Success factors for repeat behaviour in technology reward-based crowdfunding : an empirical study

(1)

SUCCESS FACTORS FOR

REPEAT BEHAVIOUR IN

TECHNOLOGY

REWARD-BASED CROWDFUNDING

AN EMPIRICAL STUDY

ABSTRACT

Crowdfunding is a relatively new concept, and repeat behaviour in crowdfunding has received limited academic attention. In this study I propose that a lower average amount pledged per backer, a high amount of backers, and a high amount of money pledged compared to the initial funding goal are success factors for repeat projects in technology crowdfunding on Kickstarter. I found that repeat projects reach, on average, more backers per project and reach a higher pledged-to-goal ratio, both indicators of success. Moreover, the results also suggest that a low average amount pledged per backer is associated with repeat project success, because this indicates that the project reached beyond the creators inner circle of friends and family.

Tom Lanting 10645764

Supervised by: Willem Dorresteijn

(2)

Statement of Originality

This document is written by Tom Lanting 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)

1

contributions (the crowd) instead of a small amount of large (expert) investors. Even though crowdfunding is relatively small in terms of total worldwide funding compared to traditional methods such as bank loans, it is a rapid growing market (see Massolution, 2013; 2015, as cited in Belleflamme, Omrani, & Peitz, 2015).

In 2015 technology company Pebble launched its third Kickstarter project ‘Pebble Time’. This project quickly broke all site records, reaching the funding goal of 500.000 USD within 17 minutes. At first glance it may seem logical that Pebble returned to Kickstarter for the third time after their previous successes, raising significantly more than their goals. However, repeat projects on a crowdfunding platform are not very common. Kickstarter does indicate, however, that repeat behaviour is increasing in popularity and in success on the platform.

In March 2015 popular crowdfunding platform Kickstarter reported that

approximately 22000 (12%) of its creators have used Kickstarter to acquire funding for at least two or more consecutive projects (Gallagher & Salfen, 2015). The platform also reports that these creators have success rates nearly twice the platform average, increasing after every next project launched. Moreover, the amount of backers also grows significantly with every new project, which helps creators to create a bigger community around their projects.

Kickstarter also provides insight into repeat project data for every category. From this, an interesting trend emerges when looking at the technology category specifically. Projects in the technology category have the lowest success rate of all categories (roughly 20%), and amount of repeat projects in this category is very low, at roughly 13%. However, when

technology creators do return to Kickstarter for repeat projects, their success rate increases by 75% on average, only beaten by the comic category.

These numbers suggest a large success potential for repeat technology projects that is mainly unexploited. Therefore, in this paper I will focus on answering the following question:

(5)

3 ‘What factors differentiate repeat projects from one time users in technology crowdfunding projects?’. To answer this question I will empirically analyse quantitative data on 8885

projects scraped from the technology category on Kickstarter.com by a PhD candidate at the University of Amsterdam, who provided me with the relevant dataset, to identify repeat projects and differentiate these from one time users in this category (Zhao, 2016).

Because crowdfunding is a relatively new concept, current academic literature on repeat crowdfunding is extremely limited, and focuses on cultural projects exclusively (Davidson, & Poor, 2016). Factors that differentiate repeat projects from one time users in technology projects have not been studied before. This paper aims to fill this gap by providing relevant insights in this particular realm of crowdfunding.

The second section of this paper contains an overview and review of existing relevant scientific literature. In the third section I then present a framework to link available literature to the research question and provide propositions. The fourth section explains methodological aspects on how the research question will be answered. The fifth section shows data analyses and corresponding results. Section 6 concludes and provides relevant discussion points.

2 Literature 2.1 Crowdfunding

Belleflamme, Lambert, and Schwienbacher (2010) provide a good working definition of crowdfunding: ‘’Crowdfunding involves an open call, essentially through the internet, for the provision of financial resources either in form of donation or in exchange for some form of reward and/or voting rights.’’ This definition suggests that there are multiple forms and reward structures in crowdfunding platforms, which will be discussed in the next section.

Crowdfunding is a derivative of a larger concept known as crowdsourcing, which involves using a large number of funders (the crowd) to access their resources, knowledge, or expertise (Hemer, 2011). In the case of crowdfunding the goal is to acquire funding through the crowd, often through a large amount of relatively small contributions. A major advantage of crowdfunding compared to traditional methods of acquiring funds is that it can be used as a marketing tool and to generate publicity for start-up firms (Belleflamme, Omrani, & Peitz, 2015). Crowdfunding platforms can be used to gauge future demand and build communities,

(6)

4

even before a product exists. Another advantage of using the crowd instead of a small teams or individuals is that a large amount of people may be better at solving problems because of its diverse make up (Brabham, 2008). Brabham (2008) explains this ‘wisdom of crowd’ by arguing that members of the crowd build upon others’ suggestions to come up with solutions.

Crowdsourcing and, therefore, crowdfunding, have been made possible by the emergence of and developments in what is known as ‘Web 2.0’ (Kleemann, Voß, & Rieder, 2008). Web 2.0 is a definition from 2004 by Tim O’Reilly, who defined it as the development of using the internet as a platform (Lee, DeWester, & Park (2008). Web 2.0 essentially refers to applications of the internet since 2004 that greatly enhance two-way interactive

communication (Kleemann et al., 2008). The developments of Web 2.0 have been imperative for the emergence of crowdsourcing and crowdfunding, because it facilitates the ability of companies and project creators to easily reach the crowd on a large scale (Schwienbacher, & Larralde, 2010). Lee et al. (2008) look at the phenomenon of Web 2.0 from different points of view, which can be applied to crowdfunding to identify the relevance of Web 2.0. Firstly, from a technological point of view, Web 2.0 means that computers can automatically process large amounts of information, facilitating the spread of information over the web. Secondly, from a sociological point of view, Web 2.0 is the basis for online networks that bring people with common interests together. Finally, from an economical point of view, Web 2.0 allows anyone to create, contribute to, and upload content online.

The interactions between and within funders and fundraisers are very interesting, and important to consider. In most cases funders rather fund projects on platforms that offer a large number of fundraisers and different projects, because of higher variety and, therefore, a higher chance to find a project that fits with the funders preferences (Belleflamme et al., 2015). Belleflamme et al. (2015) rightfully mention that this tendency is not always true. Funders may prefer smaller scale platforms when they fund a project that targets a niche audience, because there will be less competing projects on a smaller platform, which increases the chances that a particular project will be successful. On the other side, fundraisers prefer platforms with a large amount of funders (Belleflamme et al., 2015). A larger number of potential funders increases success chances and also increases the marketing and publicity gains.

(7)

5

2.1.1 Reward-based crowdfunding

Within the concept of crowdfunding there are 4 main forms of crowdfunding

platforms; (1) donation-based, (2) lending-based, (3) equity- and royalty-based, (3) and (4) reward-based crowdfunding platforms (Belleflamme et al., 2015; Giudici, Nava, Lamastra, & Verecondo, 2012).

Hemer (2011) rates these 4 main forms of crowdfunding platforms on complexity and uncertainty, shown in figure 1 below.

Figure 1 – The major forms of capital provision ranked by process complexity

(1) Donation-based crowdfunding platforms are popular for humanitarian and artistic related projects (Belleflamme et al., 2015). Donation-based crowdfunding is not associated with any tangible or financial rewards in return for contributing. A popular example of a donation-based platform is US-based GoFundMe (www.gofundme.com).

(2) Lending-based crowdfunding is similar to traditional bank loans. If a funded project is successful, funders are to receive an interest rate on the loan that they provide to the fundraiser (Belleflamme et al., 2015). This interest rate is determined through a contract between fundraisers and funders. However, in comparison to traditional loans from banks, a lending-based crowdfunding platform does not screen projects. Funders are free to decide

(8)

6

which projects should be funded. This is also known as peer-to-peer lending. An example of a lending-based crowdfunding platform is Prosper, based in the US (www.Prosper.com).

(3) In equity-based crowdfunding funders receive equity stakes in a project or company in exchange for funding the project (Belleflamme et al., 2015). An example of an equity-based crowdfunding platform is Crowdcube (www.crowdcube.com), based in the UK. Crowdcube said that they helped to fund over 200 projects for more than 35 million pounds as of March 2015 (Belleflamme et al., 2015).

On the other hand, equity-based crowdfunding can be problematic in certain countries (Belleflamme, Lambert, & Schwienbacher, 2014). In some countries there are serious

limitations to using the internet to attract a large number of funders in exchange for equity in the company or project. In most countries there are legal limits to the amount of private investors a firm can have, greatly limiting the potential of equity-based crowdfunding. Because of this other reward structures are more common, such as in reward-based crowdfunding. The focus in this paper lies on reward-based crowdfunding projects.

(4) It has been shown that rewards are one of the most important motivating factors for funders to contribute to a crowdfunding project (Gerber, Hui, & Kuo, 2012).. The main

difference between reward-based crowdfunding and the previously discussed forms of platforms is that funders are interested in non-financial returns for their funding activity (Belleflamme et al., 2015). Essentially, reward-based crowdfunding is when contributors receive rewards that are tangible, but non-financial, in exchange for their contributions (Kuppuswamy & Bayus, 2013). Often this means that funders play the role of ‘prosumers’, and essentially pre-purchase the product. This greatly reduces uncertainty for the fundraiser because a small market for the product is established during the funding phase. This market can also be an indicator of potential for future funding rounds (Belleflamme et al., 2015), which makes reward-based crowdfunding projects a particularly interesting field for studying repeat behaviour.

The most prominent example for a reward-based crowdfunding platform is US-based Kickstarter (www.Kickstarter.com), which will be the focus in this paper.

(9)

7

2.2 Kickstarter

Kickstarter is the most prominent example of a reward-based platform (Kuppuswamy & Bayus, 2013), and will be the focus of this paper, as it is the biggest platform worldwide (Belleflamme et al., 2015). At the time of writing this1_{, over 2.3 billion USD have been raised}

through Kickstarter, in a total of more than 100.000 successful projects since the launch of the platform in April 2009. Almost 11 million people have funded a project on Kickstarter, of which more than 30% have provided funding more than once. In total, these people have made a combined number of 30 million pledges. Roughly 36% of all projects reach their funding goal, and 86% of all dollars pledged go to successful projects.

Kickstarter makes money through applying fees to successful projects. Projects that do not reach their funding goal are not subject to any fees. On successful projects Kickstarter collects a 5% fee from the total amount funded, and a 3-5% payment processing fee. These fees and percentages are very common for most crowdfunding platforms.

Kickstarter differentiates projects between 15 different categories, mainly focusing on creative endeavours. The available categories on the platform as of May 2016 are: Art, Comics, Crafts, Dance, Design, Fashion, Film & Video, Food, Games, Journalism, Music, Photography, Publishing, Technology, and Theater. In this paper the focus will lie on the Technology Kickstarter category, for reasons explained in the next section.

2.2.1 Technology projects on Kickstarter

In this paper the focus lies on Kickstarter projects from the technology category. Kickstarter does not provide a detailed description of this category, and when a creator is starting a project he or she can freely choose which category to launch the project in. This means that any type of project could be listed as a technology project. However, to give an indication of the type of projects listed as technology projects that are used in this study, Kickstarter provides 15 subcategories within the technology category to choose from. These are 3D Printing, Apps, Camera Equipment, DIY Electronics, Fabrication Tools, Flight,

(10)

8

Gadgets, Hardware, Makerspaces, Robots, Software, Sound, Space Exploration, Wearables, and Web.

In this study all projects that are used for analysis were launched in the official Kickstarter technology category. This category is particularly interesting for several reasons. With over 22.000 projects launched it is one of the biggest categories on the site. However, only 20% of these projects reach their funding target within the time limit. It is also the category with most dollars contributed to unsuccessful projects. In terms of total successful dollars pledged the technology category is third largest, which indicated that these projects are, on average, significantly larger in terms of total funding compared to other categories. Moreover, second or later projects by the same entrepreneur in this category see a large increase in success rate of 75%. This increase is often not exploited, indicated by the low amount of repeat projects in this category, at approximately 13%. This unexploited potential makes this category especially relevant for studying repeat project success factors.

2.3 Repeat behaviour in crowdfunding

Even though crowdfunding is a new, rapidly growing phenomenon, success factors have been studied before. It has been found that quality signals play a significant role in predicting project successes (Mollick, 2014). Mollick (2014) also argues that crowdfunding can overcome physical geographic boundaries, it also creates new distance constraints. Online networks and communities are of great importance, and have a large impact on project

success. Also, a higher amount of backers is associated with project success (Davidson & Poor, 2016).

Within crowdfunding, a relatively unexplored field is repeat behaviour on

crowdfunding platforms. Repeat crowdfunding is about returning to a crowdfunding platform more than once over time (Davidson & Poor, 2016). Because it is a new phenomenon, repeat crowdfunding is a field that has not been subject to much scientific study. Studies looking at success factors in crowdfunding, among the ones mentioned in the previous paragraph, mainly use individual crowdfunding projects as the foundation of their analyses (Davidson & Poor, 2016).

(11)

9

So far, success factors in repeat crowdfunding projects have exclusively been studied by Davidson and Poor (2016). They looked at the Music, Publishing, Film, and Games categories on Kickstarter. They found that a small dollars pledged to backers ratio, a large number of supporters, and exceeding the initial funding goal within the time limit are factors that characterize repeat projects and entrepreneurs.

Davidson and Poor (2016) draw on gift-giving and social network theory to explain these findings. They argue for a comparison between gift-giving and crowdfunding, in which ‘gifts are exchanged both ways’. Douglas (1990, as cited in Davidson & Poor, 2016)

emphasized that gift-giving creates a sense of reciprocity. Applying this principle to repeat behaviour in crowdfunding, these findings imply that relying on a small circle of strong ties has serious limitations. Because of the sensation of reciprocity, a social network of strong ties can be a barrier for repeat crowdfunding, because it is often not possible to rely on this

network more than once.

Davidson and Poor (2016) draw on what Van de Rijt, Kang, Restivo, and Patil (2014) call success-breeds-success dynamics to explain the effect of exceeding the initial funding goal on repeat behaviour. Because the total amount of backers a project reached is, too, an indicator of success, success-breeds-success dynamics can also be applied to this finding. Van de Rijt et al. (2014) found that in a multitude of experiments initial success is a predictor of future success because it sends out a positive feedback signal. Van de Rijt et al. (2014) do mention, however, that success-breeds-success dynamics have decreasing marginal returns. This means that the predictive power of previous success decreases over time as more success is achieved. This finding is inconsistent with the numbers that Kickstarter reports on their blog on repeat crowdfunding behaviour (Gallagher & Salfen, 2015). The website reports that the total amount of backers increases significantly with an increasing amount of previously successfully funded projects (see Figure 2).

(12)

10

Figure 2

(Source: Kickstarter Blog: By the Numbers: When creators return to Kickstarter)

These findings by Davidson and Poor (2016) are used as a foundation for this study, and are further discussed in the next section.

2.4 Theoretical Framework

1. First project average level of contributions

The first expected factor that differentiates repeat crowdfunding projects from one time users is related to the average level of contributions. Davidson and Poor (2016) conclude that projects that are successful because of a small number of relatively very large

contributions often do not lead to a second project in the future. They explain this by stating that a small number of relatively very large contributions often indicates the help of a close relative or friend to save to project. Creators may not feel comfortable to rely on these ‘saviours’ again in future projects, and this sends a signal to the crowd that the project failed to reach beyond the social circle of the creator.

(13)

11

This prediction can be measured by looking at the Pledged-to-Backers ratio. A high ratio indicates that the project relied on a small amount of large contributions to succeed, and a low ratio indicates a low average contribution by a large number of backers. Davidson and Poor (2016) found that a higher Pledged-to-Backers (P/B) ratio decreases the chances of repeat behaviour in cultural crowdfunding projects. I predict that this finding also holds in technology projects. Therefore, I hypothesize that:

H1: A high Pledged-to-Backer ratio in the first project will decrease the likelihood of a second project in technology crowdfunding projects.

2. First project number of backers

The second important variable to look at is the total number of backers a project reaches the first time. Where a small number of backers suggests the reliance on friends or family, a large number of backers suggests that the project was able to actually reach and use the benefits of the crowd (Davidson & Poor, 2016). This implies that the project appealed to many different people, and indicates the formation of a community around this project. This is a proxy for future demand and may increase a creators confidence to launch a second project in the future. Moreover, a large amount of backers sends out a signal to the crowd indicating the success of the project. The prediction is that creators who launched multiple projects had a higher amount of backers in their first projects than one time users. Therefore, I hypothesize that:

H2: A large number of backers in the first project will increase the likelihood of a second project in technology crowdfunding projects.

3. First project performance

The final prediction is based on the idea that first project success sends out a signal to both creators and contributors. This is what Van de Rijt, Kang, Restivo, and Patil (2014) call success-breeds-success dynamics. Essentially this means that performance and success of first projects would increase chances of success of future projects. Davidson and Poor (2016) found that in most cultural categories repeat projects had higher Pledged-to-Goal ratios than one time projects. This ratio indicates the amount of total money pledged compared to the

(14)

12

funding goal. A ratio of 1 indicates that the project raised exactly its goal, a ratio of higher than 1 indicates the project exceeded its goal, and a ratio below 1 indicates the project raised less than its original goal. This ratio is used as a measure of success. I expect that success-breeds-success dynamics apply to technology crowdfunding projects, and therefore I hypothesize that:

H3: The higher the Pledged-to-Goal ratio, the higher the likelihood of a second project in technology crowdfunding projects.

3 Research Design and Methodology

Design and Sample

The hypotheses in this paper will be tested using quantitative data scraped from Kickstarter. In total 9.663 projects, both live and completed, have been scraped from the official Kickstarter technology category. This data has been scraped by Liang Zhao, PhD candidate at the University of Amsterdam, who shared the raw dataset with me for the purpose of this study (Zhao, 2016). At the time of scraping, this dataset represents the entire population of Kickstarter technology projects.

A variety of relevant information has been scraped for each project, such as project name, the funding goal, the amount pledged, the amount of backers, and the profile name of the creator. For 778 projects this data was missing or invalid for the purpose of this study, resulting in a dataset of 8885 projects valid for analysis. The profile name of the creator can be used to match creators to projects, to distinguish repeat projects from one time creators. Because not all projects were launched on Kickstarter in US dollars, actual conversion rates to USD at the time the project was live and, therefore, amounts in US dollars have been scraped as well.

Measurements

In order to perform analyses on the dataset, it is necessary to define relevant

measurements for the proposed predictions. In order to differentiate repeat projects from one time users, repeat projects will be separated from the other projects scraped. This allows for comparisons between these two groups. A simple binary variable will be created to indicate whether (1) a creator launched more than one project, or (0) not.

(15)

13

For the first hypothesis, the reliance on close social ties will be measured through a Pledged-to-Backer (P/B) ratio. This ratio looks at average contribution per backer. A reliance on close social ties such as family or friends is indicated by a high Pledged-to-Backer ratio, whereas a low Pledged-to-Backers ratio indicated a large amount of money pledged by a large amount of funders. The Pledged-to-Backers ratio can be computed by dividing the total amount of money pledged in US Dollars by the total amount of backers for each project.

For the second hypothesis, the ability of the first project to reach the crowd is measured through total amount of backers. This information is scraped directly from Kickstarter and does not require further manipulations.

For the third hypothesis, the first project’s success will be measured by the Pledged-to-Goal ratio. This ratio identifies how much money is raised in comparison to the initial funding goal. A higher ratio indicates that more money than the goal was raised within the timeframe and, therefore, represents a higher level of success. Projects that pledged less than the set funding goal are considered failed projects. The Pledged-to-Goal ratio is computed by

dividing the total amount of money pledged in US Dollars by the funding goal set by creators.

Analysis method

In order to differentiate repeat projects from one time users in this category the first step is to identify repeat projects by looking for matching project creator Kickstarter profiles, and separating these two groups. For the purpose of this study all projects with matching profile names are classified as repeat projects. This includes both finished and ongoing projects, because live projects also indicate repeat behaviour by the creator. Because the data represents the entire population at the time of scraping Kickstarter, a combination of

descriptive statistics and regression analyses are used to identify the proposed effects in these groups. Descriptive statistics can identify differences between repeat projects and one time users in terms of mean Pledged-to-Backer ratio, amount of total backers, and Pledged-to-Goal ratio for technology Kickstarter projects. Also, median values will be calculated and

compared with the mean values, because the median value may be a better representation of a typical project because a median value is more resilient to extreme values.

(16)

14

For both one-time projects and repeat projects mean and median values of the Pledged-to-Backer ratio, the amount of backers, and the Pledged-to-Goal ratio will be calculated and compared with each other.

Subsequently, regressions are then used to identify the direction of the effect of these measures on whether or not a creator launched a second project. In the regression, the binary variable indicating whether or not a creator launched more than 1 projects is the dependent variable. The independent variables are the Pledged-to-Backers ratio, the amount of backers, and the Pledged-to-Goals ratio. Because the dependent variable is binary it can therefore be classified as a categorical variable. Because of this, binary logistic regression analysis will be used to identify the direction of the effect of the variables on whether or not a creator made a second project. These direction coefficients are then compared to the previously stated predictions based on previous literature and empirical findings in order to draw conclusions and confirm or reject the hypotheses. Because the data represents the entire population at the time of scraping, p-values may be ignored (Davidson & Poor, 2016). Also, the regression analysis is used to identify the direction of the effect, without looking at the effect size.

4. Results

4.1 Descriptive statistics

In order to provide early insights into and an overview of the data used for this study, I first provide some relevant and interesting descriptive statistics for the entire dataset of 8885 projects. Table 1 shows frequencies and percentages of the state of all projects. These results give insight into the success and failure rates of all projects in the Kickstarter technology category.

Table 1: Frequencies and percentages (Total)

Frequency Percent 1. Successful 2. Failed 3. Live 4. Cancelled 5. Suspended 2097 5560 325 857 46 23,6 62,6 3,7 9,6 0,5 Note: N=8885

(17)

15

The scraped projects have a success rate of 23,6%. This is similar to the success rate of the entire technology category at 20%. This is not surprising because the sample used is a very large percentage of the entire population of Kickstarter technology projects. These numbers are not equal because the data has not been scraped at the same time of writing this paper. Therefore the dataset represents the entire population at the time of scraping, even though more projects have been launched between the time at which the data has been scraped and the time of writing this paper. Furthermore, 9,6% of the scraped projects were cancelled and 0,5% were suspended. Cancelled and suspended projects are included in further analyses, because a cancelled or suspended project still contributes to whether or not a creator has launched one or more projects. Some projects, however, were listed twice in the dataset because one version of the project was cancelled or suspended and relaunched by the creator. All duplicate projects have been removed from the dataset, along with other incomplete or invalid cases, in the initial cleaning process, and are not included in the 8885 projects that are used for analysis.

Table 2 provides some early insights into the state of repeat projects. Out of 8885 projects 774 projects have been identified as repeat projects, which is 8,7%. This is slightly lower than the percentage that Kickstarter reports. Kickstarter reports that approximately 13% of all technology projects are repeat projects. This difference is not surprising. As mentioned before, the data has been scraped at a different point in time than the writing of this paper. Kickstarter reports that the fraction of repeat projects compared to one-time users is increasing over time. Thus, it makes sense that the dataset used in this paper has a smaller percentage of repeat projects than the current population of Kickstarter technology projects. The 774 repeat projects include a creator’s first projects as well as all subsequent projects by the same creator in this category.

(18)

16

The numbers in table 2 imply that there are 8111 one time projects in the dataset. Of these 8111 projects 1858 are successful, resulting in a success rate of 22,9%.

Out of 774 repeat projects 30,9% are successful. When compared to the success rate of all projects, repeat projects are performing notably better, which is consistent with the trend that Kickstarter published on their blog (Gallagher & Salfen, 2015). However, success rate is not the only relevant measure to look at. It is also interesting to compare the amount of dollars pledged and the amount of backers for both groups, as this gives more clarification about the differences than just the success rate. This information is displayed in table 3.

The average amount of dollars pledged to projects is similar for both categories, but slightly lower for repeat projects. Although the amount of dollars pledged on average is lower for repeat projects, these projects do maintain a higher success rate. This indicates that

backers are pledging successful projects more than projects that end up failing and are, therefore, pledging more efficiently. This finding can be linked to the amount of backers reached, which is another indicator of success. Repeat projects are better able to reach the crowd by attracting, on average, 17 backers per project more than one-time projects do.

Table 2: Frequencies and percentages for Repeat Projects

Frequency Percent 1. Successful 2. Failed 3. Live 4. Cancelled 5. Suspended 239 407 23 103 2 30,9 52,6 3,0 13,3 0,3 Note: N=774

Table 3: Total and average USD pledged, and average number of backers

Total USD Avg. USD # Backers

1. Total 2. Repeat 140.010.887,30 10.859.662,36 15.758,12 14.030,57 134 151

(19)

17

In order to answer the research question the mean and median of the Pledged-to-Backer and Pledged-to-Goal ratio will be measured and compared between repeat projects and one-time users. First, however, it is relevant to look at these measures for the entire dataset of all 8885 projects. These values can be found in Table 4 below.

The average Pledged-to-Backers ratio is 74,21. However, as the median and maximum values indicate, this number does not tell everything. There is a large amount of Pledged-to-Backer ratios of 0 (N=1717), and several very large Pledged-to-Pledged-to-Backer ratios. These large values are often projects where only 1 backer was reached who pledged a very large sum, for example 5000 dollars. This results in a very high ratio, even though this suggests that 1 friend of family member pledged this money to either save the project or send a signal to the crowd that 5000 dollars have already been raised. For these reasons it is wise to compare both mean and median values, which will be done in the next section of this paper.

The average Pledged-to-Goal ratio is 0,9601. Here the same situation as above applies. Again, 1717 projects have a P/G ratio of 0. As the median and maximum in comparison with the mean indicate there are several very large P/G ratios, for example 1155. In this case the goal of the project was set at 1 dollar. As these cases indicate, it is again wise to look at the median as well as the mean values.

In the next section the mean, median, and maximum values for the P/B ratio, the amount of backers, and the P/G ratio will be compared between repeat projects and one-time users. Subsequently, regression analysis will be used to identify the direction of the effect of these variables on whether or not a project is a repeat project.

Table 4: Mean and Median P/B ratio and P/G Ratio

Median Mean Max

1. P/B Ratio 2. P/G Ratio 34,63 0,0245 74,21 0,9601 5000 1155 Note: N=8885

(20)

18

4.2 Empirical Results 4.2.1 Correlations

Because of possible high correlations between variables in the model I calculate the correlation between all variables to ensure validity and reliability of the results (See Appendix A). The correlation between the amount of backers and the P/G ratio (0,053), and the P/B ratio (0,025) were low. The correlation between the P/G ratio and the P/B ratio was also low (0,010). These low correlations are indications of low multicollinearity, which is good for the power of the model.

4.2.2 Results

A comparison of the median, mean, and maximum values of the P/B ratio, the amount of backers, and the P/G ratio can be found in Table 5.

Firstly I hypothesized (H1) that a higher Pledged-to-Backer (P/B) ratio decreased the chance of the creator coming back to Kickstarter for a second project. This means that I expect the median and mean values of the P/B ratio to be lower for repeat projects than for one-time users. The mean and median P/B ratio are, however, almost the same value for repeat projects as for one-time users. The medians are 33,2051 and 34,8571 for repeat projects and one-time users respectively. The mean values are 74,7647 and 74,1620 respectively

(Table 5). However, as the maximum values suggest, the ratios change significantly when controlling for outliers. The median and mean values between brackets represent results after eliminating outliers. In this case, outliers are defined as all P/B ratios higher than 1000 with less than 10 backers. For repeat projects 6 outliers were eliminated. For one-time projects 20 outliers were eliminated. The adjusted median and mean values are indeed lower for repeat projects than for one-time projects.

Also, based on theory I predicted the direction of the effect of the P/B ratio on whether or not a creator launched a second project later to be negative. However, the regression

coefficient was positive in relation to whether a project is a repeat project (Table 6). The adjusted median and mean values are lower for repeat projects than for one-time projects, but the direction of the effect of the P/B ratio on whether or not a creator launched a second project is not negative. These results only partially confirm the first hypothesis (H1)

(21)

19

Secondly I hypothesized (H2) that reaching the crowd beyond the circle of close ties, indicated by a higher amount of backers, increases the likelihood of a creator launching another project. This means that I expect that the median and mean values of the amount of backers are higher for repeat projects than for one-time users. The median amount of backers is slightly higher for repeat projects than for one-time users. The mean amount of backers are 151,25 and 132,64 for repeat projects and one-time users respectively (Table 5), indicating that repeat projects reach, on average, 19 more backers.

Also, as I expected, the direction of the effect of the amount of backers on whether or not a creator launched a second project is positive (Table 6). These findings indicate that repeat projects tend to have a higher amount of backers than one-time projects, confirming the second hypothesis (H2).

Table 5: Descriptive comparisons between Repeat projects and one-time users

2nd_Project? _Median _Mean _Maximum

1. P/B Ratio 2. Backers 3. P/G Ratio Yes No Yes No Yes No 33,2051 (32,74) 34,8571 (34,70) 7 6 0,0612 0,02284 (0,0226) 74,7646 (62,20) 74,1620 (69,63) 151,25 132,64 1,2131 0,9358 (0,599) 2545 5000 21975 105857 45,338 1155 Note: N=8111 (one-time users), N=774 (Repeat Projects)

Table 6: Binary logistic regression coefficient directions for relevant variables

Variable Theory predicts Result

1. P/B Ratio 2. Backers 3. P/G Ratio Negative Positive Positive Positive Positive Positive Note: Dependent variable is whether or not a project is a repeat project. See Appendix A for effect sizes

(22)

20

Thirdly, I hypothesized (H3) that initial success, measured by the ratio of the amount pledged compared with the funding goal, is an indicator of repeat behaviour. This means that I expect the P/G ratio to be higher for repeat projects than for one-time users. The median P/G ratio values are 0,0612 and 0,02284 for repeat projects and one-time users respectively, and the mean values are 1,2131 and 0,9358 respectively (Table 5). For the P/G ratio outliers were also eliminated. In this case, all ratios higher than 10 with a goal lower than 50 dollars were

identified as outliers. For repeat projects there were no such cases, and for one-time projects 12 outliers were eliminated. After eliminating these outliers, the mean P/G ratio for one-time projects dropped from 0,9358 to 0,599, indicated by the mean value between brackets in Table 5.

Also, as expected based on theory, the direction of the effect of the P/G ratio on whether or not a project is a repeat project is positive (Table 6). These results indicate that repeat projects tend to have a higher Pledged-to-Goal (P/G) ratio than one-time projects, confirming the third hypothesis (H3).

5 Discussion 5.1 Summary

The main goal of this study was to identify success factors for repeat behaviour in technology reward-based crowdfunding projects from Kickstarter. Specifically, I analysed whether success factors found by Davidson and Poor (2016) for repeat behaviour in music, publishing, film, and games categories also applied to projects in the technology category on Kickstarter.

Three main hypotheses were proposed. Firstly I expected that one-time projects have a higher average donation per backer than repeat projects (H1). This was measured by the Pledged-to-Backer (P/B) ratio. A high P/B ratio indicates dependence on a small number of large donations, which sends a signal that the project did not reach the crowd. This was partially supported by the data. After controlling for outliers repeat projects did indeed have a slightly lower P/B ratio than one-time projects. However, the direction of the effect on

whether or not a creator launched a second project, although small, was positive, thus the first hypothesis (H1) is only partially confirmed. Secondly I expected that a large number of backers increased the chances of a creator launching another project. Repeat projects did

(23)

21

indeed have on average 19 more backers per project, and the direction of the effect on whether or not a creator launched a second project was positive, confirming the second

hypothesis (H2). Thirdly, I expected that initial success would breed further success. This was measured by how much money was raised in comparison to the funding goal in the Pledged-to-Goal (P/G) ratio. I predicted that repeat projects would have a higher P/G ratio than one-time projects. This was confirmed by the data. Repeat projects did indeed have a significantly higher P/G ratio, especially after controlling for outliers. This indicates that exceeding the funding goal on the initial project increases the chances of the creator launching another project later on.

5.2 Discussion Points and Future Research

What do these results mean? In this study I assumed, based on the work by Davidson and Poor (2016), that crowdfunding shares characteristics of gift giving, reliance on close social ties, and success-breeds-success dynamics (Van de Rijt et al., 2014). A high Pledged-to-Backer ratio indicates that the project relied on a circle of close social ties such as good friends or relatives to save or kick-start the project. The results found in this study suggest that one-time projects rely on a circle of close social ties more than creators of repeat projects, indicated by a higher Pledged-to-Backer ratio for one-time projects. This sends a signal that the project did not truly reached the crowd, and that the project depended on gift-giving by a circle of close social ties. Moreover, a large amount of backers indicates that the project was indeed able to truly reach the crowd. The results in this study show that repeat projects reach the crowd better by reaching, on average, more backers than one-time projects. This suggests that having a large number of backers and truly reaching the crowd sends out a signal of success to the public, which encourages repeat behaviour in crowdfunding. This is what Van de Rijt et al. (2014) call success-breeds-success dynamics. Another indicator of success that distinguishes repeat projects from one-time users is the Pledged-to-Goal ratio. Repeat projects have, on average, higher P/G ratios than one-time projects, which also sends a signal of success to the public.

All of these results emphasize the importance of networking and reaching the crowd. However, there seems to be a difference in the composition of the crowd reached. One-time projects more often rely on close ties such as family and friends, whereas repeat projects tend to reach the crowd better. Also, previous research has established the importance of

(24)

22

geographical location on the type of project proposed and success (Mollick, 2014). Agrawal, Catalini, and Goldfarb (2011) have found that online crowdfunding platforms overcome most distance related costs and frictions. However, not all spatial frictions are eliminated through crowdfunding (Agrawal, Catalini, & Goldfarb, 2015). which is consistent with other literature on entrepreneurial finance. Moreover, research has established significant strategic investment differences based on culture. Carr and Tomkins (1998) have shown significant differences between the Anglo-American, the Continental European, and Japanese strategic approaches It would be very interesting for future research to further examine the differences of the

composition of the crowd in repeat crowdfunding projects and one-time projects based on geographical and cultural differences. Future research should combine crowdfunding project data with survey or qualitative measures to gain better insight in these differences.

Moreover, the relatively large differences between median and mean values found in this study in addition to several large outliers indicate substantial differences within the technology Kickstarter category. It has been beyond the scope of this study to investigate differences within the technology category. It would be a valuable addition to repeat

behaviour in crowdfunding literature for future research to build on the results in this study by investigating differences within the technology category by distinguishing subcategories based on different parameters such as, among others, whether a project has been successful or not, the type of technology, or the size of the project based on the initial funding goal, the amount pledged, or the amount of backers reached.

5.3 Strengths and Limitations and Future Research

In this study I investigated success factors for repeat behaviour in technology crowdfunding projects. This is a field that has received very little academic attention previously. Crowdfunding in itself is a relatively new concept; Kickstarter was founded in 2009. Specifically, repeat behaviour in technology crowdfunding has not been studied before. Therefore, the biggest strength of this study is that I provide insights for a relatively

unexplored field of crowdfunding. Moreover, these insights are based on a large dataset of 8885 technology crowdfunding projects from Kickstarter, representing the entire population of technology projects on Kickstarter.com at the time of scraping the data. Because of this, the results and conclusions based on this data are very powerful.

(25)

23

However, this study has also been subject to several limitations which create opportunities for future research.

First of all, I made no distinction between successful and failed projects. As my second and third hypotheses suggest initial success does play a significant role in predicting repeat behaviour. For this reason it would be very interesting for future work to create this distinction to further explain the reasons and factors behind the motivation for repeat

behaviour. For example, some creators who launched a successful project will come back to crowdfunding, whereas others will not. Vice versa, some creators who failed the first time will come back to try crowdfunding again, whereas others will not. Future work should make this distinction and investigate the reasoning and motivation behind each of these scenarios.

Secondly, it has been beyond the scope of this study to investigate cross-category repeat projects. Some categories have some overlap in terms of classification. For example, certain aspects of the gaming or photography categories have projects with many technology-heavy characteristics. Creators who launched a second project in a different category have not been included in this study. Also, some categories are very different in nature. For example, Davidson and Poor (2016) looked at success factors for repeat behaviour in four different Kickstarter categories. Their results suggest substantial differences between categories. They found that that a higher P/B ratio is related with an increased likelihood of repeat behaviour in the music category. They explain this by stating that, on average, music projects have

relatively small funding goals, and that music has had a very long history of support by ‘wealthy patrons’. They also mention that these explanations are speculative and that future work should investigate this further. Nonetheless, these findings indicate differences between categories. It would be a very valuable contribution to crowdfunding research for future work to identify cross-category repeat behaviour.

Thirdly, there are some relevant limitations due to the nature of repeat crowdfunding. Firstly, some creators may launch a second project on a different crowdfunding platform. This may be more likely for creators who failed the first time, and want to try crowdfunding again. Because of the failure of their initial attempt, these creators may be more tempted to use a different platform. This kind of repeat behaviour would be extremely hard to research on a large scale for several reasons. It would, however, be interesting for future research to investigate this type of repeat behaviour with a qualitative research design. Surveys and

(26)

24

interviews could be used on a smaller scale to gain insights into cross-platform repeat behaviour in crowdfunding. Secondly, the intentions of project creators are generally unknown. Some creators may have intended to use crowdfunding only once, regardless of how well they did. Also, some creators may have intentions to come back to crowdfunding at a later point in time. These creators have now been labelled as one-time users, when in reality their projects should be classified as repeat projects. To account for this problem, future research should aim to complement quantitative measures of repeat behaviour with surveys or interviews to gain a better understanding of these future intentions.

Finally, a minor limitation is that Kickstarter does not give a detailed description of the technology category and does not set any project requirements for the creator to list the project in this category. For these two reasons it is very difficult to give a detailed and accurate description of the type of project that is analysed in this study. Kickstarter does list 15 possible subcategories to choose from, but this is optional and does not have come with any requirements. Because of these reasons the generalizability to other platforms or types of projects may be slightly limited.

5.4 Contributions and implications for Practise

Until now, there has only been one study in which the authors looked at repeat behaviour in crowdfunding, which is the work by Davidson and Poor (2016). The predictions tested in this study have not been studied before for technology projects from Kickstarter. Therefore, the results from this study contribute to existing literature by providing new valuable insights into this realm of crowdfunding, as well as a foundation for future research to expand on.

Also, the results from this study have several implications for creators and backers of repeat crowdfunding projects. For creators, these results can be used as a tool for assessment. This study suggests that if the first project of a creator had a high P/B ratio, a low amount of backers, and a low P/G ratio, it is less likely that subsequent projects will be successful. On the other hand, if a creator truly reached the crowd the first time and realized a low P/B ratio, a high amount of backers, and a high P/G ratio, the likelihood for success is higher for subsequent projects. This mechanism of assessment may create a system of self-selection, in

(27)

25

which creators who score well on these factors will more often come back to crowdfunding than creators who do not perform well. This system may lead to even higher success rates among repeat projects, because creators who are less likely to succeed again will attempt repeat projects less often. For backers, these results may be used as a set of suggestions. A creator who has reached a low P/B ratio, a high amount of backers, and a high P/G ratio in the past is more likely to succeed. This means that backers should pledge their money more in projects that meet these criteria, to increase the success rate of their pledges and to be more likely to receive the rewards. Projects that do not meet these criteria are less likely to succeed, thus pledging money to these projects may not be optimal.

6 Concluding thoughts

Previously, success factors for repeat behaviour in technology reward-based crowdfunding has been undiscovered academic terrain. This study shows empirical results from 8885 Kickstarter technology projects, both live and completed. Using these Kickstarter projects I aimed to answer the following question: ‘What factors differentiate repeat projects

from one time users in technology crowdfunding projects?’. Based on limited existing

literature I proposed three main hypothesis. The data confirmed hypothesis 2 and 3, and partially confirmed hypothesis 1. This proves that, on average, repeat projects tend to reach more backers per project and also raise more money in relation to the initial funding goal. Moreover, the results suggest that a success factor for repeat projects is to have a lower average donation per backer, which would indicate that the project did not rely on a limited amount of very large pledges by family or friends. However, although the data points in this direction, this is not completely confirmed. Based on these results I provided several

(28)

26

References

Agrawal, A. K., Catalini, C., & Goldfarb, A. (2011). The geography of crowdfunding (No. w16820). National bureau of economic research.

Agrawal, A., Catalini, C., & Goldfarb, A. (2015). Crowdfunding: Geography, social networks, and the timing of investment decisions. Journal of Economics & Management Strategy, 24(2), 253-274.

Belleflamme, P., Lambert, T., & Schwienbacher, A. (2010, June). Crowdfunding: An industrial organization perspective. In Prepared for the work shop Digital Business Models:

Understanding Strategies’, held in Paris on June (pp. 25-26).

Belleflamme, P., Omrani, N., & Peitz, M. (2015). The economics of crowdfunding platforms. Information Economics and Policy, 33, 11-28.

Brabham, D. C. (2008). Crowdsourcing as a model for problem solving an introduction and

cases. Convergence: the international journal of research into new media technologies , 14(1), 75-90.

Brabham, D. C. (2008). Moving the crowd at iStockphoto: The composition of the crowd and motivations for participation in a crowdsourcing application. First monday, 13(6). Carr, C., & Tomkins, C. (1998). Context, culture and the role of the finance function in strategic

decisions. A comparative analysis of Britain, Germany, the USA and Japan. Management

Accounting Research, 9(2), 213-239.

Davidson, R., & Poor, N. (2016). Factors for success in repeat crowdfunding: why sugar daddies are only good for Bar-Mitzvahs. Information, Communication & Society, 19(1), 127-139.

Gallagher, D., & Salfen, J. (2015, March 24). By the Numbers: When Creators Return to Kickstarter. Retrieved from https://www.kickstarter.com/blog/by-the-numbers-when-creators-return-to-kickstarter

Gerber, E. M., Hui, J. S., & Kuo, P. Y. (2012, February). Crowdfunding: Why people are motivated to post and fund projects on crowdfunding platforms. InProceedings of the International

Work shop on Design, Influence, and Social Technologies: Techniques, Impacts and Ethics .

Giudici, G., Nava, R., Rossi Lamastra, C., & Verecondo, C. (2012). Crowdfunding: The new frontier for financing entrepreneurship?. Available at SSRN 2157429.

Hemer, J. (2011). A snapshot on crowdfunding (No. R2/2011). Working papers firms and region. Kleemann, F., Voß, G. G., & Rieder, K. (2008). Un (der) paid innovators: The commercial utiliza-tion of

consumer work through crowdsourcing. Science, technology & innovation studies, 4(1), PP-5. Kuppuswamy, V., & Bayus, B. L. (2015). Crowdfunding creative ideas: The dynamics of project

backers in Kickstarter. UNC Kenan-Flagler Research Paper, (2013-15). Lee, S. H., DeWester, D., & Park, S. R. (2008). Web 2.0 and opportunities for small

businesses. Service Business, 2(4), 335-345.

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

(29)

27

Schwienbacher, A., & Larralde, B. (2010). Crowdfunding of small entrepreneurial ventures. Handbook

of entrepreneurial finance, Oxford University Press, Forthcoming.

van de Rijt, A., Kang, S. M., Restivo, M., & Patil, A. (2014). Field experiments of success-breeds- success dynamics. Proceedings of the National Academy of Sciences , 111(19), 6934-6939. Zhao, L. (2016). [Kickstarter technology]. Unpublished raw data.

(30)

28

Appendix A

Correlations between relevant variables, and direction of effect regression output

Correlations

Binary Regression output. Direction of effect on whether or not a creator made another project B Step 1a PG_Ratio ,0006538171 backers_count ,0000067989 PB_Ratio ,0000210889 Constant -2,352

Note: Dependent variable is whether (1) or not (0) a creator made another project

backers_count PG_Ratio PB_Ratio backers_count Pearson Correlation 1 ,053** _,025* Sig. (2-tailed) ,000 ,017 N 8885 8881 8885 PG_Ratio Pearson Correlation ,053** ₁ _,010 Sig. (2-tailed) ,000 ,323 N 8881 8881 8881 PB_Ratio Pearson Correlation ,025* _,010 ₁ Sig. (2-tailed) ,017 ,323 N 8885 8881 8885

Success factors for repeat behaviour in technology reward-based crowdfunding : an empirical study