THE DRIVERS OF FAIR TRADE CONSUMPTION

(1)

THE DRIVERS OF FAIR TRADE CONSUMPTION

The influence of the vice and virtue nature of a food category and personal values on fair trade purchase behaviour.

By

Suze van der Lijke

(2)

THE DRIVERS OF FAIR TRADE CONSUMPTION

The influence of the vice and virtue nature of a food category and personal values on fair trade purchase behaviour.

Master Thesis

Msc. Marketing Intelligence & Marketing Management Faculty of Economics and Business, University of Groningen

By

Suze van der Lijke S2396998 Scharreweersterweg 12

9902 CG Appingedam

(06) 21852008 suzevanderlijke@gmail.com

(3)

ABSTRACT

Purpose – The purpose of this paper is to assess the drivers of fair trade consumption for the Dutch consumer. As such the vice and virtue nature of a product category, and the consumer’s biospheric, altruistic and egoistic value orientation are included in the framework.

Value – There is a distinct lack of literature that focuses on actual fair trade purchase behaviour. The results of this study are relevant for fair trade practitioners interested in enhancing their fair trade market share within The Netherlands.

Methodology – To capture consumer purchase incidence regarding fair trade products, a binary logistic regression was performed including 10,927 purchases by 1,244 individual households. A multiple regression analysis was performed to determine variables that influence the purchase quantity of fair trade products that included data from 749 purchases by 542 households. Findings – Contrary to our expectations, we find that the virtue nature of a product category has a stronger positive influence on consumer purchase incidence of fair trade products than the vice nature of a product category. Biospheric values positively influence the purchase incidence, whereas the egoistic value orientation has the opposite effect. To increase the quantity of purchased fair trade products, availability is most important.

Research limitations/implications – Our findings are relevant to practitioners of the fair trade field depending on whether their goal is to develop (e.g. focus on virtue nature of product category, biopheric value orientation, availability and education level) or to penetrate the fair trade market (e.g. focus on availability of the ‘neither’ nature product category). Because of our model’s limited predictive capacity, it is probable that future research reveals other variables that explain actual fair trade purchase behaviour.

(4)

MANAGEMENT SUMMARY

The fair trade concept has received increasing attention in the market place and marketing literature. Consumers are becoming aware of the social impact of their consumption patterns, and are getting conscious that the fair trade food system brings less negative social externalities than the conventional food system. However, despite the fact that fair trade is growing worldwide, this appears different with regards to the Dutch market (Nederlanders & Fairtrade, 2013). Therefore, assessing factors that influence fair trade purchase behaviour can provide valuable insights that can be used to improve the marketing and targeting of-, and eventually increase the market share of sustainable foods. Because there is still decidedly little known about the drivers that shape the fair trade market, this paper focuses on exploring the drivers of fair trade consumption.

Literature indicates that personal values (e.g. biospheric, altruistic and egoistic) and the vice or virtue nature of a product category influences purchase behaviour. Vice products are relative vices ‘wants’, and contribute to negative long-term effects, and relative virtues are ‘shoulds’ and have fewer long-term consequences. Consequently, the research question of this paper is defined as the following: “To what extent does the vice or virtue nature of a product

category plus the personal values of an individual drive fair trade purchacing behaviour”.

To be able to answer this research question, this study observes a total of 23,365 purchases by 1,244 households in 12 food categories observed over two time periods. Data used in this study originates from an existing data-set, which stems from GfK and was accumulated and studied by Van Doorn & Verhoef (2015a).

To assess the drivers of fair trade purchase behaviour two different analyses were performed. First, a binary logistic regression analysis was carried out to observe the purchase incidence of fair trade products. Second, an Ordinary Least Squares (OLS) model was performed to predict the share of purchases regarding households that purchase fair trade products.

Contrasting the expectations it is the virtue nature of a food category rather than the vice nature that has the stronger positive influence on a consumer purchase incidence of fair trade products. Surprisingly, it is not the altruistic, rather the biospheric value orientation that has an important effect on a person’s fair trade purchase incidence. Regarding purchasing quantity, it is the availability of virtue food products in combination with a reasonable price that plays the greatest role.

(5)

1. INTRODUCTION

Consumers have the ability to express their concerns about ethical behaviour of firms by means of their purchase behaviour (De Pelsmacker, 2005). They can do so by deciding to purchase products for their positive qualities (e.g. green products), and boycotting products featuring negative externalities (e.g. products manufactured by children). As such, it is the consumer who has the power to boycott products of firms (Connolly & Shaw, 2006). An example of a consumer boycott is the multinational corporation Nike back in 1990, that endured a global boycott because of their alleged violation of human rights (Jackson & Schantz, 1993; De Pelsmacker, 2005). This kind of consumer behaviour can be defined as ethical behaviour, which refers to the purchase of products that contain a certain ethical issue, varying from human rights, labour conditions, animal welfare and the environment (Doane, 2001). Ethical consumption comprises several dimensions, therefore in this paper we will focus on ethical behaviour that benefits other people, with which we refer to the fair trade concept (De Pelsmacker, 2005). To define fair trade the following definition is used as proposed by FINE1, an informal association of four international fair trade networks:

‘‘Fair trade is a trading partnership, based on dialog, transparency and respect, that seeks greater equity in international trade. It contributes to sustainable development by offering better trading conditions to, and securing the rights of, marginalized producers and workers - especially in the South”.

The fair trade concept has received increasing attention in the market place and marketing literature, partially due to the notion of ethical issues which resulted in the consumer’s willingness to boycott large multinationals such as Coca-Cola, Esso and Nestlé (Connolly & Shaw, 2006; Sirieix, 2008). Moreover, consumers have a growing interest in fair trade because of their increasing awareness of the social impact of their consumption pattern (Giddens, 1991, 1994). Consumers are getting conscious that the fair trade food system brings less negative social externalities than the conventional food system (Getz & Shreck, 2006). These developments contributed to the fact that fair trade no longer serves a niche market that solely appeals to the socially aware and middle class consumer (Cailleba & Casteran, 2010). This finding is supported by increasing sales figures, as world-wide shoppers spent €5.5 billion on fair trade products in 2013, an increase of 15% to the previous year. The purchase of a fair trade product can be seen as a solidary commitment, as consumer’s concerns are mainly related to the well-being of workers and farmers in developing countries (Sirieix, 2008).

More than 1,210 producer organizations and 1,4 million farmers and workers in 74 countries were good for the sales of 30,000 fair trade products in 2013 (“Fair Trade International,” 2015). A product can receive the fair trade label when manufacturers or retailers comply to the terms of the licensing agreement offered by a labeling organization specialized in fair trade products (“Max Havelaar,” 2015a). These type of organizations solely control fair trade products and do not own brands or specialty stores (De Pelsmacker, 2005; “Max Havelaar,” 2015a).

1_{Fair Trade Labeling Organizations International (FLO-Int), IFAT renamed World Fair Trade Organization}

(7)

Fair trade products are currently marketed via several channels varying from catalogues and webshops, to specialty stores and general merchandising retailers in over 70 countries. Various products in several product categories are for sale, ranging from agricultural products, to handcrafted clothing, household items and decorative art (Ma, Littrell, & Niehm, 2012). According to some experts, fair trade could follow the growth of the multi-billion organic food and drink industry (Wright & McCrea, 2007). However, despite the fact that fair trade is growing worldwide with a growth percentage outperforming that of the organic market (Didier & Lucie, 2008; Wright & McCrea, 2007), this seems to be different with regards to the Dutch market (Nederlanders & Fairtrade, 2013).

Overall, the sustainable food market in The Netherlands grew by 10.8% in 2013. The majority of this growth is attributable to the organic market, which grew by 40.1% in 2013 (Monitor Duurzaam Voedsel, 2013). In contrast, growth of the fair trade market seems to have stabilized2 (Nederlanders & Fairtrade, 2013). Observation of factors that influence fair trade purchase behaviour can provide valuable insights that could be used to improve the marketing and targeting of, and eventually increase the market share of sustainable foods (De Pelsmacker et al., 2006; Doran, 2009). Whilst such a study already has been conducted for the organic market (Van Doorn & Verhoef, 2015a), there is still decidedly little known about the drivers that shape the fair trade market. Therefore, this paper focuses on exploring the drivers of fair trade consumption.

Multiple studies in the field of fair trade have investigated purchase intentions (Auger, Devinney, Louviere, & Burke, 2008; Ma et al., 2012). However purchase intentions and similarly self-reported sustainable behaviour are not sufficient enough in ensuring high sales levels (Vermeir & Verbeke, 2006), because consumers have a habit of overestimating their willingness to pay for products (Ajzen, Brown, & Carvajal, 2004). Moreover, objective measurement of willingness to pay prove highly variable between studies. Results generated by surveys where consumers are directly asked about their willingness to pay may be biased by social desirability (Didier & Lucie, 2008; Canavari et al. 2003; Auger,

Devinney, Louviere, & Burke, 2008; Auger & Devinney, 2007). Therefore these surveys are

regarded as poor predictors of actual purchase behaviour (Chandon, Morwitz, & Reinartz, 2005).

In contrast, scanner data prove very valuable in measuring actual purchase behaviour. Doran (2009) investigated fair trade purchase behaviour by making use of this method. However her findings are based on data derived from Internet-based fair trade retailers. Other studies that researched fair trade purchase behaviour using scanner data, were limited in that they solely investigated the purchase behaviour of coffee (Cailleba & Casteran, 2010; Stratton & Werner, 2013). An exception is the study of Yamoah, Fearne, & Duffy (2014). By using data derived from loyalty cards, they were able to study actual fair trade purchase behaviour of several food categories. However, only demographics were used to identify purchase behavior whereas the literature shows that personal values are a more effective way to profile consumers and segment markets (Boote, 1981; De Pelsmacker, Driesen, & Rayp, 2005; Kennedy, Best, & Kahle, 1988; Prakash & Munson, 1986).

(8)

Several studies have reported that personal values will influence consumption behaviour (Rohan, 2000; Lowe & Corkindale, 1998). Over time consumption behaviour has shifted from a more self-centric to values-centric stance; this implies that consumers seek satisfying values that encompass themselves and others (Nicholls, 2002). While a large body of literature has studied the effect of values on consumption, varying from consumption of genetically modified food (Honkanen & Verplanken, 2004), to ethical consumption (Shaw, Grehan, & Shiu, 2004), fair trade consumption (De Pelsmacker et al., 2005), food choices (Goldsmith, Freiden, & Henderson, 1995), consumption of nutritional food (Homer & Kahle, 1988), and pro-environmental attitudes and behaviour (Dietz, Kalof, & Stern, 2002; Karp, 1996; Schultz & Zelenzy, 1998; Shean & Shei, 1995), there is a distinct lack of literature that focuses on actual fair trade purchase behavior.

Consumer responses regarding products are also influenced by the vice and virtue nature of a product category (Hui, Bradlow, & Fader, 2009; Milkman et al., 2008; Mishra & Mishra, 2011; Okada, 2005; Wertenboch, 1988). By vice we refer to products that can be classified as ‘wants’ and provide negative effects on the long-term (e.g. weight gain and alcoholism), whereas by virtue we refer to products that are ‘shoulds’ and therefore are less appealing on the short term but have fewer negative long term consequences (Milkman, Rogers, & Bazerman, 2008; Okada, 2005; Van Doorn & Verhoef, 2011; Wertenboch, 1988). There is an increased likelihood that consumers purchase more vice food products than virtue products, due to the moral licensing effect. This effect increases a consumer’s self-concept after purchasing in the virtue category, which provides the consumer a ‘license’ to purchase vice food products (Conway & Peetz, 2012; Khan & Dhar, 2006). However, in literature there are no studies on how the vice or virtue nature of a product category affects actual fair trade purchase behaviour. Therefore this study aims to investigate actual fair trade purchase behaviour, instead of self-reported behaviour such as willingness to pay or purchase intentions (Auger, Devinney, Louviere, & Burke, 2008a; Ma, Littrell, & Niehm, 2012). In summary, this study seeks to investigate (1) the effect of the vice and virtue nature of a food category, and (2) how consumer values influence fair trade purchase behavior. To be able to investigate these variables, we address the following research question:

To what extent does the vice and virtue nature of a product category and personal values drive fair trade purchase behaviour?

To the best of the author’s knowledge, this paper is the first in the field of sustainable consumption to investigate not only the influencing effect of the vice and virtue nature of a food category, but also consumer’s biospheric, altruistic and egoistic value orientation on fair trade purchase behaviour by making use of panel data from Dutch households. In addition, this paper further extends the current body of literature on actual fair trade purchase behaviour (see table 1 for an overview). Valuable insights of use in marketing fair trade products to- and the targeting of fair trade customers will be gained. Finally, findings on sustainable consumption within this paper will hopefully contribute to a more sustainable development of this world.

(9)

The remainder of this paper is structured as follows: Chapter 2 contains the theoretical framework including the conceptual model and explanation of the variables included in our study. In chapter 3 we justify our hypotheses, followed by the 4th chapter on the chosen methodology. Thereafter we present our results, followed by a discussion and finalized by our conclusion.

Table 1: Overview of literature that investigates sustainable consumption

Author(s) Independent variables Dependent variables Data Values Vice/Virtue Organic

consumption Fair trade consumption Panel data/scanner data

Van Doorn & Verhoef

(2015a) X X X X

Van Doorn & Verhoef

(2015) X X X X

Van Doorn & Verhoef

(2011) X X

Verhoef (2005) X X

Bezawada & Pauwels

(2013) X X

Van Herpen et al.

(2012) X X

Didier & Lucie (2008) X X X

Hauser et al. (2013) X X X

Doran (2009) X X X

Ngobo (2011) X X

Yamoah et al. (2014) X X

Stratton & Werner

(2013) X X

Cailleba & Casteran

(2010) X X

(10)

2. THEORETICAL FRAMEWORK

2.1 Size of the fair trade market

In 2013 shoppers spent globally €5.5 billion on fair trade products, leading to an increase of growth by 15% (“Fair Trade International,” 2015). This growth rate corresponds with consumption trends regarding social responsibility, as the percentage of consumers that have purchased a cause-related product has almost doubled between 1993 (20%) and 2008 (38%) (Ma et al., 2012). However, despite the global growth rates of the fair trade market, the fair trade market in the Netherlands seems to have stabilized (Nederlanders & Fairtrade, 2013).

Parties that have a key role on developing the fair trade market are fair trade labeling organizations (Connolly & Shaw, 2006). The largest labeling organization within The Netherlands is the foundation Max Havelaar. This foundation is the owner of the Dutch fair trade hallmark, which is known by Dutch consumers as ‘Max Havelaar’. Max Havelaar grants his hallmark to other companies in The Netherlands and controls whether they comply to their fair trade criteria (“Max Havelaar,” 2015b). Another well-known fair trade hallmark goes by the name UTZ Certified. This fair trade label stands for sustainable farming and improves opportunities for their farmers (UTZ). The UTZ Certification can be received for products such as coffee, tea and cacao (“UTZ Certified,” 2015)

The first product to receive a fair trade certification within The Netherlands was coffee, back in 1988. Fair trade certified products that followed were cacao, bananas and tea. In 2005 consumers were introduced to products of the now popular fair trade brands Tony Chocolonely and Ben & Jerry’s. The German supermarket chain Lidl was the first supermarket that launched their own fair trade brand ‘Fairglobe’. In 2008, chocolate brand Verkade, switched to sell only fairly traded chocolate. Two years later Superunie, a Dutch buying group, declared to switch all private labels for tea to fair trade for all the supermarkets they deliver to (“Max Havelaar,” 2015b).

Within The Netherlands a total of 4,5 million households purchased fair trade food products in 2013, which accounts for 59% of all Dutch households. In previous years the percentage of buying households increased, however in 2013 this seemed to have stabilized for the first time in years (Nederlanders & Fairtrade, 2013; Stichting Max Havelaar, 2013). Whilst the fair trade market seemed to have stabilized, the Dutch sustainable market is growing. Much of this growth is attributable to the organic market (Monitor Duurzaam Voedsel, 2013). Overall, fair trade sales in the Netherlands totaled an estimated 197 million euros in 2013 (Stichting Max Havelaar, 2013). 93% of the fair trade turnover occurred in supermarkets (Nederlanders & Fairtrade, 2013), and almost 60% of Dutch consumers bought a fair trade product in 2013. The four most popular product categories were chocolate, fruit, tea and coffee (Nederlanders & Fairtrade, 2013).

2.2 Determinants of fair trade purchase behaviour

(11)

nowadays is more values-centric, inferring that consumers seek satisfying values that encompass themselves and others (Nicholls, 2002). This could be seen as another factor implying that the concept of fair trade has a high appeal to the consumer. However, as promising as this seems, literature finds that it is the attitude towards fair trade that is a deciding factor in determining consumer’s intentions to purchase fair trade products (Ma. et al., 2012). Furthermore, personal values play an important role in the purchase of fair trade goods as well. Doran (2009), finds universalism to be the most important value amongst loyal fair trade consumers, with the concept of universalism meaning the following:

“When individuals and society do not accept all others unconditionally and do not treat all people fairly, this will ultimately result in everyone’s downfall”.

However, we have to emphasize that her findings are based on data derived from Internet-based fair trade retailers, whereas 93% of the fair trade turnover in The Netherlands takes place in supermarkets (Nederlanders & Fairtrade, 2013). Moreover, cautiousness is required when interpreting the aforementioned results, as they are derived by studies that are based on self-reported behaviour, and as such might have been biased due to social desirability (Didier & Lucie, 2008; Canavari et al. 2003; Auger, Devinney, Louviere, & Burke, 2008; Auger &

Devinney, 2007).

A more reliable data-type to determine actual purchase behaviour is scanner data. Studying this data-type, Stratton & Werner (2013) find that consumers have an increased interest in purchasing products that promote societal well-being. As such, they support the notion of Auger (2008), that social features affect a person’s likelihood of purchasing fair trade. The findings of Doran (2009) and Ma et al. (2012) are supported by Hauser et al. (2013), being that they find sustainability to have a strong positive effect on the attitude and purchase of fair trade products. That personal values are a more effective tool to profile consumers and segment markets (Boote, 1981; De Pelsmacker, Driesen, & Rayp, 2005; Kennedy, Best, & Kahle, 1988; Prakash & Munson, 1986), is supported by Doran (2009), as she finds demographics to be useless in creating consumer profiles. Yamoah, Fearne, & Duffy (2014), confirm this finding as they find that fair trade has the most appeal to all life stages, varying from older adults, and pensioners to young adults and students.

2.3 Sustainable consumption in vice and virtue food categories

Food products can be classified into relative vices and relative virtues. Relative vices, which are wants, provide an immediate pleasurable experience in the short term, but contribute to negative long term effects (such as weight gain and alcoholism),whereas relative virtues, or ‘shoulds’, are less satisfying and appealing in the short term, but have fewer long term negative consequences (Milkman, Rogers, & Bazerman, 2008; Okada, 2005; Van Doorn & Verhoef, 2011; Wertenboch, 1988). According to literature, the vice or virtue nature of a product category affects a consumers’ response with regards to products, promotions and packages (Hui, Bradlow, & Fader, 2009; Milkman et al., 2008; Mishra & Mishra, 2011; Okada, 2005; Wertenboch, 1988).

(12)

for this difference in preference is that consumers are not able to generate good justifications with regards to the vice food category. Consumers are aware of the fact that whenever they purchase a vice food in a bonus pack, they would have to consume more of this vice food. However, when price discounts are applied to vice foods, this can be justified by the consumer (Mishra & Mishra, 2011). Consumers are even willing to pay more for smaller packages of vice foods, as this avoids having too many vice foods lying around (Milkman et al., 2008). Another situation when consumers are more likely to purchase products in the vice category, is when they already have shopped at locations that carry virtue categories (Hui, Bradlow, & Fader, 2009).

The aforementioned examples indicate that the consumption of vice food is associated with feelings of guilt, which require special justification (Khan & Dhar, 2006). The purchase behaviour consumers display with regards to vice food categories are consistent with licensing behaviour. Moral licensing theorists find that a consumer’s moral self-perception decreases pro-social behaviour and that it induces compensatory behaviour (Conway & Peetz, 2012). Khan & Dhar (2006), have studied this effect and find that a consumer’s self-concept increases once they have purchased in a virtue category, which results in an increased likelihood of a vice purchase as this provides a consumer the ‘license’ to do so. Hence, the vice nature of a product category is likely to increase the purchase of fair trade products. Another study that confirms this finding is that of Strahilevitz & Myers (1998) regarding donations to charity as purchase incentives. In their study they compared the effectiveness of promised donations to charity regarding practical necessities (e.g. box of laundry detergent), and luxury goods (e.g. hot fudge sundae), which can be further classified as relative virtue and vice goods respectively. What they find is that charity incentives are more effective in promoting vice products than virtue products.

2.4 Personal value orientation and sustainability

Purchase of fair trade products can be best described as a solidary commitment, with consumer concerns being related to the well-being of workers and farmers in developing countries. The reason why consumers are motivated to purchase, is linked to their personal values (Sirieix, 2008), as such consumer attitudes and behaviour towards fair trade can be traced back to their personal value system (Allport, Vernon, & Lindzey, 1960; Feather, 1995; Rohan, 2000; Rokeach, 1973; Schwartz, 1992). Food consumption is a valuable domain to study relationships between values and behaviour as eating is an essential requirement (Hauser et al., 2013). How a person’s values influence food consumption has been studied by several researchers (De Pelsmacker et al., 2005; Dietz et al., 2002; Goldsmith et al., 1995; Homer & Kahle, 1988; Honkanen & Verplanken, 2004; Karp, 1996; Schultz & Zelenzy, 1998; Shaw et al., 2004; Shean & Shei, 1995). Understanding such values is a prerequisite for strategic marketing and product development (Hauser et al., 2013).

(13)

of others (Batson, 1987), it is likely that altruistic values strongly influence fair trade purchase behaviour. Moreover it is plausible that the altruistic value orientation has a stronger influencing effect on fair trade purchase behaviour in the virtue food category than in the vice food category, due to the licensing effect. This effect causes people that engage in an altruistic task, such as purchasing fair trade products, to eventually purchase vice foods, because the altruistic task created a moral license to be self-indulgent (Khan & Dhar, 2006; Mishra & Mishra, 2011; Strahilevitz & Myers, 1998).

The concept of fair trade contributes to a more equal (Giddens, 1991, 1994), and more sustainable world (Robins & Roberts, 1997). However, despite growth of the global fair trade market, the Dutch fair trade market seems to have stabilized (Nederlanders & Fairtrade, 2013). Marketing research can contribute to gain insights in the factors that influence fair trade purchase behaviour, and eventually improve the marketing to-, and targeting of fair trade customers. However, there is a distinct lack of literature on actual fair trade purchase behaviour, therefore in this paper actual fair trade purchase behaviour is observed by studying scanner data.

Based on previous conducted research, we expect that the nature of a food category affects the purchase of fair trade products. We anticipate an increase of the vice nature of a food category due to the moral licensing effect (Conway & Peetz, 2012; Khan & Dhar, 2006), and the fact that charity incentives are more effective in promoting vice products as opposed to virtue products (Strahilevitz & Myers, 1998). Moreover, we expect that a consumer’s value system influences fair trade purchase behaviour as well (Steg, Dreijerink, & Abrahamse, 2005; Van Doorn & Verhoef, 2015b; Van Doorn & Verhoef, 2015a).

(14)

Figure 1: Conceptual model illustrating the drivers of fair trade purchase behaviour

Price and availability are included as control variables as they have a very straight- forward effect on purchase behaviour. Fair trade consumers are willing to pay a price premium and are willing to do so, because this premium guarantees that specific input will be paid at a socially acceptable rate (Yanchus & de Vanssay, 2003). However, this price premium should not increase beyond what consumers are willing to pay, as this results in behaviour which, even when consumers value ethical consumption, is no longer coherent with their attitude (Wathieu & Bertini, 2007). Based on this argumentation, and because of standard micro economic theory (Van Doorn & Verhoef, 2015a), we expect that price will negatively affect fair trade purchase behaviour. The availability of products has a positive effect on purchase behaviour (Ataman, Mela, & Heerde, 2008; Prasad, Mahajan, & Bronnenberg, 2003), simply because when a product is not available it cannot be sold (Steinhart, Mazursky, & Kamins, 2013). Moreover, the convenience of availability ensures that consumers do not have to exert significantly more effort to buy the products they want (De Pelsmacker et al., 2006); that is, availability has a rather straight-forward effect on purchase behaviour. No hypotheses will be put forward for these existing and well documented effects of price premium and availability.

3. HYPOTHESES

(15)

lower share of organic purchases in the vice category. Despite the fact that fair trade and organic both make up the sustainable food market, we reason that the vice nature of a food category has a different effect for fair trade purchase behaviour.

We attribute this difference to the concept of moral licensing, that is when moral self-perceptions decrease pro-social behaviour and increase compensatory behaviour (Conway & Peetz, 2012). Khan & Dhar (2006) studied this effect with regards to consumer choice and found that when participants volunteered to do a community service in their first task, they were more likely to choose a hedonistic product over a utilitarian in their second task. In literature the terms hedonistic and utilitarian are another way to categorize respectively vice and virtue types of purchases (Dhar & Simonson, 1999; Wertenboch, 1988; Wertenbroch, 2000). The finding by Khan & Dhar (2006) implies that when consumers engage in an altruistic task, their self-concept of altruism gets activated and he or she will be more self-indulgent.

Generally, the consumption of vice food is associated with feelings of guilt, which requires special justification (Khan & Dhar, 2006). These feelings of guilt can get mitigated when linking a vice product to a good cause (Strahilevitz & Myers, 1998). Mishra & Mishra (2011) find that when participants donate, this significantly reduces their search for guilt-mitigating justification for vice food. Hence, donating to charity, can serve as a moral license to be self-indulgent (Khan & Dhar, 2006; Mishra & Mishra, 2011; Strahilevitz & Myers, 1998).

Based on theory, we reason that when people engage in an altruistic task, such as fair trade purchasing, their moral self-concept gets activated allowing them to engage in compensatory behaviour, such as buying a vice product. Moreover consumer’s feelings of guilt for purchasing a vice product are further mitigated because the fair trade product is purchased out of an act of charity. Hence, when consumers purchase a fair trade product, this product is likely to be purchased from a vice food category, due to moral licensing. Based on this argumentation we put forward our first hypothesis:

H1: Consumers are less likely to purchase fair trade products in virtue than in vice product categories.

3.2 The effects of the personal value system on fair trade purchase behaviour

Biospheric Values

Biospheric values reflect a person’s concern for the biosphere and non-human species (Steg et al., 2005; Stern, Dietz, & Kalof, 1993). Consumers with high biospheric values consider the environment and animal welfare important, and are more likely to display sustainable behaviour (Steg et al., 2005; Stern et al., 1993).

(16)

together. Hence, consumers that purchase fair trade products, are doing so because they are concerned about the environment as well. Therefore it is likely that fair trade consumers also have high biospheric values (Steg et al., 2005; Stern et al., 1993). Moreover there is a positive relation between biospheric values and social sustainable consumption, which includes organic and fair trade products (Van Doorn & Verhoef, 2015b). This indicates that biospheric values are also likely to have a significant positive effect on fair trade purchase behaviour.

Furthermore, when consumers are asked about what they consider to be sustainable consumption, their top three responses are (1) throw away as less food as possible, (2) eat meat from animals that enjoyed better living conditions, and (3) eat fair trade products. Interestingly, eating organic was 5th_{after the desire to eat sustainable seafood (Witte & Scholtz, 2014). Prior}

research found significant effects of biospheric values on organic purchase behaviour (Van Doorn & Verhoef, 2015a). Thus, as eating fair trade products is more associated with sustainable behaviour than eating organic food, we reason that it is likely that biospheric values also have a positive significant effect on fair trade purchase behaviour. Based on these insights we propose our second hypothesis:

H2: Biospheric values have a positive effect on fair trade purchase behaviour.

Altruistic Values

An altruistic value orientation is based upon a person’s concern and welfare for other human beings (Dietz et al., 2002; Steg et al., 2005). Meaning that when a person’s behaviour is induced by altruistic factors his or her objective is to increase another person’s welfare (Batson, 1987). Moreover, consumers that have high altruistic values are more likely to display sustainable behaviour (Stern et al., 1993). Loreiro & Lotade (2005) confirm these findings in their study on sustainable labeling. They find that consumers whom express their concerns regarding general working conditions in developing countries, are also very receptive to ethical labeling. This indicates that altruism plays an important role in evaluating fair trade practices. Besides, the fair trade consumer adds more importance to altruism, equality, peace, and a beautiful, environmentally secure world (Littrel & Dickson, 1999).

The fair trade concept stands for offering better trading conditions and securing rights of peripheral workers and producers (World Trade Organization, 2015). By purchasing a fair trade product, consumers can directly contribute to another person’s welfare as fair trade promotes a direct relationship between producers and consumers (Barret, 1993; Jaffee et al., 2004; Nicholls, 2002). It seems likely that this concept especially appeals to people that are concerned about other people’s welfare. Therefore, it is plausible that a high altruistic value orientation has a positive effect on fair trade purchasing. Based on this argumentation, the third hypothesis is as follows:

H3: Altruistic values have a positive effect on fair trade purchase behaviour. Egoistic Values

(17)

purchasing this value did not yield a significant effect (Van Doorn & Verhoef, 2015b), whilst for the organic market it did (Van Doorn & Verhoef, 2015a). Nonetheless, with regards to the fair trade market, the effect of an egoistic value orientation has not been studied.

In general, consumers with a high egoistic value orientation are not likely to benefit the collective interest, as they are trying to maximize their own individual outcomes (Steg et al., 2005). Egoistic consumers value authority, wealth, social power and influence; these individuals are more likely to care about their own welfare than that of society. An egoistic value orientation seems to contrast with the concept of fair trade, were consumers are offered the opportunity to directly contribute to another person’s welfare (World Trade Organization, 2015). People tend to act more favorably towards others that share similar important attributes of their own identity (Akerlof & Kranton, 2000; Tajfel & Turner, 1979). Therefore, we reason that it is highly unlikely that consumers with a high egoistic values will support vulnerable demographic groups by purchasing fair trade products. Following this line of reasoning, we expect consumers with a high egoistic value orientation to have a significant negative effect on fair trade purchase behaviour. Our fourth hypothesis is as follows:

H4: Egoistic values have a negative effect on the purchase of fair trade products.

3.3 The interplay of the altruistic value orientation and the virtue food category

We expect that in general, consumers are more likely to purchase fair trade products in the vice food category, due to the moral licensing effect. Moral licensing is expected to contribute to the effect that consumers that are not so altruistically orientated are still likely to purchase fair trade products. When a consumer purchases a fair trade product, his or her moral self-concept gets activated which provides the person a ‘license’ to purchase a vice product (Khan & Dhar, 2006). Moreover, the feelings of guilt consumers endure when purchasing a vice food, are further mitigated by moral licensing. Because, by purchasing a fair trade product, the consumer donates to charity, leading to compensatory behaviour such as purchasing a product from the vice category (Khan & Dhar, 2006; Mishra & Mishra, 2011; Strahilevitz & Myers, 1998). Hence, we reason that the influence of altruistic value orientation on fair trade purchase behaviour in the vice nature of a food category is less profound than for the virtue food category, as for the virtue food category there is no moral licensing effect.

Moreover, according to literature, consumers that purchase sustainable products, such as organic and fair trade, are likely to consider future consequences (Van Doorn & Verhoef, 2015b). This implies that the altruistically oriented fair trade consumer, is more likely to purchase products from the virtue category, as opposed to the vice category. Following this line of reasoning we expect the altruistic value orientation to have a stronger influencing effect regarding fair trade purchase behaviour in virtue food categories, which leads to our 5th hypothesis:

H5: Altruistic values have a stronger effect on fair trade purchase behaviour in virtue categories than in vice categories due to moral licensing.

3.4 Control variables

Gender

(18)

1997; De Pelsmacker et al., 2005; Schwepker & Cornwell, 1991; Shrum, Mccarty, & Lowrey, 1995), whereas other studies found that women are more likely to behave ethically (Blend & Van Ravenswaay, 1999; Loureiro & Lotade, 2005; Mainieri, Barnett, Valdero, Unipan, & Oskamp, 1997; Roberts, 1996). Findings of gender on actual fair trade purchase behaviour are very scarce as the majority of these studies do not include gender (Cailleba & Casteran, 2010; Stratton & Werner, 2013; Yamoah et al., 2014). The study of Doran (2009) is an exception, although they did not find any significant differences between consumers and non-consumers of fair trade products in terms of gender.

An explanation for the mixed results of the effect of gender, could be that originally, fair trade products were only available in craft and apparel stores which appealed more to women. As products such as cocoa, sugar, bananas, coffee and tea were sold as fair trade and became available in conventional retail locations, fair trade products have also recently increased in appeal to men (Doran, 2009). Taking into account the increased appeal of fair trade products to men, and enhanced availability of fair trade products in the conventional retail locations, we expect that gender differences do not significantly affect fair trade purchase behaviour.

Income

Several studies have documented significant effects of income on consumer preferences and attitudes towards fair trade products (Cowe & Williams, 2000; De Pelsmacker et al., 2005). This also accounts for studies that investigate actual purchase behaviour based on scanner data. Cailleba & Casteran (2010) found that traditional non-fair trade coffee customers had a lower income compared to the exclusive fair trade customer. Another study, conducted by Yamoah et al., (2014), included alongside coffee, banana, tea, chocolate, drinking chocolate and sugar and, similar to Cailleba & Casteran (2010), they found that fair trade appeals most to wealthy customers. An explanation for this effect could be that households with a higher income are more likely to donate to charity (Smith, Kehoe, & Cremer, 1995). Based on this argumentation, we expect income to positively drive fair trade purchase behaviour.

Age

In the 70’s most research was aimed at finding positive relationships between age and

(19)

Education

Studies measuring the willingness to pay for fair trade products, claim that education has a significant effect (Cowe & Williams, 2000; De Pelsmacker et al., 2005). Fair trade purchase behaviour is said to be driven by a relatively high educational status (Carrigan & Attalla, 2001; Littrell & Dickson, 1999; Maignan & Ferrell, 2001; Roberts, 1996); however, despite the fact that De Pelsmacker et al., (2005) found a significant effect of education on fair trade purchase behaviour, they only find this for high school and higher educated respondents. Contrary, Doran (2009) did not find education to have an influence on fair trade purchase behaviour and reasoned that this might be the case due to more widespread marketing and availability of fair trade products. However their finding might be biased as their study is built on data derived from Internet-based retailers, whilst it is known that online purchasing is affected by education (Goolsbee, 1998; Lohse, Bellman, & Johnson, 2000; Narayanan, Koo, & Cozzarin, 2012; Punj, 2011). For this study we expect education to drive fair trade purchase behaviour. This is because the complexity surrounding fair trade is likely to be better understood by higher educated people (Dietz et al., 1998; Ngobo, 2011).

Household size

Household size has a negative effect on fair trade purchase behaviour, as larger households are less likely to purchase fair trade products (Van Doorn & Verhoef, 2015a). This can be explained by the fact that household size affects purchase behaviour as this correlates with price sensitivity (Richardson, Jain, & Dick, 1996). Some ‘green’ products have price premiums as high as 30-50% (National Consumer Council, 1996; Rushe & Lynn, 1999), with large households being less likely to afford these premiums.

In this study, we expect to find a negative significant effect for small households (1 to 2 people) and large households (more than 4 people) with respect to purchase behaviour. In contrast, we expect to find a positive significant effect on fair trade purchase behaviour for moderate households sized between 3 and 4 people. This is because fair trade has the most appeal to families than to older adults, pensioners, young adults and students (Yamoah et al., 2014).

4. RESEARCH METHODOLOGY

4.1 Data collection

(20)

In this study 1,246 households of the 4,000 households in the Gfk panel are included, as solely data of the former are linked to consumer values and socio-demographics. Each of these households scanned their food purchases by making use of an in-home scanning device3. Of the 29 food categories scanned by these households, only 12 categories included fair trade purchases or UTZ labelled purchases (see table 3). Hence in this study we include a total of 23,365 observations from 1,244 households in 12 food categories assessed over two time periods.

Dependent variables

Fair trade purchase behaviour is observed by two different dependent variables. First, we observe whether a household purchases fair trade products at all within the observation period. This will provide us with insights on which variables determine whether or not a household will purchase fair trade products. Second, we observe the purchase quantity of households that already purchase fair trade products, which will provide us with information on variables that affect the share of fair trade purchases.

Independent variables

Personal values are assessed by a survey deduced from values originating from the Schwartz Value Survey (1992). Values from Schwartz (1992) were shortlisted by Dietz et al. (2002), and further developed by Steg et al. (2005), into a short survey measuring biospheric, altruistic and egoistic values (see appendix A). This survey has previously been used to investigate organic purchase behaviour by Van Doorn & Verhoef (2015a). It was administered to part of the GfK panel in November 2007, which resulted in 1,246 usable responses. The reliability and descriptive statistics for our measures and their data source are presented in table 2.

The alpha’s for our values have been calculated previously in the study on organic purchase behaviour by Van Doorn & Verhoef (2015a), using the same data-set. As can be inferred from table 3, the alpha scores exceed the critical threshold of .7 (Nunnally & Bernstein, 1994). To test whether the items from the survey did not have any overlap, a principal components analysis was conducted.

Table 2: Measures and data source variables

Variables Cronbach’s alpha Mean SD R with

Category level (Hui et al., 2009; Van Doorn & Verhoef, 2015a)

3_{The panel used in this paper is operated under the ISO 9001 quality procedure, which means that GfK}

(21)

Virtue category (dummy variable) Price premium Availability Availability n.a. n.a. n.a. .15 .48 .01 .36 .85 .02 -.053*** -.099*** .428***

Consumer level (Steg et al., 2005; Van Doorn & Verhoef, 2015a)

Biospheric values Altruistic Values Egoistic values .88 .81 .76 4.36 4.99 2.07 1.36 1.14 1.17 .010 -.001 -.005 ***significant at p <.01.

To determine the effect of the vice and virtue nature of a food category on fair trade purchase behaviour, we classified our data into virtue, vice and neither categories (see table 3). This classification is based on the distinction by Hui, Bradlow, & Fader (2009) and Van Doorn & Verhoef (2015a). As can be inferred from table 3, a total of 12 different food categories are included in our study.

4.2 Descriptive Statistics

Independent variables

In table 3, an overview of the statistics of share of purchase, price premium and availability can be found of all categories included. The average SOP of fair trade products is 0.65% in period 1 and in period 2 this accumulated to 0.72%. However, as table 3 indicates, the average share of fair trade purchases per category are very small, with an exception of the category ‘coffee and tea’.

The price premium is calculated as the difference (as a percentage) between the average price of fair trade products and conventional products (purchased by the whole household sample of N > 4,000) in a category using GfK data. The average price premium of fair trade products, relative to the price of other products in the food category, varies with the highest price premium for ‘cookies and pastries’ (275.15%), and the lowest price premium for the category ‘sweets and candy’ (-17.76%).

(22)

Table 3: Overview categories included in study

Subdivision Food categories Average share of fair trade purchases Average price premium Average availability Virtue categories

Vegetables and fruit Canned fruit 0.17% 0.10% 16.84% -10.17% 0.16% 0.17% Vice categories Alcohol Chocolate

Cookies and pastries Soft drinks

Cheese

Sweets and candy

0.003% 0.24% 0.01% 0.06% 0.01% .001% 12.51% -1.23% 275.15% -15.64% 70.22% -17.76% 0.39% 0.64% 0.05% 0.19% 0.02% 0.12% Neither

Coffee and tea Rice and pasta Sandwich filling Seasoning 11.16% 0.1% .41% 0.01% -11.67% 32.23% 40.03% 157.19% 8.57% 0.61% 0.93% 0.36%

The correlation matrices of all variables assessed over period 1 are reported in appendix B. Unsurprisingly, the share of fair trade products correlated positively with availability (.43, p < .01), indicating that when a product’s availability increases, it is very likely that its share of purchase will increase as well. As expected, the share of fair trade purchases correlated positively with the vice nature of a product category (.098, p < .01). Furthermore, share of fair trade purchases correlated negatively with the virtue (-.053, p < .01), and neither (-.0.67, p < .01) nature of a food category. This indicates that fair trade products are purchased more in vice categories than in virtue. Unexpected is that all values included in our research do not correlate with share of fair trade purchases, which infers that these values do not affect fair trade purchase behaviour. However, the biospheric value orientation does correlate with the altruistic value orientation (.62 p < .01), indicating that these values are somewhat related to each other.

The virtue nature of a food category and the altruistic value orientation, neither revealed significant correlating effects. The availability of fair trade products correlated positively with the vice nature of a product category (.21, p < .01), whilst it correlated negatively with the virtue (-.61, p < .01), and neither (0.10, p < .01) nature of a food category. This indicates that there are more vice options available regarding fair trade products than there are virtue and neither food options. Moreover, availability correlated negatively with price premium (-.29, p < .01), suggesting that there are not many fair trade options available in categories that are highly priced.

Control variables

(23)

than € 2,500, whereas 26.6% earned more than € 3,100 For education, we found that 35.6% had an associate’s/BA/BS degree, and 35.9% went to graduate school.

The only significant correlation we found regarding socio-demographics, was for household size (-.10, p < .05)., which indicates that large households purchase less fair trade products.

4.3 Method

First, a binary logistic regression analysis was conducted to observe the purchase incidence purchase of fair trade products for all 1,244 households. The dependent variable in this choice model is either (1), which indicates that the respondent purchases fair trade products, or (0) meaning that there were no fair trade products purchased. Our model follows a Cumulative Distribution Function (CDF) (Blattberg, Kim, & Neslin, 2008). The beta values of our independent variables are used to estimate each variable’s utility (U). Subsequently these utilities are translated into probabilities, with which we will calculate the probability P that a consumer i purchases fair trade products:

= ( ) = + + ℎ + ℎ + + + + + + + + + ℎ + × + ( ) ( = 1) = (2) ( = 0) = 1 − (3)

where is the virtue nature of a food category c and ℎ is the variable of the neither category for a food category c. Furthermore, ℎ is the biospheric value orientation of household i, is the altruistic value orientation of household i, and is the egoistic value orientation of households i. Subsequently, the marketing

variables and , both varying over category c and the

socio-demographics ( , , and ℎ ), which differ per

household i. Additionally the moderating effect × , measuring the

interaction between the vice nature of a food category c and the altruistic value orientation of household i. The term describes the unobserved value of the disturbance term.

(24)

Table 4: Results of χ test and descriptive statistics of fair trade purchase and time

Time period Fair trade purchase

No fair trade At least one product

shopping basket

November 2007-March 2008 10178 (93.1%) 749 (6.9%)

November 2008-March 2009 11600 (93.3%) 838 (6.7%)

Note: = .13, df = 1. Numbers in parentheses indicate column percentages. *p <. 05.

Second, an Ordinary Least Squares (OLS) model was conducted to predict the share of purchase (SOP ) of households that purchase fair trade products:

,

=

,

(4)

where SOP _, is the share of purchases of fair trade products of a household i in category c; items , describes the number of fair trade items purchased by household i

in category c; and defines the number of items purchased by household i in category c i. To determine whether SOP differs over the two time periods we performed an independent samples t-test. The independent samples t-test was not significant, t (1585) = .685,

p = .289. The average SOP in period 1 (M = .32, SD = .38), was not significantly different from the average SOP in period 2 (M = .31, SD = .37). As such, in our second analysis we will observe the data of period 1.

= + + ℎ + ℎ + + + + + + + + + ℎ + × + ( ) = + (6)

5. RESULTS

(25)

5.1 Variables that influence purchase incidence of fair trade products

At first we will observe a household’s purchase incidence regarding fair trade products by conducting a binary logistic regression. As time had no significant impact on the purchase of fair trade products, our model is estimated over period 1, which consists of 10,927 cases. We checked our data for missing values, and as there were none, we did not have to exclude any cases from our analysis. To validate the predictive capacity of our final model, we will run our model over the data from period 2, which consists out of 12,438 observations (Guadagni & Little, 1983). In order to compare our results, we standardised all our measures. Furthermore, we created a dummy for gender, and dummy’s for the nature of a food category (e.g. virtue, neither and vice as reference category). To be able to test our 5th hypothesis an interaction effect between the altruistic value orientation and the virtue nature of a food category was created. In order to test whether the interaction effect contributed to our model, we ran our model several times with and without this effect to determine whether its predictive capacity would improve. To determine the goodness-of-fit of our model we used the Akaike Information Criteria (AIC4), which penalizes for additional variables (Blattberg et al., 2008). Inclusion of the interaction effect resulted in the lowest value of the AIC (2695.22). Therefore, we decided to include the interaction effect, as this improved our model’s predictive capacity (Imori, Yanagihara, & Wakaki, 2014). The betas of this model are summarized in table 5.

Table 5: Estimates binary logistic regression model

Variables β Standard deviation Wald-statistic p- Exp(β) VIF Constant -4.698 .157 900.927 .000 .009 Altruistic -.085 .074 1.293 .256 .919 1.813 Biospheric .267 .068 15.226 .000*** 1.306 1.643 Egoistic -.134 .054 6.194 .013** .875 1.030 Price Premium -.260 .152 2.951 .086* .771 1.158 Availability 1.147 .051 502.974 .000*** 3.150 1.130 Age .064 .058 1.226 .268 1.066 1.180 Income .044 .057 .605 .437 1.045 1.177 Education .179 .057 9.885 .002 1.197 1.168 Household size -.069 .060 1.307 .253 .933 1.229 Gender (female) -.013 .167 .006 .939 .987 1.088 Virtue 2.027 .205 97.875 .000*** 7.590 1.167 Neither .405 .249 2.646 .104 1.500 1.109 Altruism*Virtue -.181 .124 2.112 .146 .835 1.181

a***significant at p <.01. **significant at p <.05. *significant at p <.10.

Overall our model is significant (Wald = 4406.84, p <. .01), indicating a significant difference between the households that purchase fair trade products in period 1, versus people

4

(26)

that do not. The likelihood ratio test = 1406.07 is significant (p < .01) as well, indicating that the joint significance of all variables are not redundant and the fitted model is better than a naïve, intercept-only model. Furthermore we checked the Variance Inflation Factor (VIF), the VIF scores quantify overlap among our independent variables and detect possible multicollineary issues in our data. Whenever variables have a high VIF score there is an issue with multicollinearity (Leeflang, Bijmolt, Pauwels, & Wieringa, 2014). However, as can be inferred from table 5, all VIF scores are below the critical value of 4, with a minimum VIF of 1.030 and a maximum VIF of 1.813. As such there are no signs of multicollinearity in our model and we can further analyse the model estimations presented in table 5.

The effect of the vice and virtue category

We reasoned that consumers are less likely to purchase fair trade products in virtue product categories than in vice categories. However, as can be seen in table 5, the virtue nature of a product category has a stronger positive significant effect (β = 7.590, p < .01), relative to the vice nature of a product category on a consumer’s purchase incidence regarding fair trade products. Moreover, the virtue nature of a product category has the strongest positive effect on the purchase incidence of fair trade products (see table 5). Hence, we are not able to confirm H1, however, we can conclude that consumers are less likely to purchase fair trade products in vice categories than in virtue product categories.

Values

The biospheric value orientation has a positive significant effect on the purchase incidence of fair trade product (β = 1.306, p < .01), which is in line with H2 (see table 9). Unexpectedly the altruistic value orientation did not have a significant effect on the purchase incidence of fair trade products (β = .919, p > .1). Hence we are not able to confirm H3 (see table 9). On the contrary, the egoistic value orientation did have a significant negative effect (β = .875, p < .05), and as such we are able to confirm H4. However the expected interaction effect of the altruistic value orientation and the virtue nature of a product category on the purchase incidence of fair trade products was not significant (β = .835, p > .1). Therefore we are not able to confirm H5 (see table 9).

Control variables

In line with our expectations, gender and age do not have a significant effect on the purchase of fair trade products. However, where we expected income and household size to both have a significant effect on the purchase of fair trade, they did not. The only expected significant effect amongst socio-demographics was found for education level (β = 1.197, p < .01), which indicates that higher educated people are more probable to purchase fair trade products. As expected price had a negative significant effect on the purchase of fair trade (β = 0.771, p < .10), whereas availability had a positive significant effect (β = 3.150, p < .01).

Predictive capacity model

(27)

is a high percentage, this result is rather disappointing. Our estimated model was intended to improve predictions relative to the naïve model, instead its predictions decreased by .1%. However when we look at the percentage of predictions regarding the purchase incidence of fair trade products, we see that these predictions have increased by 3.2%, which means that our model predicts 3.2% better than the naïve model.

Table 6: Hit-rates binary logistic regression period 1

Null-model Predicted % correct Logit model Predicted % correct

Observed purchase fair trade product

No Yes Observed purchase

fair trade product

No Yes

No 10425 0 100% No 10399 22 99.8%

Yes 502 0 0% Yes 487 16 3.2%

Overall hit-rate 95.4%* Overall hit-rate 95.3%*

*Percentage of correct predictions relative to all predictions

To validate the predictive capacity of our final model, we will use the outcomes of our model, which is estimated over period 1, to predict the values for the data of period 2 (Guadagni & Little, 1983). The betas of the constant and our explanatory variables are translated into utilities for period 2. Subsequently, these are translated into probabilities by making use of the following formula:

( ) 1 + ( )

(7) Of all the 12,438 cases in period 2, the values of each of the explanatory variables (table 5) are multiplied with the corresponding parameters and cumulated together with the constant (see formula 7). Subsequently we calculated the probability of purchasing a fair trade product, which was translated into a prediction of adoption :

(28)

Where the cut-off value for is a probability of .5, meaning that when a case has a probability lower than .5, no purchase is predicted, and when a case has a probability higher than .5 purchase of a fair trade product is predicted.

Decile analysis

Figure 2 represents a graphical distribution of the mean probabilities of purchasing a fair trade product, which is assessed by our estimated model over the data of period 2.

The probabilities of all cases are ranked and divided into 8 deciles. As can be seen, the first decile has the highest probability of purchasing a fair trade product, whereas 8 is the decile with the lowest probability. The steep downward slope in our first decile shows a clear distinction in probabilities captured by our model.

Cumulative lift

Figure 3 shows the cumulative lift curve which is assessed by our model over the cases of period 2. The straight line through the origin represents predictions based on random selection. The model based line is a graphical representation of the probabilities of purchasing fair trade products. What can be inferred from this graph is that our model predicts better than random selection, as the first decile contains 69.17% of fair trade purchasers, whereas the random selection only predicts

12.5%. Subsequently the lift for the customers in each decile is calculated5. The top decile lift is 5.53, meaning that customers in this first decile are 5.53 times more likely to eventually purchase fair trade products compared to the entire sample. The lift in the second decile measures 1.41, indicating that in this decile customers are still more likely to purchase fair trade than the average. However, the remaining deciles show a lift of less than 1, indicating that customers in these deciles are less likely to purchase fair trade products. Hence the customers of interest, as they are most likely to purchase fair trade products, reside in the first and second decile.

5_{Lift is calculated as following:} ₌ _{, where} _{= lift for the kth decile,} _{= probability for the kth decile and}

r = probability across the entire sample (Blattberg et al., 2008).

Figure 2: Mean probability purchase fair trade

Figure 3: Cumulative lift curve

0,00 0,05 0,10 0,15 0,20 0,25 0,30 0,35 0,40 0,45 0,50 1 2 3 4 5 6 7 8 Mean pro ba bilit y Number of deciles 0 12,5 25 37,5 50 62,5 75 87,5 100 0 1 2 3 4 5 6 7 8

(29)

The hit-rates of our model are calculated again over the cases in period 2. As can be seen in table 7, the hit-rate of the naïve model and our estimated model are similar (94.7%). When comparing the rate of period 2, to the rate of period 1 (95.3%), we see that our hit-rate decreased by .6%. However, the percentage of correct prediction for adoption of fair trade products increased by .1%, as this first was 3.2% and now this is 3.3%. Hence, our model predicts 3.3% better than the naïve model when predicting the purchase of fair trade products.

Table 7: Hit-rate binary logistic regression period 2

Null-model Predicted % correct Logit model Predicted %

correct

Observed purchase fair trade product

No Yes Observed purchase

fair trade product

No Yes

No 11777 0 100% No 11757 20 99.8%

Yes 661 0 0% Yes 639 22 3.3%

Overall hit-rate 94.7%* Overall hit rate 94.7%*

*Percentage of correct predictions relative to all predictions Implications binary logistic mode1

Our model has little predictive capacity. The hit-rate captured over period 1 is 95.3%, whereas in the hold-out sample of period 2 this is 94.7%. Although this is a high percentage, the model does not predict any better than a naïve model.

The purchase of fair trade products is marginally explained by the explanatory variables in the model (3.3%). It is likely that the explanatory variables are not strongly contributing to the correct predictions of purchase of fair trade products, suggesting that additional variables are needed in order to improve these predictions.

5.2 Variables that influence purchase quantity of fair trade products

To determine which variables affect a consumer’s quantity of fair trade purchases, we

estimated an OLS model. The SOP (see formula 4) was captured over 542 households

that purchase fair trade products, as such our analysis included 749 cases.

(30)

Overall our model is significant, F-test (12,735) = 14.464, p < .01, indicating that the independent variables in our model explain 17.8% of the SOP . Furthermore we checked the VIF scores of the variables in our model to possible detect multicollinearity issues (Leeflang et al., 2014). As can be seen in table 8, all VIF scores are below the critical value of 4, and with a minimum VIF of 1.033 and a maximum of 2.497 there are no signs of multicollinearity.

Table 8: Estimates multiple regression

Variables Unstandardised β Standardised β p- VIF

Constant .039 .063 Altruistic .001 .006 .895 1.624 Biospheric -.004 -.019 .643 1.595 Egoistic .001 .006 .854 1.033 Price Premium -.008 -.015 .713 1.459 Availability .056 .453 .000*** 2.355 Age -.011 -.049 .181 1.218 Income -.009 -.039 .290 1.209 Education .003 .015 .677 1.176 Household size -.015 -.063 .091* 1.243 Gender (female) .012 .018 .603 1.095 Virtue .002 .003 .939 1.867 Neither .087 .107 .009*** 1.515

a***significant at p <.01. **significant at p <.05. *significant at p <.10. The effect of the vice and virtue category

We reasoned that consumers are less likely to purchase fair trade products in virtue product categories than the vice category. However, as can be seen in table 8, the virtue nature of a product category does not have a significant effect (β = .002, p > .01), relative to the vice nature of a product category on a consumer’s purchase quantity of fair trade products. Surprisingly, the neither nature of a product category has a significant effect (β = .156, p < .01), relative to the vice nature of a product category. Therefore we cannot confirm H1 (see table 9).

Values

THE DRIVERS OF FAIR TRADE CONSUMPTION

THE DRIVERS OF FAIR TRADE CONSUMPTION

THE DRIVERS OF FAIR TRADE CONSUMPTION

ABSTRACT

MANAGEMENT SUMMARY

TABLE OF CONTENTS

1. INTRODUCTION

2. THEORETICAL FRAMEWORK

3. HYPOTHESES

4. RESEARCH METHODOLOGY

=

5. RESULTS