Master thesis MSc BA. Strategic Innovation Management
The impact of innovation promoters on the diffusion of social innovation
models – evidence from “Bioenergy Villages” in Germany
Niklas Deistler
S3782018
University of Groningen
Faculty of Economics and Business
January, 20
th2020
Supervisor: Dr. B. C. Mitzinneck
Co-Assessor: Prof. Dr. J.D.R. Oehmichen
Word count: 11.314 (excluding references & appendices)
Abstract:
Modern diffusion theory assumes a perspective in which diffusing innovations are subject to
change and adaption. Extant literature has constantly pointed out the importance influence
individuals assume in innovation diffusion processes. Utilizing quantitative methods, this
thesis investigates the influence of two types of such individuals, authority- & expert
promoter, on the extensiveness of diffusing innovative practice. Survey data has been
gathered resulting in a sample of 78 distinctive observations. The applied hierarchical
regression analysis reveals significant negative relationship between both innovation
promoters and the extensiveness of diffused innovative practice. Additionally, the utilized
moderator analysis with cross-functional working groups remained slightly insignificant but
gave hints to a potential positive moderation. Theoretically, the findings suggest to further
investigate the effect of individuals including additional contextual factors and a larger data
set. Practically, the obtained findings imply that a unilateral approach of key individuals
harm the degree to which innovations are adopted in local communities.
1. Introduction
Energy transition represents one of the biggest challenges to society. Starting in the early 2000s, the German Federal Agency of Regrowing Resources (Fachagentur für Nachwachsende Rohstoffe e.V., hereafter: FNR) in collaboration with researchers from the University of Göttingen developed strategies to foster the use of renewable energy in rural areas. The village Jühnde has been selected as a demonstration project and in 2005 was granted the status as the first Bioenergy Village (BEV) in Germany (FNR, 2014). Since then, the community has been regarded as a showcase model of an innovative practice. The project promotes the energy transition and sustainability. The model was developed to show the diffusion of the practice to local communities all across Germany. In 2019, 176 communities have achieved the status as BEVs. The practice, hence, can be considered a successful example of a diffusing social innovation.
Classic diffusion theory assumes that a diffusing innovation remains unchanged from prior versions when adopted in other organizations (Ansari et al., 2010; Rogers, 1995; Westphal et al., 1997). Theories describing the diffusion of innovations have altered over the course of the last decades (Dearing, 2008). A modern approach of diffusion theory states that innovation adoption is a more vivid process allowing for adaptations of the innovation as it diffuses (Fiss et. Al, 2012; Ansari et al., 2014). In this context, Ansari et al. (2010) present a two-dimensional framework that measures the variations as an innovative model diffuses; featuring fidelity and extensiveness as the key dimensions of variation. Extensiveness measures degree a practice has been implemented by the adopter and fidelity measures how similar a diffused practice is to the prior version.
individuals who commit enthusiastically to the Innovation and help to overcome barriers within the adopting organization.
Rogers (1995) first acknowledged the impact that key individuals can have on the adoption of innovative models, but the focus of relevant studies has been on the mere acceptance or rejection of innovations in an organization. Under modern diffusion theory innovation diffusion is considered as a contested process that allows for practice variations. Fiss et al. (2012) highlighted the potential impact key individuals could have, but did not elaborate on these factors further. Additionally, Brandyberry (2003) and Gondo & Amis (2013) highlighted several contextual factors that lead to variations of diffusing innovations, such as the contribution of individuals, but did not differentiate further between the types of individuals. This represents a gap in the existing literature. While the impact of different key individuals on the diffusion of innovation has been found; research into the impact of traits and characteristics of key individuals on diffusing innovations has not been conducted.
Therefore, this study intends to contribute to modern diffusion theory by investigating what impact key individuals can have on the extensiveness of diffusing innovations. Extensiveness represents a particularly interesting domain of the analyzed phenomenon. It may be considered the most relevant measure in the energy transition because as more individuals adopt the BEV model the amount of fossil fuels consumed decreases and results in increased contribution to counteract climate change.
On a managerial level, this macro-lens perspective can offer interesting implications for potential adopters in channeling their innovation efforts. Moreover, innovators can gain valuable knowledge on how to streamline the crucial process of diffusing their innovation and to gain economic rents from them (Schoemaker, 1990). Ultimately, this study will elaborate on the following research question:
What is the impact of innovation promoters on the extensiveness of diffusing innovations?
phenomena is presented. Second, the study context is described. Following by third how the respective hypotheses were derived. In the third section, the applied methodological approach of this research is covered and includes the data collection and sample, the obtained measurements and the analytical instruments used. In fourth section, I will present the results obtained from the analysis and evaluates the hypotheses including assumption- and robustness checks. The fifth section will derive theoretical and managerial implications followed by the sixth section presenting minor limitations and future research directions. Lastly, the I will present several conclusions obtained from this study.
2. Theoretical background
The following section will provide a theoretical foundation of the thesis. The shift in diffusion theory from a classic to a modern perspective will be briefly elaborated and condensed in the discussion of the adoption stage of diffusing innovations. A framework to measure variations in innovations as they are adopted will be introduced featuring extensiveness as one of the dimensions to interpret the results of such variations. Accordingly, section two will highlight several factors influencing variations in the adoption process leading to multiple hypotheses
2.1 Diffusion of innovations
The research on diffusion of innovation has evolved significantly throughout the last decades. Rogers (1995) defines the diffusion of an innovation as the process in which the innovation‘s characteristics are subject to communication through different channels by members of an organization over a certain period of time. An underlying assumption of this definition is that the diffusing innovation does not change during the process. Further, the definition follows a linear path because the diffusion process concludes with either acceptance or rejection of the diffusing innovation. Findings from Westphal at al. (1997) and Tolbert & Zucker (1983) provide additional support stating that within classic diffusion theory the diffusing innovation remains invariable and uniform while the speed of diffusion represents the main research interest in the field.
and their impact with the success of an innovation adoption. Further, they argue that innovations can be subject to modification and alteration during the diffusion process.
Damanpour (1991) has been amongst the first to investigate organizational antecedents influencing the adoption of innovations. His findings showed positive relationships between key organizational characteristics and the adoption of different types of innovations. For instance, functional differentiation, management attitude towards change and specific knowledge resources all positively impacted the way innovations were adopted. Additionally, he analyzed the impact of organizational characteristics in the most crucial stages in the adoption process; the initiation- and implementation stage (Zaltman et al. 1973). Brandyberry (2003) followed this line of reasoning in his study on different factors influencing the adoption of organizational innovations. He found support for the existing assumptions that innovation themselves can be subject to variation among the adoption process and change over the lifecycle of the innovation. In addition to the previous arguments, Ansari et al. (2010) argued that diffusing innovations can be subject to adaption and reconfiguration among the adoption process. They point out that there is a lack of fit between a diffusing practice, such as an innovation, and the adopting organization‘s characteristics being the main reason why adaption occurs during the adoption process.
2.2 Variations in innovation adoption
Several scholars have considered innovation diffusion as a contested process yet they failed to point out what affects such contests at an organizational level (Schneiberg and Soule, 2005; Sanders and Tuschke, 2007). Ansari et al. (2010) analyze the adoption of innovations at an organization & population level and identify characteristics at both levels within the adopting organization. The challenge regarding the adoption of innovations lies in reducing a ―perceived misfit‖ of organizational and innovation characteristics. The perceived misfit represents barriers that have to be overcome in order to create value from adopting an innovation. To overcome these barriers, an innovation becomes subject to variation when being adopted by an organization.
Within their study, they developed a two-dimensional framework that provides a measure of such variations. The framework introduces extensiveness and fidelity as the underlying dimensions through which the adoption of innovation may vary. Extensiveness describes the degree by which an innovation is adopted in comparison to the prior example. For example, this can describe how many business units are utilizing a new technology following the introduction. Fidelity, on the other hand, measures how similar a practice is to the prior version. Here, a low degree of fidelity does not imply success or failure because a low fidelity might also indicate variation among the adoption process.
Among others factors, Ansari et al (2010) account for a technological and political fit of the innovations and incorporate temporal differences which result in different adaption patterns when adopting. These characteristics include, for example, technological knowledge, individuals‘ backgrounds, power relationships and organization members‘ interests. According to Fiss et al. (2012), the advantage of the framework of Ansari et al. (2010) is characterized by the fact that it considers both the nature of the practice and its respective adoption at the same time. The framework enables its adopters to analyze practice variations in detail to understand the motives affecting the relative spread.
2.3 Involvement of key individuals (Innovation promoters)
participation of key individuals in an organization is a crucial factor when overcoming barriers of innovation adoption. He identifies the individual promoting the adoption of an innovation requires access to certain local networks and power structures. Further complex innovation characteristics need to be translated into a common language in order to be adopted extensively by an organization. Brandyberry (2003) supports these findings as variations in the adoption of innovative practices might stem from different individuals leading the adoption process.
2.3.1 Authority promoter
In this context, Witte (1999) first termed individuals who promote innovations in organizations as innovation promoters. These individuals possess certain traits and characteristics which help them to overcome organizational barriers. Promoters with executive authority, for instance, are able to push an adoption decision forward and decide it autonomously (Eisenhardt & Zbaracki, 1992).
Authority promoters are typically situated in top-management or decision-making positions. The decision-making process typically follows a top-down-approach. The promoter acts in the best interest of the organization towards goals such as to establish a sustainable advantage in the long run (Quinn, 1977). Due to the decision-making power an authority promoter has and the responsibility to act in the organization‘s best interest, we can expect that the involvement of an authority promoter results in a higher extensiveness of an adopted innovative practice.
create acceptance among potential adopters. Combining the power stemming from both, his position and social capital, helps the authority promoter to create acceptance and, ultimately, legitimacy for an innovation to be adopted. Gondo & Amis (2013), in this context, found that the variations in the extensiveness of adopted innovations are relative to the perceived acceptance within an organization. Hence, the more people perceive an adopted innovation to be useful, the more people will employ the innovation as part of their activities.
Combined with the help of the authority promoter‘s efforts, trust and acceptance can be established and organizational barriers can be overcome. The authority promoter‘s strong social ties, access to resources and decision-making authority allows him to gather collective support for an innovation. Ultimately, this leads to a higher extensiveness of an innovation adoption.
2.3.2 Expert promoter
Besides the authority promoter, Witte (1999) identifies another key individual that can have significant impact on overcoming barriers of innovation adoption. Witte (1999) describes this type of innovation promoter as the one with distinct technological expertise regarding a certain innovation that enables one to identify advantages for potential adoption. For clarification purposes, this study will use the term ‗expert promoter‘ when referring to this type of innovation promoter.
Unlike the authority promoter, the expert promoter is mostly situated within the members of the organization. The promoter distinguishes itself from the other members through its advanced capabilities and access to resources. The expert promoter has access to physical resources to adopt an innovative practice and the technological knowledge to use them effectively (Greene et al., 1999). These distinctive features allow the expert promoter to overcome barriers regarding the applicability and functionality within the organization. Following Teece (1986), the expert promoter possesses specialized complementary assets that allow him to obtain value from the adoption of an innovative practice. This allows the expert promoter to make efficient and thorough use of his resource. Additionally, the promoter can take personal benefits from the adoption such innovative practice (Lawless & Price, 1992).
the legitimacy needed in order to overcome organizational barriers. In fact, contrary to the authority promoter, the expert promoter lacks the social capital needed to establish trust and, respectively, acceptance among the members of an organization (Rogers, 1995). In contrast, the other members perceive his advancements as acting in his own best interests. His social ties within the organization are not strong enough to convince others of the need to adopt because they only see the expert promoter benefiting in the first place (Hewes & Lyons, 2008).
Furthermore, the promoters‘ distinctive technological knowledge and skills may lead to information asymmetries and, consequently, result in problems of perceived control. Power inequalities may lead to distrust in the potential benefits of the adopting innovation and, thus, reduce the likelihood of an individual adoption (White, 1985, Fiss et al., 2012). Members of the organization could perceive the need for the innovation to be unilateral which leads to a lower acceptance and consequently to a lower extensiveness (Gondo & Amis, 2013; Rogers 1995).
In summary, the expert promoter possesses the necessary skills, technological know-how and access to the respective physical resources needed in order to realize adoption of an innovative practice. However, he lacks the social ties throughout the organization needed to mitigate risk and generate enough acceptance among the members of the organization. The expert promoter is not able to establish the organizational legitimacy needed to adopt an innovative practice thoroughly. Consequently, this leads to a lower degree of extensiveness of the adopted innovation.
2.3.3 Moderation effect of working groups
that gathering cross-functional support among organization members helps facilitate the acceptance-decision towards an innovation.
Therefore, a possible way to align the interests of both organization members and innovation promoters might be to include both of them in the decision-making process. McKauley and Kuhnert (1996) found that participation in group-decision making of an organizational decision, such as the adoption of a new practice, promotes trust among its members. Trust helps creating acceptance and mitigates the perceived risk among potential adopters (Pavlou, 2003). In turn, more organization members will adopt an innovation which results in a higher degree of extensiveness
To be more precise, both innovation promoters could benefit from gathering additional support. The expert promoter, in this case, could increase his social capital with collective interests represented in the decision-making process (Meinzen-Dick, 2004). Contrary to the theoretical suspicion elaborated in the previous section, the collective action could mitigate potential information asymmetries which in turn could help increasing trust among organization members (Monge et al., 2008). Consequently, members of the organization might not perceive the need to adopt unilaterally and may establish the acceptance which could increase the degree of extensiveness. The involvement of working groups could reverse the negative effect implied by the unilateral approach leading to a positive influence on the extensiveness when combining the efforts.
As for the authority promoter, the incorporation of cross-functional working groups in the decision-making process would have a similar effect. An increase in social capital facilitated by collective action. The extensiveness of the adopted innovation would further increase because even the non-conformity of some organization members could be overcome and trust could be gained. For instance, Draft (1978) found that more people adopt an innovative practice if top management focuses on forming professional collectives to lead the innovation adoption process.
practice. Consequently, the formation of working groups is expected to positively impact the adoption decision among organization members leading to a higher degree of extensiveness.
2.4 The case of Bioenergy Villages
This section will present the utilized unit of analysis. The Bioenergy village (BEV) model will be described in detail and, based on the theoretical advancements in the previous sections, hypotheses encompassing the impact of key individuals on the extensiveness of adopted BEV models will be derived.
In early 2000, the German government passed the Renewable Energies Act (Federal Ministry for the environment, 2000). The act contained feed-in tariffs to encourage projects featuring renewable energies. Along the promotion of a transition from fossil fuels to renewable energy sources, the act offered multiple domains for entrepreneurial endeavors.
Shortly after, a team of researchers from the University of Göttingen and collaborators of the government-funded FNR started developing a model that incorporates the advancements of the Renewable Energy Act for local communities. The model was designed to reduce the dependence on fossil fuels, promote energy independence and act against global warming (FNR, 2014, p.8f). At its core, the model plans to utilize biomass from agriculture and forestry in biogas plants to produce heat and energy supplied within a local heating grid. Besides that, the model lays out multiple ways to found a cooperative where every participating household can become part of. Ultimately, the heat and
Figure 1: Conceptual model
energy supply from the BEV initiatives is supposed to always be cheaper than depending on fossil fuels due to the feed-in tariffs imposed by the act. The village‖ Jühnde‖ has been selected as the first village to adopt the model and it was rolled out successfully in 2005. At the time of writing this paper, there are 176 villages which have adopted the BEV model.
Following the successful first BEV model adoption in Jühnde, the FNR presented a guideline describing a best-practice approach for potential adopters. The guidelines have been regularly updated and this thesis draws on data from the most relevant guideline published in 2014. The guidelines contain certain procedures and parameters which help to achieve the official BEV status. Among others, the guidelines point out that at least 50% of the local heat- and 100% of the local energy supply of the has to come from biomass sources in order to achieve the BEV status. Furthermore, the guidelines provide a desirable extensiveness of the BEV model of 80% in the respective in order to follow the relevant best practice and obtain the greatest benefits.
Besides the above-mentioned parameters, the guideline contains a number of organizational procedures that aim to ease the adoption of the BEV model. Appendix A portrays several steps that aim to guide a successful implementation of the innovative model. The illustration identifies a project initiator— an individual within the among the village community (i.e. Farmer or Mayor). Additionally, it is advised to build cross-functional working groups to carefully evaluate the needs and current circumstances. Once a general appropriateness need has been asserted, the actual planning and execution begins.
2.5 Hypothesis Development
individuals. This ultimately influences the extensiveness of an adopted innovation. Additionally, the case might offer some interesting contributions for business environments as the structures and decision-making processes stem from similar roots.
2.5.1 Authority Promoter
As elaborated in the prior theoretical background of the impact of key individuals, I expect an authority promoter to have a positive impact on the extensiveness of the adopted BEV Model. When the BEV model is initiated by the authority promoter, in this case, the mayor, I expect a high degree of extensiveness. The authority promoter possesses strong social capital embedded in trust relationships as the elected mayor. Those strong social ties build the foundation to establish organizational legitimacy of the adopted BEV model. Due to his position, the mayor has a significant administrative power to execute such an adoption and follows the village‘s best interests. When combined together, the authority promoter possesses the necessary social capital as well as the administrative authority to generate trust and acceptance among the inhabitants which will ultimately lead to more people willing to adopt the BEV model and become connected to the local grid. This brief summary put into context leads to the first hypothesis:
Hypothesis 1: The involvement of an authority promoter will have a positive impact on the extensiveness of an adopted BEV model.
2.5.2 Expert Promoter
the village. Primarily, they see the expert promoter as benefitting from a sufficient use of his resources and additional streams of revenue. This can create distrust among potential adopters as they perceive the need for the BEV model adoption to be unilateral. In consequence, fewer people agree to become connected to the grid which leads to a lower extensiveness of the adopted BEV model. Hypothesis 2 sums up the expected relationship:
Hypothesis 2: The involvement of an expert promoter will have negative impact on the extensiveness of an adopted BEV model
2.5.3 Moderating effect of working groups
As suggested by the FNR guideline, the formation of working groups is an appropriate action to incorporate various cross-functional interest and, thus, considers the need for the BEV adoption from multiple sites. The suggested approach also holds great theoretical support as discussed in the previous section. To be more precise, the formation of working groups helps to increase the social capital of the respective innovation promoters and creates trust and acceptance for the BEV model adoption. By incorporating cross-functional working groups into the decision-making processes, information asymmetries between the technological functionality and the use of resources, for example, could be mitigated. Furthermore, since the adoption of an entirely new stream of energy and heat supply represents a sensitive project to the adopters because a new connection has to be built too, the authority promoter benefits from incorporating as many stakeholders as possible in the adoption process. Both, the authority- & the expert promoter, would benefit from incorporating cross-functional working groups in the adoption process, because it creates additional trust and acceptance among the village inhabitants. Additionally, it helps to reduce information asymmetries regarding the individuals benefiting from the adoption. Collectively, the arguments lead to hypothesis three and four:
Hypothesis 3: The formation of working groups further positively impacts the positive relationship between the authority promoter and the extensiveness of the adopted BEV model
3. Methodology
As previously outlined, the impact of innovation promoters on the adoption of innovation models has not been empirically tested yet. The different dimensions of innovation adoptions previously derived from the literature (Ansari et al. 2010), helps to scrutinize obtained survey data in order to measure statistically the adoption of innovation models among the organization. This study will apply a hierarchical regression analysis that tests the impact of innovation promoters on the extensiveness of a diffused BEV‘s. The following section will describe the process of data collection and the sample data utilized for this analysis. Additionally, the data measures and their individual operation are explained. This section concludes with a description of the data analysis models applied to carry out the analyses.
3.1 Data Collection
The initiation of this study was represented by a data set compiled by Dr. Mitzinneck. This data set consisted of 147 observations of BEV‘s across Germany as listed in the online catalogue of the FNR. The data was based on self-reports from the BEV‘s responsible individuals and enriched by either direct request or website data. This data set, however, did not display the data needed to test the outlined hypotheses.
Therefore, a new data set, based on the previous one, had to be created. In an initial check, an additional 29 appointed BEV‘s were identified and added to the underlying list. The new list of potential observations consisted of 176 BEV‘s. To obtain the data necessary, a survey consisting of 26 questions was developed. The goal of the survey was to gather detailed data about the development process, the applied technology and planned endeavors in the future of the BEV‘s.
The data collection was conducted through telephone interviews using the telephone numbers available from the self-reports by the FNR and complemented by online research regarding appropriate individuals to interview for the respective villages. Furthermore, the email-addresses of the contact persons were collected in order to be able to send follow-up emails if the individual contacted did not have all the required information at hand.
figure or categorical questions. The answers to the categorical questions were obtained applying the ―in-vivo‖-approach in which the interviewee is not offered the list of answer-options, but, instead, the answer is written down and matched to the answer option with the best fit. This procedure enhances the quality of the data obtained as the interviewee is not conflicted with any choices that could influence the answer.
The interviews were executed throughout a period of 3 months. In total, observations from 79 BEV‘s were collected. To increase the likelihood of reaching the contact persons, the interviews were mostly conducted in the afternoon or early evening because most interview partners are not engaged in a full-time position with the BEV‘s. Additionally, following the interviewees consent, some of the interviews were recorded in order to eliminate potential bias as an additional person would be able to review the obtained data. In total, 25 interviews were recorded with interview partners from 8 different federal German states (Interviewees from 12 different states).
Following the interviews, the data collected was manually converted from the questionnaires into a Microsoft Excel file. The questions were separated in such a way that every answer option was represented with a single column. This procedure helped because every answer, except the numerical questions, could be answered with a ―0‖ or ―1‖ in order to establish an easier and more distinct approach for statistical analysis (Suits, 1957). Here, ―0‖ represented a negative or non-existent answer and ―1‖ represented a positive observation.
Following the conversion, additional data was obtained from a database provided by the German Federal Statistical Office. The database accessed is called ―Genesis-Online‖ and provides data about the economy, the environment, the society and the state (Statistisches Bundesamt (Destatis), 2019). Furthermore, data regarding the distance from the showcase model in Jühnde to the respective BEV‘s was collected. Here, the Google based service ―Google Maps‖ was utilized to find the distances between the postcodes of the later BEV‘s and the showcase model in Jühnde. As Google Map‘s provides multiple distances, the shortest distance presented was selected.
accurately. The last step concludes the data collection process which lead to the data set employed for this study.
3.2 Measurements
The following section will describe the variables utilized and their measurements. First, the dependent variables will be described followed by the independent variables. Lastly, the control variables will be presented.
3.2.1 Dependent variable
As previously outlined, modern diffusion theory allows for practice variation along the adoption of innovations. Ansari et al. (2010) introduced two dimensions to measure such practice variation. This study will feature extensiveness, also referred to as intensity, which measures the degree by which an innovation has been adopted by an organization (Ansari et al., 2010; Wu, Mahajan, & Balasubramanian, 2003). In order to apply it to the case of the BEV models, this study will feature the connectivity rate to the majority of cases with newly built grid system in the villages. Therefore, the higher the connectivity rate, the higher the extensiveness of the BEV model.
The operational measure for extensiveness is represented by the number of households connected to the grid divided by the total number of households in the village. Following from that is the percentage score obtained from the calculation which represents the extensiveness of the adopted BEV model.
3.2.2 Independent Variable
depiction because they consider an innovation promoter to be the individual who promotes the implementation of an innovation into an organization.
The operational use of the independent variables follows the logic previously outlined regarding the power and trust relationships within the organization. The mayor of the respective village will represent the authority promoter. The farmer, on the other hand, will represent the expert promoter—the individual in charge of operating the biogas plants as well as providing manure and wood resources. Hence, the farmer might act from a different intrinsic motivation in order to build an energy grid as extensive as possible. The data can be retrieved clearly from the dataset through question 14 in the questionnaire which asks for the individual who first presented BEV‘s model to the inhabitants of the respective village.
3.2.3 Control Variables
The adoption of innovation in organizations in most cases is influenced by a number of different factors. Damanpour et al. (2006) characterized the adoption of innovation as a multidimensional process impacted by environmental and contextual factors. This study tries to incorporate factors from all of the above-mentioned dimensions to achieve the best possible fit of control variables to help explain the investigated phenomenon and rule out alternative explanations.
First, a number of contextual factors have been integrated as predictors. The average votes in % for the party ―Bündnis 90 die Grünen‖ (i.e. green votes) per county were integrated because political stability impacts the adoption of innovation according to Berry & Berrry (1990). Again, an average has been calculated including the election results from 2005 to 2017. Additionally, the number of inhabitants per village has been included because it proved impactful on the adoption of innovations in organizations (Kimberly & Evanisko, 1981).
variable ―Extensiveness‖ as it serves a suitable predictor regarding how many households are able to become connected to the heating grid.
Furthermore, the geographical and temporal distance to the showcase model in Juehnde has been included into the analysis as control variables. Both measurements are also applied in the paper of Ansari at al. (2010). Regarding the temporal distance, the showcase model represents the zero-value counting up the year the BEV model adoption has been initiated in the respective villages. Geographical distance is utilized by obtaining the distance in km between the postcode of the showcase model and the postcode of the respective village obtaining the BEV model.
Second, the environmental factors supporting the BEV model are analyzed. Manure and wood are considered the main energy sources for production of energy and heat in every BEV (FNR, 2014). In this context, the manure availability and average forest area in km^2 per county have been incorporated. The manure availability is represented by an average of cow and pig density per km^2 per county. Furthermore, the variable has been standardized to enhance the statistical applicability (Milligan & Cooper, 1988).
Another environmental factor impacting the adoption of innovation is community wealth and access to financial support (Damanpour et al., 2006). Both factors have been included in the form of the average county GDP from 2005 to 2017 and the sum of financial support received from funding programs.
3.2.4 Moderator Variable
In order to find an effect of a third variable on the relationship between innovation promoters and extensiveness, a moderator analysis is utilized. The moderation effect of the formation of expert-groups within the adoption process – is subject to investigation in this study.
3.2.4.1 Formation of expert-working-groups
The moderator variable is represented as a dummy variable accounting for either a successful formation of multidisciplinary working-groups or not. The expert working groups consist of individuals representing different interests of members in the organization. Therefore, I expect the working groups to have a positive impact on the relationship between an innovation promoter and the extensiveness of the adopted BEV model.
3.3 Data Analysis
The data analysis is carried out applying a hierarchical regression analysis to the obtained data. Due to the continuous nature of the dependent and independent variables, the selected statistical method is suitable. All of the control variables are continuous variables. In order to enable a better interpretation (Pandis, 2016), the control variables around the village shape, ―Village Shape Similar‖ and ―Village Shape Less Similar‖, have been converted into two dummy variables from their prior categorical nature.
The applied hierarchical regression analysis models are beneficial for multivariate analysis since they show advancements from one model to another (Tabachnick and Fidell, 2014). All models test for the dependent variable ―Extensiveness‖. The first model tests for the effect of the control variables alone. The second model contains the control variables and the first independent variable ―Authority Promoter‖. The third model contains the control variables and both independent variables ―Authority Promoter‖ and ―Expert Promoter‖. The fourth model enhances the previous models by including the interaction effect of ―Working Groups‖ with the independent variable ―Authority Promoter‖. Similarly, the fifth model contains an interaction effect of ―Working Groups‖ and the other independent variable ―Expert Promoter‖.
4. Results
Third, the statistical models will be presented in order to test the hypotheses. Finally, this section concludes with several robustness checks to test the findings and eliminate potential bias.
4.1 Descriptive Statistics
Table 1 reports the data for the summary statistics and the correlation matrix of the study variables. The table includes mean and standard deviations as well as the significance levels of the correlations.
The original data set contained 79 observations applicable for the analysis. However, one observation has been dropped because it represents the showcase model ―Juehnde‖ with many variables reporting 0-values as they serve as the base for calculations.
The summary statistics offer a first glance at the data. None of the variables report specific anomalies. The variable ―Number of Inhabitants‖ displays values in the thousands since the input-data remained unchanged. Furthermore, the variable ―Green Votes‖ is reported in percentages and the variable ―Temporal Distance‖ is in years.
The correlation matrix contains information about the linear relationships of the variables selected to explain the investigated phenomenon. The significance levels were kept from 95% Confidence Interval (CI) indicating significant relationship to 99,9% CI – indicating a strong significant relationship. The star-values indicate significant relationships. Only a few variables reported significant relationships. For the dependent variable ―Extensiveness‖, two significant relationships were found. The control variables ―Village Shape 2‖ (r = 0.32) and ―Geographic Distance‖ (r = -0.27) reported significant relationships, both at a 99% CI. The variable ―Working Groups‖, utilized for the moderation analysis with the independent variables ―Authority Promoter‖ and ―Expert Promoter‖ reports significant relationships with both independent variables (r = 0.31& r =
Table 1: Descriptive Statistics and Correlations
Variable Mean Standard
Deviation (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (1) Extensiveness 0.617 0.227 1 (2) Authority Promoter 0.308 0.465 -0.07 1 (3) Expert Promoter 0.308 0.465 -0.17 -0.32** 1 (4) Working Groups 0.359 0.483 0.05 0.31** -0.26* 1 (5) Manure Availability 0.013 0.784 0.11 -0.16 -0.05 -0.15 1 (6) Forrest Area 0.381 0.115 0.01 0.02 -0.20 0.12 -0.40*** 1 (7) Number of Inhabitants 1,262 2,773 0.03 0.21 -0.06 0.17 -0.16 -0.02 1
(8) Avg. GDP per Person 0.029 0.005 0.17 -0.07 0.08 0.01 0.35** -0.02 -0.08 1
(9) Total Funding Received 0.638 0.820 0.03 0.07 -0.12 0.20 -0.30* 0.34** 0.38*** 0.05 1
(10) Green Votes 8.623 3.214 0.00 -0.08 0.00 -0.11 -0.01 0.00 -0.12 0.23* 0.08 1
(11) Village Shape Similar 0.256 0.439 -0.11 0.05 -0.01 -0.01 -0.10 0.35** 0.07 0.06 0.11 -0.19 1
(12) Village Shape Less Similar 0.603 0.493 -0.14 -0.14 -0.03 -0.05 0.12 -0.24* 0.03 -0.18 0.02 0.19 -0.72*** 1
(13) Village Shape Not Similar 0.141 0.350 0.32** 0.13 0.05 0.08 -0.03 -0.10 -0.13 0.18 -0.17 -0.02 -0.24* -0.50*** 1
(14) Geographical Distance 2.723 1.433 -0.27** -0.21 0.18 -0.40*** -0.01 -0.09 0.12 0.17 0.05 0.50*** 0.04 0.04 -0.10 1
(15)Temporal Distance 8.577 3.522 -0.21 -0.15 0.06 -0.23* 0.02 0.03 -0.26* 0.17 -0.10 0.29* -0.21 0.18 0.02 0.17 1
Note: Number of observations: 78
4.2 Assumption Checks
In the following section, the fit of the selected quantitative method will be checked through seven assumption checks. The assumption checks are necessary to carry out an evaluation into whether a hierarchical regression analysis is the suitable method to carry out the analysis. The first two assumption checks are separated into two groups. The first two checks evaluate the study design and utilized measurements. The other five assumption checks evaluate the applied detail relating to their statistical applicability. If all assumption checks result in positive outcomes, the selected statistical method fits the utilized data.
The first two assumption checks require the dependent variable to be continuous and the independent variable to be either continuous or categorical. Both cases apply in this study. The dependent variable is continuous and the independent variables are both of categorical nature.
The third assumption check requires independence of the observations. In order to validate the assumption, the variance inflation factor (vif) for each of the variables is calculated. Among all variables no cases of multicollinearity were found. Ideally, all variables should not report vif values above 2,5, but values up to 10.0 lie within the acceptance threshold. Both independent variables report vif values well below the 2,5 thresholds. ―The control variable ―Village Shape 0‖ marginally surpasses the 2.5 threshold with a value of 2.52. The complete table of the reported vif values can be found in Appendix B.
The fourth assumption check examines the dependent and independent variables in terms of their linear relationship. A linear relationship can be observed by visually checking the best line of fit in a scatterplot. Here, both independent variables have to be observed individually. The examination of both scatterplots resulted in an unambiguous outcome, supporting clear linear relationships between dependent and independent variable
CooksD. Both of the described measurements assist in identifying outliers in the data set. First, the studentized residuals were checked for observations exceeding an absolute value of three standard deviations which would characterize a potential outlier. One observation was found to meet that criteria. To further evaluate this result, CooksD was analyzed as well. The measurement has been established after criticism in research regarding the measurements of studentized residuals has come up (Cook & Weisberg, 1982). Here, an outlier can be identified if it exceeds the value of one. None of the observations reported a value above one, hence, no significant outliers were identified and the assumption was verified.
Consequently, the sixth assumption check is carried out by checking for the presence of homoscedasticity or, in other words, for equal error variances. To carry out the analysis, the standardized residuals were plotted against the unstandardized residuals. In order to confirm the presence of homoscedasticity, a random pattern of the plotted values is expected. An increasing-, decreasing- or funnel-shaped pattern serves as an indication for potential heteroscedasticity. The obtained graph showed ambiguous results. Therefore, an additional test was carried out. The Breusch-Pagan / Cook-Weisberg test for heteroskedasticity (Koenker, 1981) is executed introducing a null-hypotheses for constant variances. The reported p-value (0.23) allows to reject the null-null-hypotheses and verifies the presence of homoscedasticity in the data set.
Lastly, the seventh assumption check tests for the distribution of the residuals. The residuals, in this case, should follow a normal distribution with a bell-shaped curve. Utilizing Stata, a histogram as well as a normal quantile plot, to also consider possible skewness, are generated. Both graphs do not show any anomalies. The histogram shows a normal distribution including the bell-shaped curve. The Normal-Quantile-Plot, also, shows a normal distribution of the residuals with the inverse-normal line. Both results are reported in Appendix C.
4.3 Hypothesis testing
The following section presents the results of the hierarchical regression analysis. Below, table 2 reports the findings of such regression analysis. The statistical software Stata is utilized to carry out the regressions featuring the dependent variable ―Extensiveness‖ in five subsequent models. The first model contains the control variables. The second model contains the control variables and the predictor variable, ―Authority Promoter‖. The third model extends the second model by including the additional second predictor variable, ―Expert Promoter‖. The fourth model contains all variables of model three, but further introduces a moderation of ―Working Groups‖ on the first independent Variable ―Authority Promoter‖. Lastly, model five also includes a moderation of ―Working Groups‖ on the second independent variable, ―Expert Promoter‖.
Table 2: Regression Results
Variables Model 1 Model 2 Model 3 Model 4 Model 5
Extensiveness Independent Variables Authority Promoter -0.092 -0.126 -0.193 -0.128 (0.101) (0.029) (0.012) (0.027) Expert Promoter -0.113 -0.116 -0.089 (0.048) (0.040) (0.159) Moderator Variables 0.150 (0.175) -0.111 (0.397) Working Groups -0.090 -0.072 -0.091 -0.143 -0.073 (0.118) (0.211) (0.113) (0.039) (0.236) Control Variables Green Votes 0.014 0.015 0.014 0.013 0.013 (0.131) (0.098) (0.125) (0.130) (0.129) Village Population 0.000 0.000 0.000 0.000 0.000 (0.313) (0.187) (0.180) (0.285) (0.222)
Village Shape Similar -0.238 -0.244 -0.243 -0.224 -0.262
(0.006) (0.005) (0.004) (0.008) (0.003)
Village Shape Less Similar -0.192 -0.209 -0.215 -0.203 -0.226
(0.012) (0.007) (0.004) (0.007) (0.003) Geographic Distance -0.061 -0.066 -0.065 -0.065 -0.065 (0.005) (0.002) (0.003) (0.002) (0.002) Temporal Distance -0.017 -0.017 -0.017 -0.017 -0.019 (0.030) (0.026) (0.021) (0.019) (0.016) Forrest Area 0.196 0.170 0.040 0.067 0.045 (0.441) (0.500) (0.875) (0.792) (0.860) Manure Availability 0.029 0.024 0.004 0.013 0.002 (0.441) (0.529) (0.915) (0.739) (0.948)
Total Funding Received 0.012 0.010 0.008 0.007 0.010
(0.722) (0.763) (0.821) (0.840) (0.764)
Average GDP per Person 6.635 6.327 7.842 6.996 7.869
(0.212) (0.228) (0.132) (0.179) (0.132) constant 0.730 0.785 0.858 0.883 0.873 (0.000) (0.000) (0.000) (0.000) (0.000) R-Squared 0.297 0.325 0.366 0.384 0.373 Adj. R-Squared 0.179 0.201 0.237 0.247 0.234 N 78 78 78 78 78
4.3.1 Hypothesis 1
The first hypothesis predicts that the involvement of an authority promoter will have a positive impact on the extensiveness of the BEV model adopted by the respective village. Following the prediction, the regression coefficient is expected to be positive and significant. Model two contains only the independent variable authority promoter. The beta coefficient is negative but the reported P-value is at 0.101 and, therefore cannot be considered significant.
Surprisingly, model three portrays a different picture on the predicted relationship. Considering the results of model 3, including both independent variables, there is a significant negative relationship between the involvement of authority promoter and the extensiveness of the BEV model. The analysis shows a p-value of 0.029 with a beta coefficient of -0.126. The obtained results imply that the presence of an authority promoter in the adoption process decreases the extensiveness by 12,6%. Additionally, the model reports an increase in the explained variance (R²) of 4,1% to model two and an increase of 6,9% to model one. Hypothesis 1 is not supported.
4.3.2 Hypothesis 2
The second hypothesis predicts that the involvement of an expert promoter will have a negative impact on the Extensiveness of the BEV model adopted by the respective village. Again, the regression coefficient is expected to be negative and significant. The analysis of model three reveals this to be the case. The analysis portrays a p-value of 0.048 with a beta coefficient of -0.113. Hence, the involvement of both innovation promoters proves to be negative and highly significant. The extensiveness of an adopted BEV model decreases by 11,3 % when advanced by an expert promoter. Furthermore, the reported beta coefficients of the main effects only differ by 0.013, indicating an almost equally strong effect. In conclusion, there is significant support for the predicted relationship. Hypothesis two is supported.
4.3.3 Hypothesis 3
significant. Model four reports the results of the regression. Although providing a positive beta coefficient, the results are not significant (p=0.175). Consequently, hypothesis three is not supported.
However, model four contains some other interesting findings. The variable working groups, utilized for the interaction effect, shows a negative impact on the dependent variable featuring a beta coefficient of -0.143 at a p-value of 0.039. Both of the independent variables remain highly significant including improved p-values (authority promoter: p=0.012 & expert promoter: p=0.040) as well as slight negative increases in their impact on the dependent variable. Additionally, model four reports an increase in the explained variance (R²) of 1,8%. Also, the adjusted R-square increased by 1,0%.
4.3.4 Hypothesis 4
Hypothesis four predicts the that working groups positively influence the previously negative assumed relationship between the expert promoter and the extensiveness of the BEV model. Complementing hypothesis three, a collective engagement of the expert promoter and a working group positively impacts the number of households connected to the local heating grid. In order to find support for the hypothesis, the interaction should be positive and significant. Model five analyzes the outlined effects. Contrary to the expectations, the moderation is not significant (p=0.397). There is no support for hypothesis four.
Also, the variable working groups does not show a significant value that is any longer in comparison to model four. Furthermore, one of the main effects (i.e. Expert promoter) is not significant any longer (p=0.159). The other main effect remains significant, yet a decrease in the significance is observable (p = 0.027 vs p=0.012). This represents a change from the two prior models which reported significant relationships of both independent variables on the dependent variable. Lastly, model five reports a decrease in the explained variance of 1,1%.
4.4 Robustness Test
The guidelines published by the FNR propose an extensiveness of 50-80% to be ideal for an adopted BEV model. In the robustness checks, I transformed the dependent variable extensiveness into a dummy variable accounting for ―1‖ if the threshold of 50% has been surpassed and ―0‖, if not. To test the effect on the expected relationships, a logit regression is carried out. The logit regression utilizes a binary dependent variable. The logit regression analyzes the probability that the dependent variable equalizes ―1‖. The distribution of the logit model follows a cumulative standard logistic distribution (Stock & Watson, 1988). In order to analyze the results of the produced outputs correctly, the coefficients cannot be read as ordinary least squares (OLS) coefficients. In order to analyze the effects of the independent variables on the dependent variable, the Odds Ratio (OR) has to be examined. The OR represents the odds of the dependent variable to be ―1‖ when the independent variable increases by one unit. To be more precise, if the OR > 1 the odds that the dependent variable is ―1‖ increases. Vice versa, if the OR < 1 the odds that the dependent variable is ―1‖ decreases. Furthermore, to test whether the analyzed models contain some explanatory power, the measure ―Prob > chi2‖ is observed. If the measure reports values < 0.05, it can be assumed the model contains some explanatory power. Additionally, the logit regression reports a ―Pseudo R2‖ which is comparable to the R-Squared known in the regular OLS regressions. In order to test the fit of the variables, similarly to the assumption tests, I applied the Hosmer-Lemeshow test with an analysis of the classification table and sensitivity checks which tells how well the obtained results were predicted by the predictors. None of the tested models reported anomalies.
Table 3: Robustness Test (Extensiveness > 50% Threshold)
Variables Model 1 Model 2 Model 3 Model 4 Model 5
Extensiveness (Threshold > 50%) Independent Variables Authority Promoter 0.127 0.864 0.124 0.841 (0.008) (0.004) (0.034) (0.004) Expert Promoter 0.361 0.364 0.406 (0.176) (0.176) (0.255) Moderator Variables 0.293 (0.466) 0.435 (0.656) Working Groups 0.720 1.185 0.947 1.796 1.094 (0.638) (0.831) (0.948) (0.647) (0.922) Control Variables Green Votes 1.001 1.025 1.028 1.029 1.029 (0.959) (0.824) (0.804) (0.802) (0.801) Village Population 1.000 1.000 1.000 1.000 1.000 (0.925) (0.601) (0.555) (0.522) (0.575) Village Shape Similar 0.542 0.336 0.345 0.298 0.328
(0.569) (0.350) (0.370) (0.320) (0.345) Village Shape Less Similar 0.591 0.361 0.347 0.311 0.344
(0.585) (0.333) (0.327) (0.287) (0.319) Geographic Distance 0.624 0.519 0.516 0.609 0.515 (0.065) (0.027) (0.027) (0.025) (0.027) Temporal Distance 0.825 0.817 0.820 0.807 0.819 (0.086) (0.096) (0.107) (0.096) (0.100) Forrest Area 5.767 5.524 1.989 1.478 2.170 (0.573) (0.607) (0.839) (0.910) (0.821) Manure Availability 1.442 1.240 1.086 1.007 1.079 (0.495) (0.700) (0.886) (0.991) (0.896) Total Funding Received 1.020 0.975 0.951 0.9627 0.952
(0.961) (0.954) (0.916) (0.934) (0.916) Average GDP per Person 5.390 2.040 6.060 6.670 1.020
(0.598) (0.676) (0.598) (0.555) (0.576) constant 20.502 98.349 195.602 239.304 173.409 (0.219) (0.094) (0.062) (0.057) (0.067) Prob > chi2 0.400 0.070 0.059 0.072 0.080 Pseudo R2 0.127 0.219 0.24 0.246 0.242 N 78 78 78 78 78
Note: p-values in parentheses
5. Discussion
The following section will discuss the obtained empirical findings and interpret the information to draw theoretical and managerial implications. To be more precise, the results of the analysis will be interpreted in context of the stated hypotheses and the derived research question. Additionally, limitations of this study and propositions for future research endeavors will be presented.
In the context of the transition of diffusion theory from a classic- to a more modern perspective that allows for variations of innovations in the adoption process; this study investigated the impact key individuals in the adoption process have on such variations. In short, significant relationships were found between the two predictors, the authority promoter and the expert promoter, and the extensiveness of an adopted practice.
5.1 Theoretical implications
Diffusion theory is one of the most researched fields in contemporary research. Nonetheless, due to the continuous development and advancement, such as the shift to modern diffusion theory, this field still offers many unexplored domains. Variations in the adoption process of organization offers an interesting area that requires more attention as there are a multitude of contextual factors influencing them. Scholars have made advancements regarding the influence individuals have on the adoption of innovations (Rogers, 1995; Weijnert, 2002). Ansari at al. (2010) provided a framework which allows to first quantify variations within the dimension‘s extensiveness and fidelity. This study combines both as it investigates the impact of key individuals, namely, innovation promoters (Witte, 1999) have on the extensiveness of an adopted innovative practice.
Surprisingly, the obtained results show a significant negative relationship between both authority and expert promoters and the extensiveness of an adopted practice. The findings enhance the existing literature as they show that key individuals possessing distinctive social capital, power and knowledge capabilities are unable overcome barriers of acceptance in a unilateral approach. The innovation promoter‘s involvement was not sufficient in order to increase the level of acceptance; leading to a more extensive adoption. In fact, their presence had a negative effect.
extensiveness remained (slightly) insignificant. From this thesis, it can be derived that even collective efforts cannot lead to a more extensive adoption of an innovative practice. Since the moderation effect of working groups on the authority promoter was only slightly insignificant, perhaps a larger sample size would answer this question more thoroughly. The findings enhance the existing modern innovation diffusion literature because variations in the adoption of innovations are influenced by key individuals, but in a negative way. Further, it shows that there are distinctive differences to what classic diffusion theorists such as Rogers (1995) pointed out regarding the involvement of such key individuals. Rogers (1995) predicted that the involvement of key individuals would enhance the understanding among the organization members and, in turn, lead to higher adoption rates of innovation. Damanpour (1991), for example, found positive relationships if a higher administrative intensity has been employed to promote the adoption of innovations. Alliances of different innovation promoters could yield interesting results such as outlined by Hausschildt & Kirchmann (2001) who suggest a dyadic promoter model to overcome organizational barriers. It includes not only one promoter but two or three promoters, each with different roles attached. There can be an authoritative promoter, a technology-versed promoter and additionally a process promoter who functions as project manager combining the traits and characteristics of the latter innovation promoter and facilitates their actions. Hence, there are a multitude of unexplored factors that still can be addressed in order find more thorough answers on how key individuals impact the extensiveness of innovation adoptions.
Concluding the above-mentioned arguments, the study enhances the understanding regarding the impact of key individuals in the form of authority and expert promoters on the extensiveness of an adopted practice. Utilizing the macro-lens perspective of the modern diffusing theory enabled the author to incorporate a multitude of influencing factors for a finely-detailed analysis that enhances the explanatory power of this study. However, this field of research still offers manifold of dimensions that allow for more detailed analyses utilizing the macro-lens perspective of modern diffusion theory. Future research directions address this point in detail.
5.2 Practical implications
The nature of this study and the utilized data provides several implications for practitioners. The gathered survey data represents, to the authors best knowledge, the first time that quantitative data regarding the spread and development of the BEV models has been gathered. The obtained data contains a multitude lot of contextual factors that helped in testing several relationships despite the relatively small observation size. The practical implications are, for the most part, directed towards the innovation and the government bodies that have developed the model of the BEV‘s.
First, significant negative results were found for hypothesis one and hypothesis two in assessing whether the influence of key individuals within the adopting villages impacted the extensiveness of the BEV models. These findings raise the question who the policymakers should address when engaging in diffusing affairs. On the other hand, these findings provide an indication about the dynamics of the adopting communities. The expected positive moderation utilizing the formation of working groups remained insignificant, although, a positive significant correlation can be obtained from the descriptive statistics for the authority promoter (r=0,31 at p<0.01). The correlation results for the expert promoter showed a negative but significant correlation (r=-0.26 at p<0.05). The analyzed effect for the moderation draws a similar picture because positive coefficients were obtained for the authority promoter (0.150) and negative coefficients for the expert promoter (-0.111). Again, it can be argued that a higher number of observations could have provided different results.
such villages are located from the showcase model in Juehnde, the lower the extensiveness of the adopted BEV models. Again, this finding raises several questions regarding the spread of the BEV model. Future research endeavors might want to explore what accounts for those findings. Strang & Soule (1998) pointed out that cultural differences account for differences in the adoption of innovations. A majority of the adopted BEV models in the database are located in the southern states Bavaria and Baden-Wuerttemberg (54/78). An average distance of 308,1 km between the villages in the south and the showcase model located in the north can be retrieved from the data. Therefore, future research on BEV models could benefit from obtaining data on the cultural contexts in which the adopting villages are located.
In summary, implications can be derived about the negative impact key individuals have as well as the negative impact of time and distance on the extensiveness of BEV models. The results could be of particular interest to the innovation mangers at the FNR in charge of promoting the BEV model to potential adopters.
6. Limitations and Future Research
6.1 LimitationsAlthough this study is based on a comprehensive data collection and respective data set, there are minor observable limitations. In turn, those limitations allow to make suggestions leading to a fruitful prospective research in the field of modern diffusion theory.
First, to enrich the present data set, additional observations from other villages which adopted the BEV model would help to raise the reliability and validity of the study. Due to time constraints, the data collection had to be limited to a certain period.
composition of local authorities to better assess the political environment in which the BEV models are adopted since they influence how innovations are adopted (Ahn, 2011; Berry & Berry, 1990; Schaefer Morabito, 2008).
Although the above-mentioned ideas are minor adjustments, they would help to portray the investigated phenomenon more thoroughly. A number of additional control variables could potentially increase the explained variance measured in R². The explained variance ranged from 29,7% in the first model containing only the control variables to 38,4% in the fourth model incorporating all predictor variables and the moderation.
6.2 Future Research
The framework by Ansari et al (2010) that has been utilized, in part to carry out the analysis, offers a multitude of potential domains for future research. First, this study has utilized one of the two proposed dimensions in the framework. The other dimension, fidelity, measures how similar an adopted practice is in comparison to its prior version. Future research could be conducted investigating how practice variations are affected and measurable in terms of the fidelity of an adopted practice. Here, it would be particularly interesting to obtain appropriate measurements to quantify fidelity. Furthermore, the broad scope of the two-dimensional framework allows it to be applied it to different contexts and industries.
Adding to this point, future studies would benefit from a data set containing more observations. An enriched data set would allow future researchers to perhaps test both of the dimensions of the framework utilized by Ansari et al. (2010). The study of Fiss et al. (2012) manages to incorporate both and found a significant positive impact of fidelity on the extensiveness of an adopted practice. Also, it would be interesting to observe how dimensions would interact as variables.
and test if the assumptions made by extant literature (Brandyberry, 2003; Damanpour, 2006) hold up through the macro lens of modern diffusion theory.
To conclude, there are several domains within modern diffusion theory that could be explored following the results of this study. The framework of Ansari et al (2010) offers an interesting setting that allows for test variations within the adoption of innovations with sensitivity to detail. In particular, in innovation intensive times (Arkolakis et al., 2018) knowledge about the antecedents that influence the adoption among organizations and individuals could help practitioners and innovation managers to streamline their efforts.
7. Conclusion
8. References
Ahn, M. J. (2011). Adoption of e-communication applications in US municipalities: The role of political environment, bureaucratic structure, and the nature of applications. The American Review of
Public Administration, 41(4), 428-452.
Arkolakis, C., Ramondo, N., Rodríguez-Clare, A., & Yeaple, S. (2018). Innovation and production in the global economy. American Economic Review, 108(8), 2128-73.
Ansari, S. M., Fiss, P. C., & Zajac, E. J. (2010). Made to fit: How practices vary as they diffuse.
Academy of Management Review, 35(1), 67–92.
Ansari, S., Reinecke, J., & Spaan, A. (2014). How are practices made to vary? Managing practice adaptation in a multinational corporation. Organization Studies, 35(9), 1313–1341.
Berry, F. S., & Berry, W. D. (1990). State lottery adoptions as policy innovations: An event history analysis. American political science review, 84(2), 395-415.
Brandyberry, A. A. (2003). Determinants of adoption for organisational innovations approaching saturation. European Journal of Innovation Management, 6(3), 150–158.
Coleman, J. S. (1988). Social capital in the creation of human capital. American journal of
sociology, 94, S95-S120.
Cook, R. D., & Weisberg, S. (1982). Residuals and influence in regression. New York: Chapman and Hall.
Curran, P. J., West, S. G., & Finch, J. F. (1996). The robustness of test statistics to nonnormality and specification error in confirmatory factor analysis. Psychological methods, 1(1), 16.
Damanpour, F. (1991). Organizational innovation: A meta-analysis of effects of determinants and moderators. Academy of management journal, 34(3), 555-590.
Damanpour, F. (1996). Organizational complexity and innovation: developing and testing multiple contingency models. Management science, 42(5), 693-716.
Damanpour, F., & Schneider, M. (2006). Phases of the adoption of innovation in organizations: effects of environment, organization and top managers 1. British journal of Management, 17(3), 215-236. Dearing, J. W. (2008). Evolution of diffusion and dissemination theory. Journal of Public Health
Management and Practice, 14(2), 99-108.
Daft, R. L. (1978). A dual-core model of organizational innovation. Academy of management
journal, 21(2), 193-210.
Federal Ministry for the Environment, Nature Conservation and Nuclear Safety. (2000). Act on
Granting Priority to Renewable Energy Sources (Renewable Energy Sources Act) [Ebook]. Berlin.
Retrieved from https://www.lexadin.nl/wlg/legis/nofr/eur/arch/ger/resact.pdf
Fiss, P. C., Kennedy, M. T., & Davis, G. F. (2012). How golden parachutes unfolded: Diffusion and variation of a controversial practice. Organization Science, 23(4), 1077-1099.
