Crowdsourcing a Business Model
How to improve the quality and quantity of crowdsourcing contributions.
Eline van Oostveen
| 10646027 | Final Thesis Digital Business | Nick van der Meulen |
| MSc. Business Administration | Amsterdam Business School / University of Amsterdam | | 12771 words | 23/06/2017 |
Statement of Originality
This document is written by Eline van Oostveen, 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.
Abstract
Ever since the rise of web 2.0, crowdsourcing mechanisms are becoming increasingly more conventional as a source for business innovation. Many literature has been devoted to the theoretic framework of simple crowdsourcing tasks, such as product innovations. However, more complex crowdsourcing tasks such as business model innovation have received very little scientific attention so far. Although it has been shown that crowdsourcing can lead to high-quality business model innovation ideas, it was still unclear what factors could predict and manipulate the successfulness of business model innovation crowdsourcing. Thus, the primary aim of this study was to identify factors that influence the success of crowdsourcing for business model innovation. Through an experiment integrated in an online survey, the crowdsourcing contributions of 112 participants were analysed and evaluated. The main findings were that the intrinsic motivations sparked by the crowdsourcing design (either competitive or collaborative), as well as brand familiarity have a positive effect on the quality of contributions, while brand scepticism has a negative impact on the quality of contributions. Also, the positive effect of brand familiarity on the quality of contributions is negatively moderated by brand scepticism. These findings provide valuable insights for theory and practice as it adds new elements to the business model innovation crowdsourcing framework and provides guidance to managers considering crowdsourcing a new business model. However, the framework for business model innovation crowdsourcing still lacks some essential elements of knowledge and thus it is strongly emphasized that further research has to be conducted.
Table of Contents
Statement of Originality………...…2
Abstract………..3
Table of Contents………..………...….4
1. Introduction………..………….……7
1.1 The Case of Innocentive……….……..………8
1.2 Research Objectives………..………8
1.3 Scientific and Practical Contributions.……….……….………9
1.4 Paper Set-up……….……….………9
2. Theoretical framework………...………..…..10
2.1 Business Model Innovation……….………..…..…10
2.2 Crowdsourcing……….……….…..…10
2.2.1 Crowdsourcing Categories………….………..…11
2.2.2 Crowdsourcing Innovations………..…….………..11
2.3 Crowdsourcing Mechanisms for Business Model Innovation…...……...………..12
2.3.1 Crowdsourcing Contributions………...………...……13 2.4 Crowdsourcing Design………14 2.4.1 Level of Competitiveness………...……….14 2.4.2 Level of Collaboration………...…….……….14 2.4.3 Gamification………..…...……...………15 2.4.4 Intrinsic Motivation……….…...………..………...……16
2.6 Brand Bias……….……...…….……….…….18 2.7 Conceptual Model……….……...…….……….….…20 3. Methodology……….……...………21 3.1 Design………...……….……….……21 3.2 Stimuli……….……….…...…23 3.3 Description of Measures…..………….………..………25
3.3.1 Perceived Level of Competition………..…….………...………26
3.3.2 Perceived Level of Collaboration…...………..…...………26
3.3.3 Age of the Contributor…….………...………...…..…26
3.3.4 Level of Education………...…………27
3.3.5 Knowledge of the Field………...……….27
3.3.6 Brand Bias………...…….………....27
3.3.7 Quality of the Contributions………..….…….……27
3.3.8 Quantity of the Contributions.…...………..………....……28
3.4 Factor Analyses and Reliability………...……….…………..…………28
3.4.1 Perceived Level of Competitiveness………..…...………..…….28
3.4.2 Perceived Level of Collaboration….…….……….…….29
4. Results……..……….………...………30
4.1 Variable Descriptives………..………30
4.2 Initial Correlations……….……….32
4.3 Perceived Level of Competition and Collaboration………...……….…………35
4.3.1 Perceived Level of Competition on the Quality of Contributions…….………….…….35
4.3.3 Perceived Level of Collaboration on the Quality of Contributions.…...………….……36
4.4.4 Perceived Level of Collaboration on the Quantity of Contributions …..………36
4.5 Multiple Regression Model for Contribution Quality...………….……….……37
4.6 Multiple Regression Model for Contribution Quantity ………..…………..…….41
5. Discussion……….……….….………..…45
5.1 Summary of Results……….………..….………45
5.2 Implications of the Findings……….………..…...….45
5.2.1 Crowdsourcing Design…….………..…..…………45 5.2.2 Demographic Factors…………...……….………..….………46 5.2.3 Brand Bias……….………..…...….47 6. Conclusion……….………...………...49 6.1 General Conclusion……….………..……….49 6.2 Theoretical Contributions………….………..………...….50 6.3 Contributions to Practice……….………....…...…51
6.4 Strengths and Limitations………..………52
6.5 Further Research………53
6.6 Final Remarks………..………...54
References……….………..…...55
Appendix………..……….……….………...…...………61
1. Crowdsourcing a Business Model Survey………....……....…61
1. Introduction
For the last two decades, more and more businesses are reconsidering the way they do business (Chesbrough, 2010). The digitization of society has left many managers wondering whether their traditional business model will still be sufficient in the years to come and as a result, many businesses have (partly) transformed their processes and products to meet the new demands of their clients (Chesbrough, 2010). Although the embedding of digital mechanisms in business models can often lead to lower costs and less processing time, business model innovation has proven to be a fragile process with a lot of risk involved (Zott, Amit & Massa, 2011). Almost 25% of business model innovation efforts fail to succeed within the first year of implementation and this percentage increases to 35% after two years (University of Tennessee Research, 2016). This leads to the question if there is a way to make business model innovation a less uncertain process. A possible solution for this might be to change the order of decision making. Traditionally, the process of business model innovation starts by generating a new model for the value creation, value delivery and value capturing of a business, followed by implementation and finally evaluation (Chesbrough, 2010). By reorganizing the process of business model innovation to a model of evaluating different business model innovations, selecting one and only then implementing it, risks may be drastically reduced (Harvard Business Review, 2014). In this model, the business model innovation input comes directly from the (potential) customer and is generated through crowdsourcing mechanisms. This will not only limit risks by switching the order of
implementation and evaluation, it might also give a good indication of what customers wish to see innovated and therefore has a higher chance of being a successful business model
1.1 The Case of InnoCentive
A good example of the use of crowdsourcing mechanisms for business innovation (both in simple forms and complex forms such as business model innovation) could be found within the services of InnoCentive. InnoCentive is an online global platform that connects businesses with large target group panels to foster crowdsourcing mechanisms for business model
innovation. Through InnoCentive, businesses (so-called ‘seekers’) can organize business innovation competitions where individuals of their target audience (so-called ‘solvers’) can compete in generating the best business model innovation ideas. After a set amount of time, seekers select one or more winning ideas themselves and the successful solvers receive a monetary reward for their input (InnoCentives, 2016). InnoCentives is, among others, considered an open innovation pioneer (Harvard Business Review, 2014) and has proven to be very successful in it’s problem solving abilities (InnoCentives, 2016). However, specific factors that could play a role in the success of the use of crowdsourcing mechanisms for business model innovation have yet to be evaluated.
1.2 Research Objectives
West and Bogers (2014) have distinguished five major gaps in the literature focussing on crowdsourcing mechanisms for business model innovation purposes. First of all,
crowdsourcing has not yet been studied from a business model perspective. Secondly, an extensive view of commercialization processes is lacking. Also, the specific impact of crowdsourcing mechanisms on process changes needs to be evaluated and the definition of innovation needs to be cleared up. Lastly, further research needs to be done to study the potential influencers such as crowd demographics and crowdsourcing design on business model innovation. This study aims to investigate the latter and identify influencing variables in the crowdsourcing business model innovation framework. More specifically, the primary
aim of this paper is formulated in the following research question (RQ): to what extent do
crowdsourcing design, crowd demographics and business recognition influence the quality and quantity of business model innovation contributions?
1.3 Scientific and Practical Contributions
By further investigating the topic of business model innovation through crowdsourcing mechanisms and highlighting some of the possible influencing variables, this study will make a valuable contribution to existing knowledge by (partly) filling a scientific gap (West & Bogers, 2014). Through knowledge of possible influencing variables, the new framework of crowdsourcing mechanisms combined with business model innovation can be completed and possibly create new perspectives on crowdsourcing as well as business model innovation. Furthermore, the findings of this study might consult (digital) transformation agents in their considerations to use crowdsourcing mechanisms for business model innovation, which when executed correctly may have a great impact on their chances of success and even a firm’s chances at continuous competitive advantage.
1.4 Paper Set-up
In the next chapter, a theoretical framework will be provided through a literature review. Also, the research concepts and hypotheses will be introduced. This is followed up by the methods section of this paper, in which the target population, the sample, the measurements and the procedure will be specified. In the results chapter, a overview of the analyses will be presented and the hypotheses will be either confirmed or rejected. In the conclusion section, the research question will be answered based on the results. Lastly, the discussion chapter will critically address some implications, limitations and suggestions for future research.
2. Literature review
In order to make valuable contributions to the existing knowledge of crowdsourcing mechanisms used as input for business model innovation, previous literature has been evaluated.
2.1 Business Model Innovation
In the last couple of decades, business models have received growing scientific attention ( Chesbrough, 2010; Teece, 2007; Teece, 2010; Zott, Amit, & Massa, 2011). Although a clear, universal definition of business models is still lacking (Saebi & Foss, 2014), the concept of business model innovation has been introduced. There is growing consensus that business model innovation can be seen as innovation on either the value creation, value delivering or value capturing component of the business model (Sako, 2012; Zott et al., 2011). Chesbrough (2010) has shown that although vitally important to a firm, business model innovation
involves a high level of risk. Existing tools for business model innovation, often fail to change organizational processes (Sosna, Trevinyo-Rodríguez & Velamuri, 2010). In order to
successfully change these processes, Chesbrough (2010) elaborates that it is vital to a firm to be open towards business model experimentation and embrace experiences of failure.
Although Chesbrough (2010) sheds some light on the many risks and barriers surrounding business model innovation, he does not suggest the possible use of crowdsourcing to minimize the risks involved.
2.2 Crowdsourcing
Crowdsourcing is a form of open innovation that has evolved through Web 2.0 (Floyd, Jones, Rathi & Twidale, 2007). Crowdsourcing can be defined as the act of outsourcing a task to a
informal or formal team, or individual), such as a contractor, in the form of an open call (Afuah & Tucci, 2012). It has been considered to be a particularly useful way to coordinate work for tasks that can benefit from a collective intelligence or that are difficult to process by computers and are therefore outsourced to humans (Leimeister, 2010).
2.2.1 Crowdsourcing Categories
According to Geiger and Schader (2014), four categories of crowdsourcing can be recognized; crowdcreation, crowdrating, crowdprosessing and crowdsolving. Crowdcreation aims to create comprehensive artefacts based on a variety of heterogeneous contributions, such as Youtube or Wikipedia. Crowdrating systems are commonly used to show the “wisdom of crowds” (Surowiecki & Silverman, 2007) through collective assessments or predictions. With crowdrating, the large number of homogeneous “votes” is used as a source of intelligence. Third, crowdprocessing mechanisms rely on the crowd to perform large quantities tasks. Often these are relatively simple tasks that can not yet be executed by computers and are therefore performed by individuals of the crowd in exchange for a (monetary) reward. Lastly, crowdsolving approaches use the diversity of the crowd to generate a large number of
heterogeneous solutions for a certain problem. Crowdsolving is often used for very complex problems or if no pre-definable solution exists and is thus the most suitable form of
crowdsourcing for business model innovation.
2.2.2 Crowdsourcing Innovation
Although the concept of crowdsourcing is relatively new, various studies have already been focusing on the use of crowdsourcing for innovation efforts (Jeppesen & Lakhani, 2010; Poetz & Schreier, 2012; Nishikawa et al., 2013). Van der Meer (2007) studied the Dutch start-up market and found that start-start-ups that were relatively open towards their public in terms of
their business’ innovation, had a higher success rate than firms who did not, showing that crowdsourcing innovation could affect a firms performance. Also, Laursen and Salten (2006) showed early on that crowdsourcing can make a valuable contribution to business innovation, which can be complemented by a firm’s internal research and development efforts. The effectiveness of crowdsourcing mechanisms in business innovation processes have since then been shown multiple times (Jeppesen & Lakhani, 2010; Poetz & Schreier, 2012; Nishikawa et al., 2013).
2.3 Crowdsourcing Mechanisms for Business Model Innovation
While many existing literature focuses on either crowdsourcing or business model innovation, little research has investigated the role of crowdsourcing mechanisms within the business model innovation process so far (Waldner & Poetz, 2015). In their study, Waldner and Poetz (2015) aimed to test whether crowdsourcing mechanisms could be used to fuel innovation in rather complex situations, such as business model innovation. They used a case study in which online fans of a popular Swedish podcast show were asked to participate in an online crowdsourcing project, where they could share any business model innovation ideas they might come up with for the podcasting show. Waldner and Poetz (2015) found that crowdsourcing mechanisms may indeed contribute to business model innovation, as the Swedish podcast show actually implemented an idea submitted by the contributors and reported having seriously considered multiple of the ideas submitted. This finding is
particularly interesting since the possible conflict of interest of the podcast users, as many of them suggested no longer providing the podcast for free.
2.3.1 Crowdsourcing Contributions
In order to evaluate crowdsourcing efforts on their potential of leading to actual business model innovation, contributions of individuals can be evaluated based on their quantity and quality. Zhao & Zhu (2014) emphasize the value of contribution quantity, as the quantity of contributions by an individual is considered to be positively related to the success of the crowdsourcing effort. The quality of crowdsourcing contributions for business model innovation on the other hand, should be evaluated based on business model principles.
Business model innovation can be divided between activities of value creation, value delivery and value capturing (Zott & Amit, 2001). Zott and Amit (2001) distinguish four sources of value creation through business models: novelty, lock-in, complementaries and efficiency. These value creation factors all reinforce each other; for example novelty is more effective when the business model also includes lock-in, complementaries and / or efficiency, or vice versa. As Porter (1985) points out, value creation often goes beyond an update in the activity chain, a new strategic network, or a new form of exploiting a firm’s core competency. Amit and Zott (2001) found that scientists studying value creation should look beyond a firm’s or an industry’s boundaries. Hamel (2000) states that if businesses want to gain sustained competitive advantage, they must frequently innovate their entire business model, including the way they create value, deliver value and capture value. Chesbrough & Rosenbloom (2002) state that value creation is an essential step for business model innovation, but on its own is not enough. After the value chain has been innovated, a firm should look at new ways to deliver and capture this value too. The latter could be done by complementary assets on the supply side and the exploitation of network effects on the demand side (Teece, 1986).
2.4 Crowdsourcing Design
The type of crowdsourcing design seems to have a significant effect on business innovation success. Within crowdsourcing, two main types of design can be distinguished: a
collaborative design and a competitive design (Morschheuser, Hamari & Koivisto, 2016).
2.4.1 Level of Competitiveness
Competition may be conceptualized in three distinct ways (Brown, Cron, & Slocum, 1998): as a characteristic of the person (trait competitiveness), as a characteristic of the perceived situation (perceived environmental competitiveness), and as a characteristic of the actual situation (structural competition). Competition on online crowdsourcing platforms can be categorized as structural competitiveness; the design of the platform sparks a feeling of competitiveness in it’s users that is not necessarily a person’s trait, nor is it just perceived.
Structural competition represents an actual situation in which two or more people compete for a mutually exclusive achievement outcome (Johnson & Johnson, 1989). By offering
contributors some form of reward (either physical or psychological), quality and quantity of contributions within crowdsourcing for innovation have been shown to increase
(Morschheuser, Hamari & Koivisto, 2016). Nevertheless, this variable has not yet been studied within the more complex context of business model innovation (Lüttgens, Pollok, Antons & Piller, 2014; Waldner & Poetz, 2015).
2.4.2 Level of Collaboration
Crowdsourcing requires communication and cooperation from people who are
geographically-distributed and who have diverse backgrounds, resulting in a divers pool of intellectual resources (Luther & Bruckman, 2008). Crowdsourcing efforts with a collaborative
design use this diversity by allowing peer interactions to improve the quality and quantity of crowdsourcing mechanisms (Judd, Kennedy & Cropper, 2010). Collaboration can be
described as the mutual engagement of participants in a coordinated effort to solve a problem (Dillenbourg, Baker, Blaye & O’Malley, 1996). The roles of peer-based collaboration in promoting learning, understanding and achievement have been well researched. Slavin (1995) suggests that there is a broad consensus among researchers about the positive effects of cooperative learning on achievement. He outlines that collaboration has a positive impact on development, achievement, cognitive elaboration and on internal motivations. Kittur and Kraut (2008) studied the impact of a collaborative design in crowdsourcing specifically. They found that peer-to-peer communication can be very effective for crowdsourcing, as long as it’s in a free and open environment.
2.4.3 Gamification
Both competitive and collaborative forms of crowdsourcing implement techniques of gamification. Gamification can be described as the use of (video)game elements to create intrinsic motivations in individuals, by generating feelings such as flow, mastery, autonomy and suspense (Huotari & Hamari, 2012). Gamification has been shown to improve intrinsic user motivation (Ryan, Rigby & Przybylski, 2006; Sweetser & Wyeth, 2005), increased user participation (Jung, Schneider & Valacich, 2010; Witt, Scheiner & Robra-Bissantz, 2011; Von Ahn & Dabbish, 2008), a better user experience (Elatla, Gutwin, Nacke, Bateman & Mandryk, 2011; Downes-Le Guin, Baker, Mechling & Ruylea, 2012) and improved performance in non-game related tasks (Deterding, Sicart, Nacke, O’Hara & Dixon, 2011; Wilson & Birks, 2011), thus making gamification an interesting element to consider for crowdsourcing.
2.4.4 Intrinsic Motivation
Boudreau and Lakhani (2009) found that the motivations of the crowd are (among the type of platform and the type of problem) critical when deciding between a collaborative versus a competitive crowdsourcing design. Intrinsic motivations (for example community feeling, addiction and challenge) were found to be the strongest driver for quality and quantity of contributions within the crowdsourcing context. Through the use of gamification elements, different forms of intrinsic motivation can be sparked (Ryan, Rigby & Przybylski, 2006; Sweetser & Wyeth, 2005). A competitive crowdsourcing design with gamification elements leads to an increased feeling of challenge by individuals, which in turn improves the quality and quantity of contributions for innovation crowdsourcing. On the other hand, a
collaborative crowdsourcing design with gamification elements sparks the intrinsic motivation of community feeling (Slavin, 1995), which again leads to improved quality and quantity of contributions (Boudreau and Lakhani, 2009). Based on these findings, the following
hypotheses have been drafted:
H1) The perceived level of competitiveness in a competitive crowdsourcing design is positively related to (a) the quantity and (b) the quality of a participant’s
contributions.
H2) The perceived level of collaboration in a collaborative crowdsourcing design is positively related to (a) the quantity and (b) the quality of a participant’s contributions.
2.4.5 Competitive versus Collaborative Design
Kavaliova et al. (2016) show that a contest set up where contributors are competing for a reward, leads to higher effectiveness of the crowdsourcing mechanism than a collaborative contribution system. As Morschheuser, Hamari and Koivisto (2016) show, optimal
gamification techniques differ among the four categories of crowdsourcing. For
crowdsolving, reward systems tend to be the best form of gamification, as participations are heterogenous and non-emergent (Choi, Choi, So, Lee & You, 2014). However, the type of innovation problem turned out to be crucial for the distinction between competitive and collaborative crowdsourcing success as well. If the innovation problem involves cumulative knowledge, continually building on past advances, then collaborative communities have inherent advantages. If, on the other hand, the innovation problem is best solved by broad experimentation across a set of technical approaches or customer groups, then competitive markets have natural advantages (Boudreau and Lakhani, 2009). As this study will focus on the matter of crowdsolving, in which there is a strong focus on broad experimentation, the following hypothesis has been tested:
H3) Participants who have been exposed to a competitive crowdsourcing design will generate higher (a) quantity and (b) quality contributions than participants who have been exposed to a collaborative crowdsourcing design.
2.5 Demographic Factors
Crowd demographics, such as gender, age and knowledge of the specific business field, have turned out to play an important role in the effectiveness of crowdsourcing mechanisms. Being a female has a positive and significant effect on the quality of the contribution (Jeppesen & Lakhani, 2010). However, the opposite appeared to be true for the quantity of contributions, where males score significantly higher. This may be due to the phenomenon that males are generally more confident about their capabilities and will therefore easier share their ideas, Jeppesen and Lakhani (2010) reason. Poetz and Schreier (2012) found that participants who are older, have significantly more and better contributions than younger participants. This may be explained by the assumption that older participants also have more experience with
businesses and innovation. On the contrary, Jeppesen and Lakhani (2010) also found that knowledge of the specific business field is negatively related to the quality and quantity of the contributions. This rather unexpected finding is explained by Jeppesen and Lakhani (2010) by the concept of bridging knowledge fields. Participants who have a field of expertise that is rather different from the specific business field at stake, may have a novel perspective on the case that bridges different fields of knowledge. Based on this literature, the following
hypotheses on crowd demographics have been formulated:
H4) The age of the participant is positively related to (a) the quantity and (b) the quality of his or her contributions.
H5) The problem field knowledge of the participant is negatively related to (a) the quantity and (b) the quality of his or her contributions.
2.6 Brand Bias
Keller (2003) showed that pre-existing assumptions towards a brand have a strong impact on the willingness of consumers to participate in co-creation with the brand. However, Keller (2003) also showed that participating in co-creation may improve the brand perception of contributors, and that this effect is stronger when they are less familiar with the brand. Aaker (1996) showed that changes in brand perception often take some time to evolve, but when they do participants are more likely to participate in co-creation with the brand again. Poetz and Schreier (2012) showed that familiarity with the business at stake has a positive effect on both the quality and the quantity of contributions for simple business innovations. However, this effect turned out to be moderated by brand scepticism. Poetz and Schreier (2012) found that when participants are more sceptical towards the brand, brand familiarity actually gets a negative impact on the quality and quantity of contributions. As the effect of business familiarity, nor the moderating effect of brand scepticism have been shown for complex
crowdsourcing such as business model innovation, four final hypotheses have been composed:
H6) A participant’s familiarity with the brand is positively related to (a) the quantity and (b) the quality of his or her contributions.
H7) The participant’s level of scepticism towards the brand negatively moderates the positive relationship between the participant’s familiarity with the brand and the (a) quantity and (b) quality of his or her contributions, such that the effect as stated at H6 will turn more when a participant has a higher sense of scepticism towards the brand.
2.7 Conceptual Model
Figure 1: Conceptual Model
H7-) Brand Scepticism Quality Quantity
Crowdsourcing Design
H1+) Perceived Level of Competition H2+) Perceived Level of Collaboration H3-) Competition vs. CollaborationH4+) Age of the Contributor
H5+) Knowledge of the Field
H6+) Brand Familiarity
a
b
Demographic Factors
Brand Bias
Crowdsourcing Contributions
3. Methodology
3.1 DesignIn order to test the formulated hypotheses, an experiment was executed with a between-subjects design. An experimental design was believed to be most suitable for this study, as it allows for the testing of causal relationships. Based on the literature, crowdsourcing design, demographic factors and brand recognition are assumed to be causes that consistently affect the quality and quantity of crowdsourcing contributions, thus suggesting a causal relationship. This cannot truly be tested through other quantitative or qualitative techniques. However, since crowdsourcing is a digital activity, the experiment has been implemented in an online survey to create a more natural setting and to attract more participants at the same time. The population of this study were all English speaking people, as anyone could participate in crowdsourcing. A convenience sample was used to gather respondents through the use of social media. A link to the survey was posted on Facebook, reaching approximately 711 individuals. Persons who reportedly had participated were asked to share the link among their own network of friends and family, thus resulting in a snowball effect. In total, 119 persons participated in the experiment. As participants started the survey, they were randomly assigned to either one of two groups: the competitive crowdsourcing design or the
collaborative crowdsourcing design. By assigning the participants randomly amongst the two conditions, the more accurate the analysis of the manipulation (competitive versus
collaborative) could be. By making sure no pattern exists in the assignment of participants to either the competitive- or the collaborative group, the effects of these manipulations can be measured regardless of other factors that may lead some people or groups to perform
differently in terms of the quality and quantity of their contributions, as it is believed that all other factors (such as age) are distributed equally amongst both groups when assigning randomly. Both in the competitive condition, as well as in the collaborative condition,
participants were presented a fictional case of hard- and software firm IBM. IBM is one of the biggest IT firms in the world, with globally more than 400.000 employees and a revenue of $106.9 billion dollars (2011). The company operates in over 175 countries and has registered a record amount of patents since their founding in 1911. The core areas of business for IBM are Business and IT Consulting, Business Process Outsourcing, Outsourcing services,
Finance, IT Projects, Hardware, Software, Maintenance, Application Development, Education and Recovery Services (IBM, 2017). For this experiment, a case study needed to be used with a company that participants A) may or may not be familiar with, in B) an industry that
participants may or may not be familiar with, and with C) a reputation that participants may or may not be sceptical towards, so that scores on those variables would differ more greatly and the hypotheses could thus be tested easier. IBM seems to meet all these criteria, for they are: A) one of the biggest players in their field, so rather recognized by the general public, however complex enough to not be fully understood by everyone, B) part of the IT industry, which is a rather specific and complex field of industry and C) are very powerful and
influential in terms of their patents, which may provoke scepticism towards the brand. Therefore, IBM was selected to be the case study subject for the experiment. Since no actual collaboration with IBM took place for this research a fictional case study was created, in which it is claimed that IBM wants to use crowdsourcing for business model innovation. One group was asked to submit business model innovation ideas within a competitive setting (a reward of €50,- was promised to the participant with the ‘best’ contributions), while the other group was asked to do so within a collaborative setting (no reward was mentioned and a fictional list of previously made contributions by other participants was shown as inspiration). Apart from submitting their ideas, participants were also asked to indicate their gender, age, education, knowledge of the tech business field, familiarity with IBM and scepticism towards IBM. To make a fair analysis of the contributions by participants, an expert was used to rate
the contributions in terms of their inclusion of value creation, value delivery and value capturing. The expert, Rob Vreke, is a retired IBM executive IT architect who worked for the company from 1980 to 2016.
3.2 Stimuli
In both of the conditions, participants were first presented an identical introduction about crowdsourcing, IBM and their business model and a description of the case at hand, namely using crowdsourcing to innovate IBM’s business model. For the competitive condition, it was mentioned in the case description that contributions by the participants would be evaluated by experts from IBM, and that the participant with the best contribution would win €50,-. In this condition, no examples of other contributions by others were shown or mentioned. For the collaborative condition, it was mentioned in the case description that IBM had actually already launched this crowdsourcing project on their forum as well, and a snapshot of these (fictional) contributions was shown with the message that participants could use this as an inspiration to, together, come up with the best innovation. In this condition, no rewards were mentioned. The actual information presented to the participants in both conditions is shown below:
This survey will be focused on crowdsourcing. Crowdsourcing can be described as the act of outsourcing an intellectual task to an unknown online crowd. The general idea behind crowdsourcing is that when enough random persons online come up with solutions for a specific problem, at least one of those solutions should be fruitful. In this survey, you will be part of the crowd and you will be asked to come up with ideas for a real business case; the case of IBM.
IBM is a leading diversified technology company with a broad range of business offerings across IT hardware, software, and services segments. IBM was founded in 1911. IBM
operates in more than 175 countries across the globe and has over 400,000 employees. On the next page, you will be explained the current business model of IBM. Please take some time to analyse this.
Business offering and revenue IBM sells (A) services, (B) software, (C) hardware and (D) financing to their customer segments, which exists of 18 business industries (such as banking, healthcare and retail). Their main costs are in these four business offerings. Furthermore, there are additional costs in research and development, and selling, general and
administrative costs. (A) Services - The services that IBM offers include IT outsourcing, Infrastructure services, Technical support, Consulting, Application management and System integration. - By selling these services to the businesses in the 18 industries, IBM earns revenue from fixed price projects and billing on time and material basis. - Costs of services are human resource costs (B) Software - The software that IBM sells includes operating systems and other software. - By selling software in the 18 industries, IBM makes revenue in the form of licenses (one time charges and recurring license charges).- Costs of software are software developing costs, datacentre costs (C) Hardware- The hardware IBM offers include servers and storage (flash storage, disk storage and tape storage). - By selling hardware in the 18 industries, IBM earns revenue from direct sales and installation services. - Costs of hardware à manufacturing costs, shipping costs and warranty costs (D)
Financing - IBM offers client financing, commercial financing, remanufacturing and remarketing to the 18 industries. - By financing for clients, IBM earns loan interest income, refurbished - and upgraded asset sales. - There are no costs in financing.
Condition 1 – Competitive Design
IBM wants to innovate the way they do business, by finding novel ways to create value, novel ways to bring their value to the market and / or novel ways to make money from the value they brought to the market. Using the information provided above and any additional knowledge you might have, can you come up with one or more idea(s) of how IBM can innovate their business model? Please note that the ideas submitted in this survey will be shared with experts within IBM, who will evaluate the contributions. For the participant with the best idea, a price of €50,- is offered. If you wish to take part in the contest, you may leave your email-address at the end of this survey. The current business model of IBM is again shown below to consult you.
Condition 2 – Collaborative Design
IBM wants to innovate the way they do business, by finding novel ways to create value, novel ways to bring their value to the market and / or novel ways to make money from the value they brought to the market. Using the information provided above and any additional
knowledge you might have, can you come up with one or more idea(s) of how IBM can innovate their business model? IBM is involved in a similar project on an actual online crowdsourcing platform. To help get you started, some of the ideas posted on that platform are shown at the bottom of this page (see Appendix 2). Feel free to build upon these ideas, as together we might know more than we do by ourselves. Also, the current business model of IBM is again shown below to consult you.
3.3 Description of Measures
In order to test the hypotheses, two scales were used: Perceived Level of Competition and Perceived Level of Collaboration. Contribution Quality and Contribution Quality were both
measured by creating an index variable. Furthermore, Age, Gender, Education, Knowledge of the Field, Brand Familiarity and Brand Scepticism were all measured through one item each.
3.3.1 Perceived Level of Competition
The Perceived Level of Competition was measured through four seven-point items defined by Argo, Popa and Smith (2010) measuring how much feeling of challenge a participant is experiencing. The scale has a Cronbach’s alpha of 0.85 and consists of the following items:
“To what extend did you find the task challenging /difficult / hard / easy (r); To what extend did you feel challenged by the task;
3.3.2 Perceived Level of Collaboration
The Perceived Level of Collaboration was measured through four seven-point items defined by Reed (2004) to measure the degree to which a person experiences a feeling of peer-to-peer behaviour when completing a task with strangers. The scale has a Cronbach’s alpha of 0.85 and the scale includes the following items: “To what extent did you feel: 1. Inspired by the
ideas of other participants? 2. Motivated to contribute to the knowledge of the crowd? 3. Competition with other contributors (r)? 4. Helped by the other contributors?”. Participants
could answer through a 7-point Likert scale for each of the items, ranging from ‘1. Not at all’ to ‘7. Very much’.
3.3.3 Age of the Contributor
3.3.4 Level of Education
The level of Education was measured through a set of given choices: High School,
Community college (MBO), Bachelor in Applied Sciences (HBO), Bachelor (University), Master (University).
3.3.5 Knowledge of the Field
The Knowledge of the Field (in this case IT) was measured through the item created by Lakshmanan and Krishnan (2011): ‘To what extend are you knowledgeable of the IT business
field?’, after which participants could indicate their knowledge of IT through a 7-point Likert
scale, ranging from ‘1. Not at all’ to ‘7. Very much’. 3.3.6 Brand Bias
The brand recognition and brand scepticism were both measured by a single-item 7-point Likert scale, retrieved from Babin, Boles and Darden (1995): ‘To what extend are you: 1.
Familiar with IBM and the way they do business? 2. Sceptical about the business of IBM?’,
ranging from ‘1. Not at all’ to ‘7. Very much’.
3.3.7 Quality of the Contributions
The quality of the contributions was measured through a content analysis performed by an IBM expert, who rated the contributions on their inclusion of value creation, value capturing and value delivery, using a 7-point Likert scale varying from ‘1. Not at all’ to ‘7. Very much’. Then a index was composed based on the mean of these three items of each contributor, resulting in a Quality-score for each of the participants that could vary from ‘1. No quality at all’ to ‘7. Very high quality’.
3.3.8 Quantity of the Contributions
As each of the participants had the opportunity to submit multiple (unlimited) contributions, the quantity of contributions was measured by the absolute number of contributions. Since the quantity of contributions turned out to vary between 0 and 8, there was no need for grouping the data into a ordinal variable or for a quality threshold.
3.4 Factor Analyses and Reliability
In order to determine the validity of the scales composed, a Principal Components Factor analysis was performed for each of the latent variables, namely Perceived Level of Competitiveness and Perceived Level of Collaboration. A Varimax rotation was used to simplify the results, leaving each factor to have either large or small loadings for any of the items, so that the scales could be defined using a single factor each. In order to assess the results of the factor analyses, the Kaiser criterion (1960) was used. Following this criteria, any factors with eigenvalues above 1.0 can be accepted, as 1.0 is the average eigenvalue of a single item accounting for itself. Consequently, the reliability of each scale was tested using Cronbach’s alpha, to make sure that the different items used to measure the latent construct had acceptable levels of correlation between them. Based on the criteria defined by George and Mallery (2003), Kline (2013) and DeVellis (2016), any Cronbach’s alpha above 0.7 was considered acceptable, with a Cronbach’s alpha between 0.8 and 0.9 being considered good and a Cronbach’s alpha above 0.9 being considered excellent.
3.4.1 Perceived Level of Competitiveness
The Perceived Level of Competitiveness Scale has been composed of the variables Level of Challenge, Level of Ease (reversed), Level of Difficulty and Level of Worth Trying. Together these variables had an eigenvalue of 3,198 and accounted for 79,9% of the variance. Level of
Challenge has a correlation with the other three competitiveness variables of 90,5%. For Level of Ease (reversed), this is 86,2%. The Level of Difficulty correlates with 93,5% with the other variables and the Level of Worth Trying lastly correlated for 87,2%. The scale for Perceived Level of Competitiveness consisting of Level of Challenge, Level of Ease
(reversed), Level of Difficulty and Level of Worth Trying proved to be reliable with a Cronbach’s alpha of 0.91. No item had to be deleted.
3.4.2 Perceived Level of Collaboration
The Perceived Level of Collaboration Scale has been composed of the variables Level of Inspiration, Level of Contribution, Level of Competition (reversed) and Level of Helpfulness. Together these variables had an eigenvalue of 2,639 and accounted for 66,0% of the variance. Level of Inspiration has a correlation with the other three collaborative variables of 74,9%. For Level of Contribution this is 85,1%. The Level of Competition (reversed) correlates with 73,2% with the other variables and the Level of Helpfulness correlated 90,4%. The scale for Perceived Level of Collaboration consisting of Level of Inspiration, Level of Contribution, Level of Competition (reversed) and Level of Helpfulness proofed to be reliable with a Cronbach’s alpha of 0.83. Again, no item had to be deleted.
4. Results
First of all, the data was checked for missing values and errors. No data was missing, since each item of the survey required response before allowing a participant to continue. This led the data to be more complete, however has the limitation that participants might feel forced to participate and will refuse to put in a serious effort when filling out the survey. Seven cases were marked missing due to this cause, as participants entered random texts in the case study assignment box, rather than serious innovation ideas. These responses were left out of the analyses and since no further errors could be detected, the remaining data could be used for the analyses (N=112).
4.1 Variable Descriptives
To get a first impression of the data gathered, descriptive statistics were collected for all ten of the variables (see Figure 2). Half of the participants (N = 112) were 24 years old or older, while 25% were 44 years old or older, M=31,07, SD=13,75. The youngest participant was 18 years old, while the oldest participant was 61 years old. As for gender, 57 participants
(50,9%) identified themselves as male while 55 participants identified as females (49,1%). The most occurring highest level of education by participants was a Master’s Degree,
indicated by 43 participants (38.4%), while a Bachelor of Science and a Bachelor of Applied Sciences were both indicated by 31 participants (27.7% each). Only 7 participants (6.3%) indicated that their highest level of education was High School, and none of the participants stated Community College to be their highest level of education. Since there was no PhD level classified, it is assumed that those participants (if any) indicated a Master’s Degree to be their highest level of education. Based on a scale of 1 to 7, participants on average stated to have moderate knowledge of the IT field, M= 3,75, SD=0,16. On average, the knowledge of
have very little knowledge of the IT field of business, while 84 participants (75%) had slightly above moderate knowledge of the IT field or less. On a scale of 1 to 7, participants of the competitive condition (N=57) on average perceived the case study to be rather competitive, M=4.84, SD=1.41. 14 Participants (25%) perceived the case study to be just moderately competitive or less (Q1=4.5), and another quarter of the participants experienced a very high sense of competition (Q3=5.75). Based on a scale of 1 to 7, participants in the collaborative condition (N=55) on average stated to perceive the case study to be rather collaborative, M= 5.29, SD=0.96. On average, the perceived level of collaboration varied 0.13 from the mean knowledge of the field. 13 Participants (25%) indicated to experience a moderate level of collaboration or less (Q1=4.75), while another 13 participants (25%) perceived a very high sense of collaboration (Q3=6.00). As for the quality of contributions, participants (N=112) on average scored to have created contributions of moderate quality, M= 3.94, SD=0.50, based on a scale of 1 to 7. On average, quality of contributions varied 0.14 from the mean quality of contributions. 28 Participants (25%) created contributions of very little quality (Q1=2.33), while the 28 best scoring participants (25%) created contributions of slightly above moderate quality or higher (Q3=5.00). The quantity of contributions ranched from 0 to 8 contributions per participant, M=2.09, SD=0.90. The most occurring quantity of contributions was 2 per participant, which occurred 48 times (42.9%). 28 participants generated less than 1.25 contributions per participant, while the middle 56 participants (50%) generated between 1.25 and 3.0 contributions. Brand familiarity amongst participants was rather high, M=4.54, SD=1.67, based on a scale of 1 to 7. On average, participants indicated to be rather familiar with the brand IBM. 28 Participants (25%) indicated to be moderately familiar with IBM or less (Q1=4.00), while the 28 participants most familiar with IBM (25%) indicated to be very familiar with the brand (Q3=6.00). Lastly, participants appeared to be little sceptical towards IBM, M=2.65, SD=1.90, on a scale of 1 to 7. 28 participants (25%) were not at all sceptical
(Q1=1.00) and another 28 participants (25%) we’re moderately sceptical towards IBM or more (Q3=4.00).
Figure 2: Variable Descriptives
Variable N Minimum Maximum Mean Std. Deviation
1. Age 112 18,00 61,00 31,0714 13,74983
2. Gender 112 1 2 1,49 ,502
3. Highest Level of Education 112 1 5 3,92 1,108
4. Knowledge of the Field 112 1,00 7,00 3,7500 1,69525
5. Perceived Level of Competition 57 1,50 6,75 4,8421 1,41313
6. Perceived Level of Collaboration 55 2,50 6,75 5,2909 ,95593
7. Brand Familiarity 112 1,00 7,00 4,5357 1,67078
8. Brand Skepticism 112 1,00 7,00 2,6518 1,90168
9. Quality of Contributions 112 1,00 6,67 3,9375 1,49509
10. Quantity of Contributions 112 ,00 8,00 2,0893 ,89597
4.2 Initial Correlations
After assessing the descriptive statistics of all ten variables, a correlation matrix was created to see whether each of the variables significantly correlated to any of the other variables (see
Figure 3). First of all, Age of the participant was found to significantly positively correlate
with the Level of Education and the Brand Familiarity of participants. Both are not very surprising, as a majority of the younger participants may not have finished their education yet and older participants have simply had a bigger chance of encountering the IBM brand at some point in their lives. However, the expected correlation between Age of the Participant and the Quality- and Quantity of Contributions could not be found. Secondly, Gender was found to be positively and significantly correlated with the Perceived Level of Competition and the Perceived Level of Collaboration and significantly negatively correlated with the Level of Education, Knowledge of the Field, Brand Familiarity and Brand Scepticism, even though no hypotheses were formulated based on the gender of participants. These correlations
mean that female participants tend to experience stronger internal motivations triggered by the crowdsourcing design (either competitive or collaborative), while male participants tend to be higher educated, indicate to have more knowledge of the IT business field, be more familiar with the brand IBM and also tend to be more sceptical towards IBM. However, since Gender is represented rather equally (57 male participants versus 55 female participants), these correlations were not expected to distort the results. Next, the Level of Education was found to be significantly and positively correlated with the Knowledge of the Field, the Perceived Level of Competition and the Brand Familiarity, whilst being negatively correlated with the Perceived Level of Collaboration. The correlation of the Level of Education with the
Knowledge of the Field and Brand Familiarity could be explained by the assumption that a higher education and / or intelligence are required to understand the complexity of the IT field and IBM in specific better. However, the opposite correlations with the Perceived Level of Competition and the Perceived Level of Collaboration are surprising. Knowledge of the Field was, apart from earlier mentioned correlations, found to be significantly positively correlated with the Brand Familiarity and Brand Scepticism. It makes sense that participants who indicate to have more knowledge of the IT field, are more familiar with IBM. However, there was no significant correlation between the Knowledge of the Field and the Quality or
Quantity of Contributions, in contrast to the hypothesis. The Perceived Level of Competition is significantly positively related to the Quality of Contributions, which is in line with
expectations, yet surprisingly there was again no significant correlation with the Quantity of Contributions. Brand Familiarity was found to be positively correlated with the Brand Scepticism, which could be explained by the notion that persons who are not familiar with IBM, have encountered fewer triggers for scepticism towards the brand. Also, Brand Familiarity was significantly positively related to the Quality of Contributions, which is in line with the expectations. Brand scepticism significantly negatively correlated with the
Quality of Contributions, which leads to the suggestion that there might also be a direct negative effect of Brand Scepticism on the Quality of Contributions. Lastly, the Quality of Contributions was significantly positively correlated with the Quantity of Contributions. Surprisingly, the Quantity of Contributions had no significant correlation with any of the variables, apart for the Quality of Contributions. The latter could be explained by the assumption that participants who come up with a higher quality of contributions, are likely more talented in business model crowdsourcing and thus more easily come up with a higher quantity of contributions as well.
Figure 3: Initial Correlations
1 2 3 4 5 6 7 8 9 10
1. Age 1
2. Gender -0.14 1
3. Education 0.22* -0.27** 1
4. Knowledge of the Field 0.18 -0.60** 0.36** 1
5. Competitive Scale 0.19 0.37** 0.55** -0.25 1
6. Collaborative Scale 0.08 0.32* -0.46** -0.30 N/A 1
7. Brand Familiarity 0.34** -0.50** 0.24* 0.74** -0.15 -0.04 1
8. Brand Scepticism 0.16 -0.41** -0.03 0.36** -0.68 -0.09 0.43** 1
9. Quality -0.10 -0.03 -0.05 -0.17 0.39** 0.54** 0.04** -0.35** 1
10. Quantity -0.06 0.06 -0.07 -0.12 0.14 0.13 0.07 -0.09 0.54** 1
*. Correlation is significant at the 0.05 level (two-tailed). **. Correlation is significant at the 0.01 level (two-tailed).
4.3 Perceived Level of Competition and Collaboration
In order to test H1 and H2, separate simple regression analyses were performed. Since participants were either asked for their Perceived Level of Competition or for their Perceived Level of Collaboration, a multiple regression analysis with both those variables was
impossible to execute. As participants were divided among both conditions, both the Perceived Level of Competition (N=57) as well as the Perceived Level of Collaboration (N=55) have only half the sample size of the all the other variables (N=112). Therefore, it was decided that the other variables would not be added to the regression model for H1 and H2, but would be measured in a separate multiple regression model instead.
4.3.1 Perceived Level of Competition on the Quality of Contributions(H1a) The regression model with the Quality of Contributions as dependent variable and the
Perceived Level of Competition as independent variable was found to be significant, F (1, 55) = 9.98, p < 0.05. The regression model can thus be used to predict the quality of contributions in business model crowdsourcing practices. However, this prediction is weak: 17% of
variation in contribution quality can be predicted based on the Perceived Level of
Competition (R2 = 0.17). The Perceived Level of Competition, β = 0.39, t = 3.16, p = 0.003, 95% CI [0.161, 0.720], has a significant, but weak positive effect on the Quality of
Contributions. When someone’s perceived level of competitiveness of a crowdsourcing design increases by 1, the contribution quality increases by 0.44 (both on a scale of 1 to 7). Therefore, hypotheses H1a was accepted.
4.3.2 Perceived Level of Competition on the Quantity of Contributions (H1b) In order to test H1b, ‘The perceived level of competitiveness in a competitive
crowdsourcing design is positively related to the quantity of a participant’s contributions’, a
Quantity of Contributions as dependent variable and the Perceived Level of Competition as independent variable did not turn out to be significant, β = 0.14, F (1, 55) = 1,10, p = 0,300. The regression model cannot be used to predict the quality of crowdsourcing contributions and thus hypothesis H2a could not be accepted.
4.3.3 Perceived Level of Collaboration on Contribution Quality (H2a)
In order to test H2a, ‘The perceived level of collaboration in a collaborative crowdsourcing
design is positively related to the quality of a participant’s contributions’, a simple regression
analysis was performed (N=55). The regression model with the Quality of Contributions as dependent variable and the Perceived Level of Collaboration as independent variable was found to be significant, F (1, 53) = 21.81, p < 0.001. The regression model can thus be used to predict the quality of contributions in business model crowdsourcing practices. However, this prediction is moderately weak: 29% of variation in contribution quality can be predicted based on the Perceived Level of Collaboration (R2 = 0.29). The Perceived Level of Collaboration, β = 0.54, t = 4.67, p < 0.001, 95% CI [0.443, 1.109], has a significant, but moderately weak positive effect on the Quality of Contributions. When someone’s perceived level of collaboration of a crowdsourcing design increases by 1 (on a scale of 1 to 7), the contribution quality increases by 0.78. Based on these results, H2a could thus be accepted.
4.4.4 Perceived Level of Collaboration on Contribution Quantity (H2b) In order to test H2b, ‘The perceived level of collaboration in a collaborative
crowdsourcing design is positively related to the quantity of a participant’s contributions’, a
simple regression analysis was again performed (N=55). The regression model with the Quantity of Contributions as dependent variable and the Perceived Level of Collaboration as independent variable did not turn out to be significant, β = 0.13, F (1, 53) = 0,951, p = 0,334.
The regression model cannot be used to predict the quality of crowdsourcing contributions and thus hypothesis H2b could not be accepted.
4.5 Multiple Regression Model for Contribution Quality (H3A - H7A)
In order to test Hypothesis 3A, a multiple regression analysis was conducted with control variables Gender and Level of Education, and dummy variable Crowdsourcing Design (0 = competitive condition, 1 = collaborative condition) as independent variables and the Quality of Contributions as dependent variable (N=112). The multiple regression analysis (see Figure
4) was hierarchical (or sequential), meaning that the independent variables were added to the
model in blocks. Through this method, each independent variable could be assessed in it’s addition to the prediction of the Quality of Contributions, while the previously added
independent variables are being controlled for. This thus method provides more clear insights in the autonomous contributions of each independent variable to the model. Preliminary analyses were conducted to ensure that there was no violation of the assumptions on
normality, linearity, multicollinearity and homoscedasticity, which was not the case. Gender, β = -0.04, p = 0.663, and Highest Level of Education, β = -0.06, p = 0.535, were entered in block 1, but these control variables did not significantly predict the Quality of Contributions, R2 = -0.01, F (2, 109) = 0.23, p = 0.792 (see Figure 4, Model 1). In Model 2, the
Crowdsourcing Design dummy variable was added to the model while controlling for Gender and the Highest Level of Education, ΔR2 = 0.02, ΔF (1, 108) = 2.13, p = 0.148. The multiple regression model showed that there is no significant effect of Crowdsourcing Design on the Quality of Contributions, β = -0.14, p = 0.148, when correcting for the Gender and the Highest Level of Education. Again, the Multiple Regression Model could not significantly predict the Quality of Contributions, R2 = -0.00, F (3, 108) = 0.87, p = 0.461 (see Figure 4,
crowdsourcing design will generate higher quality contributions than participants who have been exposed to a collaborative crowdsourcing design’, could not be accepted.
In the third step, the demographic factors were entered in the model in order to see if the Age of the Participant and the Knowledge of the Field had an effect on the Quality of Contributions (see Figure 4, Model 3). The addition of Age and Knowledge of the Field significantly changed the predictability of the model, ΔR2 = 0.05, ΔF (2, 106) = 3.12, p = 0.048. However, the model could still not be used to significantly predict the Quality of Contributions, R2 = -0.03, F (3, 108) = 1.79, p = 0.461. The Age of the Participant was not
found to significantly influence the Quality of Contributions, β = -0.08, p = 0.383, when all other independent variables in the model were kept constant. Therefore H4A, ‘The age of the
participant is positively related to the quality of his or her contributions’, could not be
supported. However, the Knowledge of the Field was found to have a significant, negative effect on the Quality of Contributions, β = -0.273, p < 0.01. When a participant’s Knowledge of the Field increases by 1, their Quality of Contributions decreases by 0.24 (both measured on a scale of 1 to 7). Thus, H5A, ‘The problem field knowledge of the participant is negatively
related to the quality of his or her contributions’, could be accepted.
In block 4, the Brand Familiarity of participants was added to the model. The predictive power of the multiple regression model increased significantly with 2.2%, ΔR2 = 0.02, ΔF (1,
105) = 2.60, p < 0.05 (see Figure 4, Model 4). The complete model now had a significant prediction power of the Quality of Contributions of 5%, R2 = 0.05, F (6, 105) = 1.95, p < 0.05. The Brand Familiarity was found to have a significant, positive effect on the Quality of
Contributions, β = 0.24, p < 0.05, when controlling for the other independent variables in the model. When a participant’s familiarity with IBM increased by 1, their quality of
participant’s familiarity with the brand is positively related to the quality of his or her contributions’, could be accepted.
In order to test the last hypothesis (H7) for the Quality of Contributions, Brand Scepticism was first added to the analysis in step 5. Although Brand Scepticism was only expected to have a moderating effect on the effect of Brand Familiarity on the Quality of Contributions, a significant, negative direct effect of Brand Scepticism on the Quality of Contributions was identified through this block (see Figure 4, Model 5), β = -0.52, p < 0.001, when other independent variables were controlled for. With the addition of Brand Scepticism, the predictive power of the multiple regression model for Quality of Contributions increases with 18.9%, ΔR2 = 0.19, ΔF (1, 104) = 27.63, p < 0.001. Together, the independent variables in Model 5 can be used to predict 24.1% of the variance in Quality of Contributions, R2 = 0.24, F (7, 104) = 6.04, p < 0.001. Lastly, the 6th block with interaction variable Brand Familiarity * Brand Scepticism was added to the analysis (see Figure 4, Model 6), to test whether Brand Scepticism also has a moderating effect on the effect of Brand Familiarity on the Quality of Contributions. The model again significantly improved by 6% by the addition of the interaction variable, ΔR2 = 0.06, ΔF (1, 103) = 10.12, p < 0.01. The final model could be used to predict 30.2% of the variance in the Quality of Contributions, R2 = 0.30, F (8, 103) = 7.01, p < 0.001. It was found that, when all other independent variables were kept constant, Brand Scepticism indeed has a significant, negative interaction effect on the effect of Brand Familiarity on the Quality of Contributions, β = -0.34, p < 0.01. To illustrate the negative moderating effect of Brand Scepticism on the positive effect of Brand Familiarity on the quality of contributions, a simple slope was plotted (see Figure 5). This figure shows that the moderating effect of Brand Scepticism indeed changes the valence of the effect of brand familiarity on the Quality of Contributions. Thus, H7a, ‘The participant’s level of scepticism
familiarity with the brand and the (a) quantity and (b) quality of his or her contributions, such that the effect as stated at H6 will turn more when a participant has a higher sense of
scepticism towards the brand.’, could be accepted.
1 2 3 4 5 6 7 Low (-25%) High (+25%) Q u al ity of C on tr ib u ti on s (M =3.94, S D =1.5) Brand Familiarity Brand Scepticism -1 SD Brand Scepticism +1 SD
Figure 4: Multiple Regression Models for the Quality of Contributions
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
1. Control Variables Gender Level of Education -0.129 -0.083 -0.045 -0.048 -0.198 0.027 -0.177 0.044 -0.328 -0.052 -0.336** -0.037 2. H3 Crowdsourcing Design -0.139 -0.135 -0.142 -0.217 -0.173 3. H4, H5 Age
Knowledge of the Field
-0.084 -0.273* -0.137 -0.429** -0.110 -0.408** -0.035 -0.492** 4. H6 Z Brand Familiarity 0.235* 0.384** 0.208** 5. Z Brand Scepticism -0.516** -0.313** 6. H7
Z Brand Familiarity x Z Brand
Scepticism -0.341** R2 Change Adjusted R2 F 0.004 -0.014 0.233 0.019 -0.004 0.866 0.054* 0.034 1.789 0.022* 0.049 1.948* 0.189** 0.241 6.040** 0.064** 0.302 7.013**
*. Significant at the 0.05 level (two-tailed). **. Significant at the 0.01 level (two-tailed).
4.6 Multiple Regression Model for Contribution Quantity (H3B - H7B)
After hypotheses 3 to 7 were tested for their effects on the Quality of Contribution (a), the same had to be done for the Quantity of Contributions (b). Therefore a hierarchical multiple regression analysis was again conducted with control variables Gender and Level of
Education, and dummy variable Crowdsourcing Design (0 = competitive condition, 1 = collaborative condition) as independent variables and the Quantity of Contributions as dependent variable (N=112). The multiple regression model did not violate the assumptions on normality, linearity, multicollinearity and homoscedasticity and therefore the conditions for multiple regression were met. Gender, β = 0.05 p = 0.650, and Highest Level of
Education, β = -0.06 p = 0.531, were entered in block 1, but these control variables did again not significantly predict the Quantity of Contributions, R2 = 0.01, F (2, 109) = 0.41, p = 0.666 (see Figure 5, Model 1). In Model 2, the Crowdsourcing Design dummy variable was added to the model while controlling for Gender and the Highest Level of Education, ΔR2 = 0.03,
ΔF (1, 108) = 3.29, p = 0.073. The multiple regression model showed that there is no significant effect of Crowdsourcing Design on the Quantity of Contributions, β = -0.17 p = 0.073, when correcting for the Gender and the Highest Level of Education. Again, the
Multiple Regression Model could not significantly predict the Quantity of Contributions, R2 = 0.04, F (3, 108) = 1.37, p = 0.255 (see Figure 5, Model 2). Therefore H3B, ‘Participants who
have been exposed to a competitive crowdsourcing design will generate higher quantity contributions than participants who have been exposed to a collaborative crowdsourcing design’, could not be accepted.
Next, the demographic factors were added to the model in order to see if the Age of the Participant and the Knowledge of the Field had an effect on the Quantity of Contributions (see Figure 5, Model 3). The addition of Age and Knowledge of the Field did not significantly change the predictability of the model, ΔR2 = 0.01, ΔF (2, 106) = 0.50, p = 0.609. Also, the
