The dynamics of technological substitution: the case of eco-innovation diffusion of surface cleaning products

Accepted Manuscript The Dynamics of Technological Substitution: the Case of Eco-Innovation Diffusion of Surface Cleaning Products Edgars Vīgants, Andra Blumberga, Lelde Timma, Ivars Ījabs, Dagnija Blumberga PII:

S0959-6526(15)01388-8

DOI:

10.1016/j.jclepro.2015.10.007

Reference:

JCLP 6238

To appear in:

Journal of Cleaner Production

Received Date: 24 June 2014 Revised Date:

24 September 2015

Accepted Date: 2 October 2015

Please cite this article as: Vīgants E, Blumberga A, Timma L, Ījabs I, Blumberga D, The Dynamics of Technological Substitution: the Case of Eco-Innovation Diffusion of Surface Cleaning Products, Journal of Cleaner Production (2015), doi: 10.1016/j.jclepro.2015.10.007. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

1 2 3

ACCEPTED MANUSCRIPT THE DYNAMICS OF TECHNOLOGICAL SUBSTITUTION: THE CASE OF ECO-INNOVATION DIFFUSION OF SURFACE CLEANING PRODUCTS

4 5

Edgars Vīgantsa, Andra Blumbergaa, Lelde Timmaa*, Ivars Ījabsb, Dagnija Blumbergaa

8 9

a

Institute of Energy Systems and Environment, Riga Technical University, Latvia b Department of Political Science, University of Latvia, Latvia

RI PT

6 7

*Corresponding author: e-mail: [email protected], phone number: +371 67089943; +371 29705879, Fax: +371 67089908, Address: Āzenes iela12/1, Riga -1048, Latvia. Abstract

11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

Some eco-innovations have already reached a mature state, but they diffuse slowly. There is little known about the importance of various factors on the diffusion process of ecoinnovations. The aim of this research is to propose a conceptual model for eco-innovation diffusion, where particular case study of eco-innovation – micro-fibre cloth (MFC) for surface cleaning purposes – is given. The work will study the relative importance of various motivations on diffusion processes as well as the influence of information campaigns. This novel methodology has been developed under the system dynamics framework and is coupled with the logistic regression model. Structural validation tests were done for the proposed model. The methodology is proposed, then the integrated empirical study and system dynamics model are completed. The results of the logistic regression models show that the overall attitudes and social influence has the most dominant impact on intention to use the studied eco-innovation products. The most sensitive parameters in the system dynamics model were found to be the adoption faction of information received by word-of-mouth, households adopting micro fibre cloth at the beginning of the simulation, and the contact rate per household. The proposed methodology, and developed mathematical model can also be used as the basis for other studies of other diffusion processes, since system dynamics modelling is a white-box modelling approach, and the structure of the model can be enhanced with the latest results from other studies.

M AN U

TE D

EP

30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47

1. INTRODUCTION

The exponential growth of the population has generated similar patterns in final consumption. As a result, this leads to boosts in the economy, resource consumption, energy consumption and, consequently, environmental degradation. As the founder of system dynamics Jay Forrester (1961) pointed out: growth cannot be limitless in a finite system; when the system reaches its carrying capacity, a collapse occurs. Therefore, we believe that sustainable reductions in both production and final consumption are the key challenge for modern society. Eco-innovations are related to sustainable production and consumption; they are increasingly used to substitute existing products or services (EIO, 2011). In our work we review one particular eco-innovation: micro fibre cloth for surface cleaning purposes. In substitution theory, buyers’ decisions are based on both a generic function and additional functions performed by the product (Porter, 1985). In our studied case, when the substitution of a product is made by an eco-innovation, it performs the same generic function as the product being replaced; however, the product itself is very different. Micro-fibre clothes have the same generic function as other wipes and an important addition property substitution of cleaning agents. In the context of this research, the traditional scope of eco-innovations, which is to reduce the material inputs and waste outputs, has a broader and more comprehensive view – absolute reductions.

AC C

29

SC

10

ACCEPTED MANUSCRIPT

EP

TE D

M AN U

SC

RI PT

A recent review by Karakaya et al. (2014) states that understanding the diffusion of eco-innovations is critical as some of them have now reached a mature state, but the diffusion rate is slow and the path unclear. Consequently, eco-innovations have required a long-time period to be adopted. One of the theories on innovation diffusion is Rogers' general innovation theory (2003). This theory defines five stages of innovation adoption, and shows how innovations are diffused among members in a social system. Recently, Rogers’ theory has been applied to study the diffusion of electric vehicles by Plötz et al. (2014), smart meters by Rixen and Weigand (2014), and broadband by Lin and Wu (2013). The factors promoting or inhibiting pro-environmental behaviour have been studied from various theoretical perspectives. In environmental psychology, there are multiple approaches to explain pro-environmental behaviour. For example, from cognitive behavioural studies the theory of reasoned action (TRA) by Ajzen and Fishbein (1980) and the theory of planned behaviour (TPB) by Ajzen (1991). The norm activation model is widely used (Schwartz, 1977) to explain personal normative influences on altruism as well as feelings of moral obligation. Stern et al. (1999) proposed the value-belief-model, and Stern et al. (2000) the attitude-behaviour-context model. Goal-framing theory (Lindenberg and Steg, 2007) attempts to integrate multiple behavioural motivations. Moreover Bamberg (2003) studied how environmental concerns can be translated to the environmentally sound actions. The diffusion of eco-innovations from the perspective of the manufacturer has been studied by Sushandoyo and Magnusson (2014), where the need for policy support was identified as a crucial element. The need for support has also been found by Ghisetti and Rennings (2014), where it was concluded that the benefits for companies to reduce externalities (harmful material outputs to the environment) is insufficient. A study by Govindan et al. (2014), which focuses on the instigators for the introduction of green manufacturing, states that two of the three most important drivers are stakeholders and consumers. Schwarz and Ernst (2009) have attempted to use agent-based modelling to study the diffusion of water-related technologies in households. Here, the authors distinguished between various consumer groups based on the dominant behaviour. However, the theory to explain the diffusion of eco-innovations has not been defined, and there is a little known about the importance of the various factors in this process. Rennings (2000) states that eco-innovations cannot be treated like other innovations and require a specific theory and policy. Ozaki (2011) has shown that market mechanisms alone cannot provide an appropriate diffusion speed for eco-innovations to be adopted in households. Zeppini et al. (2013) outline various threshold models for technological transitions, and argue that transitions can be caused by different processes and most probably by the combination of those processes. Abrahamse et al. (2005) review various intervention models to change consumption patterns in households. As concluded by de Medeiros et al. (2014), there is a clear gap in the current research on decision-making processes when consumers are faced with the choice between ‘green’ vs traditional products. Moreover, there is a need to propose models which are able to forecast the diffusion processes of new products with little or no data available (Meade and Islam, 2006). Our proposed methodology deals with the multiple motivations – in our case, for using micro-fibre clothes (MFC). Since we are interested in different factors that might influence an individual’s decision to start using MFC the various motivations adopted from Ozaki (2011) provides us with a useful framework. Unlike other theories that concentrate mainly on one set of factors, our study tries to integrate factors, as well as find the interconnectedness of multiple motivational levels. This can help to study eco-innovation diffusion and proenvironmental behaviour as it results from multiple conflicting /non-conflicting motivations that change over time, and often more than one motivation is active at any given time leading to a situation where an individual’s behaviour results from multiple motivations. The aim of this research is to propose a conceptual model for eco-innovation diffusion. This can help to study eco-innovation diffusion and pro-environmental behaviour as it results from multiple conflicting/non-conflicting motivations that change over time, as well as more

AC C

48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102

ACCEPTED MANUSCRIPT

than one motivation often being active at any given time. Consequently, behaviour is a result of multiple motivations. To achieve this aim, the dynamics for the diffusion of eco-innovations will be outlined. The work will study the relative importance of various motivations on diffusion processes as well as the influence of information campaigns. We apply a system dynamics model to account for the avoided environmental pollution based on the decisions made by the consumer. The case study covers micro-fibre clothes for surface cleaning. The paper starts with the background information concerning the eco-innovations in the case study – micro-fibre cloth. Following that, the methodological framework will be outlined. This will be followed by an empirical study which will be based on various motivations, factor analysis, and logistic regression analysis. After that, a section devoted to the system dynamics model will be presented. The final section will present and discuss the results. Although the system dynamics model used was based on one specific eco-innovation’s diffusion process, its general application to other products and services is proposed and discussed.

119

2. BACKGROUND INFORMATION ON THE CASE STUDY

120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149

The introduction of micro-fibre clothes (MFC) for the purpose of cleaning surfaces is an eco-innovation (in a broad definition) since the MFC can substitute both conventional wipes and cleaning agents. MFC have the same generic function as other wipes as well as an important additional property: the substitution of cleaning agents. As a result, this ecoinnovation reduces pressure on the environment through both the production and consumption of cleaning agents. MFC were introduced in the early 1990s; they are made from synthetic micro-fibres (the weight to length ratio of fibres < 1 g/10000 m). In the 2000s, ultra-micro fibre clothes were commercialized with a smaller weight to length ratio: < 1/3 g/10000 m. This ratio is acquired by splitting synthetic fibres into thinner ones with chemical and mechanical techniques (Nilsen et al., 2002). Various studies showed the competitive cleaning performance of MFC compared to traditional wipes and cleaning agents: a study by Nilsen et al. (2002) showed that MFC and ultra-MCF removed over 90% of the deposits from surfaces, Wren et al. (2008) conclude that ultra-MFC performed better than conventional wipes at removing bacteria from surfaces in hospitals, Smith et al. (2011) report that MFC is an effective tool for dealing with microorganisms in hospitals on various surfaces. The cleaning abilities of MFC were shown by Humphreys et al. (2012) to not differ significantly after 350 conventionally laundered wash cycles and 250 wash cycles with sterilization. It has been reported that the average use of cleaning agents is 91 g/week, or 4.7 kg/year, using the traditional housekeeping method in a US household (Bennett et al., 2012). It has also been reported by Nilsen et al. (2002) that the benefits of MFC can be estimated as: reduced discharge of chemicals from cleaning agents into the environment (50%), reduced water consumption (20%) and energy (30%), and reduced waste streams from the consumption of cleaning wipes and cloths (50%). A pilot project carried out by UC Davis Medical Center (2006) confirmed these projections by reducing purchases of chemicals for floor cleaning by 46%. Moreover, the economic benefits associated with the introduction of MFC was a 60% cost-saving for cleaning cloths (mops), 95% for chemicals needed for mopping, and 20% saved in relation to the labour invested per day.

150

3. METHODOLOGY

151 152

This work combines an empirical study, given in Section 4, with system dynamics modelling, given in Section 5, see Figure 1 for the conceptual scheme.

AC C

EP

TE D

M AN U

SC

RI PT

103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118

ACCEPTED MANUSCRIPT EMPIRICAL STUDY

SYSTEM DYNAMICS START

Define dynamic problem

Put forward dynamic hypothesis

RI PT

Design and distribute questionnaire

Collect responses

NO

Develop logistic regression model

YES

SC

Do results correspond to dynamics hypothesis?

Process data with correlation analysis

M AN U

Formulate the system dynamics model

Validate the model

TE D

Do results correspond to logistic regression?

153 154 155 156 157 158 159 160 161

AC C

EP

YES

NO

Simulate reference behaviour Identify influential parameters Evaluate policy scenarios

Communicate results and futher research

END

Figure 1. Conceptual scheme of methodology used Initially, the dynamic problem and hypothesis is developed, this is followed by the design of a questionnaire (Section 4.1.1). After this, the responses are processed by factor and item reliability analysis and next logistic regression model is developed (Section 4.1.2). The results of the empirical study are used for the formulation of the system dynamics model (Section 5.1.1 and 5.1.2) and presentation of assumptions (Section 5.1.3). Later, the system dynamics model undergoes validation (Section 5.1.4) and the validation results are crosschecked with the results from the logistic model (Section 6.1.1). After that, the reference

ACCEPTED MANUSCRIPT

162 163 164 165 166

behaviour of the system dynamics model is simulated (Section 6.1.2), a sensitivity analysis (Section 5.1.5) is used to state the influential parameters (Section 6.1.3), information campaigns are applied to perform the scenario analysis (6.1.4). The major results and discussion are provided in Section 6, while conclusions and further research is summarized in the conclusion section.

167

4. EMPIRICAL STUDY

168 169

The framework used for the empirical study was study by Ozaki (2011) where various motivations were outlined.

170

4.1.

171 172 173 174

Various motivations in this study are used to explore the diffusion of an ecoinnovation. We explore and model the motivations behind the decision to start using microfibre clothes (MFC). This is done by the logistic regression model, which determines the probability and its changes for the event of using or rejecting to use MFC.

189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211

RI PT

SC

M AN U

We surveyed both those respondents who knew about MFC and used them, and those who were aware of the existence of MFC but did not use them. This allowed for the establishment of the differences between the motivations of users and non-users, as well as to suggest possible further ways to stimulate the adoption of this type of eco-innovation. Questions regarding all these motivations were integrated into the questionnaire design which was adopted and adjusted from a questionnaire about the use of green electricity tariffs by Ozaki (2011); the author based his questionnaire on multiple environmental psychology theories. Furthermore, as the author suggests, that his research methodology can be used to explore the adoption behaviour of consumers for innovative and tangible consumer goods. An anonymous on-line questionnaire (www.survs.com) was placed on co-author’s Mr. Ivars Ījabs Twitter account (with more than 9000 followers) and was open for filling in for 3 weeks during May, 2014. The questionnaire was in Latvian. 4.1.2.

Questionnaire data analysis

TE D

188

Questionnaire design

All responses were screened, and partly-completed questionnaires were removed from further analysis. Based on the answer about knowledge and use of MFC, all respondents were divided into three groups: (1) those who knew about MFC and used them in households (hereinafter “Users”), (2) those who knew about MFC, but did not use them in households (“Non-users”) and (3) those who have never heard of MFC before and those who have used MFC before but stopped using them. The answers of the third group where excluded from the statistical analysis, since the aim was to study the motivation behind the decision to use or not to use MFC. The analysis was done by interactive statistical data analysis tool Statgraphics Centurion 16.1.15.

EP

176 177 178 179 180 181 182 183 184 185 186 187

4.1.1.

AC C

175

Methods

4.1.2.1. Factor analysis

Factor analysis was preformed to reduce the amount of variables used in further data analysis by extracting the optimal number of common factors from initial variables. The model of orthogonal common factors by Johnson and Wichern (2002) was used. The variables were grouped based on the correlation matrix. Principal components were used as the type of factoring. As rotation method the Varimax was applied. The factors where extracted based on the Eigenvalue, scree plot, estimated communality and specific variance. Two types of principal components analysis can be done: exploratory and confirmatory. In the case of exploratory analysis, the factors are obtained from the computer software based on the information embedded in variables; in the contract for the confirmatory analysis the variables included in factors are defined by the researcher. In this research confirmatory analysis was performed by dividing variables based on the motivations from Ozaki (2011) research.

ACCEPTED MANUSCRIPT

EP

TE D

M AN U

SC

RI PT

The factor analysis was the pre-condition for obtaining logistic regression model, since with factor analysis the dimensionality of independent variables is reduced down to optimal number of factors.

AC C

212 213 214

ACCEPTED MANUSCRIPT

4.1.2.2. Item reliability analysis

215 216 217 218 219

Item reliability analysis was done to test the consistency of the variables included in the factor analysis. The scale reliability was expressed by Cronbach’s alpha. As the rule-of-thumb Cronbach’s aplha’s values above 0.7 are considered the minimal threshold. Item reliability analysis in this study was performed for all studied motivations. 4.1.2.3. Logistic regression analysis

220

The logistic regression analysis was preformed to fit a regression model in which the dependent variable characterizes an event with only two possible outcomes: to use MFC or not to use MFC. The use of MFC was expressed as the answer “yes” to the question “do you know about MFC and us it” (given as “Users” in the Section 4.3.1) and the rejection of MFC was expressed as the answer “yes” to the question “do you know about MFC and do not us it” (given as “Non-users” in the Section 4.3.1). These answers were coded in binary for as 1 for “yes, I do know about MFC and us it” and as 0 for “yes, I do know about MFC and do not us it”. The model assumes that users of MFC will continue to use this product further on. As the independent variables factors obtained from principal components analysis were used. The fitted model relates predictor variables, and assumes the probability of an event through a logistic function, see Eq. (1).

232

log[P ( X ) (1 − P ( X ))] = exp( β 0 + β1 X1 + β 2 X 2 + ... + β k X k )

233 234 235 236 237 238

Where P is the probability of the occurrence of the outcome Y to the predictor variables X, β0 is the constant of the fitted model and β1, β2, …βk are estimates. Stepwise backward selection factor selection was chosen with P-value to enter 0.05. To estimate the accuracy of the fitted model, the percentage of deviance explained by the model (R2) was calculated in a similar manner to R-squared statistics in a multiple regression; see Eq. (2).

239

R 2 = λ ( β 1, β 2 ,Kβ k β 0 ) λ ( β 0 ) 2

adj)

2 R adj = [λ ( β 1, β 2 ,Kβ k β 0 ) − 2p] λ ( β 0 )

241

(1)

(2)

in a similar manner to R-squared adjusted; see Eq. (3).

TE D

And adjusted deviance (R

240

M AN U

SC

RI PT

221 222 223 224 225 226 227 228 229 230 231

(3)

242 243

Where p equals the number of coefficients in the fitted model, including the constant term. Statistical significance of the fitted model was estimated using Chi-squared value.

244

4.2.

245 246 247

In total 235 questionnaires were filled in; 170 were completed fully and used for further analysis. From the completed questionnaires, 130 were users of MFC, and 39 were non-users intending to use MFC. Only one respondent had used MFC before, and stopped using them.

249 250 251 252 253 254 255 256 257 258

EP

AC C

248

Main results of empirical study

4.2.1.

Demographics

The sample group was biased towards female respondents, people with higher education, and the average income per household member was higher than the national average, see Table 1. Biases might have several explanations: (1) the questionnaire was voluntary; (2) MFC use is part of housekeeping, traditionally being a women’s duty, (3) the questionnaire was in Latvian, as the Twitter account is in Latvian, and this is a barrier for the Russian speaking population. A sample group also had a strong bias towards a pro-environmental attitude. These biases affect the representativeness of the questionnaire, but in this study the intention was to explore the attitudes only of those people who had at least some idea what MFC was, i.e., who had either used MFC themselves or had some previous information about them.

ACCEPTED MANUSCRIPT

Table 1: Demographics of surveyed sample and average population in Latvia (N = 170) Statistics from

Parameter, unit Age, average years Gender, % females Income per household member, EUR/month Household size, persons per household Education % higher % secondary % primary % below primary or none Housing type, % living in a flat Language spoken at home, % speaks Latvian

269

71.76 27.65 0.59 0.00 68.24 93.53

23.10 (2011) 53.99 (2011) 18.72 (2011) 4.19 (2011) 70.00 (2013) 56.27 (2011)

SC

The major part of the respondents did fall into the category “Users” (those who knew about MFC and used them in households), which can be explained by the environmentally inclined persons interests in completing survey related to the environmental issues. Similar issue is described by the Tonglet et al. (2004) in the study on waste recycling behaviour, where majority of study population already did take part in the waste recycling. Tonglet et al. (2004) agrees that results of the study may be biases by pro-environmental respondents, nevertheless the purpose is to evaluate to motivation behind waste recycling, therefore it is reasonable to use obtained data set. We would argue that in our case study, the same motivation behind use of obtained data set is valid. 4.2.2.

M AN U

260 261 262 263 264 265 266 267 268

32.62 81.76 628.99 3.30

Average in Latvia, (year) (CSB, 2014) 41.60 (2011) 54.18 (2014) 319.90 (2012) 2.43 (2013)

questionnaire

RI PT

259

Factor analysis and item reliability

The results of the factor analysis and item reliability tests for motivations are given in Table 2.

272 273

Table 2 Factor analysis and item reliability statistics for the model “Intention to use MFC (mirco-fibre clothes)” Factors Cumulative Minimal Sample Cronbach’s Factors KMO extracted percentage Eigenvalue size alpha Overall attitudes & 1 73.7 2.2 0.7 137 0.8 Social Influence Green Norms 2 75.9 1.4 0.8 166 0.8 Functionality & 1 74.7 1.5 0.6 147 0.7 Controllability Self-Efficacy 1 72.8 2.2 0.6 142 0.8

274 275 276 277 278 279 280

The item reliability for all motivations showed Cronbach’s alpha of above usually used threshold value of 0.7. These factors explained from 72.8 % to 75.9 % of cumulative percentage in particular studied motivations. The total amount of motivations where bigger, including access to information, green expectations, green beliefs, green values and consequential beliefs, but these factors are not given here, since they are found to be not statistically significant in following logistic regression model.

282 283 284 285

AC C

281

EP

TE D

270 271

4.2.3.

Logistic regression model

The final model “Intention to use MFC” (which combines all motivations given in Table 2 together) explained 34.24 % of deviance and 28.45 % of adjusted deviance. If compared to a study by Tonglet et al. (2004) where intention to recycle was studied, the adjusted deviance in our research is in the same range (34.38 % vs. 33.3 %); see Table 3.

286 287

ACCEPTED MANUSCRIPT

289 290 291 292

Since the P-value for Chi-square test is greater than 0.05, therefore there are no reason to reject the adequacy of fitted model and it can be concluded that the logistic function adequately fits observed data. The statistical model of “Intention to use MFC” (MFCUSE) is given as Eq. (4).

MFCUSE = exp(η ) /(1 + exp(η ))

293 294 295 296

RI PT

288

Table 3: Analysis of deviance, residuals and goodness of fit for the model “Intention to use MFC (micro-fibre clothes)” (N = 126) Deviance explained Adjusted Chi-square Model Df MSE by the model, % deviance, % with 1 Df 1.97·10-2 0.87 (P >0.05) Intention to use MFC 41.74**** 34.38 4 **** Significant at P < 0.0001.

(4)

Where the η is calculated as given in Eq. (5), where four motivations were found statistically significant.

η = 2.33 + 1.26OASI − 0.31GN − 0.65FC − 0.73SF

(5)

In the Eq. (5) OASI is the overall attitudes and social influence, GN is the green norms, FC is functionality and controllability and SF is self-efficacy. Detailed results of the statistical tests for these factors are given in Table 5.

300 301

302

Table 5: Maximum likelihood and likelihood ratio tests for the model “Intention to use MFC (microfibre clothes)” (N = 126) Estimate Estimated odds ratio Factors S.E. Chi-square Df (confidence interval) (confidence interval) Constant 2.33 (1.47; 3.20) 0.44 Overall attitudes & 1.26 (0.71; 1.81) 0.28 3.53 (2.04; 6.10) 40.89**** 1 Social Influence (OASI) Green Norms (GN) − 0.31 (− 0.60; − 0.03) 0.14 0.73 (0.55; 0.97) 4.92* 1 Functionality & − 0.64 (− 1.11; − 0.16) 0.24 0.53 (0.33; 0.85) 8.29** 1 Controllability (FC) Self-Efficacy (SF) − 0.73 (− 1.18; − 0.28) 0.23 0.48 (0.31; 0.75) 14.92**** 1 * Significant at P < 0.05; ** at P < 0.01; *** at P < 0.001; **** at P < 0.0001.

303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321

As the results in Table 5 shows, that the strongest impact on the choice is by the constant term and overall attitudes and social influence (where questions like “Use of MFC was/would be good idea” and “My closest support use of MFC” where evaluated); therefore the stronger the social influence the high probability of adopting MFC. We would like to interpret the constant term as the inertia of the system or the current position of product. It also can be regarded as the effect of habits in household cleaning practices. The overall attitudes and social influence is given with the positive sign, therefore the model’s value for the intention to use MFC increases, when this motivation grows. The estimates in the logistic regression do not depict the magnitude of the changes in a dependent variable, instead for these purposes estimated odds ratio or the exponential form of the estimate can be used. The odds ratio or 3.53 shows increase in the probability of the event to use MFC given expose to this motivation. The remaining motivations – green norms, functionality and controllability, and selfefficacy – show similar statistics, where the estimates are negative and from − 0.31 to – 0.73 and the odds ratio are smaller then 1, therefore exposure to these motivations are associated with lower odds of outcome. Since the confidence level for odds ratio do not includes value 1, these factors are still statistically significant, but plays marginal role in the model (Hair et al. 1998).

AC C

EP

TE D

M AN U

SC

297 298 299

ACCEPTED MANUSCRIPT

5. SYSTEM DYNAMICS MODEL

323 324 325 326 327 328 329 330 331 332 333

The model is constructed using system dynamics methodology, an effective tool in dealing with dynamic problems in complex systems with delays, non-linearities, and feedbacks. This methodology was introduced by Jay Forrester (1958) and allows for the outlining of relations between physical activities, information flows, and policy measures. All model equations link together these elements in total, and this constitutes the structure of the model while generating dynamic behaviour over the time of the system under study. The purpose of a system dynamics study is to explore how and why problematic behaviour is generated, and to discover leverage points to abate the causes of these problems. System dynamics converts results obtained from the empirical study into a mathematical model for the diffusion of eco-innovations to allow for the prediction of a system’s behaviour over time. Causal loop diagram

Information feedback plays an important role in the system dynamics modelling process of complex dynamic systems. A causal loop diagram provides the main structure of the system dynamics model and illustrates the feedback mechanisms in the system that generated the dynamic behaviour (Forrester, 1961). ‘Causal’ refers to the cause-and-effect relationship while ‘loop’ denotes the closed chain of cause and effect creating the feedback. In positive feedback, both variables change their values in the same direction, while in negative feedback the variables change their values in opposite directions. Positive or reinforcing loops are labelled with the letter R in the middle of the loop, negative or balancing loops with the letter B, see Figure 2.

SC

335 336 337 338 339 340 341 342 343

5.1.1.

self-efficacy + green norms B4

B5

+

overall attitudes and social influence B3

functionality and controllability

pop ulation not aware of M FC -

AC C

information campaign effectiveness

EP

information camp aign + +

++ + + intention to use

TE D

+

M AN U

334

RI PT

322

B2

B1

time to adopt M FC

+ + M FC adoption rate +

contacts p er M FC adopter

R1

pop ulation adopted M FC +

adoption fraction

344 345 346

Figure 2. Causal loop diagram for the proposed innovation diffusion model. R – positive or reinforcing loops, B – negative or balancing loops, MFC – micro-fibre clothes.

347 348 349 350 351 352 353 354 355

The reinforcing loop R1 drives the word-of-mouth effect. The more households which have adopted MFC, the higher the word-of-mouth effect, and thus the higher the rate of MFC adoption. This loop is balanced by the negative loop B1 which depletes the stock of population not aware of MFC. The higher the MFC adoption rate, the fewer households which are not aware of MFC. Four motivations overall attitudes and social influence, green norms, functionality and controllability and self-efficacy each represent one balancing loop (B2, B3, B4 and B5). The higher the number of households which are not aware of MFC, the stronger the information campaign needed to make people aware, and to increase the values of the motivations. When

394 395 396 397 398

RI PT

Stock and flow diagram

M AN U

SC

The diagram of the causal loop is converted into a mathematical model given with a stock–flow diagram, see Figure 3. While the causal loop diagram presents 6 causal loops, the stock and flow model contains 40 variables. All variable are mathematically described. The main building blocks of a system dynamics model are stocks, flows, and auxiliary variables. Stocks (represented as rectangles in Figure 3) represent accumulation in the system. Flows or rates (represented as valves in Figure 3) change the value of stocks. Stocks, in turn, influence the value of flows. Auxiliary variables (represented as circles in Figure 3) are calculations, and are computed from stocks, flows, constants, data, and other auxiliary variables. General diffusion model (Bass, 1969) is used as the basis for MFC diffusion among households where the cumulative number of MFC adopters (substituting traditional cleaning agents with MFC) follows S shape growth and reaches market saturation. Two main stocks in the diffusion model (Figure 3) are households that have and have not adopted MFC. This is where the substitution of traditional cleaning agents with MFC is modelled. Both the outflow of households who have not adopted MFC, and inflow of those who use MFC, are governed by the MFC adoption rate measured as households per year. The rate of adoption of the MFC depends on the number of households that have not adopted MFC and on the time needed for them to adopt MFC, in years. The adoption time is calculated by dividing the stock by the adoption time. The adoption time depends on the intention to use MFC – the lower the intention, the longer the adoption time. The intention to use MFC is calculated using an empirical relationship from Eq. (4) and Eq. (5). The values of estimates are taken from the logistic regression model (Table 3). In the absence of any external diffusion drivers, word-of-mouth is the only variable affecting the adoption rate. In the market-driven Bass model, the adoption from word-ofmouth is most important in the long run. It depends on the number of households who have and have not adopted MFC, the rate by which contacts are made between non-adopters and adopters (communicating positive experiences of MFC use), the fraction of those nonadopters that become MFC adopters as a result of that contact, and is calculated as Eq. (6).

TE D

366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393

5.1.2.

EP

365

ACCEPTED MANUSCRIPT

a motivation is low, the intention to use MFC is low as well. This in turn increases the time needed to adopt MFC and reduces the rate of MFC adoption. Information campaigns are leading to an increased adoption rate, and a depleted stock of the unaware population. Ozaki (2011) states that innovations and social norms affect each other – the more people adopt an innovation, the more the innovation becomes a norm. This leads to more people adopting it. Also, Peres et al. (2010) argue that social pressures affect consumers regardless of their knowledge. This link (of social norms and social pressure) is not represented in the system dynamics model as data were not available to describe it mathematically (a dotted line in Figure 2).

AC C

356 357 358 359 360 361 362 363 364

WOM = AF ⋅ CH ⋅ HH NA ⋅ HH A /( HH NA + HH A )

(6)

Where WOM – adoption from word-of-mouth, household, year; frAD – adoption fraction; CH – contacts per MFC users household, household/household/year; HHNA – households who have not adopted MFC, households; HHA – households who have adopted MFC, households.

ACCEPTED MANUSCRIPT

W eekly consumption of cleaning agents

Average annual consumption of cleaning agents per household

Adoption rate of innovations

RI PT

Households adopting innovations

Households not adopting innovations

Contacts per innovation users household Adoption fraction

W ord-of-mouth Info campaign strength overall attitude & social influence

Awareness rate overall attitude & social influence

Info campaign effect on green norms inverse

Overall attitude & social influence questions estimate

AC C

Overall attitude & social influence estimate

EP

Green norms questions estimate

Green norms estimate

Avoided cleaning agents

Avoided cleaning agents rate

SC

TE D

Awareness green norms

Info campaign effect on overall attitude & social influence

399 400

Awareness rate green norms

Adoption fraction information

Awareness overall attitude & social influence

Intention to use innovation - round

Info campaign Info campaign strength green strength selfnorms efficacy

M AN U

Time to adopt innovation

Cleaning agents used parallel to MFC

Households not adopting innovations

Awareness rate self-efficacy

Info campaign strength functionality

Awareness rate functionality Adoption fraction information Awareness self-efficacy Info campaign effect on self-efficacy inverse Self-efficacy questions estimate

Self-Efficacy estimate

Awareness functionality

Info campaign effect on functionality inverse Functionality questions estimate Functionality estimate

Intention to use innovation

Figure 3. The stock and flow diagram for the proposed innovation diffusion model (MFC – micro-fibre clothes)

ACCEPTED MANUSCRIPT The word-of-mouth effect is not sufficient to spread the news of a good product in the short term. By means of information campaigns about MFC, awareness of the product is created in the market, and some people will become aware and start adopting MFC without having encountered other customers who already use MFC and are happy with the product. The awareness of MFC is accumulated for each motivation separately, and each of them depend on the strength of the information campaign, see Eq. (7). AW =

407

438 439 440 441 442 443 444 445 446 447 448

init

(7)

RI PT

∫

t =1

t =0

i nit

(Cweek ⋅ CA ⋅ HH A )t ⋅ dt + AWCA

(8)

Where ACA – accumulated avoided traditional cleaning agents, tonnes; Cweek – weekly consumption of traditional cleaning agents, tonne/household/week; CA – fraction of traditional cleaning agents used parallel to MFC; ACAinit – initial value of accumulated avoided traditional cleaning agents, tonnes. A system dynamics model is developed using the graphical program Powersim. The simulation time step was set as one year. Major assumptions of the model

TE D

5.1.3.

Some of the major assumptions within the developed model are: • the adoption fraction of information campaigns – assumed that an information campaign can reach 10% of its target audience; • when an information campaign is highly effective, the perceived information is delayed by 6 years; • in 1% of households, MFC have been adopted at the beginning of simulation; • total number of households is 800,000; • average contacts per household which have been using MFC is 0.5 per year; • fraction of MFC adoption in households (not aware of MFC) is assumed to be 0.5 upon meeting MFC users. • average use of cleaning agents 91 g/week (Bennett et al., 2012). • average time to adopt MFC is 5 years.

EP

426 427 428 429 430 431 432 433 434 435 436 437

[WOM +AFinf ⋅ HH A /(S ⋅ T IC )]t ⋅ dt + AW

ACA =

AC C

425

t =0

Where AW – households aware of MFC, households; frINF – information adoption fraction by those non-adopters that become aware of MFC after receiving information by means of information campaign; S – strength of information campaign (ON or OFF); TIC – perceived information delay time, years; AWinit – initial value of households aware of MFC, households. Information campaigns are used as the only leverage point in the system. Other policy tools are not modelled, and the reasons are explained in Section 6.1.4. Information campaigns target four motivations identified in the logistic regression model. Therefore, the diffusion of MFC is triggered and environmental pollution is avoided due to the use of MCF. The accumulated avoided traditional cleaning agents are calculated as a given in Eq. (8).

418 419 420 421 422 423 424

t =1

SC

408 409 410 411 412 413 414 415 416 417

∫

M AN U

401 402 403 404 405 406

5.1.4.

Validation of model

The main concerns for modelling studies are data quality and availability, since no model can fully represent the system under study. In order to determine whether a model can be used for its intended use, a validation of the model must be done (Forrester, 1961). Nevertheless, the aim of system dynamics modelling is to depict the patterns of wide dynamic behaviour of the real system, not to give “point” projections (Sterman, 2000). Both structural and behavioural validation tests can be performed to validate a model. To test the reasonability of the equations within a model, a structural validation is completed where each individual equation is tested and a model is evaluated under extreme conditions. Behaviour validity tests are performed to compare patterns of the major variable as compared to the historical data (Barlas, 1996).

ACCEPTED MANUSCRIPT 449

5.1.5.

Sensitivity analysis

A sensitivity analysis was carried out to determine if the developed model is sensitive to different variables, and what variables have the highest influence on the model. In order to prepare a sensitivity analysis, a risk assessment tool within the Powersim software was used. The Latin Hypercube method was used to model variances within selected variables in the range of ± 25 % of the existing value under truncated normal distribution.

455

6. RESULTS AND DISCUSSION OF THE SYSTEM DYNAMICS MODEL

456

6.1.1.

467 468 469 470

Validation of the system dynamics model

SC

The model was validated based on structural validation protocol: the reasonability of the equations included in the model was found valid for the purposes of the study. Parameters were tested under extreme conditions, and the model was found robust. The behaviour validity tests could not been done since no historical data on the diffusion of MFC in the market were found. We also contacted the resellers of MFC; they were not able to provide data needed for a behaviour validation test. However, we did compare the results obtained from the correlation analysis and logistic regression analysis with the outputs from the model; the behaviour of consumers simulated in the system dynamics model matched with the equations obtained from the questionnaire data analysis. 6.1.2.

M AN U

457 458 459 460 461 462 463 464 465 466

Reference behaviour of the model

A reference behaviour test studies a system’s behaviour for 30 years under the initial set of data, and without any intervention of information campaigns; see Figure 4. Avoided environmental pollution, 1000 t

Households, 1000 units

800

10

5

0

473 474 475 476 477 478 479 480 481 482 483 484

10

AC C

0

EP

TE D

15

471 472

RI PT

450 451 452 453 454

600 400

Households using MFC

200 0

20 Simulation year

Avoided envionmental pollution

Households not using MFC

30

Figure 4. Reference behaviour of the proposed innovation diffusion model

The only driving force for the diffusion of MFC in households is the effect of word-ofmouth – a household member sharing their experience in a social cycle. This diffusion model is classified as one of the technology transition models by Zeppini et al. (2013). The results show a relatively weak diffusion process at the beginning of the simulation because the wordof-mouth effect alone is only effective in the long run. The total accumulated avoided environmental pollution by not using traditional cleaning agents reaches almost 13,000 tonnes in the 30th year. 6.1.3.

Sensitivity analysis

A sensitivity analysis shows which variables have the highest influence on the model. In total, 9 variables were chosen for examination where variables were set to vary according to a statistical distribution. These variables are: adoption fraction of information from campaigns, the strength of information campaigns, adoption fraction of information received by word-of-

ACCEPTED MANUSCRIPT 485 486 487 488 489

mouth, the amount of cleaning agents used in parallel to MFC, the weekly consumption of cleaning agents, contacts per MFC users’ household, and the amount of households already using MFC at the beginning of the simulation. These variables account for those that were introduced as the assumptions. The number of simulations was set to 50 simulation sets. The results of the sensitivity analysis are given in Figure 5. Avoided environmental pollution, 1000 t 250 High 90 Percentile

RI PT

200

75 Percentile

150

25 Percentile 10 Percentile

100

50 Percentile

50 0 0

10 20 Simulation year

SC

Low

30

Figure 5. The results of the sensitivity analysis for ± 25% changes in the value of parameters the proposed innovation diffusion model

493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508

The sensitivity analysis for the selected parameters changed the total cumulative amount of avoided environmental pollution to the environment by – 29.88% at the lowest estimate and by + 50.59% at the highest estimate in the 30th simulation year. As given in Figure 5, the models showed expected behaviour with increasing uncertainty in the future. The adoption faction of information received by word-of-mouth, households adopted MFC at the beginning of the simulation and the contacts per household were found to be the most sensitive parameters in the short term, with decreasing importance in the long term, since the effect of word-of-mouth is the main driver for the diffusion of MFC. The opposite was found for the sensitivity of the adoption fraction of information from campaigns, the strength of information campaigns, and the weekly consumption of cleaning agents in households: these parameters have a low impact in the near term with increasing uncertainty in the future. This trend is due to the small number of households willing to adopt MFC at the beginning of the simulation. The least sensitive variable was cleaning agents used in parallel with MFC, due to the fact that the number of households adopting MFC is of greater importance, the remaining cleaning agents represent relatively low amounts in the households using MFC.

510 511 512 513 514 515 516 517 518

TE D

EP

AC C

509

M AN U

490 491 492

6.1.4.

Scenario analysis

Only information campaigns were analysed under a scenario analysis, since Steg et al. (2014) state that activities targeted to reduce the costs of some particular proenvironmental activities (for example, giving out MFC for free or subsidizing their introduction into households) or to increase the costs of environmentally harmful behaviour (increasing the tax on cleaning agents) have only a short-term effect – in the long term, households will return to their previous cleaning practices. Therefore, we have studied the effects of information campaigns targeted to the specific motivations: overall attitudes and social influence (OASI), functionality and controllability (FC), self-efficacy (SE) and green norms (GN); see Figure 6.

Households 807878

OASI+FC+SE

807878

OASI+FC+GN

807841

OASI+SE

807841

OASI+FC

807841

OASI+SE+GN

807763

OASI+GN

807763

GN+SE

376683

GN+FC

376683

SE+FC

376683

OASI

376683

GN

376683

SE

376683

FC

376683

None

376683

Households using MCF, 1000 units 800 600

RI PT

Scenarios All

400 200 0 0

10 20 Simulation year II trend III trend

SC

IV trend

III trend

II trend

I trend

ACCEPTED MANUSCRIPT

I trend

30

IV trend

Figure 6. The results of the scenario analysis under various information campaigns the proposed innovation diffusion model. Overall attitudes and social influence (OASI), functionality and controllability (FC), self-efficacy (SE) and green norms (GN).

523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555

In our case study, the overall attitudes and social influence is found to have the most influence on the diffusion rate of MFC, given by the combination of the functionality and controllability and self-efficacy. The most prominent is overall attitudes and social influence, which also includes social acceptance. This finding is directly supported by Ozaki (2011), where signing for “green” electricity is also strongly dictated by social pressure. Moreover this result is indirectly supported by the Eurobarometer opinion polls. When thinking about factors influencing their quality of life, 67 % of Latvians think about the environmental issues (the European average being 75 %). When asked about the factors influencing an individual’s choice of a particular good, 54 % of Latvians report that ‘brand’ or ‘brand-name’ is a significant factor – vis-à-vis only 39 % for the European average (Flash Eurobarometer, 2009). These results demonstrate that social acceptance is an important source of motivation for consumer choice in Latvia. These data correspond with our findings that social influence, play a significant role in the use of MFC. This conclusion also has noteworthy consequences for governmental policy. In order to promote eco-friendly products, it would be necessary to emphasize the element of social acceptance. The use of MFC might be increased by emphasizing their socially accepted and well-regarded image in the eyes of one’s own significant others, e.g., “it is important for your family and friends that you are using this product.” The all motivations alone is found not effective, the same conclusions have been found by Abrahamse et al. (2005) – the rewards have encouraged a change in the behaviour, but only in the short term. Regarding the strategy aimed to increase the level of knowledge alone, it should be mentioned that information alone is not an effective strategy to promote changes of behaviour; since information leads to higher knowledge levels, but not necessarily to changes in behaviour (Abrahamse et al., 2005). The effect of information campaigns on the diffusion speed has also been found in a study by Schwarz and Ernst (2009), but only as the third most effective measure, preceded by regulations and subsidies. There is also evidence that environmental concerns do not directly lead to pro-environmental actions (Bamberg, 2003). We see that the main value of the proposed modelling framework is a conceptual model for the studies on eco-innovation diffusion and, consequently, the reduction of environmental pollution based on consumer choice. The proposed approach can also be used as the basis for other studies of other diffusion processes, since system dynamics modelling is a white-box

AC C

EP

TE D

M AN U

519 520 521 522

ACCEPTED MANUSCRIPT modelling approach and the structure of the model can be enhanced with the latest results of studies. The current work has a set of limitations that we were not able to address in our study: firstly, the behavioural validation with historical data was not possible as no data was found in the literature or at the resellers of MFC; secondly, the effect of information campaigns on motivations was expressed mathematically with an error function, since specific mathematical expressions from previous studies were unavailable’; finally, in the literature a link between social pressure and diffusion speed is discussed, but to our knowledge no mathematical expressions of this phenomena have been developed. To sum up, our model is able to shed light on the questions – how eco-innovation diffusion happens and what impacts the speed of diffusion, but further research is needed to address stated limitations. We see that our developed system dynamics model could be a platform for the further incorporation of the results from those fields of research.

569

CONCLUSIONS

570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606

The ultimate goal of this research was to develop a conceptual model for eco-innovation diffusion based on consumer choice. This study sheds light on the explanation of proenvironmental behaviour. The diffusion of micro-fibre clothes (MFC) in Latvia for cleaning purposes in households was selected as a case study. In this research, eco-innovations were studied in their broader and more comprehensive view: the substitution of a product is made by an ecoinnovation, which performs the same generic function, but the product itself is very different – thus achieving absolute reductions. The proposed conceptual model studies the influence of information campaigns to reinforce adoption of the studied eco-innovation products. The questionnaire was carried out to obtain empirical data for the conceptual model based on questionnaire answers of microfibre clothes’ users and aware non-users. This novel methodology has been developed under the system dynamics framework and is coupled with the logistic regression model. The model allows for the outlining of the dynamics under technological substitution within the context of the eco-innovations’ diffusion. Structural validation tests were done for the proposed model. While the questionnaire allowed for a detailed assessment of the respondents motivations, the system dynamics model provided the conceptual model. This model is used to analyse the structure of the problem under study; therefore, understanding the causes of the system’s behaviour as well as helping to determine the action plan for improvements. The results for the logistic regression models show that the strongest impact on the choice is by the constant term and overall attitudes and social influence, which also includes social acceptance. This result is indirectly supported by other data concerning Latvians’ attitude towards the environment. Therefore, in order to promote eco-friendly products, it is necessary to emphasize the element of social acceptance. The simulation results of information campaigns (used as policy tools) shows that campaigns targeting the overall attitudes and social influence in combination with the functionality and controllability and self-efficacy have triggered the intention to use MFC with the highest values. Therefore, by including aspects of social acceptance within information campaigns, a new habit of using micro-fibre clothes can be introduced. The most sensitive parameters in the system dynamics model were found to be the adoption faction of information received by word-of-mouth, households adopting micro-fibre cloth at the beginning of simulation, and the contacts per household. The proposed methodology and developed conceptual system dynamics model can also be used as the basis for other studies of various diffusion processes, since system dynamics modelling is a white-box modelling approach, and the structure of the model can be enhanced with the latest results from other studies.

AC C

EP

TE D

M AN U

SC

RI PT

556 557 558 559 560 561 562 563 564 565 566 567 568

ACCEPTED MANUSCRIPT ACKNOWLEDGEMENTS

608 609 610 611 612

First of all, the authors would like to gratefully acknowledge anonymous reviewers for their constructive comments on improving an early version of this manuscript. The authors also express an acknowledgement to Mr. Geoffrey Thorpe for his help with proofreading. Support for this work was provided by the Riga Technical University through the Scientific Research Project Competition for Young Researchers No. ZP-2014/5.

613

REFERENCES

614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664

Abrahamse W., Steg L., Vlek C., Rothengatter T. 2005. A review of intervention studies aimed at household energy conservation, Journal of Environmental Psychology 25, 273-291. Ajzen, I. 1991. The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50, 179-211. Ajzen I., Fishbein M. 1980. Understanding Attitudes and Predicting Social Behaviour. PrenticeHall:Englewood-Cliffs, NJ. Bamberg S. 2003. How does environmental concern influence specific environmentally related behaviors? A new answer to an old question, Journal of Environmental Psychology 23, 21-32. Barlas, Y. 1996. Formal aspects of model validity and validation in System Dynamics, System Dynamics Review 12, 183–210. Bass F.M. (1969). A new product growth model for consumer durables. Management Science 15: 215227. Bennett D.H., Wu M.X., Teague C.H., Lee K., Cassady L.D., Ritz B., Hertz-Picciotto I. 2012. Passive sampling methods to determine household and personal care product use, Journal of Exposure Science and Environmental Epidemiology 22, 148-160. CSB (Central Statistical Bureau of Latvia) 2014.Databases. EIO (Eco-innovation observatory). 2011. The Eco-Innovation Challenge: Pathways to a resourceefficient Europe. Eco-Innovation Observatory. Funded by the European Commission, DG Environment, Brussels. Flash Eurobarometer 2009. Flash Eurobarometer 256. Europeans’ Attitudes towards Sustainable Consumption and Production. ec.europa.eu/public_opinion/flash/fl_256_en.pdf Forrester, J.W. 1961. Industrial dynamics. Pegasus Communications, Waltham, USA, 646 pp. Forrester, J.W. 1958. Industrial dynamics: a major breakthrough for decision makers, Harvard business review 36 (4), 37–66. Ghisetti C., Rennings K. 2014. Environmental innovations and profitability: how does it pay to be green? An empirical analysis on the German innovation survey, Journal of Cleaner Production, Article in Press. Govindan K., Diabat A., Shankar K.M. 2014. Analyzing the drivers of green manufacturing with fuzzy approach, Journal of Cleaner Production, Accepted Manuscript. Hair J., Anderson R., Tatham R., Black W. 1998. Multivariate Analysis. Prentice Hall. Humphreys P.N., Hook J., Rout S. 2012. Evaluation of the cleaning efficiency of microfibre cloths processed via an ozonated laundry system, Journal of Infection Prevention 13(4), 104-108 Johnson R.A., Wichern D.W. 2002. Applied multivariate statistical analysis, Prentice Hall, p. 767 Karakaya E., Hidalgo A., Nuur C. 2014. Diffusion of eco-innovation: A review, Renewable and Sustainable Energy Reviews 33, 392-399. Lin M.S., Wu F.S. 2013. Identifying the determinants of broadband adoption by diffusion stage in OECD countries, Telecommunications Policy 37, 241-251. Lindenberg S., Steg L. 2007. Normative, Gain and Hedonic Goal Frames Guiding Environmental Behavior, Journal of Social Issues 63(1), 117-137. Meade N., Islam T. 2006. Modelling and forecasting the diffusion of innovation – a 25-year review. International Journal of Forecasting 22, 519-545. de Medeiros J.F., Duarte Ribeiro J.L., Nogueira Cortimiglia M. 2014. Success factors for environmentally sustainable product innovation: a systematic literature review, Journal of Cleaner Production 65, 76-86. Nilsen S.K., Dahl I., Jørgensen O., Schneider T. 2002. Micro-fibre and ultra-micro-fibre cloths, their physical characteristics, cleaning effect, abrasion on surfaces, friction, and wear resistance. Building and Environment 37, 1373-1378. Ozaki R. 2011. Adopting sustainable innovation: what makes consumers sign up to green electricity? Business Strategy and the Environment 20, 1-17. Peres R., Muller E., Mahajan V. 2010.Innovation diffusion and new product growth models: A critical review and research directions. International Journal of Research in Marketing 27, 91-106.

AC C

EP

TE D

M AN U

SC

RI PT

607

ACCEPTED MANUSCRIPT

EP

TE D

M AN U

SC

RI PT

Plötz P., Gnann T., Weitschel M. 2014. Modelling market diffusion of electric vehicles with real world driving data — Part I: Model structure and validation. Ecological Economics 107, 411-421. Porter M.E. 1985. Competitive Advantage: Creating and Sustaining Superior Performance. The Free Press: New York. Rennings K. 2000. Redefining innovation – eco-innovations research and the contribution from ecological economics, Ecological Economics 32, 319-332. Rixen M. and Weigand J. 2014. Agent-based simulation of policy induced diffusion of smart meters. Technological Forecasting & Social Change 85, 153-167. Rogers EM. 2003. Diffusion of innovations, 5th edn. Free Press: New York. Schwartz, S. H. 1977. Normative influences on altruism. In L. Berkowitz (Ed.). Advances in experimental social psychology (Vol. 10, pp. 221–279). New York: Academic Press] Schwarz N., Ernst A. 2009. Agent-based modeling of the diffusion of environmental innovations – An empirical approach, Technological Forecasting & Social Change 76, 497-511. Smith D.L., Gillanders S., Holah J.T., Gush C. 2011. Assessing the efficacy of different microfibre cloths at removing surface micro-organisms associated with healthcare-associated infections, Journal of Hospital Infection 78, 182-186. Steg L., Bolderdijk J.W., Keizer K., Perlaviciute G. 2014. An Integrated Framework for Encouraging Pro-environmental Behaviour: The role of values, situational factors and goals, Journal of Environmental Psychology 38, 104-115. Sterman, J.D. 2000. Business dynamics: systems thinking and modelling for a complex world, Irwin/McGrw-Hill, Boston 982 pp. Stern P.C., Dietz T., Guagnano G.A., Kalof L. 1999. A value-belief-norm theory of social support for social movements: the case of environmentalism. Human Ecology Review 6(2): 81-97. Sushandoyo D., Magnusson T. 2014. Strategic niche management form a business perspective: taking cleaner vehicle technologies from prototype to series production, Journal of Cleaner Production, Article in Press. Tonglet M., Phillips P.S., Bates M. 2004. Determining the drivers for householder pro-environmental behaviour: waste minimization compared to recycling, Resources, Conservation and Recycling 42, 27-48. UC Davis Medical Center 2006. UC Davis pioneers use of microfiber mops in hospitals, UC Davis Health Systems, News. Available: http://www.ucdmc.ucdavis.edu/publish/news/newsroom/2379 Zeppini P., Frenken K., Kupers R. 2013. Thresholds models for technological transitions, Environmental Innovation and Societal Transitions, Article in Press. Wren M.W.D., Rollins M.S.M., Jeanes A., Hall T.J., Coën P.G., Gant V.A. 2008. Removing bacteria from hospital surfaces: a laboratory comparison of ultramicrofibre and standard cloths, Journal of Hospital Infection 70, 265-271.

AC C

665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700