the complexety of ('green') material selection by companies

(1)

THE COMPLEXITY OF (‘GREEN’)

MATERIAL SELECTION BY

COMPANIES

AN INTERDISCIPLINARY RESEARCH ON (SUSTAINABLE) PRODUCTION,

CONTAINING INSIGHTS FROM THE FIELDS OF POLITICAL SCIENCE,

ECONOMICS, PSYCHOLOGY AND ARTIFICIAL INTELLIGENCE

FIGURE. THE BIRTH OF ARTIFICIAL INTELLIGENCE. RETRIEVED NOVEMBER 16, 2016, FROM: HTTP://BLOG.TIMESUNION.COM/OPINION/FILES/2011/10/1101_WVROBOTS.JPG

NAME

STUDENT NUMBER

Marian Kes 10746439 Emma Voncken 10738339 Tim van Loenhout 10741577

(2)

ABSTRACT

In this interdisciplinary research report will be discussed how governmental regulations, consumer demand and costs contribute to the complexity of a company’s green material choice, and how machine learning could facilitate in making this choice. In the theoretical section is argued that the three aforementioned main elements influencing material choice form a complex system. Afterwards, these three elements are further discussed. First of all, the influence of governmental regulations on recycling costs for a firm is elaborated on. As recycling costs are not the only costs that the firm has to bear in mind when choosing a particular material. In the second section of the theoretical framework, the other costs of a firm that influence material choice are discussed. In the last section the values which cause consumers to buy certain products or want particular materials are analyzed. Subsequently, the value of machine learning for facilitating the complexity of material choice for companies is discussed. Lastly, a material ranking tool is developed. This tool can help companies to make informed decisions on which materials they should choose, depending on how important they value complying to ‘legislation’, ‘consumer demand’ and ‘economic costs’.

INTRODUCTION

In 2050, 9 billion people need enough resources to live in prosperity, whereas resources are becoming scarcer (Ministerie van Infrastructuur en Milieu, 2016). On September 14, 2016 the programme of the Dutch government for a circular economy, the ‘Rijksbrede programma Circulaire Economie’, was published (ibid.). It states that resources should be reused and used utmost efficiently (ibid.): The government’s ambition is to have a circular economy by 2050. Over the past years, diverse companies, sometimes with governmental assistance, have developed effective and at the same time profitable ways to deal with scarcity of resources (Ministerie van Infrastructuur en Milieu, 2016). According to the ‘Rijksbrede programma Circulaire Economie 2016’ a change in business models can bring about opportunities for companies. Yet, a company who considers ‘greening’ its business also faces investment risks. As a consequence, companies fear for losing their competitiveness when they design more sustainable products (Nidumolu et al., 2009). This is because they think the costs will outweigh the benefits. Therefore, it is of utmost importance for a sustainable transition that corporations are able to recognize the benefits of implementing ‘eco-design’ measures for their business model.1_{In order to get an overview of this cost-benefit analysis, a firm needs to know the costs} and benefits of a certain material. Consequently, it can make an informed decision on whether it should or should not choose ‘green’2_{materials for their products.} The drivers for material choice by companies can be seen as a complex system. This will be elaborated on in the section (‘Complexity’). In this report three main elements of material choice will be discussed: governmental policies which attempt to promote ‘eco-design’ by 1_{‘Eco-design’ means that products and the used materials can be easily recycled or have longer} durability. 2_{The Organization for Economic Cooperation and Development (OECD, 2009 in Lin et al., 2012) states} that ‘green’ products ‘’prevent, limit, reduce or correct harmful environmental impacts on water, air, and soil’’.

(4)

companies; economic costs of a firm and consumer demand. In order to facilitate the complexity of material choice by corporations, this research report will portray the value of machine learning and a developed material-ranking tool. Therefore, the research question is: ‘How do governmental regulations, consumer demand and costs contribute to the complexity of a company’s green material choice, and how could machine learning facilitate in making this choice?’ In this report, first of all, the question of complexity, and why it applies to material choice will be discussed. The three important drivers of material choice by companies and their interconnections will be mentioned briefly in this section. After the first (short) complexity analysis, there will be an in-depth discussion of the three important drivers of material choice by companies. This theoretical section consists of three sub-questions: 1.1 What is the influence of (supra) national legislation on (‘green’) material selection by firms? 1.2 How do economic costs influence the material selection of firms? 1.3 How does consumer demand influence material selection of firms? Subsequently, a more elaborate complexity analysis of the drivers will be made, derived from insights of the previous in-depth analyses of the three main drivers. Thereafter, the value of machine learning and its potential to cope with this complexity will be discussed elaborately in the method-section. This leads to the setup of a tool that could facilitate companing during their material selection process. So this section attempts to answer the following sub-question: ‘1.4 How could machine learning facilitate the material choice by firms?’. Finally, conclusions are stated on how machine learning and the material-selection tool can facilitate the complexity of material selection by firms.

INTERDISCIPLINARITY

Insights in why and how the government intervenes, why a consumer wants a certain product/material and what makes up the economic costs of materials for a firm can not be derived from one single discipline. Therefore, the analyses are derived from theories from the fields of Political Science, Psychology and Economics. The influence of (supra) national legislation will be analyzed from political science’s and economical insights. Furthermore, the influence of consumer demand on material selection of firms and why consumers have certain demands, will be explained by economical and psychological factors. Moreover, the question of how economic costs influence the material selection firms, are answered through theories from the field of Economics. The different disciplines each thus answer all of the first three sub-questions, and insights from the field of Artificial Intelligence will take the conclusions from these FIGURE 1. INTEGRATION OF DISCIPLINES

(5)

disciplines into account when addressing the sub question on how machine learning can facilitate the complexity of material choice by corporations. The data and the theories from the different disciplines are necessary, since they complement each other. Moreover, the integration of the results of the disciplines helps to gain understanding of the elements that make up the complexity of material choice and their interactions. As to address the complexity of material choice, this research is thus conducted in an interdisciplinary way.3_{The picture on the left} portrays how the integration of insights of the different factors, through machine learning, helps to gain understanding in how to chose materials. Machine learning is a technique from the field of artificial intelligence in which a computer learns to derive hidden patterns from a system using a self-learning algorithm. Such an algorithm could therefore potentially contribute to the understanding of this complex system of material selection. In the subsequent sections this will be discussed more elaborately. Now, first of all, why is material choice a complex problem?

THEORETICAL FRAMEWORK

COMPLEXITY

The elements which influence the material choice-process by companies can be seen as a complex system, as stated in the introduction. A system is complex when it meets four requirements: 1) there must be several different actors 2) the actors must be interconnected and they influence/are influenced by each other’s behavior 3) because of feedback- or forward loops between the actors, the actors are likely to self-organize 4) the actors should have the possibility to adapt to change or to learn (Holland, 1998; Levin 1998; and Page, 2010 in Rutting et al., 2016). When comparing the aforementioned requirements with the elements of material choice, it can be concluded that the actors and elements which determine material choice form a complex system. In the image below can be seen that the main elements are: regulations by the government, economic costs for a firm, demand of consumers and environmental impact of materials. The main actors can be derived from these elements, namely: the government which imposes the legislation upon the producers, the producers, and lastly the consumer. Consequently, the complex system of material choice meets requirement 1: it consists of several different actors. Additionally, it meets requirement 2: these different actors influence each other. For instance, a government can choose to implement taxes on the end-of-life disposal of hazardous materials. Therefore, recycling costs of producers will be higher. However, a producer can make the price of a product higher to let the consumer indirectly pay for the higher taxes. This can also have effect on consumer’s purchases: when they find the functional value of a product of high importance (which means among other things its price), they might buy less of the products. As a result, producers have lower profits. This example shows that the system not only is in line with requirement 2, but also requirement 4. Namely, the actors are able to adapt in the system, according to their self-interest. 3_{Interdisciplinary research is a process in which a problem is tempted to be answered through different} disciplines, as it is too complex to be dealt with from different disciplines (Klein and Newell, 1997 in Rutting et al., 2016).

(6)

In conclusion, producers are highly dependent on governmental regulation and consumer demand. Governmental regulation and consumer demand regulate behavior and induce a firm to make a certain material choice. In the next section - the first part of the theoretical framework- the influence of legislation (and its elements ‘modulated fee’, ‘tax and subsidy’, ‘recycling norm’) on material choice will be discussed. The rest of the concepts (except ‘environmental impact of material’) will be discussed in the two subsequent sections of the theoretical framework. The influence of ‘environmental impact of material’ will be discussed in the methods section on machine learning.

1.1 WHAT IS THE INFLUENCE OF (SUPRA) NATIONAL

LEGISLATION ON (‘GREEN’) DESIGN CHOICE BY FIRMS

IN THE NETHERLANDS?

Government intervention can have a significant influence on the choices of companies for their product design. The concept of EPR will be used as an example in this research question, as it is an influential policy which tries to influence design and material choice by companies. By implementing the Extended Producer Responsibility (EPR), government's attempt to internalize environmental costs. This means that they try to influence the price of a product to make it reflect how it burdens the environment. In other words, it means that a product that is bad for the environment becomes more expensive (Lifset, 1993).The policy of EPR is applied by placing responsibility for the environmental impact of a product on the producer of the product, after the product has been used. The concept of EPR is of great importance for (‘green’ design’) choice by companies, since the OECD (2001 in Dubois et al., 2016) states that the goal of this policy is to create incentives for producers to produce more sustainable. Eco-design means that products are designed in such a way that they are easier to be recycled or reused; contain fewer hazardous materials; and have longer durability (Dubois et al., 2016). This is encouraged by FIGURE 2. OBJECTIVES INFLUENCING MATERIAL CHOICE

(7)

EPR, because when a product that meets requirements of eco-design is at end of life, the producer only has to pay a low fee. If this is not the case, firms have to pay a higher fee which might be shifted to the consumer by raising the price of the product. In this way environmental costs of a product will be internalized by implementing EPR. Firstly, an overview will be provided on how EPR is implemented in the Netherlands in comparison to other countries. Thereafter, an analysis will be made on why a policy such as EPR does not cause companies to make changes towards ‘green’ material choice. Thereafter, an inventory of policy improvements are suggested, so that companies will want to use more sustainable materials.

THE IMPLEMENTATION OF EPR IN THE NETHERLANDS V.S. OTHER

EUROPEAN COUNTRIES

The European Commission has implemented the principle of Extended Producer Responsibility (EPR) in 2003 through the EU Directive on Waste and Electrical Equipment (Dubois et al., 2016). Currently, programs of EPR include products such as batteries, paper, plastics and automobiles (Tojo et al., 2001). Some are mandated by European laws, others came about through negotiations between national governments and corporations (ibid.). The EPR- directives of the European Union are on batteries (91/157/EEC), packaging (94/62/EC), end-of-life vehicles/ ELV’s (2000/53/EC), and waste electrical and electronic equipment/ WEEE (2002/96/EC) (Mayers, 2007). The directives concerning packaging and waste state that it is the responsibility of EU-states to achieve recycling targets. The majority of the member states implemented national legislation, which forces producers to join a collective take-back scheme (ibid.). However, how the directives are further implemented varies considerably among countries (ibid.). Countries differ e.g. in the amount of products that they impose national EPR-regulation on; on the height and/or cause of imposed fees and take-back requirements. The Netherlands along with Germany, Sweden, Norway and Japan is ranked as one of the countries, which achieved the most waste minimization (Gottberg, 2016). In relation to EPR, the Netherlands is also a frontrunner (Dubois et al., 2016). Mayers and Van Rossum (2007 in Mayers, 2007) mention that the Netherlands, along with Sweden, Luxembourg, Italy and Flanders are the only states, which have wholly implemented the WEEE Directive in national laws. Yet, the scope of EPR can be improved (Dubois et al., 2016). Policies concerning products such as expired medication, needles, textiles, domestic used chemicals, are implemented in France and not in the Netherlands (Dubois et al., 2016). The products that have to comply to EPR standards in the Netherlands are consumer electronics, float glass, tires, packaging and vehicles (Dubois et al., 2016).

WHY EPR DOES NOT CAUSE IMPLEMENTATION OF ‘ECO-DESIGN’

Occasionally, producers are not influenced by policies. The next section will analyse why EPR does not necessarily cause companies to use more sustainable materials. A case study of Gottberg et al. (2006) in the European lighting sector -with small and large companies- has shown that when products are price inelastic, heightening the price of the product will not cause a decrease in demand (Turner et al. 1994 in Gottberg et al., 2006). Therefore, producers of these products are able to sell their products against higher prices. Consequently, they can pass the costs for high fees (coming from EPR-regulation) on to their customers, and the producers do not feel an incentive to design their product according to eco-design (ibid.). However, as

(8)

Gottberg et al. (2006) mention, the effects that competition can have on decreasing demand for more expensive products are also of importance. The ‘’price-elasticity argument’’ is considered to be valuable for companies which have monopoly positions or when their brand differentiates them from other companies, since there are no substitutes in that case. Moreover, producers can ‘free-ride’ with the EPR-policy because companies are able to use Collective Producer Responsibility (CPR). CPR means that producers with equivalent products manage their waste disposal jointly (Greenpeace, 2008). Consequently, ‘free riders’ can “benefit from a producer responsibility system without contributing an appropriate share of the costs” (Kroepelien, 2000: 174 in Castell et al. 2004). As a result, when companies use the CPR-system, they have fewer incentives to change their product’s design (and thus materials) (Greenpeace, 2008). This is because when a company would apply eco-design, the free riders benefit, because they have to pay a lower fee (Gottberg, 2016). After all, companies use joint collection and a shared fee. The incentive for quitting with using toxic materials, making products more durable and easier to recycle, is low (Greenpeace, 2008).

INVENTORY OF POLICY IMPROVEMENTS THAT AFFECT MATERIAL

CHOICE

There are policy improvements, which can significantly influence the choice of producers for green materials, though. Firstly, according to Monier et al. (2014) and Dubois et al. (2016), when a company returns its products to a collective system, it should have to pay a modulated Producer Responsibility Organization’s (PRO) fee. The height of the fee should be determined by assessing the environmental impact of the product (Dubois et al., 2016). The research of Dubois et al. (2016) depicts how other European countries have implemented more in-depth EPR-legislation. In France for instance, a fee for a vacuum cleaner increases with 20 percent when it contains more than 25 grams of brominated flame retardants (ibid.). If such modulated fees are widely implemented in the Netherlands, companies have a greater incentive towards eco-design. As mentioned in the previous sections, producers do not apply eco-design when the costs are that low that either customers can pay for it or that it is cheaper to just dispose the waste, and pay the taxes (Turner et al. 1994 in Gottberg et al., 2006). Thus, modulated fees should be significant in relation to the sales price (Dubois et al., 2016). Moreover, a simulation model by Brouillat and Oltra (2012) quantifies the effect of a certain taxing scheme or recycling norm on product design change. The model includes two recycling fees: Fee_A and Fee_B. The recycling fee is a tax on product sales to cover the cost of recycling. With Fee_A, the tax is identical for each firm and with Fee_B it differs between firms on the base of the recyclability of their product. Besides, the model introduces a scheme with both a tax and a subsidy, named TaxSub. The tax is proportional to the recyclability of the product, and the tax revenue is used to subsidy companies with a ‘green design’. Lastly, two recycling rate targets are introduced, named Norm_A and Norm_B. Norm_A is a strict norm: if a product doesn’t satisfy the norm, it may not enter the market. Norm_B is a less strict norm, where a firm has to pay a fine if it doesn’t satisfy the norm. Table 1 shows the simulation results.

(9)

FIGURE 3. ++ HIGH POSITIVE IMPACT, + SIGNIFICANT POSITIVE IMPACT, 0 NO SIGNIFICANT IMPACT, −− HIGH NEGATIVE IMPACT, – SIGNIFICANT NEGATIVE IMPACT. THE FIRST ROW REPORTS THE MONTE CARLO AVERAGE VALUE AND THE SECOND ROW THE WILCOXON– MANN–WHITNEY OR STUDENT As can be seen from figure 3, the combination between a tax and a subsidy, a strict norm and a less strict norm have a significant influence on ‘green’ product design by companies. The instrument that a government uses thus can be different for the goals a government has. If it only aims to increase recyclability, the combination between a tax and a subsidy, TaxSub, is the optimal choice. If a government is only interested in increasing durability, a strict recycling norm, Norm_A, is the most effective option. If it is looking to combine the two, Norm_A is also the most effective. In other words, firms choose more durable and recyclable materials for their product. In conclusion, research question 1.1 analyzes how policies both can and cannot have effect on ‘green’ material choice by companies. As a clarifying example, EPR-legislation was chosen. Analysis of the functioning of EPR showed that when producers see a chance to let consumers pay for higher fees, they would not apply more sustainable materials. However, legislation in the form of modulated fees, recycling norms, taxes and subsidies can influence the choice of a firm for a sustainable material type because it affects the costs of a firm. Thus, this research question dealt with how recycling costs for companies are affected by policies and the effects of these policies on material choice by companies. More on the costs of a firm will be discussed in the next paragraph.

1.2 HOW DO ECONOMIC COSTS INFLUENCE THE

MATERIAL SELECTION OF FIRMS?

This section provides a theoretical framework on the production costs of a firm, and what influence these costs have on material choice. Every firm tries to minimize production costs in order to be more competitive on the market.

MATERIAL COSTS

Various factors influence production costs, but the most influential factor is material costs. In the chemical industry, for instance, material costs make up 70 percent of a firm's cost structure. In assembly-oriented production, such as for instance automobile production, the material costs determine at least 50 percent of the total costs (Arnold, 1989). Therefore, it is efficient to use as little material as possible within production.

(10)

TRANSPORTATION COSTS

Furthermore, transportation costs influence material selection. According to Maranzana (1964), transportation costs are proportional to the sum of the distances from the point where the good is supplied to the city where the firm is located, each weighted by the volume of shipments. Heavy material costs more than light material to transport, so a firm tries to minimize the weight of their product, and thus tries to use the lightest materials possible.

SCARCITY

Scarcity also has influence on the material choice of firms. It is the main concept within the field of Economics, and it means that the needs of people exceed the amount of recourses that are available. With increasing evidence on the fact that some unsustainable resources, for example oil, are more finite than previously anticipated, these resources are considered to be scarce. This has influence on the price: the price of a material is positively correlated with its scarcity. Scarcity can induce ‘transmateralization’, which means substitution by other materials. Scarce materials are substituted by less scarce materials to minimize production costs (Rydén et al., 2003). This is significant because, as previously mentioned, material costs make up a large part of the total costs of a firm.

GLOBALISATION

How a country is influenced by globalisation can also induce transmateralization by firms. When a country opens up to trade, global demand has an influence on the demand of the product of a firm, as opposed to the previous situation where only national demand determined the demand of the product. According to Krugman, Obstfeld & Melitz (2015), the Heckscher-Ohlin theorem states that a country will export the good that intensively uses the factor of production that is abundant in that country. Moreover, a country will import the good that intensively uses the factor of production that is scarce in that country. This means that if a country is relatively land-abundant, its export is determined by the companies that use this factor of production in their products. From this viewpoint, it can thus also be concluded that a company prefers using materials that are abundant in that country. This preference becomes more prominent when trade is increasing.

IMPORT TARIFFS

Lastly, import tariffs on materials can change a firm’s choice for a certain material. An import tariff is a tax imposed by the government in order to discourage its import. This can have different reasons, for instance environmental reasons or to protect inland production of a material (Krugman, Obstfeld & Melitz (2015). From this expense, it can be seen that there is a link between government regulations, as discussed in the first paragraph, and the economic costs of a firm. Government regulations like an import tariff are not incentives on their own: they create economic incentives. The higher the import tariff on a material, the less attractive it is for a firm to use in its product. For instance, industrial materials and supplies are subject to an import tariff, which differs between countries. For instance, the tariff for crude oil, which can be made into plastic, ranges between 0% and 40%, with an average of 1.9% (Bowes, 2016). In the Netherlands, for instance, it is approximately 33.6% (Belastingdienst, 2016).

(11)

In conclusion, material costs make up the largest part of production costs for a firm, and together with transportation costs, scarcity and import tariffs, it influences the choice of materials for their product. If a company values low economic costs, it should consider materials which are not scarce in general, are light-weighted (as transport costs are low then) and are abundant in the country. The previous two paragraphs focussed on the supply side of the material choice of a firm. However, the incentives of a firm to choose a certain material also originate from the demand side. This will be discussed in the next paragraph.

1.3 WHAT IS THE INFLUENCE OF CONSUMER DEMAND

ON (‘GREEN’) MATERIAL CHOICE?

The former research questions (1.1 and 1.2) have explained that producers should take into account what the costs of a material are and which regulations are imposed on the materials, when choosing materials for their products. However, according to Park et al. (2008) a product must also be in line with current consumer demands. Strikingly, when a product’s environmental performance is notably high, but does not meet consumer’s demands, it will not have a good chance to become profitable (ibid.). Thus, it is for a company of high necessity that it considers all the requirements, and thereby minimizes the risks of investments. According to Park et al. a product should meet the requirements of durability, safety and performance, as consumers find this is of high importance. This is its functional value. This section will focus on consumer criteria and how they influence a company in the choice of products and materials. Since people base their consumer behaviour on the product as a whole, it may be difficult to assess how they value individual materials. Yet, what people value in a product can apply to a specific material as well. For instance, when people buy a product, as a means of to portray their identity, people who want to look wealthy, may prefer a real gold necklace. As factors, which influence consumer demand, vary between cultures, income groups and countries, this section is meant to provide general insights of the various factors.

CHANGING DEMANDS CONCERNING SUSTAINABILITY

Producers create surveys to find out if a new product/material would meet the demand of consumers (Aoe, 2007). This shows which function (e.g. environmental performance versus functional performance) a consumer finds more important in general. Vandermerwe et al. (1990) argued already in the 1990s that a ‘green tide’ is occurring. They conducted a research with 100 executives from consumer and industry backgrounds and concluded that sales of sustainable products went up. Moreover, consumers demand products (and thus materials), which are not harmful for health and the environment; are able to be re-used; and such that minimize waste and emissions (ibid.). Furthermore, Ottman et al. (2006) state that consumers desire that ‘ green’ products have a better or equal performance compared to alternative products.

DEMANDS CONCERNING COSTS

Research of Aoe (2007) has stressed that customers want high satisfaction against low costs. The materials should therefore also be of low cost. Additionally, Nidumolu et al. (2009) mention

(12)

that during a recession people are less willing to pay for environmental-friendly products. Yet, the number of people who accept to pay more for ‘green’ products is increasing (Laroche et al., 2001 in Lin et al, 2012). On the other hand, people value a product’s quality and prestige by analyzing its price: an expensive product is perceived to be of higher quality (Ljungberg et al., 2003). Besides, when the product looks attractive and the materials are perceived to be of high quality, people are generally willing to pay more money for it (Ljungberg et al., 2003).

VALUES

Sheth et al. (1991) analyse in their article a theory that elaborations on the way consumers make choices. It consists of five consumption values, which can give insight in why one brand is chosen over another; why a specific product is bought and future consumer behaviour. There are three propositions: 1) The product/material choice of a consumer comes from various values. 2) The values are independent. 3) The values explain different drivers of the specific choice. For instance, when a person decides to buy gold, it could be because of status (‘social value’) and/or as it serves as a hedge (‘functional value’). How salient the values are, varies between persons. The values are now explained in depth. First of all, functional/utilitarian and physical attributes of products, such as longer durability and price are highly valued (Ottman et al., 2006). This is the ‘functional value’ of a product. Traditionally, this value is assumed as the most important factor which influences consumer behavior (Sheth et al., 1991). Ljungberg et al. (2003) described the case of ‘the Itera’ -a bike made of polymers- to argue that materials have to be carefully selected, in order to look ‘’good’’. The Itera-project failed because people did not consider plastic as a good and prestigious material. Secondly, when products are bought for the social image a person wants to have, the choice-driver is ‘social value’. Remarkably visible products (jewelry or expensive cars) are examples of products, which are bought because of ‘social value’. Hyman (1942 in Sheth et al., 1991) also suggests that individual behavior is influenced by a social group it eagers to belong to. Additionally, there is ‘emotional value’. When a product is consumed for the emotional feeling it arouses, then ‘emotional value’ is the driver. Products that can bring about such feelings are products, which are related to childhood memories; horror movies and candles for a romantic atmosphere (Sheth et al., 1991). With regards to materials, natural materials such as wood can bring up memories (Ljungberg et al., 2003) of for instance a hike in the woods on a holiday. Moreover, ‘epistemic value’ means that a product is bought because of its novelty. The person may want to try a new brand or use a product, which is made of ‘new’ materials. Lastly, some products or materials are preferred because of the specific situation the consumer is in. Then, the choice comes from the ‘conditional value’ the consumer holds. This can be the case when it is Christmas: people, who maybe usually do not want to use paper, use it for sending postcards. But it can also relate to a specific country a person lives in. For instance, in Scandinavia villas made of wood are seen as prestigious, whilst in Germany stone has a higher status (Ljungberg et al., 2003).

VALUES IN RELATION TO SUSTAINABLE MATERIALS/PRODUCTS

Lin et al. (2012) conducted a research in which they investigated ‘green’ consumer behavior according to the previous choice theory. They applied a one-way analysis of variance and a multiple regression to the data they acquired by questionnaires. The conclusion was that

consumers who are concerned with the environment are more likely to choose ‘green products’. Knowledge is thus of great importance for consumer behavior: news reports of extreme

(13)

weather can cause people to adjust their consumption patterns (Laroche et al., 2001 in Lin et al., 2012). Additionally, consumers may want to feel positive about their influence on the environment (Arvola et al., 2008 in Lin et al., 2012). The ‘emotional value’ is then an important driver. Also, consumers may want to deal with environmental problems and thus seek for novel materials and products. This corresponds with the ‘epistemic value’. Besides, Straughan and Roberts (1991 in Lin et al., 2012) state that people can also feel that they have to buy ‘green’ products because the social group they belong to also does it. The ‘social value’ then influences consumer behavior. Price and quality are considered as less important factors in choice making, than the ‘conditional’, ‘emotional’ and ’epistemic’ values (ibid.). These results come from the multiple linear regression analysis of Lin et al. (2012), in which consumer’s choice behavior is the dependent variable and the five values are the independent variables. The adjusted R-square is 0.47. That means that these independent variables explain 47 percent of the variance in the choice behavior of a consumer. Also, the F-value is significant, which means that the model is adequate for the data and the results are valid. The results depict that the epistemic, emotional and conditional value have positive effects on consumer’s choice behavior in relation to ‘green products’. Concluding, the chance that a consumer buys a ‘green’ product/material is higher when it attributes high emotional, conditional or epistemic value to the product (ibid.).

FIGURE 4. TABLE WHICH SHOWS THAT THE EMOTIONAL, CONDITIONAL AND EPISTEMIC VALUE HAVE SIGNIFICANT P-VALUES AND HAVE POSITIVE EFFECTS ON CONSUMER BEHAVIOR REGARDING ‘GREEN’ PRODUCTS. (LIN ET AL., 2012)

DEMANDS OF CONSUMERS VERSUS PRODUCERS

Park (2008) conducted a statistical analysis to depict that the Pearson’s Correlation between product value for producers and customers is 0.201. This means that the values of consumers and producers mostly do not correspond. If a company or government wishes to use more sustainable materials it is of necessity that it informs the consumer of the benefits of the material. In this way it can change the values of consumers concerning a material.

DEMAND AND CORPORATE SOCIAL RESPONSIBILITY

A concept that is linked to demand is ‘Corporate Social Responsibility’, abbreviated as ‘CSR’. A firm is acting in a social responsible way if it is transparent towards consumers and the government, doesn’t exploit employees and tries to minimize the environmental damage of its production (Campbell, 2007). According to Campbell (2007), Corporate Social Responsibility is

(14)

more likely to happen if the demand of a product is high, consequently the profit as well. This means that a firm will tend to choose ‘green’ materials if the demand is high. This is thus also an example of how demand influences material choice by firms. There are other factors that make Corporate Social Responsibility more likely, but within this paper it has been chosen to not further elaborate on those factors, due to the scope of the research.

M E T H O D S

1.4 HOW CAN ARTIFICIAL INTELLIGENCE’ METHODS

PROVIDE A SOLUTION TO THE COMPLEXITY OF THE

AFOREMENTIONED OBJECTIVES?

COMPLEXITY: CONFLICTING ELEMENTS

The previous section has outlined the following 3 major factors influencing a company’s green material choice: legislation by the government, the company’s economic costs and consumer demand. Together these factors form a complex network of often conflicting objectives. On the one hand the company wants to reduce its financial costs in order to gain more profit. Based on this incentive as discussed in section 1.2, selecting the cheapest materials, in terms of purchase costs, transportation costs and import tariffs seem to most favourable. However, on the other hand, in order to maximise profits, not only the costs should be taken into account, but the sales as well. This is where consumer demand plays a role. From section 1.3 it can be concluded that although consumers do prefer low prices, they want quality products as well. So neglecting either of these preconditions could result in a reduction of sales. Therefore, not only the material costs should be taken into account, but also the consumers requirements concerning the material. As described in section 1.3, in order to meet these requirements, the following material characteristics should be considered: functional value, social value, emotional value, epistemic value and conditional value. Besides these two objectives, the governmental regulations should be taken into account as well. By applying taxes on environmentally harmful materials, such materials become more expensive to the company. There the costs of such materials cannot be evaluated only by taking into account the costs as described in section 1.2. On top of that consumer demand is also influenced by this environmental impact. As discussed in section 1.3, consumers prefer green products. However, they only do so at a reasonable price. So in short the degree of environmental impact also plays a role when considering the optimal materials for a product. The former objectives can be divided into sub objectives, which also can be divided into smaller components. For instance, the component consumer demand is divided into sub components such as ‘social value’. This ‘social value’ is dependent on even smaller components, for instance on the consumer’s social class (Hyman, 1942). This way the systems becomes a hierarchical organization in which the lower level components can provide a vast amount of possibilities which the system will have to adapt to. Furthermore, the connectivity does not only take place between nodes in the same tree, but also between nodes in different trees. For instance, a product's recyclability not only influences the environmental impact and thus material costs, but also the consumer’s preference concerning that material. Therefore, the system does not have a flat hierarchy, but a deep hierarchy and thus makes it a complex system. Furthermore, different materials showcase both diversities and similarities at the same

(15)

time. Subsystems of materials such as polymers can share certain characteristics, yet also maintain significant differences and thereby maintain their diversity. This also contributes to the systems complexity. On top of that the frequent abundance of candidate materials (Zhou et al., 2008), varying consumption patterns, competition for scarce resources and unexpected calamities during the production process (Helbing, Armbruster, Mikhailov & Lefeber, 2006) also cause this network to be a dynamic and evolving system in an ever changing and stochastic environment (Helbing et al., 2006). For instance, according to Nidumolu et al. (2009), people are less willing to pay during a recession. So the system is dependent on the economic prosperity of a country. So the taken all together these contradictory and dynamic objectives cause the problem of material selection to be a wicked multi criteria decision problem. Therefore, finding an optimal material is often regarded as a challenging problem (Tambouratzis, Karalekas, Mousakas, 2014). However, decisions are still generally made based on trusting experience (Zhou et al., 2008) and thus dependent on human decisions (Joseph, Sharif, Kumar, Gadkari & Mohan, 2014). However considering the inherent trade-offs and conflicts, this multi-criteria decision problem is very complex and time consuming (Joseph et al., 2014). Therefore, this problem cannot be solved by using solely this traditional experience based approach to material selection and a more accurate approach is needed (Zhou et al., 2008).

MACHINE LEARNING AS A NEW APPROACH TO THE COMPLEXITY OF

MATERIAL SELECTION

Over the past decades, developments on the field of artificial intelligence, along with the increasing popularity of big data has brought forth technologies that can collect and process large amounts of data (Joseph et al., 2014). One such a technology is machine learning. By using this technique, an intelligent machine can learn about a hidden system. This is done by deriving patterns from the data it is given and thereby it generates its own knowledge (Goodfellow et al., 2015). The derived patterns can be used to make predictions on the output when the system is in a changing environment or when new data is added to the system. The goal of the machine learning algorithm is to make the predicted output as good as an approximation of the actual output. This is done by providing vast amounts of data and adjusting the systems parameters until the predicted output is as close to the expected output as possible (Dietterich et al., 2010). Thus by providing big amounts of data on previous material selection processes, a computer program could potentially derive certain patterns and regularities that are too complex for a human to comprehend. Subsequently it could do predictions on the output when provided with new input data. Ultimately this could lead to the computation of an optimization algorithm providing the ideal (‘green’) material for any given situation. To create such a program is far beyond the scope of this research. However, we have created a tool that could be built on in future research. This tool takes into account the three major objectives (legislation, economic costs and consumer demand) concerning ‘green’ material choice and uses these to rank a collection of materials.

MATERIAL SELECTION TOOL

As the previously described incentives of company during the material selection process favor different material characteristics, compromises have to be made. For instance, consumers want quality materials, yet only for a reasonable price Aoe (2007). Cheap materials however are not necessarily quality materials. Furthermore, the government stimulates ‘green’ design by implementing fees on materials with a high negative environmental impact. However, these materials could be more expensive than their alternatives, so the material selection depends on

(16)

whether a company emphasizes on cost reduction by avoiding fees or selecting cheap but polluting materials. Besides consumers also tend to favour ‘green’ products, at least when this doesn’t compromise too much on the quality and price. In short, emphasizing too much on one objective could be at the expense of other objectives. Therefore, it is a challenging task to select the most favourable material given all the above considerations. In other words, it is hard to see which material results in the least amount of total compromise. Moreover, the characteristics that characterize a material's quality are not in the same order of magnitude. For instance, a product's recyclability is classified by a number between 0 and 1 whereas another feature such as the emotional value could take on much higher values. Therefore, the comparison can not be done by just using a straight forward first degree polynomial function in which the different objectives are simply added to each other in order to reach a total score. When using such a polynomial, emotional value for instance would contribute much more to the overall score than the recyclability does. Therefore, in order to reduce this variance in magnitude, some normalization algorithms have to be applied first. Furthermore, after this normalization the ranking is not based on solely calculating a total value for each material by adding up its sub scores for each objective. Instead for each objective, the distribution of the scores of all materials is taken into account as well. Using the standard deviation each material is thus assigned a score based on how well it does relative to the other competing materials. Finally, each objective can be assigned a different weight as one objective might be of more importance to the producer than others. This will be discussed more elaborately in the following sections. In this version of the tool, these weights (the degree of impact for each objective) can be set manually and experimented with by the user. However, in future research machine learning algorithms could be applied to have an intelligent machine calculate the optimal weights by providing examples of material choices and their resulting impact on a company’s profit. In basic terms this means the program would simulate the model over and over using different weights until it finds those weights that best represent the outcome of the current situation (given the sample data). However, in order to apply this next step a material ranking framework is required. So the next section will illustrate how this comparison is done by first elaborating on the three main objectives (legislation, costs and consumer demand) and their contribution in the tool. Finally these objectives will be combined into a material ranking tool using a combination of probability theory and the Numpy package in Python. LEGISLATION The legislation concerning material choice consists of two measures: a combination of a tax and a subsidy and a modulated fee. In order to compare how these regulations will affect the producer when using different materials, the material’s environmental impact has to be taken into account. This environmental impact of a material influences how high a tax/subsidy/modulated fee on this material will be. For instance, when the recycle fraction of a product increases, potential taxes on waste go down and moreover a subsidy could be dispensed. Furthermore, the height of a tax and subsidy are determined by the energetic usage and pollution of a material: the height of a tax is positively correlated to the environmental impact of the material. In the case that a modulated fee has to be paid, this fee would increase as well. Subsidies, on the other hand, will then decrease. So when comparing different materials, the

FIGURE 5. LEGISLATION DEPENDS ON THE ENVIRONMENTAL IMPACT

(17)

environmental impact determines the impact of governmental regulations on the producer's cost/benefit analysis. In order to assign a value to a material’s environmental impact the following characteristics (based on Zhou, Yin & Hu, 2008) will be taken into account: - recycle fraction (the fraction of a material than can be recycled) - energetic cost (MJ / kg) - CO2 emission (pollution) (kg CO2 / kg) (González & Navarro, 2006) Using these characteristics, machine learning could calculate the height of a fee/tax/subsidy for any material.

K-NEAREST NEIGHBOUR ALGORITHM

For instance, this could be done using the K-nearest neighbour algorithm. The K-nearest neighbour algorithm could use available data on existing fees, taxes and subsidies for certain materials and use these these training examples to classify the fee/tax/subsidy height for any material in the database. What this algorithm does is in essence the following: It takes a n-dimensional vector (the to be classified example / test example) and for each entry in the vector it calculates the distance to the corresponding entry of each vector in the training set (known/training examples). Eventually it uses all this information to identify the n-nearest vectors to the new vector. Finally, the new factor will be classified based on the characteristics of the n-nearest vectors surrounding it. Figure 6 illustrates the nearest neighbour of a such a test example (Beyer, Goldstein, Ramakrishnan & Shaft, 1999).

FIGURE 6. 1-NEAREST NEIGHBOUR VISUALIZATION OF A QUERY POINT (TO BE CLASSIFIED) AND THE 1-NEAREST NEIGHBOUR (TRAINING EXAMPLE) (BEYER ET AL., 1999)

So when predicting the height of a fee based on environmental properties this algorithm could do the following: By using the known examples as a training set, the environmental properties could each be assigned to an input variable (X1, X2, … ,Xn) and the corresponding (known) fees/ taxes/ subsidies could be assigned to an output variable (Y1, Y2, …, Yn). Now when inserting a test set of materials you want to classify, the algorithm searches for the n-nearest (for instance 10 nearest) neighbours of each test example. Using the Y values of these 10 nearest neighbours, the test example can be assigned that fee/ tax/ subsidy that has the highest probability.

(18)

K-MEAN CLUSTERING ALGORITHM

Furthermore, the materials could be divided into clusters based on their environmental properties. This way each member gets the same fees/ taxes/ subsidies as the other members in a particular cluster. This algorithm would be less accurate. However, fees/ taxes/ subsidies are often based on a discrete scale. In other words, they can be either for instance 5%, 10%, 15% etc… (Dubois et al., 2016: 23). Therefore, this accuracy loss could be considered insignificant. A machine learning algorithm such as the K-mean cluster algorithm is able to recognize clusters in an unlabeled data set. This way materials with similar characteristics could be assigned to the same cluster and thus be predicted to result in the same fee/tax/subsidy. Now when adding a new material, the algorithm only has to calculate the distance to the existing clusters instead of all the existing materials. Figure 7 illustrates this clustering in combination with the K-nearest neighbour algorithm as seen in Figure 6.

FIGURE 7. K-MEAN CLUSTERING IN COMBINATION WITH THE K-NEAREST NEIGHBOUR ALGORITHM. (BEYER ET AL., 1999)

These predicted fees/taxes/subsidies for each material could be used to calculate their contribution to the total financial cost of a company. This could be done by for instance using the following equation: cost = material cost *(1+tax)*(1-subsidy)*(1+modulated fee). As the application of such algorithms are beyond the scope of this research, the current tool will solely take into account the features on environmental impact and use these as a contribution to the material ranking. The previously described methods to combine these environmental impact features into new values such as a tax, fee and/or subsidy could be applied in a later stage. Besides the environmental impact of a material, the characteristics with regard to economic costs and consumer demand have to be taken into account as well. The next two subsections will outline which characteristics are used to quantify this economic cost and consumer demand.

(19)

COSTS As mentioned in section 1.2 the main contributions to the production cost are the purchase cost, transportation cost and import tariff. Furthermore, Zhou, Yin & Hu (2008) also consider the recycle cost, therefore, the main factors for the parameter ‘costs’ are: - import tariff - recycle / disposal cost - transportation cost - purchase cost CONSUMER DEMAND Based on section 1.3, the consumer demand is based on the following factors: - functional value - social value - emotional value - epistemic value - conditional value The values for these factors could for instance range from 1 tot 10. This way these values could be seen as a material’s score compared to the other materials. In this version, the user will have to assign the values for these factors. However, in the future machine learning algorithms could be implemented to have these values assigned automatically and with more accuracy. According to Joseph et al. (2014) such algorithms are able to analyse big amounts of social media data and even gather information from natural language conversations using speech recognition. These techniques could be applied to gain insight into the public opinion with regard to different types of materials. The next section will demonstrate how these objectives are stored into a material database.

MATERIAL DATABASE

As seen in figure… a list of materials is stored into a csv file. Each material is assigned different values corresponding to all the features as described in the previous section.

(20)

The tool will ultimately combine these values and calculate an overall value for each material. This value determines how beneficial they are compared to the other materials. The

mathematical methods underlying this calculation are discussed in the ‘methods’ section. Yet a short explanation will be provided in the next section.

CALCULATING THE OVERALL VALUES:

As the units for each values differ per category they cannot be merged straight away. For instance, due to these different units, the values in the column for density (m3/kg) are much higher than the values in the column for tensile strength (mPA). Consequently, a material’s density would have a much bigger contribution to the overall value than its tensile strength. Therefore, each column is normalized resulting in values between 0 and 1, yet keeping the same proportions between different materials. However, these new values do still not depict a materials true overall value. What really defines a material’s value for a given factor is its score relative to that of the other materials. For instance, when after normalizing a material has a value of 0.8 for density while the mean value for density is 0.7, this has a different meaning than when the mean would be only 0.2. Therefore, the distributions of the values for each factor is also of importance. In order to implement this degree of variance, each value is replaced by its corresponding z-score. This z-score displays how well a material scores for a certain factor compared to the other materials. Now the z-scores of relating factors are combined, forming an overall score for each category. E.g. the score for environmental impact is the sum of the z-scores of recycle fraction, energetic cost and CO2 emission. For the category quality, these scores are multiplied by -1 as a high value is positive, whereas on the other hand high values for environmental impact and costs are negative. Finally, the sum of the scores for environmental impact, costs and quality determines a material's overall score.

USING THE TOOL

When opening the tool, a user is presented the following window:

FIGURE 9. OPENING SCREEN

(21)

The upper part displays six main material types. A producer might not want to use just any type of material for a certain product. For instance, when you are a lamp producer, you would probably not want the tool to return a ceramic when designing the light bulb. Therefore, it lets the user choose which material types will be used during the analysis. According to Ashby (1989), most materials can be classified as either a metal, polymer, glass, wood, ceramic or elastomer. Therefore, the tool distinguishes these six categories. As mentioned before a producer might want to put more emphasis on one of the three objectives and see if this will result in a different proposed material. Therefore, the tool lets the user assign a weight to each objective. For instance, a weight of 2 for quality will result in a twice as big contribution of the objective quality to the materials overall score. Furthermore, a user could also for assign a value of 3 to environmental impact and a value of 2 to costs. This will result in an overall value in which the score for environmental impact is multiplied by 3 and the score for costs is multiplied by 2, whereas the score for quality remains the same. After the preferred material types are selected and the weights are assigned, the ‘run’ button provides the 3 best materials based on the user input. The next paragraph will illustrate how these weights can be used given a certrain situation. CASE As stated in section 1.1, the government might want to encourage ‘eco-design’, thus the usage of ‘green’ materials by producers. In order to provide incentives for producers, the government might choose to implement the combination of a tax and subsidy. In this case it will become more favourable for a producer to choose materials, which have less environmental impact. Furthermore, it might become a hype to buy products which contain ‘green’ materials. The consumer then buys a product because of its ‘social value’. In such a scenario, the producer might want to put more emphasis on the environmental impact compared to the financial costs. In this situation, the economic costs of the material might be of less importance, as the financial disadvantage of choosing more sustainable materials can be compensated by the resulting increase in sales, higher subsidies and decreasing taxes. The figures below illustrate how assigning a higher weight to the environmental impact results in an optimal material with a higher ranking on environmental impact, yet a lower score on financial costs. EXAMPLE WITH ALL WEIGHTS SET TO 1

(22)

(23)

INTEGRATION TECHNIQUES

This section elaborates on which integration techniques were used during this

interdisciplinary research. The goal of this section is to show how common ground was found between the sometimes conflicting approaches from the different disciplines According to Newell (2011), the integration of different disciplines plays a key role in conducting an interdisciplinary research. The reason for this is that there seems to be tension between different perspectives on a subject. When the focus lies on a certain discipline, reality is looked through by a coloured, disciplinary lens. Newell states than an integrated interdisciplinary approach creates ‘common ground’, which maximizes the understanding of reality.

Finding this ‘common ground’ can be accomplished by using several integration techniques. The main integration technique used within this paper is organisation. This concept can be clarified by looking at the figure beneath, which was introduced in the introduction of this paper. According to Somerville (2002), within organisation, different domains combine and influence each other, and form an umbrella of organisational behaviour. There is a strong link between the three factors that were discussed within this paper. If the government embraces a new policy, let’s say higher taxes on oil import, firms face higher costs when they want to use oil in their product. When these costs are shifter to the user, this leads to a drop in consumer demand. Because of the potential decrease of profits, a firm can decide to trans materialize, as was discussed in the section 1.2. In this case, consumer demand created the material choice of the firm, while it began with the government that changed policy. This example thus illustrates how the different factors influence each other and eventually form behaviour.

Another integration technique that was used within this paper is rearranging sub-systems to bring out interrelationships. One can apply this technique through different usages (Somerville, 2002). Within this paper, the discipline of politics was explored through applying knowledges of the field of Economics. From the field of politics, one could start with looking at how the government tries to encourage or discourage a certain material choice of a firm. But in order to understand better how the government handles, its incentives need to be clear. The incentives of a government to implement a certain policy can be explained by the theory of market failure, which is an economic theory. As was discussed in paragraph 1.1 it means that the environmental cost of production is internalized ehrn the government poses taxes on materials or waste. This is because then the price of a product will reflect its environmental impact, which economists plead for (Lifset, 1993). There seems to be tension between firms and the government, but the relationship between the two is that they both try to create efficiency. A firm only creates this through cost minimizing and profit maximizing and the government through internalizing environmental costs. Both behaviours seem to be leading to a more efficient society, even though their incentives seem to be in conflict at first.

The field of artificial intelligence has facilitated the transformation of opposing disciplinary axioms into a continuous variable, which is one of the integration techniques, states Somerville (2002). As can be seen from the fourth paragraph, machine learning could combine the different unit measures of the different factors that were mentioned within the paragraphs. Fir instance the different consumer values are not being measured in euros, but rather on a range from 1 to 10. So a unit measurement could be given to these values, making them comparable to the economic costs and legislation, which are measured in euros. This way the three factors can be integrated into one computer program.

This section elaborated on which integration techniques were used to conduct an interdisciplinary research that gains a maximum understanding of reality. The main

integration techniques that were used are organisation, as well as rearranging sub-systems to bring out interrelationships and the transformation of opposing disciplinary axioms into a

(24)

continuous variable. Together these integration techniques create a common ground between the seemingly conflicting disciplines.

CONCLUSION

This interdisciplinary research paper aimed to answer the following research question: ‘How do governmental regulations, consumer demand and costs contribute to the complexity of a company’s green material choice, and how could machine learning facilitate this choice?’ The research question was analyzed by theories, methods and insights derived from the fields of Economics, Politics, Psychology and Artificial intelligence. In the theoretical section is analyzed that the three aforementioned main elements influencing material choice form a complex system. This is because the system meets the requirements, which are formulated by Rutting et al., 2016. Then, the three drivers are further discussed. First of all, the influence of governmental regulations on recycling costs for a firm is elaborated on. Government interventions can influence material choice by firms, as a government can create incentives for eco-design. EPR was chosen as an example to portray the functioning and dysfunctioning of government measures to promote the usage of ‘greener’ materials. An analysis of its functioning showed that producers of price-inelastic products are not influenced by taxes. Also, Collective Producer Responsibility induces free-riding, which also creates fewer incentives for eco-design. However, research has shown that with a strict recycling norm imposed by the government, firms choose more durable and recyclable materials for their product. Additionally, modulated fees and a combination of a tax and a subsidy, will also create more stringent incentives for companies to choose ‘green’ materials. As recycling costs are not the only costs that the firm has to bear in mind when choosing a particular material, in the second section of the theoretical framework, the other costs of a firm that influence material choice are discussed. The price of a raw material greatly influences a firm in its choice because this cost makes up at least half of the costs of a firm. Also, the lightest materials possible are chosen by a firm as a consequence of transportation costs. Furthermore, scarcity and trade that emerges from it influence a firm because of its influence on prices. In the last section the values which cause consumers to buy certain products or want particular materials is analyzed. Consumers generally desire sustainable products where they derive an optimal satisfaction from against the minimal cost. Their valuation of sustainability can be influenced though, when the price is too high or when a recession occurs. A consumer buys a product on the base of functional, social, emotional, epistemic or conditional values. Moreover, if a company or government wishes to use more sustainable materials it is of necessity that it informs the consumer of the benefits of the material. In this way it can change consumers values concerning a material. Afterwards, the value of machine learning for facilitating the complexity of material choice for companies is discussed. Machine learning could contribute to this selection process by providing a more precise material selection method than the traditional approach which is often solely based on human desicions. Finally a material ranking tool is developed which, in future research, could be used to apply such machine learning methods. For now this tool can assist companies to make informed decisions on which materials they should choose, depending on how important they value complying to ‘legislation’, ‘consumer demand’ and ‘economic

(25)

costs’. The the tool can show insights into how putting emphasis on these objectives influences material choice. These insights could be of value for a company when chosing the right material for their business model.

DISCUSSION

Due to the scope of this research, it was only possible to analyze the influence of legislation, economic costs and consumer demand on material selection by corporations. Although these factors are suitable to portray the complexity of material selection, more factors could be added. Although, Corporate Social Responsibility (CSR) is mentioned, this for instance, could be analyzed more elaborately. A suggestion to scientists for further research is to elaborate on the influence of CSR on material choice. Furthermore the materials as displayed in the material database are only a small proportion of the total amount of materials available on the market. On top of that, the assigned values to these materials are currently not based on real data. We do not have acces to this kind of data as we are not in possecion of the required licences. For now, this does not feel like a major issue as this data only becomes relevant once the tool is actually applied in a case study. Therefore, we perceived this problem to be of minor importance considering the scope and focus of this research. On top of that, in this version a relatively small amount of features is taken into account when computing a total score for each material. In future research, with the acces to more data, many more features could be added in order to achieve a higher accuracy. Finally, in order to really understand the developed tool, a background in programming is required. This in depth understanding is essential to acknowledge why the programmed procedures are difficult to execute without the assistance of an automated algorithm. However, we are aware that not everyone possesses this specific technical knowledge, therefore we have tried to explain the tool’s value in the most basic terms possible. In order to compensate for this loss of preciseness in the main part of the paper, the appendix contains an in-depth elaboration on the processes applied during the tool’s development.

REFERENCES

Arnold, U. (1989). Global sourcing: an indispensable element in worldwide competition. Management International Review, 14-28. Beyer, K., Goldstein, J., Ramakrishnan, R. & Shaft, U. (1999). When Is “Nearest Neighbor” Meaningful?. Derived from http://download.springer.com/static/pdf/151/chp%253A10.1007%252F3-540-49257- 7_15.pdf?originUrl=http%3A%2F%2Flink.springer.com%2Fchapter%2F10.1007%2F3-540- 49257-7_15&token2=exp=1482154723~acl=%2Fstatic%2Fpdf%2F151%2Fchp%25253A10.1007%2 5252F3-540-49257-7_15.pdf%3ForiginUrl%3Dhttp%253A%252F%252Flink.springer.com%252Fchapter%252F10. 1007%252F3-540-49257-7_15*~hmac=864fc77e8680f4b4e80aed550942b0f778da0111694b74deb630518400f65a65 on 22 December 2016.

(26)

Brouillat, E., & Oltra, V. (2012). Extended producer responsibility instruments and innovation in eco-design: An exploration through a simulation model.Ecological Economics, 83, 236-245. Campbell, J. L. (2007). Why would corporations behave in socially responsible ways? An institutional theory of corporate social responsibility. Academy of management Review, 32(3), 946-967. Castell, A., Clift, R., & Francae, C. (2004). Extended producer responsibility policy in the European Union: a horse or a camel?. Journal of industrial ecology, 8(1-2), 4-7. Deutz, P. (2009). Producer responsibility in a sustainable development context: ecological modernisation or industrial ecology?. The Geographical Journal, 175(4), 274-285. Dubois, M., De Graaf, D. & Thieren, J. (2016) Exploration of the Role of Extended Producer Responsibility for the circular economy in the Netherlands. Retrieved from: http://hvglaw.nl/Publication/vwLUAssets/ey-exploration-role-extended-producer- responsibility-for-circular-economy-netherlands/$FILE/ey-exploration-role-extended-producer-responsibility-for-circular-economy-netherlands.pdf on October 11, 2016. European Commission (2008). Promoting innovative business models with environmental benefits. Retrieved from http://ec.europa.eu/environment/enveco/innovation_technology/pdf/nbm_report.pdf Gottberg, A., Morris, J., Pollard, S., Mark-Herbert, C., & Cook, M. (2006). Producer responsibility, waste minimisation and the WEEE Directive: Case studies in eco-design from the European lighting sector. Science of the total environment, 359(1), 38-56. Greenpeace (2008) Frequently asked Questions on Individual Producer Responsibility versus Collective Producer Responsibility. Retrieved from http://www.greenpeace.org/international/en/campaigns/detox/electronics/philips/individua l-producer-responsibi/ on October 12, 2016. Hindriks, J., & Myles, G. D. (2013). Intermediate public economics. MIT press. Krugman, P. R., Obstfeld, M., & Melitz, M. (2015). International trade: theory and policy. Pearson. Level (z.d) choosing materials. Retrieved from http://www.level.org.nz/material-use/choosing-materials/ Lifset, R. J. (1993). Take it back: extended producer responsibility as a form of incentive-based environmental policy. Journal of Resource Management and Technology, 21, 163-163. Maranzana, F. E. (1964). On the location of supply points to minimize transport costs. OR, 261-270. Mayers, C. K. (2007) Strategic, Financial, and Design Implications of Extended Producer Responsibility in Europe: A Producer Case Study. Journal of Industrial Ecology, 11 (3), 113- 131.