will increase for 0.32 units. On the other hand, if LN_CFP_created increases for one, then LN_Funding_ratio will increase for 0.18 units. Lastly, if the number of LN_KS_shares is increased with one, LN_Funding_ratio will increase with 0.07 units.

In line with the first hierarchical regression with LN_FBF_founder as predictor, the first hypothesis (H1) is not supported as backing others is insignificantly related to the success rate of a crowdfunding project. As the second regression, does not capture the relation between the number of Facebook friends of the founder and the success rate, drawing conclusions on hypothesis 2 (H2) cannot be done. However, based on the outcomes there can be concluded that connecting the personal Facebook account from the founder to the Kickstarter project so that potential backers can assess and see the number of Facebook friends of the founder is statistically insignificant related to the success rate of a project.

(34)

Hypothesis 3 (H3) is also supported by the second regression as number of created projects and number of updates are both strongly statistically significant and positive related to the success rate of a project. Also, the second regression did not found any statistically significant relation between having a Facebook page or Twitter page linked to the Kickstarter project and therefore hypothesis 4 (H4) is not supported.

Interestingly in the second regression the number of times the Kickstarter campaign is shared on Facebook becomes statistically significant related to the success rate of a project. As this relation is positive, the more times a Kickstarter project link is shared on Facebook the more likely a crowdfunding project is to succeed. Hence, hypothesis 5 (H5) is supported by this second regression. The second hierarchical regression model can be found in table 7 below.

Table 7 Second hierarchical regression LN_Funding_ratio

Predictor variable R R2 _R2_{Change B} _S.E. _ß _t

Step 1 .10 .01 Duration . -.01 -.01 -.10 -1.15 Step 2 .68 .46*** .45*** Duration -.00 .00 -.02 -.24 LN_Updates .32*** .05 0.50 6.11 LN_CFP_created .18** .07 .20 2.59 LN_backed_others -.03 .04 -.07 -.82 FB_page -.11 .09 -.09 -1.23 FBF_connected -.12 .08 -.09 -1.46 TW_page -.03 .09 -.02 -.34 LN_KS_shares .07** .02 .24 3.05

(35)

4.6. Explorative research for mediation or moderation effect

Next this research will explore the effect and relation of the social media diffusion on the different statistically significant relationships between the independent variables, also predictors, and the dependent variable success rate. Thus, the relationship between reputation of the founder and success rate is explored. To explore this effect this research uses the statistical program PROCESS, which is an add-on for SPSS designed by Andrew F. Hayes (2012). Possible moderating and mediating relationships between the different independent variables and the dependent variable funding ratio will be explored.

Moderating effects models

As in PROCESS, to examine the single moderation (model 1) effect one can only enter one independent (X), one dependent (Y) and one moderator variable (M), this research ran several analyses. Every analysis will cover one statistically significant independent (X) variable while controlling for the other independent variables. Additionally, in line with the earlier hierarchical regressions the analyses are run with the number of Facebook friends of the founder LN_FBF_founder as control variable and with FBF_connected as control variable.In sum, there will be four analyses for the moderation effect of social media diffusion. Firstly, an analysis with the number of updates (Updates) as independent X variable, which will be referred to as moderation model A, containing the number of Facebook friends of the founder (LN_FBF_founder) as one of the control variables. Secondly, the same moderation model 1 will be explored but with FBF_connected as one of the control variables instead of LN_FBF_founder. Thirdly, the moderation effect of social media diffusion on the relationship between the number of created projects by the founder (LN_CFP_created) will be explored, this model will be referred to as moderation model B. This model will be explored with LN_FBF_founder as one of the control variables, thereafter the same model will be explored but now with FBF_connected instead of LN_FBF_founder as one of the control variables.

(36)

Moderating effect model A

To check for a possible moderating effect of Social media diffusion measured as LN_KS_shares (M) on the relation between reputation measured as LN_Updates (X1) the single moderation model in PROCESS is used. The dependent variable Success is measured as LN_Funding_ratio (Y). This entails model 1 in PROCESS with the control variables LN_CFP_created, LN_backed_others, LN_FBF_founder, Duration, TW_page and FB_page. A conceptual visualization of model A is shown in figure5 in Appendix A.

The regression coefficient for X1M is c3 = -0.0587 and is statistically different from zero

(t (83) = -1.99, p<0.05). Thus, the effect of reputation as measured in number of updates (LN_Updates) on the success rate (Funding_ratio) is negatively influenced by the social media diffusion measured in number of times a Kickstarter project link is shared on Facebook (LN_KS_shares). This effect is also plotted in figure 6 (Appendix A). The outcomes of the explorative analyses for a possible moderating effect in model A with control variable LN_FBF_founder are presented below (table 8 and table 9).

Table 8 Moderating effect model A with control variable LN_FBF_founder

Coefficient SE t p Intercept i1 0.41 0.28 1.46 .147 LN_KS_shares (M) c1 0.0621 0.0336 1.85 .069 LN_Updates (X1) c2 0.25 0.07 3.60 <.001*** LN_Updates * LN_KS_shares (X1M) c 3 -0.0587 0.0295 -1.99 .050* LN_CFP_created (X2) c4 0.079 0.088 0.907 .368 LN_backed_others c5 -0.014 0.041 -0.352 .726 LN_FBF_founder c6 0.049 0.036 1.376 .173 Duration c7 0.001 0.005 0.236 .814 TW_page c8 -0.067 0.105 -0.638 .526 FB_page c9 -0.108 0.108 -0.997 .322 R2_{=0.421 p<.001***} F(9, 83) = 5.8995

(37)

Table 9 Conditional effect model A with control variable LN_FBF_founder

Conditional effect of LN_Updates (X1) on LN_Funding_ratio (Y) at levels of LN_KS_shares (M)

LN_KS_shares (M) Effect SE t p LLCI ULCI

-1.934 .365 0.102 3.570 0.0006*** 0.161 0.569

0.000 0.252 0.070 3.601 0.0006*** 0.112 0.391

1.934 0.138 0.076 1.813 0.0740 -0.014 0.290

R2_{increase due to interaction effect = 0.0314 p=0.0504*} F(1,74) = 3.956

In line with earlier hierarchical regressions this research also explores the possible moderating effect of Social Media diffusion (M) on the relation between reputation measured as LN_Updates (X1) and the dependent variable success rate (Y) as measured as LN_Funding_ratio, but now with FBF_connected instead of LN_FBF_founder as one of the control variables. Hence, the sample size went from 83 projects to 146 projects by doings so. The PROCESS model remains model 1 as this research explores the moderating effect. Controlled variables are: LN_CFP_created, LN_backed_others, Duration, TW_page, FB_page and FBF_connected.

The regression coefficient for X1M is c3 = -0.019 and is not statistically different from

zero (t (146) = -0.83, p = 0.406). Thus, the effect of reputation as measured in number of updates (LN_Updates) on the success rate (Funding_ratio) is not influenced by the social media diffusion measured in number of times a Kickstarter project link is shared on Facebook (LN_KS_shares). Therefore, the relation of reputation on success rate is not moderated by the social media diffusion. The outcomes of the possible moderating effect in model A with control variable FBF_connected are presented in table 10 and 11.

(38)

Table 10 Moderating effect model A with control variable FBF_connected Coefficient SE t P Intercept i1 0.801 0.150 5.336 <.001*** LN_KS_shares (M) c1 0.066 0.024 2.709 .008** LN_Updates (X1) c2 0.329 0.054 6.050 <.001*** LN_Updates* LN_KS_shares (X1M) c 3 -0.019 0.023 -0.833 .406 LN_CFP_created (X2) c4 0.188 0.071 2.633 .009** LN_backed_others (X3) c5 -0.034 0.038 -0.908 .366 Duration c6 -0.001 0.004 -0.164 .870 TW_page c7 -0.026 0.089 -0.296 .768 FB_page c8 -0.111 0.090 -1.228 .221 FBF_connected c9 -0.116 0.081 -1.436 .153 R2_{=0.4649 p<.001***} F(9, 146) = 13.1275

Table 11 Conditional effect model A with control variable FBF_connected

Conditional effect of LN_Updates (X1) on LN_Funding_ratio (Y) at levels of LN_KS_shares (M)

LN_KS_shares (M) Effect SE t p LLCI ULCI

-2.127 .370 0.084 4.427 <.001*** 0.205 0.535

0.000 0.329 0.054 6.050 <.001*** 0.222 0.437

2.127 0.289 0.060 4.804 <.001*** 0.170 0.408

R2_{increase due to interaction effect = 0.0027 p=.406} F(1,137) = 0.6935

(39)

Moderating effect model B

As number of created projects is a statically significant predictor of the success rate of crowdfunding projects, the possible moderation effect of social diffusion on this relationship is also explored. Figure 7 in Appendix A captures the conceptual model of this relationship also referred to as model B. Social media diffusion again measured as LN_KS_shares (M) is put into model 1 of PROCESS to check the possible effect on the relation between the reputation of a founder measured as CFP_created (X2) on the dependent variable success rate which is measured as LN_Funding_ratio (Y). Control variables are LN_Updates LN_backed_others, LN_FBF_founder, Duration, TW_page and FB_page. Note, the first analyses is with the number of Facebook friends of the founder as one of the control variables.

The regression coefficient for X2M is c3 = -.036 and is not statistically different from

zero (t (83) = -0.433, p = .659). Thus, the effect of reputation as measured in number of projects created by the founder (LN_CFP_created) on the success rate (Funding_ratio) is not statistically significant influenced by the social media diffusion measured in number of times a Kickstarter project link is shared on Facebook (LN_KS_shares). Results are presented in table 12.

Table 12 Moderating effect model B with control variable LN_FBF_founder

Coefficient SE t p Intercept i1 0.260 .254 1.021 .311 LN_KS_shares (M) c1 0.092 0.040 2.300 .024* LN_CFP_created (X2) c2 0.111 0.151 0.736 .464 LN_CFP_created * LN_KS_shares (X2M) c 3 -0.036 0.082 -0.433 .659 LN_Updates (X1) c4 0.193 0.074 2.592 .012** LN_backed_others (X3) c5 -0.002 0.043 -0.039 .969 LN_FBF_founder c6 0.040 0.037 1.087 .281 Duration c7 0.000 0.004 0.004 .997 TW_page c8 -0.088 0.130 -0.677 .501 FB_page c9 -0.098 0.121 -0.806 .423 R2_{=0.406 p<0.001} F(9, 82) = 7.149

(40)

Conditional effect of LN_CFP_created (X2) on LN_Funding_ratio (Y) at levels of LN_KS_shares(M)

LN_KS_shares(M) Effect SE t p LLCI ULCI

-1.934 0.181 0.186 0.977 .332 -0.189 0.551

0.000 0.111 0.151 0.736 .464 -0.190 0.413

1.934 0.041 0.248 0.166 .869 -0.453 0.535

R2_{increase due to interaction effect = 0.016 p=.639} F(1,74) = 0.196

Also, for model B the control variable LN_FBF_founder is interchanged with the control variable FBF_connected in order to increase the sample size to 146 projects. Hereby the number of projects created by the founder (LN_CFP_founder) (X2) is the independent variable, the success rate of a project (Funding_ratio)(Y) is the dependent variable and social media diffusion (LN_KS_shares)(M) is the possible moderator in model B.

The regression coefficient for X2M is c3 = -0.041 in model B and is statistically different

from zero (t (146) = -2.67, p<0.05). Thus, the effect of reputation as measured in number of projects created by the founder (LN_CFP_created) on the success rate (Funding_ratio) is negatively influenced by the social media diffusion measured in number of times a Kickstarter project link is shared on Facebook (LN_KS_shares). This effect appears logical, as the more the Kickstarter project link is shared on Facebook, the less important the reputation measured as the number of projects created by the founder becomes. The results are presented in table 13 and table 14 below.

Mobilizing the social network in crowdfunding

Master thesis: