Behavioral precursors in the innovation-decision process: the case of bioenergy in Ethiopia

1 Despite ample potential energy sources, most developing countries depend highly on fuelwood to 2 meet their energy needs, with repercussions on the environment and human health. Bioenergy 3 innovation is one way to combat this issue, the adoption rate of which remains low in many of them. 4 Using primary data collected from Ethiopian experts in the energy field, this study combines factor 5 analysis with ordered logit regression to identify the drivers of introduction and diffusion of bioenergy 6 innovations. Moreover, this study detects and analyzes the behavioral precursors of the respondents’ 7 intention to adopt brand new or upgraded bioenergy innovations. The results reveal differences 8 between their decision-making processes and suggest targeted research and policy strategies to boost 9 the adoption rate of bioenergy innovation. 10

This research did not receive any specific grant from funding agencies in the public, commercial, or not-34 for-profit sectors. 35 36 37 38 39 charcoal, crop residues, and animal dung) that accounts for 89% of the national total energy consumption 23 in 2010 [10,11]. As such, millions of women and children in rural areas devote their time collecting 24 fuelwood for domestic functions (e.g., food cooking and lighting [12]), while the urban poor spend a 25 sizable amount of their income on their daily energy needs [13]. Imported petroleum is an alternative 26 power source in Ethiopia, accounting for 7% of total energy use, while an important and growing source 27 is represented by the hydropower generation [14]. 28 limited evidence on the behavioral precursors that drive the adoption of novel, environmentally friendly 23 technologies [27,28,29]. Since the individual and situational diversity implies an array of behavioral 24 patterns [30, 31,32], when addressing the choice to adopt a new energy solution (especially in developing 25 societies), it is important to study the decision-making process by differentiating between categories of 26 adopters and between types of innovations [22]. Indeed, agents may have specific preferences when 27 coping with a brand new or an upgraded technology. This affects the aggregate rate of innovation 28 adoption, thus of energy access, in a society. According to the innovation diffusion theory [33], new 29 technology dissemination depicts an S-shaped curve where only a few adopters in the early stage invest 30 individuals contrast the real option with their personal expectation ("similarity judgments," [36]), thus 1 resorting to heuristics and incurring systematic biases [25]. In particular, evidence shows that when 2 dealing with a choice inherently associated with uncertainty and framed as a gain (such as the bioenergy 3 innovation-decision this study analyzes), people tend to display risk-averse behavior. This raises two 4 questions. First, to what extent can the contextual conditions influence this aversion and explain the 5 deviation of the agents' actual decision from the standard model? Second, which other behavioral 6 precursors (e.g., attitudes and abilities) contribute to the low adoption rates of cost-effective 7 technologies? To address these questions, recent studies have identified some behavioral factors, such as 8 social influence and individuals' awareness of environmental benefits [18,22,37,38] that systematically 9 affect agents' decision to adopt green technologies. 10 By referring to alternative behavioral models such as the Theory of Planned Behavior (TPB, [39]), this 11 study analyzes behavioral precursors that account for different levels of propensity for bioenergy 12 innovation ( Figure 1). The TPB is a socio-psychological model that is largely adopted in different fields 13 of behavioral analysis, such as environmental psychology [40,41,42] and innovation diffusion [43]. The 14 TPB does not assume decision-makers' rationality, but describes the human behavior as the result of a 15 structured process derived from a series of cognitive determinants (behavioral precursors). Unlike the 16 standard model that infers the decision-making process from observed behavior, the TPB analyzes the 17 process by directly assessing its constitutive elements. According to the TPB, an individual's decision is 18 a function of the intention to engage in the behavior, i.e., the motivation is the immediate antecedent of 19 the performable action and measures the interest in the option [22,30]. In turn, the individual's intentions 20 are assumed to depend on specific precursors: attitudes, subjective norms, and perceived behavioral 21 control; which are considered distant predictors of the behavior. Attitudes express beliefs and evaluations of positive or negative thoughts (i.e., knowledge) and feelings 2 (i.e., awareness and moral norms) about the possible consequences of performing the behavior. In this 3 study, attitudes are elicited by the knowledge and awareness of the expected outcomes of adopting a 4 bioenergy innovation. Specifically, these outcomes include the assessed profitability of the innovation 5 and the considered healthy and environmental benefits the technology can generate in terms of improved 6 individual and community's quality of life and reduced level of pollutant emissions. 7 Subjective norms (the second behavioral precursor affecting the individual's intentions) are determined 8 by the social customs and judgments on the considered behavior and its implications (descriptive and 9 injunctive norms, i.e., what the social referents such as customers and citizens do or approve, respectively 10 [24]). We derive the subjective norms from the respondents evaluation of what the others do (i.e., 11 imitation) or think (i.e., social acknowledgment and collaboration with customers as measure of their 12 opinion) about the bioenergy-oriented choice. 13 The third antecedent of the motivations is the perceived behavioral control that refers to the individual's 14 evaluation of the opportunities and challenges affecting the performance of the behavior. In this study, 15 the control factors concern both the decision-makers' skills and abilities to deploy and manage the new 16 technology as well as the external conditions (e.g., availability of feasible technologies and relational 17 resources) facilitating or interfering with the decision to adopt the innovation. Therefore, we measure the 18 accessibility to public financing, the capacity to design relevant organizational strategies, the availability 19 of solutions provided by the research, and the collaboration with foreign universities to study the 20 relationship between the respondents' perceived behavioral control and their intention to adopt the 21 bioenergy innovations. 22 This study derives a behavioral segmentation of Ethiopian experts based on their intention to introduce 23 alternative bioenergy innovations with different risk levels, and, according to the TPB, directly measures 24 the related behavioral precursors through surveyed evaluations of the main obstacles and drivers affecting 25 the decision. The survey is designed to test whether there is an asymmetry between the adopters' 26 decision-making processes as defined by the following research hypotheses: 27 • the intention to adopt a BNT is significantly affected by extrinsic (situational) conditions (i.e., the 28 perceived behavioral control); 29 • the intention to adopt an UBT is significantly affected by intrinsic (individual) factors (i.e., their 30 attitudes). 31 The first hypothesis is based on the assumption that adopters of a BNT are in general eager to try a new 1 solution or more likely to be open-minded, and possess abilities and skills [33] that enable them to exploit 2 the possible economic, environmental, and social benefits that sustainable technologies can provide. 3 Therefore, adopters' strong intention to use a brand new innovation is more likely affected by contextual 4 factors such as collaboration with research centers and access to cutting-edge bioenergy technologies. 5 By contrast, adopters of an UBT are assumed to react weakly to technological innovations; thus, their 6 intention to adopt a new bioenergy solution is expected to be affected by an inadequate knowledge and 7 awareness of the possible outcomes the performable behavior can produce.

Sampling and data collection 11
Purposive sampling technique was used to select and recruit the respondents among the local experts 3 12 active in the energy sector in Addis Ababa and Mekelle cities, Ethiopia. Addis Ababa is the capital of 13 Ethiopia and by far the largest city, while Mekelle is a Northern large city with flourishing bioenergy 14 sector. The two cities were chosen because they include various representative experts with direct and 15 grounded experience in the energy domain. Moreover, the sample was aimed to include entrepreneurs 16 who actively deal with innovation-centered decisions in the energy sector including entrepreneurs from 17 agriculture, processing industries, and energy services as well as private and public operators (e.g., 18 consultants and extension services), and policymakers. The experts were selected by local university 19 partners, contacted at their local address by enumerators, and invited to the local universities (Addis 20 Ababa University, and Mekelle University) to participate in the survey. 21 The primary data were collected using a pre-validated self-administered questionnaire submitted in 22 October, 2015, and in December, 2015, in Mekelle and Addis Ababa, respectively. Respondents 23 participated as representative of their organizations, were briefly introduced by the enumerators about 24 the questionnaire that even included questions specific to their organizations, and provided with 25 clarifications whenever they raise concerns. The questionnaire includes four sections, and it aims to 26 measure the respondents' evaluations about the different topics using an ordinal scale ranging from 1-9. 27 In the first section, experts are required to assess their level of interest in adopting the two types of 28 bioenergy innovations (i.e., BNT and UBT). The second and third sections focus on the respondents' 29 opinion on the obstacles and drivers affecting the introduction of bioenergy innovation (19 obstacles and 30 14 drivers: Table A.1 and A.2, respectively), while the fourth section deals with the main factors  1   motivating the diffusion of innovation (15 determinants: Table A.3). These last sections are designed to 2 elicit respondents' behavioral precursors associated with the adopters' intention to introduce the 3 bioenergy innovation. 4 A major limitation of the survey is the relatively small sample size due to the limited number of experts 5 in the energy field in Ethiopia despite the focus on the two leading areas of the country. Nonetheless, this 6 study provides specific information on the decision-making process concerning the adoption of new 7 bioenergy solutions and also offers relevant insights to researchers and policymakers regarding 8 orientation of or support for technological changes. A second limitation is the lack of information on the 9 socio-demographic characteristics (i.e., age, gender, ethnicity, and education) of the respondents. We 10 refrained from asking such detailed questions, as respondents would be less likely to participate in the 11 survey. Nevertheless, a few questions, such as respondents' sector or organization size, were included. 12 Unfortunately, the response rate was very poor and not sufficient to be reported in this study. However, 13 according to the TPB, these attributes "are considered background factors," affecting the individual 14 preferences and behavior "only indirectly," with their effect captured by the behavioral precursors this 15 study analyzes [39]. A third limitation is the possibility that respondents reveal high interest in both the 16 innovations (brand new and upgraded). For this particular class of respondents, it is challenging to 17 associate their subsequent responses (e.g., lack of knowledge) directly to BNT or UBT. In this study, this 18 was the case for a few respondents (8%) that were classified as BNT adopters. Finally, it was not feasible 19 to disentangle the respondents' personal opinion from the interest of their organization/community. 20

Data analysis 21
This study implements a two-phase data analysis using Stata/SE 15.0 to analyze the main determinants  [46,47]. Two tests are applied to check the 31 robustness of the developed FA models: Bartlett's test of sphericity that enables rejecting the hypothesis 1 that the variables are uncorrelated (1% of significance level), and the Kaiser-Meyer-Olkin (KMO) test 2 for sampling adequacy that measures the data suitability for the FA. In this study, the determinants (i.e., 3 obstacles and drivers) of the introduction and diffusion of bioenergy innovation are described by the 4 manifest variables ( " ) that FA groups into latent factors ( ! ), as in the following linear function: 5 where "! represents the factor load for each " , and " the error term. 7 After extracting the general factors ! affecting the possible evolution of the Ethiopian bioenergy sector, 8 this study develops two ordered logit regression models to scale down the analysis to the behavioral 9 precursors of the individual innovation-decision process. The related outcome variables are defined by 10 the respondents' intention to adopt a new bioenergy technology (i.e., BNT or UBT). In particular, three 11 possible degrees mirror their self-evaluated level of preference for the proposed two types of innovations. 12 If the respondent's intention to adopt the innovation is higher than the 75 th percentile (between the 75 th 13 and 50 th percentiles, or below the 50 th percentile), then the underlying motivation is assumed strongly 14 (moderately, or weakly) oriented toward that type of innovation. Afterward, the intention to deploy the 15 two types of innovations is regressed on explanatory variables ( " ) derived from the set of the The results achieved through the FA and the ordered logit models are based on the evaluations made by 2 a sample of 95 experts who completed the questionnaire. The respondents are local experts in the energy 3 field such as entrepreneurs (7 respondents), private and public consultants (64), and policymakers (12); 4 while the remaining subjects (12) belong to other professional profiles. About a half of the respondents 5 (51%) show a high or medium level of interest in adopting a BNT, whereas the equivalent share for the 6 UBT is about 64%. In general, traditional societies are more likely to have low interest in adopting 7 innovations. This generally weak propensity to adopt an innovation suggests an expected low acceptance 8 rate for new, sustainable bioenergy solutions in Ethiopia. This leads to the hypothesis that the potential 9 adopters may face numerous obstacles affecting their choice to deploy new technologies (e.g., limited 10 financial support, risk aversion, and lack of knowledge of the bioenergy domain) that are not 11 counterbalanced by adequate motivations or supportive conditions. This hypothesis finds confirmation 12 in the respondents' evaluation on the barriers to and drivers of the introduction of bioenergy innovation. 13 Table 1 shows that inadequate contributions from research and development (R&D), and lack of access 14 to information on bioenergy innovations are identified as the major obstacles to the innovation adoption. 15 Moreover, the lack of knowledge of environmental and public benefits, the limited access to public 16 financial facilities, the unavailability of skilled manpower, and risk aversion are the additional obstacles 17 the respondents recognize. Table 2 describes the drivers favoring the introduction of bioenergy 18 innovation. Accordingly, the increasing energy demand and the interest to reduce the GHGs emissions 19 stand out as the main fostering factors. Moreover, the respondents assign a high score to the contribution 20 the bioenergy technologies make to the environmental safeguard and to the quality of life. 21 22

10
This section aims at detecting and analyzing the behavioral precursors of the bioenergy-oriented 11 innovation-decision process. Firstly, from a general perspective the FA elicits the overall obstacles and 12 drivers associated with the introduction and diffusion of bioenergy innovations. Secondly, a distinction 13 between types of innovations and between adopters is made and specific regression models are developed 1 so at to identify the behavioral precursors underlying the intention to adopt a BNT and an UBT. The rotated factor matrix in Table 4 lists the factor loadings for the first FA model concerning the 4 assessed obstacles to the introduction of bioenergy innovation in Ethiopia, namely the lack of knowledge 5 and the (limited) financial facilities. Based on the modeled linear combination of the observed variables, 6 these two factors explain the 43% of the total variance of the respondents' evaluations of obstacles to 7 innovation adoption. 8 ü The first factor, knowledge and risk (F1.1, at 33%), reveals how much the respondents value the full 9 understanding of the innovation's effects in their decision-making. Limited access to information on 10 technological issues, and possible environmental and public benefits, as well as the gap between public 11 R&D and business' needs hinder the introduction of modern bioenergy solutions in the country. In 12 addition, the risk related to the new technology is moderately associated with F1.1. This prime obstacle 13 (the lack of knowledge of the innovation's opportunities, thus the lack of awareness of the implications 14 for the society) limits the strength of the behavioral beliefs (the capacity to link choice and its outcomes), 15 thus feeding (from a TPB perspective) unfavorable attitudes toward the decision to adopt the innovation. 16 ü The second factor, (limited) financial facilities (F1.2, at 9.6%), relates to the difficulties in obtaining 17 affordable capitals for investment purposes (i.e., limited access to private and public financing).  Factor loadings (i.e., measures of the relationship between the observed variable and the factor 2 F) with value > 0.75 (**), 0.75-0.5 (*), and 0.5-0.3 are considered "strong," "moderate," and 3 "weak" loadings, respectively.

5
The second FA model, based on the respondents' assessments of the innovation-decision drivers, 6 identifies two main factors that explain the 57.8% of the total variance: networking and environmental 7 concern (Table 5). 8 ü The first factor, networking (F2.1, at 49%), emerges as the major driver of innovation introduction in 9 Ethiopia emphasizing the necessity for potential adopters to establish collaborations with institutions and 10 other operators. Specifically, the results suggest that these interrelationships should be dual-goal oriented 11 and include collaborations with research centers and universities (to acquire knowledge in choosing and 12 deploying the new bioenergy solution), and various technical-support services provided by public and 13 private organizations (to develop skills and ability necessary to manage the innovation, while limiting 14 the inherent uncertainty). In addition to the collaboration with relevant stakeholders, the "economic 15 return" and "financial support to investments" variables also show a high correlation with F2.1.  Factor loadings (i.e., measures of the relationship between the observed variable and the factor F) with value > 0.75 2 (**), 0.75-0.5 (*), and 0.5-0.3 are considered "strong," "moderate," and "weak" loadings, respectively.

4
Regarding the main drivers of the diffusion of bioenergy innovation across the country, the third FA 5 model identifies two main factors (external conditions and social motivations) that explain 54.8% of the 6 total variance (Table 6). 7 ü The first factor, external conditions (at 45%), gathers a series of contextual variables that foster the 8 innovation propagation and is mainly attributable to public policies supporting the adopters' investment 9 choice (incentives and investments, F3.1). Together with these measures, a set of situational conditions 10 are identified as additional determinants of the innovation diffusion such as the availability of private 11 financing, accessibility to R&D findings, and professional skills. These elements (policy measures and 12 contextual conditions) enhance the innovators' capacity and limit the investment risks, making the 13 adopters' behavioral performance (perceived behavioral control) the crucial behavioral antecedent 1 affecting the innovation diffusion. 2 ü Moreover, FA identifies socio-economic motivations (F3.2, at 9.8%) as another driver of innovation 3 propagation. This factor links together environmental, economic, and social evaluations (from GHGs 4 reduction to imitation) that in the experts' opinion can motivate the entrepreneurs' decision to adopt the 5 bioenergy innovation, thus contributing to its diffusion. 6 7 Factor loadings (i.e., measures of the relationship between the observed variable and the factor F) with value > 0.75 (**), 10 0.75-0.5 (*), and 0.5-0.3 are considered "strong," "moderate," and "weak" loadings, respectively.

12
The results of the three FA models detect different behavioral precursors influencing the innovation-1 decision process. On the one hand, the weak individual attitude towards new bioenergy solutions (caused 2 by the lack of knowledge, thus of awareness of the consequences that the choice can generate) negatively 3 affects the motivations to adopt the innovation. On the other hand, the adopters' perceived behavioral 4 control proves to be the major behavioral driver of innovation introduction and diffusion. This ability to 5 perform the behavior is recognized not just as an individual quality the adopter innately possesses, but 6 also as a resource that strongly depends on two different contextual conditions. With reference to 7 innovation introduction, the individual capacity to deal with new solutions stems from the collaboration 8 with institutions and other operators. Regarding the innovation diffusion, the adopters' perceived 9 behavioral control relies on targeted supporting policy measures. The emerging difference between these 10 two phases of the innovation adoption path stresses the opportunity to further investigate the behavioral 11 precursors that characterize the decision to adopt a BNT or an UBT, separately. 12 13 4.1.2 Behavioral precursors of the intention to adopt a BNT and an UBT: regression results 14 The main variables that challenge and/or drive the adoption of the two types of bioenergy innovations 15 are identified by developing two distinct ordered logit models, and analyzed from a behavioral 16 perspective. According to the assumed research hypotheses, the results of this study confirm that the 17 intention to adopt a BNT is mainly and significantly correlated with extrinsic conditions (the perceived 18 behavioral control and subjective norms), whereas the intention to adopt an UBT is mainly and 19 significantly correlated with intrinsic factors such as the individual's attitude toward new technological 20 solutions and their outcomes. Moreover, the results also suggest that more complex interactions between 21 specific behavioral precursors characterize and further differentiate the two innovation-decision 22 processes. For the sake of completeness, the results include both the odds ratios and the regression 23 coefficients. Throughout this study, the odds ratio compares the probability of high intention versus the 24 combined middle and low intention to adopt the considered innovation. 25

Intention to adopt a BNT 26
Based on the results of the first ordered logit model, the intention to adopt a BNT is regressed against a 27 series of contextual determinants (Table 7). Specifically, the related odds ratios (column 3) show that 28 the probability of a high level of intention to adopt a BNT increases as the availability of R&D 29 advancements improves, the potential of reduction of GHGs emissions increases, and the opportunities 30 of establishing a collaboration with the consumers become concrete. Therefore, three main determinants 31 motivating the innovation-oriented behavioral performance are identified. First, the contribution that a 1 BNT can offer to the environmental quality is significantly and positively associated with the favorable 2 attitude to adopt it 4 . Second, the access to cutting-edge technologies (perceived behavioral control) is a 3 reliable factor directly linked to the motivation to introduce a BNT. Third, the direct relationship with 4 the closer stakeholders (i.e., the customers: subjective norm) can further contribute to orienting the 5 decision toward a BNT-centered investment. 6 On the contrary, the social acknowledgement (i.e., the overall approval or disapproval of the society for 7 an innovative solution: subjective norm) is significantly but negatively associated with the intention to 8 adopt a BNT. Accordingly, the odds ratio indicates a link between the social rejection of new 9 technologies and the innovators' propensity to introduce a BNT. This antagonistic behavioral precursor 10 reveals a gap between the mainstream idea of energy access and use in the Ethiopian society (focused 11 on providing/gaining access to conventional, traditional sources, thus on a general lack of knowledge of 12 modern, sustainable energy opportunities) and the innovators' open orientation toward the bioenergy-13 centered innovations. 14 15 2 Column (1) shows the coefficients of the ordered logit estimation; (***), (**), and (*) indicate statistical significance at the 1%, 5%, and 3 10% level, respectively. Column (2) shows the associated standard errors. Column (3) shows the odds ratio. Column (4) describes the p-4 values of the estimated coefficients. Column (5) and (6) show the 95% lower and the upper confidence intervals respectively. Column (7) 5 associates each significant explanatory variable with a behavioral precursor.

Intention to adopt an UBT 8
The results of the second regression model reveal a specific and composite set of significant variables 9 and of related behavioral precursors that explain the intention to introduce an UBT (Table 8). A first 10 group of variables concerns the outcomes the adoption of an UBT is expected or not to produce. On the 11 one hand, the UBT contribution to quality of life shows a positive correlation with the propensity for its 12 deployment and the odds ratio suggests that the probability of this decision increases as the envisaged 13 effect is valued. On the other hand, the lack of knowledge of public benefits displays a negative 14 correlation with the motivation to adopt an UBT as the odds ratio proves (the higher the unawareness of 15 the positive externalities generated by the innovation, the lower the probability of a high level of 16 intention to adopt UBT). Moreover, the reduction of GHGs emissions results negatively associated with 17 the preference for the UBT and the related odds ratio indicates that as the individuals' concern for the 18 climate change increases, their intention to adopt an UBT decreases. This first group of explanatory 19 variables describing the evaluation of the effects that an UBT can generate at individual level (quality 20 of life) or miss at societal level (public benefits and reduction of emissions), respectively, highlights the 21 role that the favorable/unfavorable attitudes play as behavioral precursors in this innovation-decision 22

process. 23
A second group of variables significantly associated with the UBT-oriented decision involves the 24 relationships with other stakeholders. Specifically, the collaboration with foreign universities as well as 25 the collaboration with customers show a positive significant correlation with the intention to adopt an 26 UBT. Coherently, the related odds ratios indicate that the probability of a high level of this intention to 27 innovate increases as the synergies with the academic world and the sympathy with the economic 28 referents (i.e., market-oriented considerations) improve. These two variables focused on the 29 collaborations the innovators can establish suggest that the intention to adopt an UBT is further 30 associated with the precursors behavioral control and subjective norms, respectively. 31 1 Note: Brant test of parallel regression assumption: chi square= 9.14; P-Value=0.519.The dependent variable is a categorical variable with 3 three levels that describes the intention to adopt an upgraded bioenergy innovation. All the independent variables are considered as 4 continuous variables. Column (1) shows the coefficients of the ordered logit estimation; (***), (**), and (*) indicate statistical significance 5 at the 1%, 5%, and 10% level, respectively. Column (2) shows the associated standard errors. Column (3) shows the odds ratio. Column (4) 6 describes the p-values of the estimated coefficients. Column (5) and (6) show the 95% lower and the upper confidence intervals .Column 7 7 associates each significant explanatory variable with a behavioral precursor.

9
As alternative estimation technique aimed at testing the robustness of the two developed ordered logit 10 models, this study implements two additional logit models 5 . The first logit model confirms that R&D, 11 reduction of GHG emissions, and collaboration with customers are significantly and positively correlated 12 with the adoption of a BNT in line with the main findings, with the only exception for social 13 acknowledgement that results not significant (Appendix B: Table B.1). Similarly, the second logit model 14 shows estimations consistent with the main obtained results except for the variables lack of knowledge 1 of public benefits and collaboration with foreign universities, which are not significantly correlated with 2 the intention to adopt an UBT (Appendix B: Table B.2). In general, while only few variables do not 3 emerge as explanatory regressors in the logit models, the robustness check validates the main significant 4 results achieved through the ordered logit models that prove to be comparatively more performing in 5 fitting the observations. 6 7

Discussion and Conclusions 8
This study relies on data collected from local experts belonging to the energy or related sectors in two 9 areas of Ethiopia, and implements a two-step approach to investigate their intention to adopt alternative 10 bioenergy innovations. First, the FA models detect from a general perspective the overall factors 11 affecting the introduction and diffusion of new bioenergy technologies. Second, we separately look at 12 the decision-making processes guiding the introduction of two different types of bioenergy innovations: 13 specifically, the ordered logit models identify the main behavioral precursors of the individuals' 14 motivations to adopt brand new and upgraded bioenergy innovations. 15 Three main orders of findings are achieved through the FA. First, the lack of knowledge stands out as 16 the major factor explaining the total variance of the respondents' evaluation of obstacles in introducing 17 bioenergy innovation. From a TPB perspective, the lack of knowledge of the technological innovation 18 and its opportunities feeds unfavorable attitudes towards the decision to adopt the innovation. Second, 19 the results indicate that networking is the most important driving factor of bioenergy innovation 20 introduction in Ethiopia. The two conditions that networking embodies-R&D and collaboration with 21 public bodies-reveal the attention that the adopters pay to the operational issues the innovation 22 introduction implies. Thus, the current capacity to deal with the innovation adoption (i.e., the perceived 23 behavioral control) emerges as the decisive behavioral precursor of the related decision-making 24 process. Third, regarding the experts' evaluation of the main drivers favoring the diffusion of the 25 bioenergy innovation in Ethiopia, a set of situational variables are identified such as the availability of 26 private financing and public supports, the accessibility to R&D findings, and the presence of adequate 27 professional skills. These elements (expressed by the factor "external conditions") are expected to 28 enhance the innovators' capacity and limit the investment risks, and confirm the crucial role that the 29 perceived behavioral control plays as behavioral antecedent of the decisions enabling the innovation 30

diffusion. 31
As per the distinction between the two types of bioenergy innovations, the regression results show that 1 the behavioral antecedents associated with the individuals' intention to adopt a BNT and an UBT let 2 emerge differences in the related innovators' decision-making processes. On the one hand, general 3 contextual conditions matter to the adoption of a BNT. Specifically, the innovators' propensity toward a 4 BNT is linked to the availability of cutting edge technologies and to the expected reduction of global 5 pollutants emission. On the other hand, more specific contextual conditions as well as idiosyncrasies are 6 crucial to the intention to adopt an UBT. In fact, the motivations generating this decision are positively 7 correlated to the collaboration with the customers and to the outcomes achievable at small scale level 8 the service providers and consultants for the technical assistance, the public institutions for their role 30 in shaping favorable external conditions, and the other enterprises for creating synergies and sharing 31 risks. Networking is the main driver that can heighten the behavioral performances in the bioenergy 1 innovation realm. The adopters' need to set up innovation-centered interrelationships calls for 2 university and public policies that include the creation/enhancement of targeted structures (e.g., 3 extension services and new decision-making bodies together with producers' associations) and the 4 implementation of tailored tools (e.g., smart systems and social events); 5 ü the adopters' innovation-decision processes reveal different behavioral patterns in function of the 6 technological characteristics of the bioenergy solution taken into consideration. Prospective research 7 and policy strategies aimed at supporting the adoption of BNTs should consider the relevant 8 underlying behavioral precursors focusing on R&D efforts, bridging the gap between research and 9 business, and giving priority to the environmentally friendly solutions. Differently, strategies oriented 10 toward the introduction of UBT-centered innovations should aim at building the adopters' abilities 11 and capacity to deploy and manage the innovative technologies, ensuring their operability and 12 scalability, and increasing the knowledge of the social benefits these solutions can generate. 13 One has to be cautious when interpreting these results because of the following limitations. This study 14 used a relatively small sample size, and it assumes that the role of socio-demographic characteristics is 15 captured (as "background factor") by the behavioral precursors. Therefore, it is a viable avenue for future 16 research to adopt large sample size, and explicitly measure the socio-demographic effects on the 17 bioenergy innovation adoption decision. Moreover, this study associates the identified variables with the 18 TPB behavioral precursors. However, there is a need for further research to directly investigate these 19 behavioral precursors through other appropriate approaches and methodologies (e.g., behavioral 20 economics experiments).         Appendix A-Description of variables 1 2 Difficulties to find qualified staffs in the local market to develop products or assist activities in bioenergy. Competition with food The potential risk related to cultivating land for biomass instead of crops.
Low benefit/cost ratio Lower net benefit/economic return from bioenergy investment.
Risk due to technology Potential risk related to lack of knowledge of the technology to generate energy from biomass. Risk due to market conditions Risk perception related with local demand for bioenergy product and or producers competition. Limited access to private financing Difficulty to get financial support from private financial sectors to invest on bioenergy.
Limited access to public financing Difficulty to get financial support from public financial sectors to invest on bioenergy.
High fiscal burden High tax rate.

Lack of information on bioenergy innovations
Imperfect information/knowledge on new/upgraded bioenergy innovations.

Lack of knowledge of environmental benefits
Imperfect knowledge on environmental benefit derives from bioenergy innovations.

Lack of knowledge of public benefits
Imperfect knowledge on the public benefits derives from modern bioenergy innovation. for example improvement of society's living standard. R&D not addressing the business' needs Research and development activities not addressing the needs of the enterprises. 2

Variables Description
Growing of energy demand An increasing of energy demand.
Entrepreneurs' imitative behavior or willingness to change Behavior of entrepreneurs (availability to change, willingness to change, imitation).
Human resources(skills) Availability of skilled man power.
Contribution to quality of life An interest to improve wellness of the local community.
Contribution to environmental quality An intention to improve environmental quality.

Reduction of GHGs emissions
An interest to reduce emission from traditional energy source.
Social acknowledgment An interest to obtain social acknowledgment as the result of diffusing the technology.
Social responsibility a responsibility to improve social well-being.
Organizational strategies Clearly defined vision/strategies, established norms for innovation promotion.

R&D
Availability of research and development activities to solve problems related with the innovation diffusion. Social norms and local partners Availability of social capital and local partnership.
Social norms and foreign partners Existence of social capital and foreign partners.
Policy incentives (subsidies, fiscal deductions) Incentives such as subsidy and fiscal deduction.

Public investments (infrastructures)
Availability of infrastructure such as road, telecommunication.
Private investments Availability of private investors in the bioenergy sector.
Credit availability Availability of financial facilities.

4
Appendix B-Robustness analysis 1 3 Note: the dependent variable describes the intention to adopt a BNT. It is a binary variable (0,1) with value 1 when the 4 respondent's intention to adopt the innovation is greater than the medium value. All the independent variables are considered as 5 continues variables. Column (1) shows the coefficients of the logit estimation; (***), (**), and (*) indicate statistical significance 6 at the 1%, 5%, and 10% level, respectively. Column (2) shows the associated standard errors. Column (3) defines the p-values of 7 the estimated coefficients. Column (4) and (5) show the 95% lower and the upper confidence intervals respectively. Column 6 8 shows the behavioral precursors category of statistically significant variables.  3 Note: the dependent variable describes the intention to adopt an UBT. It is a binary variable (0,1) with value 1 when the 4 respondent's intention to adopt the innovation is greater than the medium value. All the independent variables are considered as 5 continues variables. Column (1) shows the coefficients of the logit estimation; (***), (**), and (*) indicate statistical significance 6 at the 1%, 5%, and 10% level, respectively. Column (2) shows the associated standard errors. Column (3) defines the p-values of 7 the estimated coefficients. Column (4) and (5) show the 95% lower and the upper confidence intervals respectively. Column 6 8 shows the behavioral precursors category of statistically significant variables.