Outlining Innovation Diffusion Processes in Households Using System Dynamics

Available online at www.sciencedirect.com ScienceDirect

Energy Procedia 75 ( 2015 ) 2859 – 2864

The 7th International Conference on Applied Energy – ICAE2015 Outlining innovation diffusion processes in households using system dynamics. Case study: energy efficiency lighting.

Lelde Timmaa*, Uldis Barissa, Andra Blumbergaa, Dagnija Blumbergaa

a Institute of Energy Systems and Environment, Riga Technical University, 12/1Āzenes iela, Riga, LV1048, Latvia

Abstract

Various innovations have reached a mature state, but they diffuse slowly. The aim of our research is to propose the model of innovation diffusion for energy efficiency solutions in households. The methodology used combines empirical study with system dynamics modelling. As the case study light emitting diodes (LEDs) is studied. Although the system dynamics model was based on the one specific innovation diffusion process, its general application to other products and services is possible, since the developed model is white-box and can be used for further research of other processes.

© 20152015 The The Authors. Authors. Published Published by Elsevier by Elsevier Ltd. This Ltd. is an open access article under the CC BY-NC-ND license Selection(http://creativecommons.org/licenses/by-nc-nd/4.0/ and/or peer-review under responsibility). of ICAE Peer-review under responsibility of Applied Energy Innovation Institute

Keywords: behavioural science; demand side management, households energy use, innovation diffusion, technology adoption, system dynamics

1. Introduction

Energy efficiency in households plays a vital role, since the households in the EU-27 consumed 29.7 % of final electricity consumption in 2010 [1]. Specific electricity consumption in the households in Latvia is rising; the amount of households consuming more than 2000 kWh per month has increased from 13 % of total households in 2001 un to 38 % in 2010 [2]. To confront these trends various energy efficient technologies, energy conservation tools and policy initiatives are available at the market, nevertheless the diffusion rate of these tools and technologies is slow and downplayed by the rebound effect. These challenges to diffuse new technologies to the market is outlined by review of Karakaya [3]. The study by Frederiks et al. [4] has shown that essential variable – the behavior of consumers – has been neglected in the models of energy use in households. In the

* Corresponding author. Tel.: +371 67089943; fax: +371 67089908. E-mail address: [email protected].

1876-6102 © 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of Applied Energy Innovation Institute doi: 10.1016/j.egypro.2015.07.574 2860 Lelde Timma et al. / Energy Procedia 75 ( 2015 ) 2859 – 2864

literature general models for innovation diffusion are given by Bass [5] and Rogers [6], but these models do not study the specific factors affecting the adoption of new technologies – the coefficients for the diffusion equation are exogenous parameters. Since the affecting factors are not explored, the diffusion of innovations cannot be triggered purposefully. Therefore the aim of our research is to outline the diffusion dynamics, by the model of innovation diffusion for energy efficiency solutions in households. The methodology used is developed by the co- authors and combines empirical study with system dynamics modelling. As the case study the adoption intention of energy efficient lighting solutions – light emitting diodes (LEDs) is studied. Case study place is Latvia. The empirical part includes the development of mathematical model based on the results obtained from the questionnaire. This questionnaire is adopted for the study on LEDs diffusion from a similar questionnaire used by co-authors in the studies on the adoption of energy efficiency measures [7] and eco- innovations [8]. The system dynamics model used in this study will be adopted and upgraded from the co- authors study on the diffusion of eco-innovations [8]. In other words, we will be extending the existing model to simulate the diffusion of energy efficiency measures in households. Due to authors knowledge there are no such research done before. Moreover the novelty of our research lies in the fact that we will be also testing following hypothesis. Firstly, based on Rogers [6] the diffusion of innovations takes place when the knowledge about the product or service is obtained. Secondly, on the other side, Peres et al. [9] states that social pressure affects the consumers with or without their knowledge. Therefore we aim to search for the mathematical, statistically significant relationship between the variables in given hypothesis and the adoption intention of LEDs. These hypotheses have been discussed in previous research [8], but they have not been included into the model. The obtained mathematical relations will be incorporated into system dynamics model. Although the system dynamics model was based on the specific innovation diffusion process, its general application to other products and services is possible, since developed model is a white-box. The structure of the model can be used for other researchers; the model can be enhanced with the newest results or adopted for other case studies.

2. Methodology

The methodology used is developed by the co-authors of this paper and the first application of this methodology is presented by Vīgants et al. [8]. This work combines empirical study with system dynamics modelling. Initially the dynamic problem and hypothesis is developed. The empirical study starts with the design of questionnaire and collection of responses. The responses are processed with correlation and logistic regression analysis. Later the results of empirical study are used for the formulation of system dynamics model. This model undergoes validation, sensitivity analysis and policy tools are applied for scenario analysis.

2.1. Empirical study

Goal framing theory by Lindenberg & Steg [10] is used in this study. This framework explains the behavior motivation behind consumers’ decisions. These decisions classifies into 3 goal frames: hedonic, gain and normative. These frames serves as motives for pro-environmental action, for example, consumer can choose to use LEDs lighting because it save electricity bills or choose not to use because LEDs lighting relative cost (gain goal), because all of his friend use LEDs (hedonic goal) or because believes in the need to save energy resources (normative goal). These various goal frames can interact with each other and change the dominance in the function of time – some became focal or central, other leave in the Lelde Timma et al. / Energy Procedia 75 ( 2015 ) 2859 – 2864 2861 background. In our case study we explore the intention to adopt LED lighting solutions at the households. We study the importance of each goal frame when the decision is made on the purchase of LEDs.

2.1.1. Questionnaire design and data collection The questions regarding gain, hedonic and normative goal frames and on the demographics of respondents where included into the questionnaire. The overall design was adopted from Ozaki [11]. These three distinct study groups where: employees of the electricity generation utility in Latvia, employees of the Riga Technical University and general public. The links to the on-line questionnaire was distributed via e-mail among the employees and posted on-line for general public.

2.1.2. Questionnaire data analysis with correlation analysis The aim of correlation analysis was to explore with of goal frames might affect the willingness to adopt LEDs and are there some statistical differences between the behaviors of various surveyed groups. The correlation coefficient (R) and mean value was calculated for each independent variable at 95 % confidence level for both groups (users and non-users), thus allowing to compare statistically different answers across these two groups. The interpretation of the correlation coefficient was adopted from Ozaki [11] with the reference to Cohen [12]; where R in the boundaries ± 0.1-0.3 is referred as small correlations for behavioural studies, ± 0.3-0.5 medium, ±0.5-1.0 large correlations.

2.1.3. Questionnaire data analysis with logistic regression analysis The logistic regression analysis was preformed to fit a regression model in which the dependent variable characterizes and event with only two possible outcomes: use LED or not to use LED. The intention was expressed as answer to the question „use of LED was a good idea” (for users of MFC) and „use of LED would be a good idea” (for non-users). The fitted model relates predictor variables and assumes the probability of an event trough a logistic function, see Eq. (1).

log>@P(X) (1 P(X)) exp(E0 E1X1 E2 X2 ... Ek Xk ) (1)

Where P is the probability of 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. The obtained model was used to determine the combination of factors which explains the intention to use LED. To estimate the accuracy of the fitted model the percentage of deviance explained by the model (R2) was calculated in similar manner to R-squared statistics in multiple regression; see Eq. (2).

2 R O(E1, E2,Ek E0 ) O(E0 ) (2)

2 And adjusted deviance (R adj) in similar manner to R-squared adjusted; see Eq. (3).

2 R adj [O(E1, E2,Ek E0 ) 2p] O(E0 ) (3)

Where p equals the number of coefficients in the fitted model, including constant term. The analysis was done by interactive statistical data analysis tool Statgraphics Centurion 16.1.15.

2862 Lelde Timma et al. / Energy Procedia 75 ( 2015 ) 2859 – 2864

2.2. System dynamics model

System dynamics methodology deals with problems in complex systems having delays, non-linearities and feedbacks. This methodology was introduced by Forrester [13] and allows outlining the relations between physical activities, information flows and policy measures. System dynamics converts results obtained from empirical study into a mathematical model of innovations diffusion to allow for prediction of a system’s behavior over time. see Fig.1.

Households not Households adopting innovations adopting innovations

Adoption rate of innovations Contacts per innovation users household

Word-of-mouth Adoption fraction Time to adopt Info campaign Info campaign innovation Info campaign strength strength strength hedonic goal gain goal normative goal

Adoption fraction Awareness rate Awareness rate Awareness rate hedonic goal information gain goal normative goal Adoption fraction Awareness Awareness information gain goal normative goal Awareness Info campaign hedonic goal effect on gain goal Info campaign effect Info campaign effect on normative goal on gain goal inverse inverse Info campaign Allows Info campaign effect effect on housekeeping on normative goal hedonic goal value Costs value Standard of living Allows Water pollution value housekeeping value estimate Costs Standard of estimate living estimate Water pollution estimate Hedonic Gain goal goal value estimate

Normative goal estimate Intention to use Hedonic goal innovation estimate

Fig.1. The stock and flow diagram for the proposed innovation diffusion model [8]

General diffusion model by Bass [5] is used as the basis for LEDs diffusion among households where the cumulative number of LEDs adopters follows S shape growth and reaches market saturation. The Bass model takes into account the existence of persons, who will not adopt innovations at any circumstances; the innovation diffusion model has shape of asymptotic time series. In the absence of external diffusion drivers, word of mouth is the only variable affecting diffusion rate. In the market-driven Bass model, the Lelde Timma et al. / Energy Procedia 75 ( 2015 ) 2859 – 2864 2863 adoption from word-of-mouth is the most important in the long run. It depends on the number of households who have and have not adopted LEDs, rate by which contacts are made between non-adopters and adopters (communicating positive experiences of LEDs use), the fraction of those non-adopters that become LEDs adopters as a result of that contact and is calculated as Eq. (4).

WOM AF CH HHNA HHA /(HHNA HHA) (4)

Where WOM – adoption from word of mouth, household, year; AF – adoption fraction; CH – contacts per LEDs users household, household/household/year; HHNA – households who have not adopted LEDs, households; HHA – households who have adopted LEDs, households. Word of mouth effect is not sufficient to spread the news of a good product in a short term. By means of information campaigns about LEDs, awareness of the product is created in the market. For information campaign strength, see Eq. (5).

t 1 init AW [WOM AFinf HH A /(S TIC )]t dt AW (5) ³t 0

Where AW – households aware of LEDs, households; AFinf – information adoption fraction by those non-adopters that become aware of LEDs 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 LEDs, households. For more details on equations used in system dynamics model (Fig.1) please see the Reference [8].

3. Results and discussion

The results of the research using proposed framework for empirical study are given in Table 1.

Table 1: Logistic regression statistics for the models Adoption of energy efficiency [12] Adoption of eco-innovation [9] Deviance Model Deviance explained by Deviance explained Deviance explained by explained by model, adjusted % (pa) by model, % model, adjusted % (pa) model, % Normative 29.027 **** 5.071 (27) 20.001*** 7.012 (11) goal Gain goal 21.139 ** 2.307 (18) 31.694**** 11.365 (16) Hedonic goal 20.112 **** 5.916 (16) 28.532**** 20.342 (6) Intention to 51.369 **** 13.047 (37) 67.481**** 33.473 (27) adopt ap = the number of coefficients in the fitted model, including the constant term; bf = the number of factors in the fitted model. **Significant at P < 0.01; ***Significant at P < 0.001; ****Significant at P < 0.0001

The results for the logistic regression models shows that gain goals explained more of the deviance for the study on eco-innovation diffusion [9] on the other side normative goal explained the energy efficiency motivation [12]. The final model “Intention to adopt energy efficiency” (which combines normative, gain and hedonic goals together) explained 13.047 % of adjusted deviance at 95 % confidence level [9] and for the model “Intention to adopt eco-innovation” – 33.473 % of adjusted deviance. Study by Tonglet et al. [14] found 33.3 % of adjusted deviance for the intention to recycle. We hypothesize that the results for the logistic regression model “Intention to adopt LEDs” could be dominated by the gain goal. Further research is needed to obtain mathematical interpretation of the questionnaire results. 2864 Lelde Timma et al. / Energy Procedia 75 ( 2015 ) 2859 – 2864

As for now the questionnaire and the conceptual system dynamics model (Fig.1.) is finished, further research is carried out to obtained mathematical expressions from the questionnaire. These expressions will allow system dynamics model to simulate the adoption intention of LEDs in households. The results obtained from simulations we would like to include in publication and share at the conference.

4. Conclusions

The developed model simulates the diffusion of energy efficient lighting solution in households. The methodology used combines empirical study with system dynamics modelling. As the case study light emitting diodes (LEDs) is studied. Although the system dynamics model was based on the one specific innovation diffusion process, its general application to other products and services is possible, since the developed model is white-box and can be used for further research of other processes.

Acknowledgements

The work has been supported by National Research Program “Energy efficient and low-carbon solutions for a secure, sustainable and climate variability reducing energy supply (LATENERGI)”.

References [1] European Commision, Energy Efficiency Status Report 2012. Publications Office of the European Union: Luxembourg; 2012. [2] Central Statistical Bureau of Latvia, Databases; 2014. [3] Karakaya E, Hidalgo A, Nuur C. Diffusion of eco-innovation: A review. Renew Sust Energ Rev 2014;33:392-9. [4] Frederiks ER, Stenner K, Hobman EV. Household energy use: Applying behavioural economic to understand consumer decision-making and behaviour, Renew Sust Energ Rev 2015;41:1385-94. [5] Bass FM. A new product growth model for consumer durables. Management Science 1969;15:215-27. [6] Rogers EM. Diffusion of innovations, 5th Ed. Free Press: New York; 2003. [7] Bariss U, Dandens A, Timma L, Blumberga A, Blumberga D. How to assess involvement of electricity end user in energy efficiency improvement - Analysis of survey results. Energy Procedia, 2015; Article in Press. [8] Vīgants E, Blumberga A, Ījabs I, Timma L. The dynamics of technological substitution: the case of eco-innovations diffusion, J Clean Prod, under peer-review. [9] Peres R, Muller E, Mahajan V. Innovation diffusion and new product growth models: A critical review and research directions. Int J Market Res 2010;27:91-106. [10] Lindenberg S, Steg L. Normative, Gain and Hedonic Goal Frames Guiding Environmental Behavior. J. Soc. Issues 2007;63,117-37. [11] Ozaki R. Adopting sustainable innovation: what makes consumers sign up to green electricity? Bus Strat Environ 2011;20:1-17. [12] Cohen J. Statistical Power Analysis for the Behavioural Science, 2nd Ed. Erlbaum: Hillsdale, NJ; 1988. [13] Forrester JW. Industrial dynamics: a major breakthrough for decision makers, Harvard business review 1958;36:37–66. [14] Tonglet M, Phillips PS, Bates M. Determining the drivers for householder pro-environmental behaviour: waste minimization compared to recycling, Resour Conserv Recy 2004;42:27-48.

Lelde Timma Doctoral Researcher at Institute of Energy Systems and Environment. She works on innovation diffusion, renewable energy and artificial intelligence. She studied in Sweden, Switzerland and Lithuania; defended a double-diploma degree in 2013. She is the co-author of > 40 publications and participated in > 20 international scientific conferences.