MSc thesis

Rijksuniversiteit Groningen

Fair trade on the rise?

A comparison of innovation diffusion between

fair trade products and non-fair trade products

MSc thesis

1

A comparison of innovation diffusion between fair trade

products and non-fair trade products

University of Groningen

Faculty of Economics and Business

MSc. Marketing (Marketing Intelligence)

Author: Dennis Reitsma

Date: January 15, 2018

Address: Ijsselstraat 77a

9725 GE, Groningen

Phone number: +31651822203

E-mail address: d.reitsma.1@student.rug.nl

Student number: 2126966

1st Supervisor: Prof. Dr. T.H.A. (Tammo) Bijmolt

2

Preface

This thesis has been written in order to fulfil the final requirement of the MSc. Markting intelligence. The thesis compares the innovations diffusion patterns between fair trade and non-fair trade products. Focusing on adoption timing, potential number of adopters and internal influences in the social system. This thesis has been written in the time frame of October 2017 to January 2018.

I enjoyed working on my thesis and I’m fairly content with the end result.

I want to thank my supervisor, Prof. Dr. Tammo Bijmolt for the valuable tips and feedback which he provided during the process. Special thanks to Dr. Hans Risselada for allowing me to use one of his datasets. Furthermore, I’d like to thank my group members for the valuable feedback during group discussion. Special thanks to group member Sterre Peters for her valuable feedback during the entire process of writing this thesis.

As a last remark, I hope you enjoy reading my thesis.

Thank you,

Dennis Reitsma

3

Abstract

Nowadays, a lot of people become more aware of the origin of their products. A lot of research has been done on consumer values and attitudes when buying fair trade products. Although, no research has been done on the adoption process of fair trade products relative do non-fair trade products. This thesis explains the differences in the innovation diffusion

process of fair trade products relative to non-fair trade products. A Bass diffusion model is introduced with the advantages that it measures three important variables in innovation diffusion, namely: adoption timing, social influence and potential number of adopters. The method is applied to an aggregated GfK consumer panelist’s data with data on 8 years (from 2008 to 2015) of consumer buying behavior. The dataset contains information on relevant product categories for fair trade products with 1765 product introductions in the relevant product categories. Differences between product categories, new variants vs new products and fair trade vs non-fair trade have been examined to assess the influence the aforementioned variables have on adoption timing, social influence and potential number of adopters. The results provide managers with more insights of how new fair trade product introductions diffuse through society relative to non-fair trade product introductions. Future researchers could use the recommendations to dive further into the topic of innovation diffusion of fair trade products relative to non-fair trade products.

4

Contents

A comparison of innovation diffusion between fair trade products and non-fair trade products ... 1

Preface... 2 Abstract ... 3 1.Introduction ... 5 1.2 Relevance ... 6 1.3 Structure ... 7 2.Literature review ... 8 2.1 Consumer Innovativeness ... 8 2.2 Social influence ... 9 2.3 Market potential ... 10 2.4 Degree of newness ... 11

2.5 Interactions between product category and fair trade ... 12

2.6 Conceptual model ... 13 3. Methodology ... 13 3.1 Data description ... 13 3.2 Data Cleaning ... 14 3.3 Product selection ... 15 3.3 Newness ... 15 3.4 Method ... 16 3.5 Regression analysis ... 18 4. Results ... 20

4.1 Bass model parameters ... 20

4.2 Diffusion plots ... 21

5. Regression results ... 23

5.1 OLS assumptions and validation ... 23

5.1.1 Nonzero expectation ... 23

5.1.2 Heteroscedasticity ... 24

5.2.3 Multicollinearity ... 25

5.3.4 Normality ... 26

5.1.5 Serial correlation and constant parameters over time ... 27

5.2 Results and interpretation ... 27

6. Discussion ... 30

7. Academic implications ... 32

8. Managerial implications ... 32

9. Limitations and future research ... 33

9. References ... 35

5

1. Introduction

The Netherlands has a history of ethical consumer behavior. The origin of fair trade labels can be traced back to 1988, when a non-governmental organization started an initiative to ensure farmers in third world countries received sufficient wages. Named after Max Havelaar, the fictional character in Multatuli’s book Max Havelaar, the Dutch NGO started with what later would grow out to be a global movement.

Fair trade can be described as a trading partnership based on “dialogue, transparency and respect” (WFTO, 2017). According to Jones et al. (2003) some of the most commonly sold fair trade products are coffee, chocolate and tea and according to Krier (2005) the sales of fair trade products have increased with approximately 20% since 2000. Coffee is the highest sold fair trade product and the sales of fair trade coffee are also growing the fastest (Raynolds et al., 2004). To give a clearer overview of how fast the fair trade market is growing; in 2004 the market for fair trade products only generated 838 million dollars in revenue as opposed to 7880 million in 2016 (Statista, 2018). Thus, the fair trade market is undeniably growing and consumers are undeniably buying more fair trade products. However, due to changing consumer values and attitudes, the growth of the fair trade market is highly reliant on innovation of these products and how these products are adopted by society. This process is called innovation diffusion. Rogers (2003) defines innovation diffusion as the communication of new products over time in a social system. However, innovation diffusion does not only rely on a firm’s marketing actions. Peres, Muller and Mahajan (2010) found that innovation diffusion also entails the differences in social influence in a social system. Due to a growing fair trade market and more consumer interest in ethical consumption it is interesting to know for firms and researchers how innovations of fair trade products diffuse through society. This leads to the main research question:

How does the diffusion process of fair trade product introductions differ from the

6 1. Are there significant differences in the adoption rate of fair trade and non-fair trade

products?

Forte and Lamont (1998) found that consumers are more often basing their buying behavior on the basis of a firm’s role in society and Simon (1995) found that consumers have a more positive image of a company that is behaving ethically in the consumers’ mind. Therefore, one could argue that consumers take into account whether firms show ethical behavior when choosing a product. This gives rise to the idea of a different market potential for fair trade products compared to non-fair trade products. Leading to the next sub-question:

2. How does market potential of new product introductions of fair trade products differ from non-fair trade products in the same product category?

The NCDO (2009) found that, for 74% of the respondents to their survey, supporting farmers/producers of coffee beans was the primary reason for buying fair trade coffee. For other fair trade product categories this was 67%, suggesting there might be a difference between product categories and how people buy fair trade products. Leading to sub-question three:

3. Are there differences in the innovation diffusion process among product categories? IRI (2017) has found that Dutch consumers have become more aware of the origin of the products they buy and that consumption of ethical products is on the rise in the Netherlands. This consumption of ethical products ranges from products concerning animal welfare to fair trade. The NCDO (2009) saw the market for fair trade products grow 40% from 2008 to 2009. A trend which already started in the mid 90’s of the 20th _{century. Fair trade has become more}

widely available to the average consumer as these products are now also sold in mainstream retail outlets; 85% of total fair trade sales are made through mainstream retail outlets. (Davies and Crane, 2003; NCDO, 2009).

1.2 Relevance

7 Innovators in general have an impact on ethical consumption (Bolderdijk, Brouwer and

Cornelissen, 2018). It is interesting and relevant for firms to know how newly launched fair trade products diffuse through society relative to non-fair trade products. Due to changing consumer attitudes regarding fair trade products and consumers showing more interest in fair trade products it is expected new introductions of fair trade products will be noticed at an accelerated pace. Previous literature on ethical consumption is mostly based on consumer attitude and thus far no literature investigated the differences in the adoption process of fair trade products relative to non-fair trade products. This thesis will close the gap between consumer attitudes on fair trade products and how the innovation of these products diffuse through society.

Thus, to the best of my knowledge, this study will be the first one to investigate the

differences in the innovation diffusion process of fair trade products relative to non-fair trade products. In this study it is expected that an interaction between degree of newness and a newly launched product being fair trade will be found. Also, an interaction is expected between product category and a newly launched product being fair trade. In the literature review of this thesis a thorough explanation will be provided on why these interactions are expected.

As consumers ‘vote’ with their buying behavior firms have to continually adapt to changing consumer values and attitudes. Muster (2011) found that the consumption of ethical products is highly desirable in a society. One method of ethical consumption is buying fair-trade products (De Pelsmacker et al. 2005; Shaw and Clarke, 1999). Ethical consumers are

dissatisfied with how the global market disadvantages producers relative to retailers and try to tackle these social and environmental problems by buying fair trade products (Taylor et al., 2005; LeClair, 2003). As the fair trade market is growing each year more demand for fair trade products is expected. Since growth is dependent on innovation (OECD, 2007) it is highly relevant for firms to know how newly launched fair trade products diffuse through society. Not only is this highly relevant for firms, it is also highly relevant for society as more ethical consumer behavior leads to a more honest and sustainable future.

1.3 Structure

8 The methodology section will provide more information on the chosen methodology. Which will then be followed by the results, discussion of these results and the interpretation of the outcomes. This thesis will conclude with managerial and academic implications, limitations and directions for further research.

2. Literature review

2.1 Consumer Innovativeness

In the context of this thesis, the innovations are new fair trade products and new regular products. Mapping the differences in innovators and imitators between fair trade products and non-fair trade products can give useful insights for society and retailers. For instance,

knowing which part of consumers adopted products based on marketing messages or based on word-of-mouth effects provides valuable information on how innovations diffuse through society.

Adoption of new products relies heavily on the innovativeness of consumers. Bass argued that during the diffusion of a new product potential adopters can be classed as either innovators or imitators. Innovators are the people who are the first to adopt the new product. But what makes for an innovative consumer? Midgley and Dowling (1978) defined innovativeness as the degree to which consumers are able to make innovative decisions based on their own beliefs and independently of others. Hirschman (1980) described innovativeness as novelty seeking and the desire for newness and difference. Hirschman’s definition of innovativeness seems more hedonistic than how innovation is defined by Midley and Dowling. Although, both definitions refer to a consumers’ individual beliefs for adopting innovations. A study conducted in Michigan by Taylor and Boasson (2014) suggests that consumers with

bachelor’s or master’s degrees bought fair trade products more often than consumers who do not have a bachelor’s or master’s degree. This points in the direction that consumers with a bachelor’s or master’s degree are more ethically minded than consumers who don’t have these degrees due to them being more sympathetic to non-market based outcomes (Taylor and Boasson, 2014).

9 Shaw (2005) found that in the United Kingdom values like curiosity, freedom, independence, and self-respect were related to ethical consumption. Doran (2009) found curiosity, freedom, independence and self-respect to positively influence ethical consumption. Both Shaw and Doran can be linked to Midgley, Dowling and Hirschmann’s definitions of innovative

consumers; a consumer who acts out of self-direction. Consumers who buy fair trade products do so out of self-direction and therefore can be attributed a certain level of innovativeness. Thus, one could argue that in the innovation diffusion of a fair trade product there are more innovators present than in the innovation diffusion of a regular product. All the earlier mentioned findings on innovativeness needed for fair trade products lead to the following hypothesis:

H1: In the innovation diffusion of new products, fair trade products attract more innovative consumers than non-fair trade products.

2.2 Social influence

The second class of consumers present during the innovation diffusion are imitators. Imitators are the consumers whose buying decision rely on influences coming from the social system. Cialdini and Goldstein (2004) found that people have the basic desire to fit in. In other words, people prefer to conform to the consumption behavior of the majority.

Buying fair trade products is form of ethical consumption. Ethical consumption or ethical consumerism can be defined as consumer behavior that reflects the consumers’ concerns with regard to problems arising from unethical and unjust trade (Uusitalo and Oksanen, 2004). What is interesting about ethical consumption or ethical consumerism is that consumers do not only consider individual values but also take into account social values (Uusitalo and Oksanen, 2004). There is evidence that the social system plays an important role in the adoption of fair trade products. In recent studies proof has been found for social contagion in the adoption process of ethical products (Iyengar, Van den Bulte and Valente, 2011;

10 This is consistent with more literature on ethical consumption which state that social norms play an important role in influencing ethical consumer behavior and that social influence is an important driver for ethical consumer behavior (Davies et al., 2003; Ryan, 2000; Lee, 2008; Smaldino et al. 2015). Summarizing the previous findings, it can be said that internal influences in the social system (or word-of-mouth effects) play an important role in the adoption process of fair trade products. This leads to the following hypothesis:

H2: In the innovation diffusion of new products, fair trade products are more affected by internal influences in the social system than non-fair trade products.

2.3 Market potential

Another important aspect of innovation diffusion is market potential (Bass, 1969). Market potential refers to the number of potential adopters a product can have during the product lifecycle.

Smith (1995), more than 10 years ago, already argued that we live in the era of ethics. This is partly due to consumers being better informed, better educated and more aware of consumer and producer rights (in Western society) (Hirschman, 1980; Barnes and McTavish, 1983). Mascarenhas (1995) found that information about a firm’s ethical behaviors influences product sales and consumers’ image of the firm. Thus, if a firm is selling fair trade products this will most likely positively influence their sales.

Availability in multiple retail outlets is very important in the adoption process of fair trade products and the success of a fair trade product introduction may rely heavily on visibility in the supermarket (Van Herpen et al., 2011). According to Desmet & Renaudin (1998) shelf-space positively influences product sales. Thus, if a newly introduced product receives the shelf space it needs, it will sell better. When multiple products in the same product category get clustered together, Van Herpen et al. (2011) argue that consumers notice these products at accelerated pace. Introducing new products can also have a positive effect on consumers noticing products (Hultink et al, 2000). Although, fair trade products are still more-or-less a niche market and receive less shelf-space in retail outlets than non-fair trade products. Thus, market potential for fair trade products will most likely be lower than for non-fair trade products.

11 They also found that some consumers are not convinced of the difference their decisions to buy fair trade products can make due to insufficient information communication about the benefits of fair trade to producers. Also, even in a time where ethically minded consumers are no longer a small pressure group, the part of society that sees themselves as ethically minded is still small (Co-op, 2000). One could argue that due the small part of society that sees themselves as ethically minded, smaller shelf-space in supermarkets and the fact that fair trade is still more-or-less a niche that market potential for fair trade products is lower than for non-fair trade products. Leading to the following hypothesis:

H3: In the innovation diffusion of new products, fair trade products have a lower market potential than non-fair trade products.

2.4 Degree of newness

There are two different kinds of product introductions, namely entirely new products and new variants. For a product to be classified as entirely new it has to have features or attributes which satisfies the user’s needs in a significantly differ manner than an older product

(Donnely and Etzel, 1973). New variants do not significantly differ from products already on the market and can be seen as product modifications or line extensions (Atuahene-Gima, 1996).

Consumers are inclined to view new products with caution and are less likely to buy products which bring uncertainty (Albayrak et al. 2011; Arts, Frambach and Bijmolt, 2011). In this context buying an entirely new product brings more uncertainty to the consumer than buying a new variant.

As stated in chapter one of the literature review, Zhou et al. (2002) found that risk averse consumers reduce risk by choosing products with a higher price. Shimp and Bearden (1982) found that risk aversion is an important factor in consumer decision making. Novelty seeking consumers take more risks and feel less threatened by entirely new products (Bao et al. 2003). For consumers with high risk aversion, new products are viewed as too risky. In section 1 of the literature review it is hypothesized that fair trade products attract more innovative

12 H4: If an entirely new product is fair trade, this attracts more innovative consumers than entirely new products which are non-fair trade.

2.5 Interactions between product category and fair trade

A difference of effects between product categories on adoption timing, number of adopters and internal influences in the social system are expected. Intuitively one could argue there is a difference between product introductions in different product categories. For instance, fair trade coffee, in absolute terms, has more sales than all the other product categories (IRI, 2017). Coffee sales have been fairly stagnate over the past few years while the “sustainable coffee” market is on the rise (Giovanunucci & Ponte, 2005; MacDonald, 2007; Taylor & Boasson, 2014). According to IRI (2017) fair trade coffee sales grew with 9.3% while the total market coffee sales decreased with 0.9% from 2015 to 2016. As opposed to the total market were tea sales grew with 2.2% and the sales of fair trade tea declined with 17.8% from 2015 to 2016. However, since the entire market for fair trade is growing each year it is

expected there are more differences than only in terms of sales. One could argue that there are also differences in innovation diffusion as new fair trade products keep entering the market and are claiming more shelf-space in mainstream retail outlets (Davies and Crane, 2003; NCDO, 2009). It is interesting to know whether this growth in sales is also due to internal or external influences and whether these influences differ among product categories.

The interaction between fair trade and product category will be tested to check for differences among product categories for fair trade products. Since the market for fair trade coffee is growing the fastest, there is reason to believe fair trade coffee differs from the other product categories. This leads to the following hypotheses:

Hypothesis 5: Fair trade coffee gets adopted faster compared to the fair trade products in the product categories tea, ice cream and chocolate.

Hypothesis 6: Fair trade coffee has a higher market potential than fair trade products in the product categories tea, ice cream and chocolate.

13

2.6 Conceptual model

The figure below provides a graphical representation of the effect of fair trade products relative to non-fair trade products on adoption timing (Hypothesis 1), social influence (Hypothesis 2) and market potential (Hypothesis 3). As well as the hypothesized interaction effects between a product being fair trade and degree of newness (Hypothesis 4) and a product being fair trade and product category (Hypothesis 5 to 7).

Figure 1: conceptual model

3. Methodology

3.1 Data description

The dataset contains information on 16936 new product introductions in 86 different product categories. The product categories ranged from non-food categories such as shampoo to food categories such as salad dressings, chocolate spread and biscuits. Information has been collected on the amount of first customers and amount of repeat customers per week. In total 365 weeks have been measured, starting in the fourth quarter of 2008 and ending in the third quarter of 2015. The data has been collected by GfK over a period of 7 years by their

14 Column name Description

barcode Barcode of the product bg_category_name Name of the product category barcode_description Detailed description of the product pafromdate Initial launchdate product

patodate Date product was taken off the market producttype New variant vs New product

pabewustekeuze Label, e.g. Fair trade

first_purch Date product was purchased for the first time last_purch Date product was purchased for the last time Table 1: Column names and descriptions

For this research four different product categories will be analyzed, namely: coffee, tea, chocolate and ice cream. The reason for this is that only in these product categories products with a fair trade label are found.

3.2 Data Cleaning

Data has been cleaned by omitting all rows and columns which were irrelevant for this research. All non-food product categories were omitted from the dataset, but also food categories in which no fair trade products were found were omitted. In total the dataset contained information on product introductions in 86 different product categories. In total there were 1765 product introductions in product categories relevant for this research. 432 (24.48%) product introductions were found in the product category coffee, 475 (26.92%) product introductions for chocolate, 480 (27.20%) product introductions for ice cream and 378 (21.41%) product introductions for tea. From these 1765 product introductions only 170 product introductions were fair trade products. 38(2.15% of the total number of product introductions) product introductions for coffee, 50(2.83%) product introductions for chocolate, 43(2.43%) product introductions for ice cream and 39(2.20%) product introductions for tea. The data has been analyzed using R.

The dataset contains information on the number of first time customers per week. For all products used in this thesis first time customers were aggregated to quarters due to the high amount of weeks (365). Aggregating the dataset to quarters instead of week provides a clearer overview of the dataset. Each quarter contains 13 weeks, resulting in 29 quarters in total. However, the 29th _{quarter only contained five weeks opposed to 13 weeks for the former 28}

15 The dataset contained no missing values. Any zeroes in the dataset are a result of a particular product not having been bought in a certain quarter.

3.3 Product selection

From the four product categories in which fair trade products were found, 10 products were chosen per product category, for fair trade and non-fair trade. Resulting in 10 products for fair trade products per product category and 10 products per product category for non-fair trade products. Chosen products range from products still active in the market and products which were pulled out of the market.

One could argue that when only investigating successful product introductions a selection bias occurs. The chosen products can be found in appendix 1. Products comparable in total first time purchases were chosen to make comparison equal.

As mentioned earlier in the data description section of this thesis, the number of customers that made a first purchase were aggregated from weeks to quarters. To be able to estimate the coefficients a, b and c with which the parameters p, q and m can be calculated a new dataset had to be made. This dataset existed of six columns. An example has been provided in the table below. ‘P_t’ refers to quarter, ‘Sales’ refers to the number of first purchases customers per quarter, ‘cumsales’ refers to cumulative sales, ‘cumsaleslag’ refers to the lagged

cumulative sales, cumsalessqrt refers to the square of the lagged cumulative sales and ‘ProdID’ refers to the barcode of the product.

P_t Sales Cumsales Cumsaleslag Cumsalessqrt ProdID

4 5 5 0 0 7,62E+12 5 7 12 5 25 7,62E+12 6 4 16 12 144 7,62E+12 7 5 21 16 256 7,62E+12 8 4 25 21 441 7,62E+12 9 2 27 25 625 7,62E+12 10 4 31 27 729 7,62E+12 11 8 39 31 961 7,62E+12 12 11 50 39 1521 7,62E+12

Table 2: Example of the new dataset

3.3 Newness

16 For instance, a product receiving a new packaging can be described as a new variant while, for instance, a new flavor can be described as a new product.

An example of a new variant is the Swiss pure chocolate bar, which slightly differs from the older version. An example of a new product is Milka naps, which is a new flavor Milka bar, thus an entirely new sub-brand of Milka and therefore can be seen as an entirely new product. Another example is Café Royal espresso caramel, which is an entirely new flavor released by coffee brand Café Royal. Café royal also released an updated version of their espresso, which can be classed as a new variant.

3.4 Method

A comparison of Bass diffusion models between fair-trade and non-fair-trade products in the same product category will be employed to investigate the difference in adoption timing, number of adopters and internal influence in the social system. The purpose of the Bass Model is to get an understanding of the speed at which new products get adopted and diffuse through the market. The Bass model employed in this research is descriptive of nature.

The Bass diffusion model is a hazard model which models adoption timing and the potential number of adopters based on three variables which will be explained in more detail later in this chapter. The reason the Bass model will be employed for analysis lies in its relative easiness of how to interpret the coefficients. Since the Bass diffusion model is a simplistic representation of reality it has a few assumptions:

1. Over the life of the product there will be a number of initial purchases of the product, m.

2. For the Bass model only first time purchases will be taken into account

3. The assumption of the Bass diffusion model is that adoption of innovations rely on internal and external influences. Internal influences refer to interactions in the social systems while external influences come from outside the social system.

17 𝑁_𝑡= [𝑝 + 𝑞 (𝑁𝑡−1

𝑚 )] (𝑚 − 𝑁𝑡−1)

Where:

Nt = number of adopters at time t.

Nt-1 = lagged number of adopters.

p = coefficient of innovation (external influences).

q = coefficient of imitation (internal influences).

m = market potential or number of potential ultimate adopters.

The innovation coefficient, p, relies on external influences, e.g. marketing messages.

Although, the innovation coefficient is called the innovation coefficient it does not necessarily refer to the first buyers. In the Bass diffusion model innovators can be present during the entire diffusion process (Mahajan et al., 1990). The imitation coefficient, q, relies on

influence coming from within the social system and can be seen as a social contagion process. Innovators are not influenced by the number of previous purchasers, although imitators are influenced by the number of previous purchasers.

The next step is an ordinary least squares regression to estimate a, b and c. An ordinary least squares regression was applied due to the relative easiness of implementation compared to other regression methods such as nonlinear least squares. The number of adopters at time t (𝑛_𝑡) is the dependent variable while the cumulative lag sales (𝑁_𝑡−1) and cumulative lag sales squared (𝑁²_𝑡−1) are used as independent variables. This leads to the following regression equation:

𝑛𝑡 = 𝑎 + 𝑏𝑁𝑡−1+ 𝑐𝑁²𝑡−1

Or in words

18 Where: 𝑎 = 𝑝 ∗ 𝑚 𝑏 = (𝑞 − 𝑝) 𝑐 = −(𝑞 𝑚)

With the parameter estimates for a, b and c it is possible to calculate p, q and m. To calculate m the following equation has to be used:

𝑚 =−𝑏 ± √𝑏

2_{− 4𝑎𝑐}

2𝑐

The quadratic equation produces a value for m from which the parameters p and q can be derived. This is done as follows:

𝑝 = 𝑎 𝑚

𝑞 = (𝑏 + 𝑝)

There are a few more useful equations which can be derived from the Bass Model equation. For instance, sales at time T, as given by the following formula:

𝑆(𝑇) = 𝑝 [𝑚 − 𝑌(𝑇)] + 𝑞

𝑚𝑌(𝑇)[𝑚 − 𝑌(𝑇)]

Where the first part of the equation(𝑝 [𝑚 − 𝑌(𝑇)]) refers to the adoptions made by innovators (external influence) and the second part (𝑞

𝑚𝑌(𝑇)[𝑚 − 𝑌(𝑇)]) refers to the adoptions made by

imitators (internal influence).

The research will focus on comparison of Bass diffusion models between fair trade and non-fair trade products. An average innovation diffusion curve will be plotted for every product category visualize the differences in the innovation diffusion between fair trade and non-fair trade products.

3.5 Regression analysis

19 A new dataset was created with the estimates of p, q and m as well as dummy variables for product category, fair trade and whether a product introduction is a new variant or a new product. In this dataset fair trade products will receive the code ‘1’ while non-fair trade products will be coded as ‘0’.

This is the same for new product and new variant products, where new product will receive the code ‘1’ and new variant will receive the code ‘0’. In total there are 15 new products (Three for fair trade and 12 for non-fair trade) and 65 new variants (37 for fair trade and 28 for non-fair trade). Product category will be coded from 1 to 4, where chocolate is 1, ice cream is 2, coffee is 3 and tea is 4. In R product category was transformed to a factor with chocolate as the reference category. By transforming product category to a factor R automatically identifies the different levels as dummies.

This allows for an OLS regression with which the differences in adoption timing (p), number of adopters (m) and internal influences in the social system (q) can be tested. Leading to the following three regression equations:

𝑝 = 𝛽₀+ 𝛽₁𝑥_{𝐹𝑇𝑁𝐹𝑇}+ 𝛽₂𝑥_{𝑁𝑃𝑁𝑉}+ 𝛽₃𝑥_𝑃𝐶+ 𝛽₄𝑥_{𝐹𝑇𝑁𝐹𝑇∗𝑁𝑃𝑁𝑉}+ 𝛽₅𝑥_{𝐹𝑇𝑁𝐹𝑇∗𝑃𝐶}+ 𝜀

𝑞 = 𝛽0+ 𝛽1𝑥𝐹𝑇𝑁𝐹𝑇+ 𝛽2𝑥𝑁𝑃𝑁𝑉+ 𝛽3𝑥𝑃𝐶+ 𝛽4𝑥𝐹𝑇𝑁𝐹𝑇∗𝑁𝑃𝑁𝑉+ 𝛽5𝑥𝐹𝑇𝑁𝐹𝑇∗𝑃𝐶+ 𝜀

𝑚 = 𝛽₀+ 𝛽₁𝑥_{𝐹𝑇𝑁𝐹𝑇}+ 𝛽₂𝑥_{𝑁𝑃𝑁𝑉}+ 𝛽₃𝑥_𝑃𝐶 + 𝛽₄𝑥_{𝐹𝑇𝑁𝐹𝑇∗𝑁𝑃𝑁𝑉}+ 𝛽₅𝑥_{𝐹𝑇𝑁𝐹𝑇∗𝑃𝐶}+ 𝜀

Where:

FTNFT = Fair trade vs non-fair trade products

NPNV = New product vs new variant

PC = Product category with four levels: chocolate, coffee, tea and ice cream

FTNFT * NPNV = interaction between fair trade vs non-fair trade products and new product vs new variant

20

4. Results

4.1 Bass model parameters

One of the assumptions of the Bass diffusion model is that parameters ‘p’ and ‘q’ have a value between zero and one (Bass, 1969). Not only is it necessary that these parameters have a value between zero and one, the parameters should also be positive for the model to be applicable (Bass, 1969; Jiang et al. 2006). There are different reasons for why the parameters cannot be estimated, namely: for some products there are less than four periods of first time purchases or there were less than 25 purchases in total. In both cases the parameters ‘p’ and ‘q’ receive a negative value or no value at all. In table 3 the average values for parameters ‘p’, ‘q’ and ‘m’ per product category are provided.

Product

category Average p Average q Average m

FT chocolate 0.0606 0.2346 603.5722 NFT chocolate 0.0077 0.1750 390.4106 FT Ice cream 0.0818 0.4171 256.8832 NFT Ice cream 0.0040 0.3899 420.3008 FT Tea 0.0420 0.1687 117.9438 NFT tea 0.0102 0.2242 168.8753 FT Coffee 0.0345 0.1505 62.6978 NFT coffee 0.0072 0.1575 302.0292

Table 3: average Bass parameter scores.

Three One-way ANOVA’s were conducted to compare the effects of fair trade products relative to non-fair trade products on p, q and m. Also, three more One-way ANOVA-tests were conducted to compare the effects of product categories on p, q and m.

The one-way ANOVA showed that, using a confidence interval of 5%, among product categories there are no significant differences in the parameter ‘p’ (F = 0.77, p = 0.515). Suggesting that there are no significant differences in adoption timing among product categories.

There are significant differences in internal influences in the social system (q) for product categories when tested using an One-way ANOVA (F = 7.436, p = <0.001). Using a

21 No significant differences were found between product categories chocolate and coffee (p = 0.651) and also no significant differences were found between product categories chocolate and tea.

The TukeyHSD test was employed again to test for significant differences between product categories for market potential (m). No significant differences were found between product categories chocolate and ice cream (p = 0.944), between ice cream and coffee (p = 0.206), between ice cream and tea (p = 0.13) and between coffee and tea (p = 0.996). However, there are significant differences in market potential between product categories chocolate and tea (p = 0.0381) and marginally significant differences between chocolate and coffee (p = 0.067).

Product Average p Average q Average m

FT 0.0547 0.2427 260.2743

NFT 0.0073 0.2367 320.404

Table 4: average value FT/NFT

The One-way ANOVA showed that, using a confidence interval of 5%, there are significant differences in adoption timing for fair trade products relative to non-fair trade products. (F = 24.29, p = <0.001).

There are no significant differences in the parameter ‘q’ between fair trade products and non-fair trade products when tested using an One-way ANOVA (F = 0.061, p = 0.806).

Suggesting, there are no significant differences in internal influences in the social system for fair trade products relative to non-fair trade products.

There is a marginally significant difference in market potential between fair trade products and non-fair trade products when tested using an One-way ANOVA (F = 3.268, p = 0.0745). And there are significant differences among product categories when tested using an One-way ANOVA (F = 3.806, p = 0.014).

A more thorough explanation of the differences in p, q and m and possible reasons for these differences will be provided in section 5.

4.2 Diffusion plots

Figure 2 to 5 show a graphical representation of the different p, q and m values in table 3. The vertical axis represents the number of products sold at time t while the horizontal axis

22 As can be seen in the graphs, fair trade products seem to be adopted at a faster rate than non-fair trade products as is in line with the higher p values for non-fair trade products. Although, market potential seems lower for fair trade than for non-fair trade. Figure 6 to 9 show the cumulative sales curves as estimated by the Bass diffusion model. Again, the blue line

represents non-fair trade products while the green line represents fair trade products. Similarly as with the inverted u-shape diffusion curves the cumulative diffusion curves also depict that adoption timing for fair trade products in the aforementioned product categories is higher. And again, these lines suggest that market potential is significantly lower for fair trade products than for non-fair trade products.

23 Figure 6 to 9: Average cumulative sales curve per product category

5. Regression results

5.1 OLS assumptions and validation

Before interpreting the OLS results, a few tests should be performed to investigate whether the model adheres to model assumptions for OLS. The assumptions to test for are the nonzero expectation, heteroscedasiticy, multicollinearity and non-normality test (Leeflang, Bijmolt, Pauwels & Wieringa, 2015). After testing for the OLS assumptions the regressions can be conducted.

5.1.1 Nonzero expectation

An assumption of OLS is that the errors in the regression should be exogenous. Thus, the error terms should have a conditional mean of zero. If this assumption fails to hold the regressors are endogenous and the OLS estimates become invalid. A method to test for the nonzero expectation is the Ramsey RESET test (Leeflang, Bijmolt, Pauwels & Wieringa, 2015). Dep.Var. RESET test P-value p 0.537 0.587 q 0.996 0.375 m 0.331 0.720

24 As can be seen in table 7, none of the p-values for the Ramsey RESET-test fall below the critical value of 0.05 and therefore it can be concluded there the conditional mean of the error terms is zero.

5.1.2 Heteroscedasticity

Another assumption of OLS is that the residuals should be homoscedastic and thus, there should be no heteroscedasticity present in the model. When heteroscedasticity is present in the models the parameter estimates become less efficient. One method to check for

heteroscedasticity is plotting the residuals. In figure 10 to 12 the plots of the residuals are given. Where the top left figure is the model with m as dependent variable. The top right figure the model with p as dependent variable and the bottom figure the model with q as dependent variable.

Figure 10 to 12: plot of residuals of independent variables, m, p and q.

25 In table 5 the results of the Breusch-Pagan test are given for each model. The interpretation of the Breusch-Pagan test is as follows: if the p-value falls below the confidence interval of 5%, there is heteroskedasticity present in the model (Breusch and Pagan, 1979).

Dep. Var. Breusch-Pagan P-value p 10.625 0.302 q 11.404 0.249 m 13.226 0.153

Table 6: BP-test results

As can be seen in table 5 none of the p-values falls below the confidence interval of 5%. Therefore, it can be concluded that there are no problems with heteroskedasticity in the dependent variables, p, q and m.

5.2.3 Multicollinearity

Independent variables should be independent from one another. In a model were interaction effects are hypothesized it might be possible to run into multicollinearity (Danaher & Dagger, 2013). Testing for multicollinearity is done by comparing the variance inflation factor, or VIF, and to check whether these values exceed the threshold of five or by calculating 1/VIF which is not allowed to fall below 0.2 (Leeflang, Bijmolt, Pauwels & Wieringa, 2015).

For the following three models a test for multicollinearity will be performed:

𝑝 = 𝛽0+ 𝛽1𝑥𝐹𝑇𝑁𝐹𝑇+ 𝛽2𝑥𝑁𝑃𝑁𝑉+ 𝛽3𝑥𝑃𝐶+ 𝛽4𝑥𝐹𝑇𝑁𝐹𝑇∗𝑁𝑃𝑁𝑉+ 𝛽5𝑥𝐹𝑇𝑁𝐹𝑇∗𝑃𝐶+ 𝜀

𝑞 = 𝛽₀+ 𝛽₁𝑥_{𝐹𝑇𝑁𝐹𝑇}+ 𝛽₂𝑥_{𝑁𝑃𝑁𝑉}+ 𝛽₃𝑥_𝑃𝐶+ 𝛽₄𝑥_{𝐹𝑇𝑁𝐹𝑇∗𝑁𝑃𝑁𝑉}+ 𝛽₅𝑥_{𝐹𝑇𝑁𝐹𝑇∗𝑃𝐶}+ 𝜀

𝑚 = 𝛽₀+ 𝛽₁𝑥_{𝐹𝑇𝑁𝐹𝑇}+ 𝛽₂𝑥_{𝑁𝑃𝑁𝑉}+ 𝛽₃𝑥_𝑃𝐶 + 𝛽₄𝑥_{𝐹𝑇𝑁𝐹𝑇∗𝑁𝑃𝑁𝑉}+ 𝛽₅𝑥_{𝐹𝑇𝑁𝐹𝑇∗𝑃𝐶}+ 𝜀

Variable VIF Score 1/VIF

FTNFT 4.710 0.212 NPNV 1.599 0.625 Coffee 3.411 0.293 Tea 3.192 0.313 Ice cream 3.192 0.313 NPNVxFT 1.590 0.630 FTxCof 3.971 0.252 FTxTea 3.639 0.275 FTxIce 3.639 0.275

Table 7: VIF scores and 1/VIF scores

26

5.3.4 Normality

The errors in a model should be normally distributed for hypothesis testing to make sense. The Normal Q-Q plots in R provide a visual check for normality in the residuals. Figure 13 to 15 show the Normal Q-Q plots as provided by R for the three models. Where the top left figure is the model with m as dependent variable. The top right figure the model with p as dependent variable and the bottom figure the model with q as dependent variable. As can be seen in the plots non-normality seems to be an issue for all three models.

Figure 13 to 15: Normal qq plots

Another method to check for non-normality in the residuals is by employing a Shapiro-Wilk test.

Dep. Var. Shapiro-Wilk P-value

p 0.635 <0.001

q 0.933 <0.001

m 0.810 <0.001

27 As can be seen in table 10 all of the p-values falls below the confidence interval of 5%.

Therefore, it can be concluded the residuals of the three models with dependent variables, p, q and m show signs of non-normality.

Although, Lumley et al. (2002) found that for linear regression the assumption of normally distributed residuals is not required for sufficiently large samples. They found that sample sizes less than 100 are already sufficiently large enough. Field (2009) states, based on the Central Limit Theorem (CLT), an approximate normal distribution of the residuals can be assumed with a sample larger than 30. In this thesis the sample size is larger than 30, namely 80.

5.1.5 Serial correlation and constant parameters over time

Serial correlation and constant parameters over time are not relevant in this case as these assumptions are only important when dealing with time series data. In this case the interest lies with testing whether there are any significant differences in Bass Model parameters.

In conclusion, all assumptions for OLS regression are met and the regressions can be conducted.

5.2 Results and interpretation

The next step is examining the model outputs for all three dependent variables. In table 10 to 12 all parameter estimates are presented. In every regression product category ‘chocolate’ was used as the reference category and thus the intercept can be interpreted as the effects of

product category ‘chocolate’ on the dependent variables. p Estimate Std. Error t-value P-value Intercept 0.006 0.016 0.381 0.704 FTNFT 0.062 0.021 2.915 0.005 NPNV 0.004 0.015 0.243 0.809 Coffee 0.0003 0.021 0.019 0.985 Tea 0.003 0.020 0.152 0.880 Ice cream -0.003 0.020 -0.163 0.871 NPNVxFT -0.023 0.028 -0.828 0.412 FTxCof -0.032 0.028 -1.137 0.259 FTxTea -0.025 0.028 -0.888 0.377 FTxIce 0.017 0.028 0.625 0.534

28 The R-square of the model with ‘p’ as dependent variable is 0.325 (F = 3.692, p = 0.001), meaning the model is significant at a 1% confidence interval. Meaning 32.5% of the variance in the dependent variable is explained by the independent variables. Only one parameter estimate is significant at a 5% confidence interval. Fair trade products compared to non-fair trade products have a higher ‘p’. More precisely, p is 0.062 higher for fair trade products than non-fair trade products on average.

Thus, a product being fair trade attracts more innovative consumers than non-fair trade products and therefore hypothesis 1 can be accepted. All other variables p-values are above 0.05 and therefore it can be concluded the coefficients of these variables do not significantly differ from zero and thus do not influence the dependent variable, p. The hypothesized interaction effect between new products and fair trade products does not hold and thus hypothesis 4 has to be rejected. Therefore, it can be said that entirely new fair trade products do not attract more innovative customers than entirely new non-fair trade products and new variants are not being adopted at a faster pace than entirely new products.

q Estimate Std. Error t-value P-value Intercept 0.135 0.067 2.012 0.048 FTNFT 0.121 0.089 1.347 0.182 NPNV 0.089 0.064 1.385 0.170 Coffee 0.002 0.087 0.027 0.978 Tea 0.062 0.084 0.737 0.463 Ice cream 0.228 0.084 2.701 0.009 NPNVxFT -0.066 0.118 -0.560 0.577 FTxCof -0.113 0.118 -0.954 0.343 FTxTea -0.127 0.118 -1.083 0.283 FTxIce -0.067 0.118 -0.561 0.576

Table 10: q as dependent variable

29 To facilitate a more thorough understanding of the parameter estimates one could say that q is 0.135 higher for chocolate and 0.135 + 0.228 = 0.363 higher for ice cream on average,

holding all other parameters equal.

A product being fair trade does not lead to a higher q and therefore hypothesis 2 has to be rejected. Fair trade products are not more affected by internal influences in the social system than non-fair trade products.

m Estimate Std. Error t-value P-value Intercept 445.92 98.019 4.549 <0.001 FTNFT -58.25 129.947 -0.448 0.655 NPNV -124.90 94.147 -1.327 0.189 Coffee -116.14 127.088 -0.914 0.364 Tea -239.58 122.935 -1.949 0.054 Ice cream 11.85 122.935 0.096 0.924 NPNVxFT 80.95 172.063 0.470 0.640 FTxCof -204.79 172.200 -1.189 0.238 FTxTea -6.563 171.647 -0.038 0.970 FTxIce -142.639 171.647 -0.831 0.410 Table 11: m as dependent variable

The model with ‘m’ as dependent variable returns an R-square of 0.221 (F = 2.172, p = 0.035), meaning this model is significant at a 5% confidence interval. Again, two parameter estimates are significant, the intercept and tea. Where tea is only marginally significant due to its p-value being 0.058 and thus above the confidence interval of 5%. Chocolate has an average market potential of 445.92 products whereas tea has an average market potential of 445.92 – 239.58 = 206.34 products. This could be due to the more accessible nature of chocolate, it can eaten anywhere as for tea hot water is needed. Thus, chocolate could benefit more from impulse purchases leading to a higher market potential. Fair trade products do not have a significantly lower market potential than non-fair trade products and the interaction between entirely new products and fair trade products does not have a significantly lower market potential. Therefore, hypothesis 3 has to be rejected. The reason the market potential for fair trade products could be due to the fact that fair trade chocolate on average has a high market potential compared to non-fair trade chocolate.

30

6. Discussion

This study investigated the research question whether there are any significant differences in the adoption process of fair trade products versus non-fair trade products as well as the question whether fair trade products have a higher market potential than non-fair trade products. Not only is it important to know whether there are any significant differences, it is also interesting to know if there are any differences across product categories and if there are any differences in the adoption process if a product is a new variant or an entirely new product.

1. Are there significant differences in the adoption rate of fair trade and non-fair trade products?

In this study the parameter ‘p’ or the innovation coefficient has been very important in determining adoption timing for a product, whether fair trade or non-fair trade. The results show there is indeed a significant difference between fair trade and non-fair trade products regarding adoption timing. Thus, Dutch consumers adopt products faster when a product is fair trade, which is in line with previous literature on ethical consumption in which it was stated that people buying ethical products score higher on innovativeness (Doran, 2009; Shaw, 2005). One could think of multiple reasons for fair trade attracting more innovative customers. One such reason could be that fair trade, due to its relative newness in the market, brings too much risk for the average consumer and thus only more risk-taking consumers buy these products. Which is in line with previous findings by Bao et al. (2003) who state that innovative consumer feel les threatened by risks associated with entirely new products. Another reason could be that some consumers are not convinced of the positive impact fair trade has on producer rights due to a lack of information on fair trade. While more innovative consumers choose to inform themselves on the aforementioned issue and thus choose to buy fair trade products. Which is in line with previous findings by Carrigan and Attalla (2001).

When looking at the effects of internal influences in the social system it can be said there are no differences between fair trade products and non-fair trade products in this study.

31 It could also be due to consuming fair trade products not being the norm yet. This explanation would be in line with previous literature stating social norms play an important role in

influencing ethical consumer behavior (Davies et al., 2003; Ryan, 2000; Lee, 2008; Smaldino et al. 2015).

2. How does market potential of new product introductions of fair trade products differ from non-fair trade products in the same product category?

There are no significant differences in market potential for new fair trade products relative to new non-fair trade products. From these results one could argue that fair trade products have become more mainstream over the years. This is in line with a report on labels by IRI (2017), in which they have found that the market for fair trade products is still growing each year. When looking at the difference in sales from 2015 to 2016 in the product categories ‘coffee’, ‘tea’ and ‘chocolate’, the product category fair trade chocolate alone grew 116.7% (IRI, 2017). These results confirm that the Dutch consumer is more conscious about the origin of their products and whether everyone in the production chain receives a fair compensation.

3. Are there differences between product categories?

Interestingly, no differences were found for adoption timing for product categories. This suggests that among product categories adoption timing does not differ. A reason for this could be that coffee and tea require low involvement of the consumer. Coffee and tea,

32 The most notable differences between product categories lies in market potential. Chocolate has a significantly higher market potential than tea. One could argue that this is due to chocolate being more prone to impulsive purchases compared to tea, as chocolate can be consumed on the go. One the other hand, it could again be due to heuristic buying behavior. Consumers will most likely not buy chocolate heuristically and thus are more open to new chocolate products.

7. Academic implications

Findings of this thesis provide a basis for researching innovation diffusion of fair trade products or ethical products in general further. This thesis made an attempt at combining existing literature on adoption timing, social influence and consumer attitudes regarding fair trade products and hereby add to the existing research on ethical consumption. Relationships supported by previous academic literature have been found. However, relationships which were not found in previous literature were also found and for these relationships possible explanations are provided. Moreover, this study adds to the existing literature by mapping the innovation diffusion of fair trade products relative to non-fair trade products and providing possible explanations for these differences.

8. Managerial implications

From discussion of the results of this paper some interesting implications for marketers and firms come forward. This paper its aim was to investigate whether there are significant differences in the innovation diffusion process of fair trade and non-fair trade products. Some significant difference have been found. First, the difference in adoption timing between fair trade and non-fair trade products. The data has shown that fair trade products are adopted at a significantly faster pace than non-fair trade products. This opens up possibilities for younger companies looking to enter the market. If they enter the market with a product that carries a fair trade label their product will get adopted faster leading to higher revenues in the short run and making it easier to launch new products in the long run. Not only do younger companies profit from this, so to the producers of the raw materials for these products. Producers in developing nations receive fairer compensations thanks to fair trade initiatives.

33 Even if there are no differences in social influence, positive word-of-mouth should be

promoted, as earlier findings (Iyengar, Van den Bulte and Valente, 2011; Kulviwat, Bruner and Al-Shuridah, 2009) indicate social influence play an important role in the adoption process of fair trade products.

Secondly, market potential for fair trade products and non-fair products does not significantly differ. Proving that the market potential of fair trade products is comparable to the market potential of non-fair trade products. Suggesting that Dutch consumers are ready for more ethical products to enter the market and that fair trade has risen out of the niche market it once was. Fair trade products are no longer a risky investment as Dutch consumers adopt these products at a faster pace while market potential is not significantly lower. This implies that when more companies switch to fair trade products the working conditions of producers in developing nations can make a change for the better.

Moreover, the results suggest that there are no differences between new products and new variants in the adoption process, not only for fair trade but also for non-fair trade. New products and new variants are equally successful forms of innovations in the four product categories studied here. The risks associated with entirely new products did not prove to play such an important role as was hypothesized. Thus, marketers looking to promote fair trade products do not have to have worry about whether a new product is a new variant of an already existing product or an entirely new product.

Also, no differences in the innovation diffusion were found among fair trade product categories. Marketers can benefit from this knowledge as more-or-less the same marketing strategy can be used among different product categories for fair trade products as external influences do not pay a significantly differing role among products categories. However, managers have to keep in mind that in this thesis only a relatively small amount of products were investigated per product category.

9. Limitations and future research

This research also has some limitations and there is still room for future research. First of all, the dataset has its limitations. There is a lot of data but only a handful of new fair trade

34 Furthermore, only a limited part of all product introductions could be analyzed as not all products were suitable for estimating a Bass model. Suggesting that the Bass diffusion might not be the most suitable hazard model for measuring differences in the adoption process of fair trade products relative to non-fair trade products.

In future research a more elaborate dataset could be used to investigate what factors influence the adoption process of fair trade products relative to non-fair trade products. For instance, a multinomial logit model could be employed with fair trade as a dependent variable to

investigate what factors increase the probability of buying fair trade products relative to non-fair trade products.

Secondly, there are a few restrictions to the Bass model employed in this research. The Bass model used in this paper was Frank Bass its first Bass model developed in 1969 while numerous extensions have been made to the Bass model. One of the most notable, of course, is an extension developed in 1994 by Frank Bass, Trichy Krishnan and Dipak Jain. A version of the Bass diffusion model which also includes pricing. Browne et al. (2000) argued that 80% of the world population would buy fair trade products if there were no price premiums on fair trade products. Creyer and Ross (1997) found that consumers are willing to pay higher prices for products produced by firms that show ethical behavior. Also, price is an important quality indicator for the consumer (Erdem, 1998; Erdem and Swait, 2004; Ngobo, 2011) Giving rise to the idea that price promotions are beneficial for the sales of fair trade products.

And according to De Pelsmacker & Janssens (2007) excessive price is still the most important reason for not buying fair trade products. Although, this contradicts previous literature where consumers were willing to pay a price premium for fair trade products (Creyer and Ross, 1997; Nimon and Beghin, 1999; Johnston et al., 2001). Not only price could be an important variable influencing adoption timing so could advertising and distribution be. Van Herpen et al. (2011) argued that distribution is an important driver of success for the sales of fair trade products. A model in which more marketing mix instruments can be implemented could explain more of the variance in the dependent variable.

This gives rise to the idea that more marketing mix instruments should be included in future research on innovation diffusion of fair trade products compared to non-fair trade products.

35 data on multiple countries could be used to compare results across countries. This could also make the results and managerial implications more generalizable.

9. References

Albayrak, T., Caber, M., Moutinho, L. & Herstein, R. (2011). The influence of skepticism on green purchase behavior. International Journal of Business and Social Science. Vol. 2. No. 13. Pp. 189 – 197.

Arts, J. W. C., Frambach, R. T. & Bijmolt, T. H. A. (2011). Generalizations on consumer innovation adoption: A meta-analysis on drivers of intention and behavior. International Journal of Research in Marketing. Vol. 28. No. 2, pp 134 – 144.

Atuahene-Gima, K. (1996). Differential potency of factors affecting innovation performance in manufacturing and services firms in Australia. Journal of Product Innovation Management. Vol. 13. No. 1, pp. 35 – 52.

Bao, Y., Zhou, K. Z. & Su, C. (2003). Face Consciousness and Risk Aversion: Do They Affect Consumer Decision Making? Psychology and Marketing. Vol. 20. No. 8, pp. 733 – 755. Barnes, J. G. & McTavish, R. (1983). Segmenting Industrial Markets by Buyer Sophistication. European Journal of Marketing, Vol. 17. No. 6, pp. 16 – 33.

Bass, F. M., Krishnan, T. V. & Jain, D. C. (1994). Why the Bass Model Fits without Decision

Variables. Marketing Science. Vol. 13. No. 2, pp. 203–223.

Bass, F. (1969). A new product growth for model consumer durables. Management Science. Vol. 15. No. 5, pp. 215–227.

Breusch, T. S. & Pagan, A. R. (1979). A Simple Test for Heteroscedasticity and Random Coefficient Variation. Econometrica. Vol. 47. No. 5, pp. 1287 – 1294.

Browne, A. W., Harris, P. J. C., Hofny-Collins, A. H., Pasiecznik, N. & Wallace, R. R. (2000). Organic Production and Ethical Trade: Definition, Practice and Links. Food Policy. Vol. 25, pp. 69–89.

36 Costa, D. P., Herter, M. M., Rossi, P. & Borges, A. (2014) Going green for self or for others? Gender and identity salience effects on sustainable consumption. International Journal of Consumer Studies, Vol. 38, No. 1, pp 540 – 549.

Creyer, E. H. (1997). The influence of firm behavior on purchase intention: do consumers really care about business ethics? Journal of Consumer Marketing. Vol. 14. No. 6, pp. 421 – 432.

Danaher, P. J. & Dagger, T. S. (2013). Comparing the relative effectiveness of advertising channels: A case study of a multimedia blitz campaign. Journal of Marketing Research. Vol. 50. No. 4, pp. 517 – 534.

Davies, I. A. & Crane, A. (2003). Ethical Decision Making in Fair Trade Companies. Journal of Business Ethics. Vol. 45, pp. 79.

De Pelsmacker, P., Janssens, W., Sterckx, E. & Mielants, C. (2006). Fair-Trade Beliefs, Attitudes and Buying Behaviour of Belgian Consumers. International Journal of Nonprofit and Voluntary Sector Marketing. Vol. 11, pp. 125 – 138.

De Pelsmacker, P. & Janssens. W. (2007). A Model for Fair Trade Buying Behavior: The

Role of Perceived Quantity and Quality of Information and of Product-Specific Attitudes.

Journal of Business Ethics. Vol. 75, No. 4, pp. 361 – 380.

Desmet, P., & Renaudin, V. (1998). Estimation of product category sales responsiveness to allocated shelf space. International Journal of Research in Marketing. Vol. 15. No. 5, pp. 443 – 457.

Donnelly, J. H. & Etzel, M. J. (1974). Degrees of Product Newness and Early Trail. Journal of Marketing Research, Vol. 10. No. 3, pp. 295.

Doran. C. J., (2009). The Role of Personal Values in Fair Trade Consumption. Journal of Business Ethics. Vol. 84. No. 4, pp. 549 – 563.

Erdem, T. (1998). An empirical analysis of umbrella branding. Journal of Marketing Research. Vol. 35. No. 3, pp. 339 – 351.

37 Field, A.P. (2009). Discovering statistics using SPSS: and sex and drugs and rock ‘n’ roll (3rd edition). London: Sage

Fornari, D., Grandi, S., & Fornari, E. (2009). The role and amangement of product innovation in retailer assortments: evidence from the Italian FMCG market. The International Review of Retail, Distribution and Consumer Research. Vol. 19. No. 1, pp. 29 – 43.

Forte, M. & Lamont, B.T. (1998). The bottom line effects of greening: implications of environmental awareness. Academy of Management. Vol. 12. No. 1, pp. 89-90.

Fourt, L. A. & Woodlock, J. W., (1960). Early prediction of market success for new grocery products. Journal of Marketing. Vol. 25, pp. 31 – 38.

Gelb BD. 1995. More boycotts ahead? Some implications. Business Horizons. Vol. 38. No. 2, pp. 70 –77.

Giovannucci, D. & Ponte, S. (2005). Standards as a New Form of Social Contract? Sustainability Initiatives in the Coffee Industry. Food Policy. Vol. 30, No. 3, pp. 284-301. Hirschman, E. C. (1980). Innovations, novelty seeking and consumer creativity. Journal of Consumer Research. Vol. 7. No. 3, pp. 283 – 95.

Hultink, E.J., Hart, S., Robben, H.S.J. & Griffin, A. (2000). Launch decisions and new product success: an empirical comparison of consumer and industrial products. Journal of Product Innovation Management. Vol. 17, pp. 5-23.

IRI, (2017). Monitor Keurmerken retail 2017.

Jiang, Z., Bass, F. M. & Bass, P. I., (2006). Virtual Bass Model and the left-hand data-truncation bias in diffusion of innovation studies. International Journal of Research in Marketing. Vol. 23, pp. 93 – 106.

Johnston, R.J., Wessells, C.R., Donath, H. & Asche, F., (2001). Measuring consumer preferences for eco-labeled seafood: an international comparison. American journal of Agricultural Economics. Vol. 26 No. 1, pp. 20 – 39.

38 Krier, J. M. (2005). Fair Trade in Europe 2005. The Heinrich Boll Foundation, Retrieved October 15th, 2017, from http://www.ifat.org/downloads/marketing/Fair

TradeinEurope2005.pdf.

Kulviwat, S., Bruner, G. C. & Al-Shuridah, O. (2009). The role of social influence on adoption of high tech innovations: The moderating effect of public/private consumption. Journal of Business Research. Vol. 62. No. 7, pp. 706 – 712.

LeClair, M. S. (2003). Fighting Back: The Growth of Alternative Trade. Development. Vol. 46. No. 1, pp. 66 – 73.

Leeflang, P. S.H., Bijmolt, T. H.A., Pauwels, K. & Wieringa, J. E. (2015). Modeling Markets. New York: Springer.

Lumley, T., Diehr, P., Emerson, S. & Chen, L. (2002). The Importance of the Normality Assumption in Large Public Health Data sets. Annual Review of Public Health. Vol. 23, pp. 151 – 169.

Macdonald, K. (2007). Globalising Justice within Coffee Supply Chains? Fair Trade,

Starbucks and the Transformation of Supply Chain Governance. Third World Quarterly. Vol. 28, No. 4, pp. 793 – 812.

Mahajan, V. & E. Muller (1995), Timing, Diffusion and Substitution of Successive Generations of Technological Innovations: The IBM Mainframe Case. Working Paper, Department of Marketing, The University of Texas of Austin.

Mascarenhas, O. A. J. (1995). Exonerating unethical marketing behaviors: a diagnostic framework. Journal of Marketing. Vol. 95. No. 2, pp. 43 – 57.

Mason, T. (2000). The Importance of Being Ethical. Marketing. Vol. 26, pp. 27.

Midgley, D. F. & Dowling, G. F. (1978). Innovativeness: The Concept and Its Measurement. Journal of Consumer Research. Vol. 4, pp. 229 – 242.

NCDO. (2009) ’Jaarverslag’, NCDO, Retrieved on October 15th

https://www.ncdo.nl/sites/default/files/Jaarverslag%202009%20Reflectie.pdf

39 Nimon, W. & Beghin, J. (1999), Are eco-labels value? evidence from the apparel industry. American Journal of Agricultural Economics. Vol. 81 No. 4, pp. 801-11.

OECD: 2007, ‘Innovation and growth’, OECD USA, Retrieved on February 14th, http://www.oecd.org/sti/inno/39374789.pdf

Raynolds, L. T., Murray, D. & Taylor, D.P. L. (2004). Fair Trade Coffee: Building Producer Capacity Via Global Networks. Journal of International Development. Vol. 16. No. 8, pp. 1109 – 1121.

Rogers, E. M. (2003). Diffusion of Innovations. 5th ed., New York: The Free Press.

Salazar, H. A., Oerlemans, L. & van Stroe-Biezen, S. (2013). Social influence on sustainable consumption: evidence from a behavioural experiment. International Journal of Consumer Studies. Vol. 37. No. 2, pp. 172 – 180.

Shaw, D. & Clarke, I. (1999). Belief Formation in Ethical Consumer Groups: An Exploratory Study. Marketing Intelligence and Planning, Vol. 17. No. 2, pp. 109 – 120.

Shaw, D., Grehan, E., Shiu, E., Hassan, L. & Thomson, J. (2005). An exploration of values in ethical consumer decision making. Journal of Consumer Behavior. Vol. 4. No. 3, pp. 185 – 200.

Shimp, T. A. & Bearden, W. O. (1982). Warranty and Other Extrinsic Cue Effects on Consumers’ Risk Perceptions. Journal of Consumer Research. Vol. 9, pp. 38 – 46. Simon, F.L. (1995). Global corporate philanthropy: a strategic framework. International Marketing Review. Vol. 12. No. 4, pp. 20 – 37.

Smaldino, P. E., Janssen, M. A., Hills, V. & Bednar, J. (2015). Adoption as a Social Marker: The Diffusion of Products in a Multigroup Environment. eprint arXiv.

Smith, N. C. (1995). Marketing Strategies for the Ethics Era. Sloan Management Review. Vol. 36. No. 4, pp. 85 – 97.

40 Taylor, P. L., Murray, D. L. & Raynolds, L. T. (2005). Keeping Trade Fair: Governance Challenges in the Fair Trade Coffee Initiative. Sustainable Development. Vol. 13. No. 3, pp. 199 – 208.

TransFair: 2006, ‘Fair Trade Almanac’, Transfair USA, Retrieved on October 23rd, http://www.trans fairusa.org/pdfs/2005FTAlmanac3.17.06.pdf.

Uusitalo, O. & Oksanen, R. (2004). Ethical consumerism: a view from Finalnd. International Journal of Consumer Studies. Vol. 28. No. 3, pp. 214 – 221.

Van den Bulte, C. & Joshi, Y. V. (2007). New-Product Diffusion with Influentials and Imitators. Marketing Science. Vol. 26. No. 3, pp. 400 – 421.

van Herpen, E., van Nierop, E. & Sloot, L. (2011). The relationship between in-store

marketing and observed sales for organic versus fair trade products. Marketing Letters. Vol. 23, No. 1, pp. 293 – 308.

WFTO: 2017, ‘Definition of Fair trade’, WFTO, Retrieved on October 15th, from https://wfto.com/fair-trade/definition-fair-trade