‘Drivers of environmental innovation performance of SMEs: the role of direct
experiential learning and observational learning’
Master Thesis Strategic Innovation Management & Marketing Management
Marlijn Botter
S2724561
Supervisor: Dr. Thijs Broekhuizen
Co-assessor: Prof. dr. Jenny van Doorn
17-01-2020
2
Abstract
Firms’ interest in sustainable business solutions has sparked research on what drives firms to develop environmental innovations. To date, literature has mainly focused on the role of regulations and internal firm characteristics and motivations, but has ignored the role of learning and observing from the firm’s direct environment. Drawing on the social learning theory, this study investigates the role of learning through direct experience and learning by observation in explaining SMEs’ environmental innovation performance. By using datasets containing information on SMEs’ environmental innovation performance and innovation characteristics as well as the local network’s environmental and innovation performance, this research examines the effect of direct experience (active network participation in innovation networks) and the effect of observation of (eco-) innovative firms in the geographical network of the focal SME. This study finds that direct learning through active network participation in formal innovation networks does not influence the environmental innovation
performance of SMEs, however direct learning through collaborating with a diverse set of partners does positively influence the environmental innovation performance. The role of observational learning is dynamic. While, having product innovative firms and many eco-innovative firms in close proximity to the SME reduces the next year’s environmental innovative performance of the SME, it finds a weak positive two-year lagged effect of the local network’s environmental innovation performance on the environmental innovation performance of the focal SME, indicating that observational learning might play a role in explaining an SME’s environmental innovation performance in the long term.
3
Table of content
1. Introduction ... 4
2. Literature review and hypothesis development ... 6
2.1 Conceptual model ... 6
2.2 Eco-innovations ... 6
2.3 Social learning theory ... 7
2.3.1 Active network participation ... 8
2.3.2 Observing and imitating surrounding network ... 8
3. Methodology... 10 3.1 Data collection ... 10 3.2 Measurements ... 10 3.2.1 Dependent variable ... 10 3.2.2. Independent variables ... 10 3.2.3 Control variables ... 12 3.3 Data analysis... 13 4. Results ... 13
4.1 Descriptive statistics & correlations ... 13
4.2 Regression results ... 18
4.3 Robustness checks and additional tests ... 21
5.Conclusion and discussion ... 22
5.1 Conclusions ... 22
5.2 Theoretical implications ... 23
5.3 Managerial implications ... 24
5.4 Limitations and further research ... 25
References ... 26
4
1. Introduction
In a time where the concentration of C02 in the atmosphere reaches an all-time high and natural resources are depleting, the need for ecological innovations is rising rapidly. In response to external pressures from the market, stakeholders, or regulations, firms are increasingly introducing
eco-innovations (Rennings, 2000). Several global climate change agreements like Europe 2020 or the Paris agreement of 2015 aim to make production and consumption more sustainable on a large scale. To contribute to reaching the challenging climate objectives, organisations are incentivised to introduce innovative products and processes that reduce firms’ ecological footprint via so called
eco-innovations. While a focus on large firms makes sense given their high overall environmental impact, the creation of a long-term sustainable eco-system cannot ignore the importance of small- and medium sized companies (SMEs) in contributing to a more sustainable society (Jamali, Zanhour & Keshishian, 2009).
Eco-innovations are defined as ‘the production, application or exploitation of a good, service, production process, organisational structure, or management or business method that is novel to the firm or user and which results, throughout its life cycle, in a reduction of environmental risk, pollution and the negative impacts of resources use (including energy use) compared to relevant alternatives.’ (Kemp & Foxon 2007, p.4). According to Klewitz, Zeyen & Hansen (2012) eco-innovations could be advantageous for SMEs as they can result in cost reductions or a strategic advantage over competitors. However, there are several barriers for SMEs to properly develop eco-innovations. First of all, SMEs experience a greater resource scarcity than larger firms (Lee, 2009). Furthermore, SMEs have different ways of capturing value from innovations (Spithoven, Vanhaverbeke, & Roijakkers, 2012) and might have different motivations to engage in eco-innovations (Andries & Stephan, 2019). Therefore, it is important to further investigate what drives SMEs to invest in the development of eco-innovations and enhance their environmental innovation performance.
5
learning) will be investigated as possible drivers of environmental innovation performance of SMEs. To investigate to what degree SMEs acquire the required knowledge via these two ways of learning and improve their environmental innovation performance, the following research question was formulated:
‘What are the effects of learning through direct experience and by observational learning on the environmental innovation performance of SMEs?’
The research question will be answered by using survey data from an annual survey that is held by the SNN: the Innovation Monitor Northern Netherlands. The survey data spans four years (2016-2019) which allows to test for the direct and lagged effects of the innovation efforts of the focal firm’s network of surrounding firms on the environmental innovation performance of the focal firm.
The research contributes to the existing literature on SME eco-innovations by adopting a broader view of the possible factors that can influence environmental innovation performance. Instead of taking a top-down approach and looking at the effect of regulations and policies, a bottom-up perspective is central to this research by looking at the different ways of learning that are in line with the social learning theory. By looking at the effect of direct experience of active network participation of a firm and the observational learning from the firm’s surrounding network of firms a new light is shed on what enables a high environmental innovation performance for SMEs. Moreover, by taking into account different aspects of innovation when investigating the observational learning aspect instead of just looking at the environmental performance of the firms surrounding the focal firm, a more complete picture is painted to explain the environmental innovation performance of SMEs.
Managerial implications of this research provides new insights for SMEs in improving their environmental innovation performance. Furthermore, it can help policy makers gain insights on the effectiveness of measures aimed at stimulating eco-innovation at SMEs.
6
2. Literature review and hypothesis development
2.1 Conceptual model
This research investigates the effect of the two different learning types as proposed by the social learning theory, direct experience (H1) and observational learning (H2) on the environmental innovation performance of SMEs. Previous literature has emphasised the positive effects of knowledge sharing and creation through networks (Doloreux, 2004; Peters, Johnston, Pressey & Kendrick, 2010), therefore it is expected that direct experience gathered by the focal firm through active network participation in innovation networks has a positive effect on its environmental innovation performance. Furthermore, based on the work by Audretsch & Feldman (2004), it is expected that observational learning based on the innovative characteristics and activity of firms in close proximity to the focal firm has a positive effect on the environmental innovation performance of the focal firm.
2.1 Eco-innovations
2.2 Eco-innovations
Eco-innovations pertain to a large variety of innovations (Horbach, Oltra & Belin, 2013), because every process, product, or organizational change that leads to a reduction in pollution or use of
resources, whether it is intentional or not, can be labelled an eco-innovation (Kemp & Pearson, 2008). Research on innovations often stresses the role of regulation as one of the major drivers of eco-innovation in companies (Del Río, Romero-Jordan & Peñasco, 2014; Horbach et al., 2008). Although this tends to be true for larger firms, regulation does not seem to be a driver for SMEs to develop
7
environmental innovations (Hansen, Sondergaard & Meredith, 2002; Andries & Stephan, 2019). Instead, Andries & Stephan (2019) find that consumer demand and a proactive, voluntary attitude from an SME are better determinants in explaining the development of eco-innovations. A qualitative paper by Klewitz et al. (2012) identifies cost savings and risk management as the main drivers of environmental innovations in SMEs. The presence of eco-innovations is thought to generate a so-called double knowledge spill-over, as the knowledge is easy to reproduce and hard to appropriate by the innovating firm (Rennings, 2000). This may demotivate SMEs to invest in environmental innovations. Nonetheless, Mazzanti & Zoboli (2006) find that one of the drivers of eco-innovations is firm participation in networks and groups. This research focuses on how SMEs can learn from its local network.
2.3 Social learning theory
Social learning theory was introduced by Bandura (1971) as a reaction to earlier behavioural theories that mainly focused on intrinsic factors as determinants of observed behaviour. In contrast, the social learning theory posits that behaviour is often the result of outside stimuli. Bandura (1971) proposes two types of learning: learning by direct experience or learning by observation. The first type, direct experience, is often the result of the immediate consequences of an action which are either positive or negative. Once an individual or firm gains experience and learns which behaviour results in a
satisfactory outcome, it will select this type of behaviour in similar situations in the future. The gained experience of behavioural effects in different type of situations also enables individuals or
organisations to predict what outcome is expected in completely new circumstances based on a particular type of behaviour that is exerted. However, this trial-and-error approach of learning might be inefficient in some situations. Therefore, most learning is done based on the second type of
learning, which is observational or modelling learning. By imitating the observed behaviour of others, it lowers the chance of choosing the ‘wrong’ type of behaviour and subsequently, an unwanted outcome.
8
2.3.1 Active network participation
Innovation is increasingly seen as a result of reciprocal interactions among actors in a network (Doloreux, 2004). Networks are seen as a place where informational resources are exchanged from which firms can obtain valuable knowledge that can strengthen their innovative capabilities (Peters et al., 2010). Innovation performance of firms that participate in a network depends on the effectiveness of the knowledge transfer system within a network (Kang &Kang, 2009; Von Hippel, 1986). Firms can join formal network forms (joint ventures, strategic alliances or R&D collaborations) that are based on a contractual agreement or join informal networks that include collaborations between different actors like customers and suppliers that are not based on formal contracts (Van Aken & Weggeman, 2000). Those informal networks are perceived to be beneficial to firms in the form of knowledge sharing and creation, especially in the beginning stages of the development of radical innovations (Laursen & Salter, 2006)
Bougrain & Haudeville (2002) argue that informal networks are particularly effective for SMEs. First of all, informal networks provide SMEs with knowledge on, such as, technological change, competitor decisions and market demands. Secondly, the transfer of tacit knowledge is made easier because of network through personal interactions. Lastly, the knowledge gained from networks could substitute internal knowledge creation which might be very costly in terms of irreversible investments. Because resource scarcity is an important barrier to the successful implementation of (eco-) innovations (Klewitz et al., 2012; Lee, 2009), this could be overcome by SMEs in actively participating in innovation networks. In those networks SMEs can actively experiment, share knowledge resources with other SMEs, and create new knowledge needed for eco-innovations. Furthermore, Wagner (2007) concludes that cooperating with environmentally concerned stakeholders has a positive effect on successful eco-innovation. This leads to the following hypothesis regarding learning via direct experience:
H1: Active innovation network participation has a positive effect on a firm’s environmental innovation performance
2.3.2 Observing and imitating surrounding network
Looking at innovation on a regional level, past research uncovered that inter-networking processes and the ease and timeliness of spreading knowledge among the different actors are considerable
9
show that knowledge can also spill between geographically close companies due to job switching of employees (Almeida & Kogut, 1999).
Innovation diffusion literature has investigated the effects of observing and imitating individuals in your social environment regarding the adoption of new products or services (Rogers, 2003). The likelihood of consumers to adopt new consumer products can be partially explained by the influence of previous adopters via observational learning and word-of-mouth (WOM) messages (Chen, Wang & Xie, 2011). A large body of research exists that investigates the influence of WOM and distinguishes between the effects of volume, which is the quantity of messages received by the consumer (Chen, Wu & Yoon, 2004) and valence, which reflects how positive or negative a particular message is (Lee, Rodgers & Kim, 2009). Various studies have found a positive effect of WOM volume on product sales (Chen et al., 2004; Liu, 2006), however, the influence of WOM valence remains unclear, with some studies reporting a positive influence while others report a negative influence (Yang, Kim, Amblee & Jeong, 2012). Observational learning differs from WOM in two ways (Chen et al., 2011). First of all, it is less informative than WOM, it only reveals the actions of the consumer and not the reasoning behind it. Nonetheless, the credibility of observational learning might be higher compared to WOM as it is based on actions rather than words. Applying this logic to SMEs implementation of eco-innovations, this study seeks to understand whether firms that imitate or model their innovations from neighbouring firms do so mainly when a lot of firms implement eco-innovations (i.e. high volume), or that the focal firm imitates more strongly when the neighbouring firms score high on environmental innovation performance (i.e. positive valence). Furthermore, as the definition given of eco-innovations earlier in this literature section already suggests, the concept of what constitutes an environmental innovation is fairly broad. Firms might not only observe eco-innovations from neighbouring firms and imitate them, the ‘regular’ innovation activities of firms in close proximity can also influence the environmental innovation performance of SMEs.
This study applies marketing theory to this innovation context, and hypothesizes that both constructs (volume and valence) positively influence the environmental innovation performance of the focal SME. Moreover, this study hypothesises that a higher general innovation performance of
neighbouring firms positively influences the environmental innovation performance of SMEs. This leads to the following hypotheses regarding observational learning:
10
3. Methodology
3.1 Data collection
This research studies firms located in the Northern part of the Netherlands that have participated in the Innovation Monitor Northern Netherlands in the period between 2016-2019. This innovation
benchmark has been developed by SNN and the University of Groningen and is sent by email on an annual basis to SMEs that are located in the provinces of Friesland, Groningen and Drenthe. The survey includes, among others, questions on how innovation is organised, the firms’ financial situation and subsidies, and firms’ collaborating activities. Because of the regional spread of the survey in it provides a unique opportunity to study firms that are located closely together and observe if SMEs influence each other in terms of environmental innovation performance. As the data spans multiple years, it also allows for the investigation of lagged effects of learning through observation or imitation.
Information about the geographic location is limited to the participating companies’ zip code for the years 2016, 2017 and 2019. For the year 2018, the zip codes were retrieved and manually added based on the participating companies’ names with help of Orbis.
The questions included in the survey slightly differ from year-to-year. Considering that data from multiple consecutive years is available, the influence of the surrounding network from the year
t-1 on the environmental innovation performance of the focal firm in year t will be tested.
3.2 Measurements
3.2.1 Dependent variable
Environmental innovation performance: This measure is tested using six items from the Innovation
Monitor. Respondents indicated whether they have introduced a product, process or organizational innovation in the past 2 years, resulting in (1) lower material usage per unit, (2) less energy usage per unit, (3) smaller CO2 footprint, (4) substitution of materials by less toxic or dangerous materials, (5) less pollution of surface, water, air of noise and (6) recycling of waste, water and materials. A factor analysis was performed to ensure that all six aspects can be combined into one variable representing environmental innovation performance, which led to a single solution. Each question can be answered with ‘yes’ or ‘no’ from which 6 dummy variables were created that can be transformed into one overall score with a minimum value of zero (every item answered with ‘no’) and a maximum value of six (every item answered with ‘yes’).
3.2.2. Independent variables
Network participation: The Innovation Monitor contains three questions related to network
11
with incubators or innovation accelerators by selecting the particular networks, test surroundings, accelerators and incubators. For each of the three questions, a variable was created from which the number represented the number of selected networks, test surroundings, accelerators or incubators. After standardizing the three variables, a factor analysis was conducted which confirmed the unidimensional nature of the three variables, such that they could be combined into one composite variable representing active network participation. For the years of 2018 and 2019, the same three questions regarding network participation have been asked; however, participants only had to answer with ‘yes’ or ‘no’ instead of indicating the specific network, test surrounding, accelerator or incubator. For both years, the three dummy variables with 0 representing ‘no’ and 1 representing ‘yes’ were combined into one composite variable representing network participation after conducting a factor analysis.
Cluster environmental innovation performance (CEIPval): Dutch zip codes range from 1000 to 9999
and are hierarchic. Defining regions for firms based on zip codes can be done by looking at the first number of the zip code of every firm, however, this still represents quite a large region. If the first two, three or respectively first four numbers of the zipcode are similar the regions become smaller with each similar, additional number. Following the methodology of Beugelsdijk & Cornet (2002), the first two numbers of every zip code are used to define the geographic clusters to ensure a region size that includes a sufficient number of firms within one cluster. This leads to average cluster sizes of 19 (2016), 31 (2017) and 32 (2018). By grouping clusters of firms based on the first two digits of their zip codes for the years 2016, 2017 and 2018 an average score of a cluster’s environmental innovation performance could be computed. This measure represents the valence of the influence on
environmental innovation adoption.
Cluster size of SMEs with environmental innovations (CEIPvol): The clusters are calculated the same
way as the CEIPval variable. This number represents the amount of firms in a cluster that score at least 1 on the environmental innovation performance measure. This measure represents the volume of the influence on environmental innovation adoption.
Cluster’s product innovation (Cprod): The Innovation Monitor contains two questions regarding
product innovations which were the same in all three years. The first one asks participants to indicate whether they have introduced new or strongly improved products and the second one asks if
participants have introduced new or strongly improved services. For both questions two dummy variables were created with 0 representing ‘no’ and 1 representing ‘yes’. The variables were combined into one composite score with a minimum value of 0 (both questions answered with ‘no’) and a maximum score of 2 (both questions answered with ‘yes’).
Cluster’s process innovation (Cproc): The Innovation Monitor contains three questions regarding
12
firm has introduced new or strongly improved production methods, the second one asks if the firm has introduced new or strongly improved logistics, delivery or distribution methods and the last one asks participants to indicate whether the firm has introduced new or strongly improved supporting activities for its processes e.g. administrative systems. For all three questions dummy variables were created with 0 representing ‘no’ and 1 representing yes’. The variables were combined into one composite score with a minimum value of 0 (both questions answered with ‘no’) and a maximum score of 3 (all three questions answered with ‘yes’).
Cluster’s radical innovation (Crad): This variable is measured by the amount of revenue generated
from introduced products and services that were new to the market the firm is operating in, in percentages of the total revenue that is generated by the firm.
3.2.3 Control variables
Firm size: As the size of a firm can influence a firm’s need for innovation (Damanpour, 1992) and
ability of the firm to innovate (Laforet, 2008), this study will control for firm size. Specifically, for the topic of environmental innovations, Andries & Stephan (2019) find that larger firms have different drivers than smaller firms for developing environmental innovations. This construct will be measured by the number of employees working in the firm.
Firm age: The age of the firm can have a significant impact on its innovativeness and responsiveness
to its environment. Sørensen & Stuart (2000) find that older firms generate more innovations. Consequently, this research will control for the firm age, as measured in number of years.
Green subsidy: Research suggests there is a positive relationship between the introduction of
environmental innovations and the availability of green subsidies (Yalabik & Fairchild, 2011). This study will control for the reception of the VIA-subsidy (Accelerator Innovation Ambitions). This subsidy is aimed at SMEs in the provinces of Friesland, Drenthe and Groningen that wish to contribute to the reduction of C02 emissions. A dummy variable is created with 1 representing firms that did apply for a VIA-subsidy and 0 representing the firms that neither applied for the subsidy nor had ever heard of the subsidy before.
Intellectual Property Protection (IP-protection): Literature states that IP-protection mechanisms can
influence the innovation rate and performance of the firm (Pisano & Teece, 2007). This study will therefore control for the level of innovation protection a firm uses. For all the three years, the
Innovation Monitor includes similar questions on intellectual protection measures. Participants need to indicate whether the firm makes use of (1) patents, (2) design/model rights, (3) trademarks, (4)
13
with 0 representing ‘no’ and 1 representing ‘yes. To create one overall score for IP-protection the dummy variables were combined into one composite variable with scores ranging from 0 (no protection mechanisms were used) to 5 (all protection mechanisms were used).
Degree of urbanity: In the literature on agglomeration economics a central topic is the geographic
concentration of firms and how it relates to knowledge spill overs. Urban centres are related to knowledge diversity and access to a highly educated workforce (Magrini & Galliano, 2012). Thus, it could be the case that there is a difference in the environmental innovation performance between firms having a headquarter in a urban area and firms that are headquartered in rural areas. Hence, this study controls for the urbanisation degree of the participating firm’s headquarter. A measure from the organisation Statistics Netherlands (CBS) which has ranked Dutch zip codes from 1 to 5 based on their degree of urbanity was used, with 1 representing the highest degree of urbanity and 5 representing the lowest degree of urbanity. After recoding, a score of 1 represents the lowest degree of urbanity, and 5 the highest degree of urbanity.
Family ownership: Previous research suggests that family-owned companies might be more willing to
invest in environmental innovations compared to ‘regular’ firms (Craig & Dibrell, 2006). This might be explained by their long-term strategic vision or by personal values and motives of the founders (Harris, Martinez & Ward, 1994; McCrea, 1997). Hence, this research will control for family owned companies. A dummy variable is created with 0 representing a firm that is not family owned and 1 representing a firm that is family owned.
3.3 Data analysis
This research uses a hierarchal regression analysis in which the variables will be entered in a specific order. It first includes the control variables, followed by the independent variables. Before entering the variables to the regression, all variables are mean centred. The analysis will be performed for all the three years separately.
4. Results
4.1 Descriptive statistics & correlations
14
were respectively 5.784, 7.479 and 4.42 suggesting little variance in this percentage. Interestingly, in 2017, the variable CEIPvol, the volume measure for environmental performance of the cluster has a negative significant correlation (r=-.177, p<.01) with the dependent variable EIP indicating that the more firms in a cluster have environmental innovations, the lower the environmental innovation performance of the focal firm. For the years 2018 and 2019 this correlation is also negative, however not significant. Another interesting observation is the significant negative correlation between
CEIPval and CEIPvol I (r=-.581, P<.01; r=-.174, P<.01) in the years 2018 and 2019. It seems that in
clusters where CEIPval is high, indicating a diverse set of environmental solutions, less firms are involved in developing environmental solutions and vice versa.
In table 1 the highest Variance Inflation Factor scores of every regression is shown, all scores are below the suggested cut off of 10 therefore, suggesting that there is no sign of multicollinearity for any variable in the regressions (Hair, Anderson, Tatham & Black, 1995).
Table 1: Highest Variance Inflation Factor scores
Year Variable VIF
2017 (valence) Cluster radical innovation 2.715
2017 (volume) Degree of urbanity 1.770
2018 (valence) Cluster radical innovation 1.592
2018 (volume) Degree of urbanity 1.564
2019 (valence) Cluster process innovation 1.877
15
Table 2: Descriptive statistics & correlation matrix 2017
Variable M SD 1 2 3 4 5 6 7 8 9 10 11 12 13 1.EIP 1.910 2.023 .093 .018 -.177** -.049 -.014 -.022 .171** .164** .117* .217** -.098* .218** 2.NP (Z-score) 0.011 .958 -.095 .073 -.024 -.013 .078 .023 -.121* .287** .208** .134** .013 3. CEIPval 1.516 .452 -.094 .343** .156** -.521** .117* .070 -.033 -.062 -.314** .081 4. CEIPvol 16.073 11.015 -.168** -.172** .031 -.063 -.171** -.077 .023 .517** -.228** 5.Cprod .921 .155 .420** .171** .025 .047 .032 .018 -.046 .076 6.Cproc .889 .201 -.322** .038 -.010 -.002 .004 .021 -.130** 7.Crad 18.355 5.784 -.124* -.116* .107* .098* .309** -.130** 8.Firm size 18.029 32.642 .434** .037 .073 -.055 .038 9.Firm age 22.850 25.497 -.099* -.133** -.117* .246** 10.Green subsidy .085 .279 .213** .011 -.031 11.IP-protection 1.097 1.331 .145** -.049 12.Degree of urbanity 2.350 1.352 -.201** 13.Family ownership .507 .501
16
Table 3: Descriptive statistics & correlation matrix 2018
Variable M SD 1 2 3 4 5 6 7 8 9 10 11 12 13 1.EIP 2.130 2.063 -.080 .084 -.096 .027 -.007 .042 .102 .111 .152** .108 -.131* .141* 2.NP (Z-score) -.018 1.017 .113* -.135* .086 -.005 -.032 .014 .134* -.141** -.136* -.047 .063 3. CEIPval 1.468 .448 -.581** .335** .370** .271** .006 .074 .040 -.082 -.246** .177** 4. CEIPvol 24.500 13.926 -.052 .086 .251** -.073 -.087 -.084 .124* .517** -.211** 5.Cprod 1.024 .131 .193** .175** -.060 .047 .087 -.067 .151** -.003 6.Cproc .930 .179 -.499** -.016 .086 -.082 .019 .050 .014 7.Crad 18.787 7.479 -.011 -.001 -.024 .098 .307** -.012 8.Firm size 35.650 103.392 .249** -.071 .024 .055 -.004 9.Firm age 26.059 27.822 -.034 -.154** -.059 .291** 10.Eco-subsidy .174 .380 .240** -.019 .052 11.IP-protection 1.178 1.217 .094 .019 12.Degree of urbanity 2.395 1.313 -.207** 13.Family ownership .510 .501
17
Table 4: Descriptive statistics & correlation matrix 2019
Variable M SD 1 2 3 4 5 6 7 8 9 10 11 12 13 1.EIP 2.090 2.072 .044 -.008 -.115 -.110 -.079 .093 .117 .058 .090 .184** -.140* .078 2.NP (Z-score) .018 1.015 -.022 .198** -.087 -.057 .137* -.043 -.093 .188* .158* .190** -.085 3. CEIPval 1.235 .285 -.174** .345** .596** .225** .035 .034 .053 .055 -.132* .068 4. CEIPvol 22.870 16.778 -.117 .032 .632** -.025 -.058 -.018 -.006 .603** -.173** 5.Cprod .822 .121 .443** .248** -.046 .028 -.042 .003 -.055 -.087 6.Cproc .677 .166 .132* .091 .029 -.117 -.025 .012 .006 7.Crad 16.899 4.420 -.041 -.021 .028 .001 .366** .036 8.Firm size 25.366 41.270 .518** -.045 -.043 -.035 -.002 9.Firm age 27.400 33.625 -.046 -.151* -.018 .216** 10.Green subsidy .160 .363 .158* -.034 .089 11.IP-protection 1.191 1.249 .044 -.062 12.Degree of urbanity 2.354 1.318 -.105 13.Family ownership .450 .498
18
4.2 Regression results
Tables 5, 6 and 7 present the results of the multiple regression tests. In the first model, only the control variables are entered; in the second model the independent variables are added. On the left side of the tables the valence measure for the environmental innovation performance of the cluster CEIPval is used and on the right side the volume measure CEIPvol.
Model 1: This baseline model only included the control variables. The results slightly differ
across the three years. In general, IP-protection has a positive significant effect (= .219, p<.01; = p<.01) for the years 2017 and 2019, but not in 2018 (= p>.10). Furthermore, firm size had a significant positive effect in 2017 and 2018 at the 10 percent significance level (=.100, p<.10; =.102, p<.10), however, it did not reach significance in 2019. Similarly, green subsidy has a positive significant effect in 2017 and 2018 (=.085, p<.10; =.135, p<.05), though it is not significant in 2019. The variables firm age (=.115, p<.10) and family ownership (=.194, p<.01) were only positively significant in 2017. The degree of urbanity was the only factor that is negatively significant in 2018 and 2019(=-.133, p<.05; =-.142, p<.05), in 2017 it was not significant in the baseline model.
Model 2: In this model, the independent variables network participation, cluster
environmental innovation performance (valence) (CEIPval), cluster size of SMEs with environmental innovations (CEIPvol), cluster product innovation (Cprod), cluster process innovation (Cproc, and cluster radical innovation (Crad). In none of the three years network participation shows a significant
effect, thereby rejecting hypothesis 1. For the second hypothesis, significant results can only be found in the year 2017. In both the valence and the volume model the variable cluster product innovation (=− p<.05; =-.126, p<.05) shows a significant negative relationship with the dependent variable. Moreover, in the valence model of 2017 the variable cluster radical innovation (=.134, p<.10) shows a positive significant relationship, even though it is only significant at the 0.1 level. In the volume model, the CAEI (=-.113, p<.10) shows a significant negative relationship with
environmental innovation performance, although, again only significant at the 10 percent significance
level. The second hypothesis is therefore also largely rejected, with the exception of cluster radical
innovation which is the only innovation characteristic of surrounding firms having a positive effect on
19
Table 5: Regression results 2017
Hypothesis Valence Volume
Model 1 Model 2 Model 2
Control variables Firm size .100* (.055) .098* (.055) .103* (.055) Firm age .115* (.059) .133** (.059) .112* (.059) Green subsidy .085* (.049) .069 (.051) .063 (.051) IP-protection .219*** (.047) .215*** (.047) .209*** (.047) Degree of urbanity -.078 (.049) -.102* -.044 (.062) Family ownership .194*** (.050) .208*** (.051) .188*** (.051) Independent variables Network participation .050 (.054) .050 (.053) CEIPval .087 (.073) - CEIPvol - -.113* (.061) Cprod -.187** (.078) -.126** (.062) Cproc .091 (.068) .027 (.065) Crad .134* (.080) .041 (.063) Adjusted R² .128 .132 .136 F value 11.022 6.674 6.897
Unstandardized betas, standard error between brackets; *=p<0.1, **=p<0.05, ***=p<0.01
Table 6: Regression results 2018
Hypothesis Valence Volume
Model 1 Model 2 Model 2
Control variables Firm size .102* (.061) .105* (.061) .104* (.062) Firm age .086 (.072) .095 (.073) .095 (.073) Green subsidy .135** (.059) .119** (.060) .119** (.060) IP-protection .093 (.058) .086 (.058) .086 (.059) Degree of urbanity -.133** (.063) -.171** -.169** (.077) Family ownership .093 (.064) .088 (.065) .088 (.065) Independent variables Network participation -.084 (.061) -.084 (.061) CEIPval .018 (.075) - CEIPvol - -.018 (.075) Cprod -.051 (.070) .055 (.067) Cproc -.072 (.076) -.067 (.073) Crad .107 (.073) .113 (.073) Adjusted R² .058 .058 .058 F value 4.09 2.694 2.693
20
Table 7: Regression results 2019
Hypothesis Valence Volume
Model 1 Model 2 Model 2
Control variables Firm size .108 (.066) .104 (.067) .103* (.067) Firm age .012 (.073) .017 (.073) .017 (.073) Green subsidy .056 (.063) .039 (.065) .040 (.065) IP-protection .189*** (.061) .185*** (.062) .185*** (.062) Degree of urbanity -.142** (.064) -.144* -.129 (.081) Family ownership .072 (.066) .084 (.067) .078 (.068) Independent variables Network participation .041 (.066) .042 (.067) CEIPval .017 (.088) - CEIPvol - -.048 (.105) Cprod -.106 (.078) -.118 (.082) Cproc -.044 (.090) -.031 (.075) Crad -.023 (.075) .007 (.093) Adjusted R² .057 .055 .055 F value 3.567 2.344 2.362
21
4.3 Robustness checks and additional tests
As the network participation variable had no direct effect on the dependent variable I examined if there could be a moderating effect of network participation influencing the relationship between respecitively the CEIPval and CEIPvol and the environmental innovation performance of the focal SME. The results only show a positive significant effect in the volume model of 2017 (=.081, p<.10), indicating that there might be a positive influence of active participation in networks on how firms transform observed eco-innovative behaviour from their surroundings into a high environmental innovation performance.
To see if another proxy for the direct experience of learning of the focal firm could produce significant results, this study looks at the active collaboration with other partners which was measured in the innovation benchmark by six items. Firms indicated whether they collaborated with different kinds of partners by answering with either ‘yes’ or ‘no’ for: (1) Clients, (2) Consultancy firms, (3) Suppliers, (4) Competitors, (5) Universities or other knowledge centres, and (6) Firms operating in a different industry. From these six dummy items, I created an overall composite score with a minimum value of zero (representing no collaboration with any partner) and a maximum score of 6 (representing collaboration with every partner). After re-running the multiple regression analysis with the variable
collaboration instead of network participation, the findings show a significant positive effect of
network participation on the dependent variable in all the three years. This indicates that such ties with partners that have a different position in the vertical value system than the firm, are much more valuable in realising environmental innovations than partnering through innovation networks with only horizontal partners.
To see if a different way of calculating the geographical clusters confirmed the results of the regression analysis for the second hypothesis, a Google API distance calculator was used to estimate the actual driving distance (in kilometres) between two zip codes. By using this tool, I calculated every distance between the zip codes of the year t and the zip codes of the clusters t-1, for all the three years the analysis was performed. In this way every zip code had a unique cluster of zip codes that fell within a radius of 15 kilometres driving distance. For all these separate clusters, the averages for the variables were calculated, and in case of the CEIPvol variable the firms within a cluster that had developed environmental innovations were counted. Next, a multiple regression analysis was
performed using the new distance measure. None of the independent variables had a significant effect on the dependent variable. This indicates that observational learning, at least in this sample, does not explain the environmental innovation performance of SMEs over the time period of a year.
22
scores of 2016 for the year 2018 of the focal firm, and cluster scores of 2017 for the year 2019. The results show that for the comparison 2016-2018 there is a positive significant relationship of cluster
radical innovation for both the valence and volume model. A bit more complicated is the 2017-2019
comparison, with all the original control variables entered into the model there are no significant relationships of the independent variables. However, once the control variable degree of urbanity is excluded, the CEIPval variable and the Cprod show significant relationships with the dependent variable. The CEIPval variable in this case has a positive significant effect on the environmental
innovation performance and the Cprod has a negative significant effect.
To decrease the chance of omitted variable bias and a subsequent misinterpretation of the effect of Crad and Cprod on the EIP I first used the radical innovation of the focal firm as a dependent variable to see if it had a significant relationship with the Crad variable. Similarly, I used the product innovation of the focal firm as a dependent variable to see if there was a significant relationship with the Cprod variable. Both relationships were not significant so there is no strong indication that there is omitted variable bias in this case.
5.Conclusion and discussion
5.1 Conclusions
As SMEs can contribute significantly in working towards a more sustainable society it is of great interest to find out what drives SMEs to develop eco-innovations, and how they can learn via reinforcement (direct experience), and by merely observing others (observational learning). Current literature on eco-innovations focuses mainly on what drives large firms to invest in environmental solutions (Kesidou & Demirel, 2010; Przychodzen & Przychodzen, 2015). It has been argued by Andries & Stephan (2019) that those findings do not pertain to SMEs per se because SMEs often experience resource scarcity, have different ways of capturing value organize their innovations differently than large firms (Rennings, 2000; Spithoven et al., 2012).
23
between the radical innovation of the cluster and the focal firm’s environmental innovation performance was found in both the one- and two-year time period. Furthermore, negative relations were found in the one-year time period between the product innovation and absolute environmental innovations of the firm clusters on the environmental innovation performance of the focal SME.
5.2 Theoretical implications
These findings have several theoretical implications. First of all, it seems that collaborating with a diverse set of partners offers a better chance of being successful in implementing environmental innovations in SMEs than collaborating in formal innovation networks. Literature on open innovation, like the work of Laursen & Salter (2006), supports the view that collaborating with a variety of external partners has a positive effect on innovation performance in general.
Looking at the effect of the innovation practices of the surrounding firms on the focal SME the findings for the lagged effect of one year seem contradictory at first sight. There is a negative effect of the clusters’ average level of product innovation and the number of firms in a cluster that develop environmental innovations on the environmental innovation performance of the focal SME. However, if the percentage of revenue generated from radical innovations as part of the total revenue by the firms in a cluster is high, this positively influences the environmental innovation performance of the SME. Conceivably, the negative relations of the clusters’ product innovation and the number of firms that develop eco-innovations could be explained by the framework of Utterback & Abernathy (1975) which looks at the dynamics between product and process innovation (Appendix 6). The authors argue that in the first stage of innovation product innovation tends to be high, while process innovation is low, over time the focus on product innovation diminishes and shifts to process innovation. In the first stage the focus is on fulfilling the needs of the consumer in the best way possible, the process
innovation in this stage is low as the focus is not on technologies and follows un uncoordinated path as it follows the demands from the market which can change rapidly, stimulating radical innovation. Later on, when a dominant design is set, the process innovation of the firm becomes more structured and the focus switches to technological solutions that improve the efficiency of the product, leading to more incremental innovations. Going back to the results of this study, after the first year of developing environmental innovations the SME the dominant design might already be in place, corresponding to the second stage of the model by Utterback & Abernathy (1975), which explains the negative relationship between the product innovation of the cluster and the number of environmental innovations of the cluster with the environmental innovation performance of the focal SME.
24
radical innovations of the firms in the cluster are measured through the relative revenue they generate means that its success is visible to the focal SME. This visibility might explain the positive direction and strength of the relationship found in this research. Taken together with the fact that there is no positive effect of the environmental innovation performance of the cluster on the environmental innovation performance of the focal SME it seems that there is not a set guideline on how to develop successful eco-innovations. There might be very different pathways towards a high environmental innovation performance and, in line with the findings of this research, environmental innovation performance might therefore be determined more by observing general (radical) innovation characteristics of neighbouring firms than specific environmental innovation characteristics.
Examining the lagged effect over two years, the results show that when the average environmental performance (valence) of the cluster is high, there is a positive effect on the
environmental innovation performance of the focal SME. However, this positive effect only occurs in a certain research setting and is therefore considered weak. Nonetheless, this finding supports the statements of Audretsch & Feldman (2003) that geographical location and the resulting knowledge spill overs between firms are a determinant of innovation.
5.3 Managerial implications
This research provides some interesting insights for managers and policy makers. First of all, the results of this research suggest that collaborating with a diverse set of partners increases the chances of successful environmental innovations. Managers that are interested in developing such innovations can actively try to engage with different partners in order to access the necessary knowledge and
capabilities. Another interesting finding is that SMEs located in urban areas are less likely to have a high environmental innovation performance compared to SMEs located in rural areas. This can be explained by the fact that firms in rural areas simply have more room to expand their operations physically in order to develop innovations. Managers that are keen to invest in environmental
innovations might therefore consider to start operating in more rural areas. Lastly, as the results show, a positive, albeit weak, effect between the environmental innovation performance of the surrounding firms and the environmental performance of the focal SME only occurs after two years. This suggests that it takes time to successfully develop environmental innovations, hence, managers should have a long-term focus when investing in environmental solutions without expecting direct positive results on the short term.
25
collaboration in innovation networks does not influence the environmental innovation performance of SMEs. Instead, policy makers could incentivise SMEs to collaborate with a diverse set of partners.
5.4 Limitations and further research
This research has several limitations which lead to suggestions for further research. First of all,
although the dataset covers multiple years, the data being used is cross-sectional. This makes the study of the lagged effects of the firms’ innovative characteristics less reliable as it does not allow for comparing the innovative activities of a single firm over the years, as very few firms have participated every year the survey was held. Secondly, the available survey data only covers three years, which only allows for testing the lagged effects over a one- and two-year time period, even though, the development of successful environmental innovations might take longer than those two years. It also remains unclear what the innovative activities of the participating firms were before 2016, the first year the survey was issued. Thirdly, in this research only a distinction is made between the volume and valence of environmental innovation activities of the cluster. However, there might also be interesting differences between the volume and valence of the product, process and radical innovation characteristics of the cluster. Lastly, the radical innovation of the cluster is measured by the revenue it generates in the market, which gives a good indication of the success of the radical innovations that were introduced by the cluster. However, the product and process innovations of the cluster are measured in the amount of innovations realised, which does not indicate whether the innovations were successful in the market.
26
References
Almeida, P. & Kogut, B. (1999). Localization of Knowledge and the Mobility of Engineers in Regional Networks. Management Science, 45(7), 905-917.
Andries, P. & Stephan, U. (2019). Environmental Innovation and Firm Performance: How Firm Size and Motives Matter. Sustainability, 11(13), 3585.
Anselin L., A Varga and Z. Acs (1997), Local Geographic Spillovers Between University Research and High Technology Innovations, Journal of Urban Economics, 42, 422-448.
Audretsch D. and Feldman M. (1996), R&D Spillovers and the Geography of Innovation and Production, American Economic Review, 86, 631-640.
Audretsch, D. B., & Feldman, M. P. (2004). Knowledge spillovers and the geography of innovation. In
Handbook of regional and urban economics (Vol. 4, pp. 2713-2739). Elsevier.
Bandura, A., & Walters, R. H. (1977). Social learning theory (Vol. 1). Englewood Cliffs, NJ: Prentice-hall.
Beaudry, C., & Schiffauerova, A. (2009). Who's right, Marshall or Jacobs? The localization versus urbanization debate. Research policy, 38(2), 318-337.
Beugelsdijk, S., & Cornet, M. (2002). 'A Far Friend is Worth More than a Good Neighbour':
Proximity and Innovation in a Small Country. Journal of management and governance, 6(2), 169-188.
Bougrain, F. & Haudeville, B. (2002). Innovation, collaboration and SMEs internal research capacities. Research Policy, 31, 735-747.
Carbonara, N. (2005). Information and communication technology and geographical clusters: opportunities and spread. Technovation, 25(3), 213-222
Chen, Y., Wang, Q., & Xie, J. (2011). Online social interactions: A natural experiment on word of mouth versus observational learning. Journal of marketing research, 48(2), 238-254.
Chen, P. Y., Wu, S. Y., & Yoon, J. (2004). The impact of online recommendations and consumer feedback on sales. ICIS 2004 Proceedings, 58.
Craig, J. and Dibrell, C. (2006). The Natural environment, innovation, and firm performance: A comparative study. Family Business Review 19(4): 275–288.
Damanpour, F. (1992). Organizational size and innovation. Organization Studies, 13, 375–402.
27
Doloreux, D., (2004). Regional networks of small and medium sized enterprises: evidence from the metropolitan area of Ottawa in Canada. European Planning Studies 12 (2), 173–189.
Hair, J. F. Jr., Anderson, R. E., Tatham, R. L. & Black, W. C. (1995). Multivariate Data
Analysis (3rd ed). New York: Macmillan.
Hansen, O.E., Sondergaard, B. & Meredith, S.(2002). Environmental Innovations in Small and Medium Sized Enterprises. Technology Analysis & Strategic Management, 14(1), 37-56.
Harris, D., Martinez, J.I. & Ward, J.L. (1994). Is strategy different for the family-owned business?
Family Business Review, 7, 159-174.
Horbach, J., Oltra, V., & Belin, J. (2013). Determinants and specificities of eco-innovations compared to other innovations—an econometric analysis for the French and German industry based on the community innovation survey. Industry and Innovation, 20(6), 523-543.
Jamali, D., Zanhour, M., & Keshishian, T. (2009). Peculiar strengths and relational attributes of SMEs in the context of CSR. Journal of Business Ethics, 87(3), 355-377.
Kemp, R., Foxon, T., 2007. Typology of eco-innovations. Deliverable 2. EU FP6 funded project 044513: 24. Maastricht.
Kesidou, E., & Demirel, P. (2012). On the drivers of eco-innovations: Empirical evidence from the UK. Research Policy, 41(5), 862-870.
Klewitz, J., Zeyen, A. & Hansen, E.G. (2012). Intermediaries driving eco-innovation in SMEs: a qualitative investigation. European Journal of Innovation Management, 15(4), 442-467.
Laforet, S. (2008). Size, strategic, and market orientation affects on innovation. Journal of business
Research, 61(7), 753-764.
Laursen, K., & Salter, A. (2006). Open for innovation: the role of openness in explaining innovation performance among UK manufacturing firms. Strategic management journal, 27(2), 131-150.
Lee, M., Rodgers, S., & Kim, M. (2009). Effects of valence and extremity of eWOM on attitude toward the brand and website. Journal of Current Issues & Research in Advertising, 31(2), 1-11.
Liu, Y. (2006). Word of mouth for movies: its dynamics and impact on box office revenue. Journal of
Marketing, 70(3), 74-89.
Magrini, M. B., & Galliano, D. (2012). Agglomeration economies, firms' spatial organization and innovation performance: some evidence from the French industry. Industry and innovation, 19(7), 607-630.
Mazzanti, M., & Zoboli, R. (2006). Examining the factors influencing environmental innovations.
28
McCrea, B. (1997). Growing a family business, one step at a time. Industrial Distribution, 86: FA7-FA10.
Peters, L.D., Johnston, W.J., Pressey, A.D. & Kendrick, T. (2010). Collaboration and collective learning: Networks as learning organisations. Journal of Business & Industrial Marketing, 25(6), 478-484.
Pisano, G. P., & Teece, D. J. (2007). How to capture value from innovation: Shaping intellectual property and industry architecture. California management review, 50(1), 278-296.
Przychodzen, J., & Przychodzen, W. (2015). Relationships between eco-innovation and financial performance–evidence from publicly traded companies in Poland and Hungary. Journal of Cleaner
Production, 90, 253-263.
Rennings, K. (2000). Redefining innovation - eco-innovation research and the contribution from ecological economics. Ecological Economics, 32(2), 319–332.
Rogers, E.M. (2003). Diffusion of Innovations, 5th ed., New York: Free Press
Sørensen, J. B., & Stuart, T. E. (2000). Aging, obsolescence, and organizational innovation.
Administrative science quarterly, 45(1), 81-112.
Spithoven, A., Vanhaverbeke, W., & Roijakkers, N. (2013). Open innovation practices in SMEs and large enterprises. Small Business Economics, 41(3), 537-562.
Thompson, P., & Fox-Kean, M. (2005). Patent citations and the geography of knowledge spillovers: A reassessment. American Economic Review, (95), 450–460.
Utterback, J. M., & Abernathy, W. J. (1975). A dynamic model of process and product innovation.
Omega, 3(6), 639-656.
Van Aken, J. E., & Weggeman, M. P. (2000). Managing learning in informal innovation networks: overcoming the Daphne‐dilemma. R&d Management, 30(2), 139-150.
Von Hippel, E. (1986). Lead users: a source of novel product concepts. Management science, 32(7), 791-805.
Wagner, M. (2007). On the relationship between environmental management, environmental innovation and patenting: Evidence from German manufacturing firms. Research Policy, 36(10), 1587-1602.
Yalabik, B., & Fairchild, R. J. (2011). Customer, regulatory, and competitive pressure as drivers of environmental innovation. International Journal of Production Economics, 131(2), 519-527.
29
Appendices
Appendix 1: 2017- Factor Analysis Rotated Component Matrix (Varimax), N=504
Factor 1 Factor 2
Material reduction .726 .022
Energy reduction .763 -.074
C02 reduction .739 -.149
Less toxic materials .746 .016
Less noise pollution .705 -.014
More recycling .720 -.128
Innovation network -.093 .860
Innovation test surrounding -.123 .833
Innovation
accelerator/incubator
.026 .571
Appendix 2: 2017 – Cronbach’s alpha
Cronbach’s alpha Number of items N
Environmental performance
.830 6 504
Network participation .646 3 625
Appendix 3: 2018- Factor Analysis Rotated Component Matrix (Varimax), N=326
Factor 1 Factor 2
Material reduction .695 -.040
Energy reduction .785 .058
C02 reduction .749 -.013
Less toxic materials .711 -.024
Less noise pollution .726 .161
More recycling .677 .110
Innovation network .049 .785
Innovation test surrounding .077 .815
Innovation
accelerator/incubator
-.001 .654
Appendix 4: 2018 – Cronbach’s alpha
Cronbach’s alpha Number of items N
Environmental performance
.820 6 348
30
Appendix 5: 2019- Factor Analysis Rotated Component Matrix (Varimax), N= 269
Factor 1 Factor 2
Material reduction .734 .073
Energy reduction .793 -.102
C02 reduction .765 -.016
Less toxic materials .717 -.001
Less noise pollution .686 -.045
More recycling .663 -.117
Innovation network -.054 .788
Innovation test surrounding -.078 .824
Innovation
accelerator/incubator
.015 .658
Appendix 6: 2019 – Cronbach’s alpha
Cronbach’s alpha Number of items N
Environmental performance
.840 6 325
Network participation .633 3 270
