Do Consumers Care about Ethics? Willingness to Pay for Fair-Trade Coffee


  • Financial support of the University Development Cooperation of the Flemish Interuniversity Council is gratefully acknowledged. The authors wish to thank the participants of the 33rd European Marketing Academy (EMAC) conference for their useful comments and suggestions. All remaining errors are ours.


Consumers’ buying behavior is not consistent with their positive attitude toward ethical products. In a survey of 808 Belgian respondents, the actual willingness to pay for fair-trade coffee was measured. It was found that the average price premium that the consumers were willing to pay for a fair-trade label was 10%. Ten percent of the sample was prepared to pay the current price premium of 27% in Belgium. Fair-trade lovers (11%) were more idealistic, aged between 31 and 44 years and less “conventional.” Fair-trade likers (40%) were more idealistic but sociodemographically not significantly different from the average consumer.

The purpose of this study was to investigate to what extent consumers were willing to pay for the fair-trade attribute when buying coffee, and how consumers differed in terms of their willingness to pay. First, we will describe fair trade within the context of ethical consumer behavior. Subsequently, the research questions used in our study will be examined.

Consumers can express their concern about the ethical behavior of companies by means of ethical buying and consumer behavior. In general, the ethical consumer feels responsible toward society and expresses these feelings by means of his or her purchasing behavior. Doane (2001) defined ethical consumption as the purchase of a product that concerns a certain ethical issue (human rights, labor conditions, animal well-being, environment, etc.) and is chosen freely by an individual consumer. There are several dimensions of ethical consumer behavior. Some forms of ethical consumption benefit the natural environment (e.g., environmentally friendly products, legally logged wood, animal well-being), while others benefit people (e.g., products free from child labor, fair-trade products). Cutting across this distinction, ethical consumption may benefit people or the environment close to home (e.g., some types of green products or organic food), or conversely in a faraway part of the world (e.g., fair-trade products or legally logged wood). Consumers can translate their ethical concerns by means of buying products for their positive qualities (e.g., green products) or by boycotting products for their negative qualities (e.g., not buying products made by children). Boycott campaigns against Nike because of alleged labor abuses and Nestlé because of the infant formula issue are among the most-cited examples of the latter (Auger, Devinney, and Louviere 2000; Carrigan and Attalla 2001; Creyer 1997; Shaw and Clarke 1999; Strong 1996). Consumers can decide to consider one or more ethical attributes when buying products.

Is ethical consumption growing? Evidence of a growing market for ethical products is often inferred from the results of opinion polls. According to a study by Hines and Ames (2000), 51% of the population had the feeling of being able to make a difference to a company's behavior and 68% claimed to have bought a product or a service because of a company's responsible reputation. On average, 46% of European consumers also claimed to be willing to pay substantially more for ethical products (MORI 2000). However, there are differences as to the reported willingness to pay a price premium for different types of ethical products. For instance, American consumers agreed with a price increase of 6.6% for green products (The Roper Organization, Inc. 1990), while French consumers wanted to pay 10%–25% more for apparel not made by children (CRC-Consommation 1998). With these studies in mind, one could expect a high demand for ethical products. However, the opposite seems to be the case. Most of the ethical labeling initiatives with respect to, for instance, organic food, products free from child labor, legally logged wood, and fair-trade products, often have market shares of less than 1% (MacGillivray 2000).

One of the main reasons for this discrepancy is the attitude–behavior gap. On the one hand, consumer perceptions and attitudes clearly influence behavior, as conceptualized and tested in several models of ethical consumption behavior (Ferrell and Gresham 1985; Hunt and Vitell 1993; Shaw and Clarke 1999; Vitell, Singhapakdi, and Thomas 2001). On the other hand, it is well documented that attitudes alone are generally poor predictors of buyer behavior (Cobb-Walgren and Ruble 1995), especially in the social marketing area (Shaw and Clarke 1999). While some consumers refuse to buy products with an unethical background (Crane 2001), the majority of people evaluate product attributes jointly in making purchase decisions. Price, quality, convenience, and brand familiarity are often still the most important factors affecting the buying decision (Boulstridge and Carrigan 2000; Carrigan and Attalla 2001; CRC-Consommation 1998; Norberg 2000; Roberts 1996; Tallontire, Rentsendorj, and Blowfield 2001). Dickson (2001) identified four segments of consumers based on the importance they attach to various product attributes of clothes. She found that only one segment, containing 16% of their sample, attached a lot of importance to the no-sweat label. The other three groups were qualified as nonusers. Moreover, often the attitudes and intentions toward ethical products are measured without explicitly taking the higher price of these products into account (Browne et al. 2000). Other explanations for the discrepancy between attitudes and ethical buying behavior can be the lack of availability of ethical products, disbelief of ethical claims, and lack of information (Carrigan and Attalla 2001; Mielants, De Pelsmacker, and Janssens 2003; Roberts 1996).

What could explain the attitude–behavior gap? In attitude research, people often give socially desirable answers. Ulrich and Sarasin (1995) somewhat cynically claimed not to do any research and not to ask the public any question on this subject because the answers are never reliable and often useless, if not misleading. Especially in situations in which respondents want to make a good impression on the researcher or want to conform to social norms, attitudes measured tend to be more positive than actual behavior (King and Bruner 2000). Moreover, attitudes are traditionally measured by means of explicit attitude measures, mostly self-reported paper-and-pencil tasks. Respondents are not always able and willing to report their attitudes and convictions accurately, especially in the case of socially sensitive issues such as ethical consumption behavior (Greenwald and Banaji 1995; Maison 2002).

If one wants to study the importance of the ethical attribute in buying decisions, a number of factors have to be taken into consideration. First of all, measuring explicit attitudes is not the most valid method to predict ethical buying behavior. Instead, measures that are closely related to the actual purchase behavior are called for. Second, a lot of buying behavior is based on multiattribute decision making in which the ethical attribute may or may not be important. In estimating the (intended) buying behavior, consumers have to be confronted with realistic multiattribute buying situations. Third, one reason for the attitude–behavior gap is the price factor. The measurement of (intended) buying behavior has to take the willingness to pay into account. Finally, not all consumers are equally likely to buy ethical products. Moreover, depending upon the characteristics and the preferences of individual consumers, different ethical dimensions may result in differences in willingness (not) to buy products incorporating ethical values. Bird and Hughes (1997) claimed that the willingness to purchase goods based on ethical credentials is limited to a minority of shoppers.

Several studies have tried to identify the socially responsible consumer in terms of demographic characteristics. Anderson and Cunningham (1972) found that younger consumers were more socially conscious, while the effect of their education level was not clear, and income was of no relevance. Dickson (2001) found that age, income, and employment status was not discriminating between socially conscious consumers who attach a lot of importance to no-sweat labels on apparel and those consumers who do not. Although in the same study it was stated that no-sweat buyers were more often female, most studies concluded that ethical buying behavior was not influenced by gender (e.g., MORI 2000; Sikula and Costa 1994; Tsalikis and Ortiz-Buonafina 1990). In his extensive literature review, Roberts (1995) found that people who did not buy from businesses that discriminated against minority groups or women were mainly women with a median age of 47 and slightly lower incomes but concluded that demographics were not very significant in identifying the socially responsible consumer. Other studies concluded that the ethical consumer was a person with a relatively high income, education, and social status (Carrigan and Attalla 2001; Maignan and Ferrell 2001; Roberts 1996).

However, demographics alone are not sufficient to define and identify the ethical consumer. People's values appear to have a significant impact on their ethical consumption behavior. Values are abstract principles that reflect an individual's self-concept (Dickson 2000). They are enduring beliefs that a given behavior or outcome is desirable or good. As such, values serve as standards that guide our behavior across situations and over time. Values are often part of our personality system and determine specific attitudes. Anderson and Cunningham (1972) found that dogmatism, conservatism, status consciousness, cosmopolitanism, personal competence, and alienation were related to ethical consumer behavior. In addition, Roberts (1996) and Dickson (2001) stressed the importance of psychographic variables such as relevant attitudes, values, and personality characteristics. The Roper Organization, Inc. (1990) and Cowe and Williams (2000) segmented consumers in terms of their degree of ethical concern. Similarly, Fritzsche (1995) concluded that the values of people behaving ethically were significantly different from the values of people behaving unethically, and in Roberts’ (1996) study, perceived consumer effectiveness, liberalism, and alienation appeared to have a significant impact on ethical consumption behavior. In addition, Dickson (2000) studied the relevance of personal values in the context of socially responsible buying behavior.

One of the best-known instruments to comprehensively measure a person's value system is the Rokeach Value Survey (Rokeach 1973). The Rokeach Value Survey contains a set of 18 terminal values that relate to “end states of existence” and another set of 18 instrumental values relating to “modes of behavior.” Some studies have tried to identify ethical values within the Rokeach scale and their effect on ethical behavior (Fritzsche 1995; Nonis and Swift 2001; Sikula and Costa 1994). Dickson (2000) used 12 Rokeach Value Survey terminal values in her study of apparel-buying behavior. Two basic dimensions could be defined: macrosocietal (socially centered) and microsocietal (self-centered) values. To an extent, the former predicted attitudes toward business intentions. However, no systematic attempt has been made to relate consumers’ value systems to ethical buying behavior.

In this study, the importance of a fair-trade label in the coffee-buying decision was investigated. In a broad sense, “fair trade” can be described as an alternative approach to trading partnership that aims for sustainable development of excluded and/or disadvantaged producers. It seeks to do so by providing better trading conditions, raising awareness, and campaigning (Krier 2001). In the broadest sense, the concept incorporates environmental as well as social issues. Littrell and Dickson (1999) developed a continuum of business practices, from minimum to maximum fair-trade practices. Apart from paying fair wages in a local context and providing a safe and clean workplace (mainstream business), they defined maximum fair-trade practices as also encompassing the development of sustainable businesses, empowerment of artisans, fostering well-being, establishing political and social justice, and developing equitable trade.

In a narrow sense, fair trade is defined based on its best-known component: fair prices for the products of farmers in developing countries. In essence, fair trade means buying products from farmers in developing countries on terms that are relatively more favorable than commercial terms and marketing them in developed countries at an ethical premium (Bird and Hughes 1997). This higher price to the consumer is warranted by the higher price that farmers receive for their products and by the fair-trade control mechanisms in the trade channel (for an extensive description of fair-trade mechanisms, see, for instance, Littrell and Dickson [1999] and Krier [2001]). Companies generally demonstrate their fair-trade behavior to consumers by means of marketing fair-trade brands or by means of cooperating with fair-trade organizations that accredit their fair-trade products and allow them to market these products using a fair-trade label. Fair-trade organizations, on the other hand, go through considerable efforts to convince companies to comply with fair-trade rules and sell fair-trade products. For instance, in April 2000, after a year-long campaign by the human rights organization Global Exchange, Starbucks decided to carry fair-trade coffee in its 2,300 stores (Straus 2000).

Fair-trade buying is a specific type of ethical consumer behavior. Based on the dimensions defined earlier, for a U.S. or a European consumer fair-trade consumption means buying items for their positive quality of supporting people in faraway developing countries. The question is to what extent conclusions from empirical research on other types of ethical consumption behavior also hold for fair-trade buying. Fair trade is an issue of particular concern for the ethical consumer. Based on a qualitative study amongst ethical consumers in the United Kingdom, Shaw and Clarke (1999) concluded that fair trade was the most important issue of ethical concern in consumer behavior (as compared with, for instance, environmental issues and vegetarianism). Fair-trade brands, or fair-trade labeled products (especially coffee), are also reasonably available. However, the relative importance of a fair-trade label in the purchase decision of consumers has not yet been studied.

In this study, conjoint measurement (see hereafter) was used to confront consumers with realistic multiattribute choice decisions. Instead of studying their attitudes or preferences, their willingness to pay was measured. As such, the importance of the price factor was explicitly taken into account. Furthermore, willingness to pay is assessed as a measurement of buying intention that can be considered a realistic proxy for actual behavior. A fair-trade coffee label needs to be efficiently monitored and subjected to third-party certification in order to become credible. This implies additional costs and a price premium for the consumer. Indeed, fair-trade coffee is more expensive than non–fair-trade coffee. Based on the willingness to pay for this label, the size of the potential fair-trade coffee–buying population was estimated.

As is the case with ethical consumption in general, not everyone is equally likely to buy fair-trade products. For instance, Littrell and Dickson (1999) found that buyers of cultural (ethnic) fair-trade products were demographically quite homogeneous and consisted of highly educated, well-off Caucasian women in their forties. A large proportion of them were teachers, health professionals, and social workers. Idea Consult (2002) concluded that the Belgian fair-trade consumer is relatively highly educated and has a relatively high income and social status. In addition, personal values appear to play a role in fair-trade buying behavior. For instance, Littrell and Dickson (1999) found that buyers of cultural fair-trade products attached more importance to altruism, equality, peace, and a beautiful and environmentally secure world, and less importance to inner-directed values such as self-respect and inner harmony. In this study, consumers are segmented according to their willingness to pay for different coffee attributes (including the fair-trade attribute). The consumer segments are then defined based on sociodemographic characteristics and their terminal and instrumental values.

In summary, this study investigates the following research questions (see also 1Figure 1):

Figure 1.

Research Model

  • 1What is the relative importance that consumers attach to a fair-trade label in their coffee-buying decision, compared with other attributes? Can segments of consumers be defined based on their relative willingness to pay for different coffee attributes? What is their willingness to pay a price premium for the fair-trade attribute? The narrow definition of fair trade (a fair price for products of producers from developing countries) is used.
  • 2To what extent are the consumer segments different in terms of demographic characteristics and personal values?

Research method

Composition of the sample

This study is based on a sample of Belgian consumers. At the crossroads of the Latin culture, with Roman Catholic roots, and the German and Nordic culture, with Protestant roots, in a strongly internationalized economy where companies share a level-playing field, because of the absence of strong national brands, the Belgian consumer market has in many aspects a profile similar to that of the Europeon Union (EU). Concerning the ethical aspects of consumption, the European survey by MORI (2000) points to results for Belgium that are very similar to the European average with respect to attitudes toward social responsibilities of firms, the importance a firm's commitment to social responsibility in buying intentions, as well as in the supplementary willingness to pay for environmentally or socially responsible products. This finding was corroborated from our exploratory survey of the market share of fair-trade coffee. The 1% market share in Belgium is similar to the market share of fair-trade coffee in France (0.9%) and Germany (1%) and is midrange between the market share in Switzerland (5%), the Netherlands (3%), and Denmark (2.5%) on the one hand, and Norway (0.8%) and Finland (0.4%) on the other.

In this study, we surveyed the total administrative and academic staff of Ghent University, which is one of the largest universities in Belgium (26,000 students) and one of the largest employers in the city of Ghent and the surrounding region. Concentrating the survey on a central spot where people gather from a large area allowed us to obtain a diversified sample in a cost-efficient way and to better monitor the data collection process. More importantly, it allowed us to quantitatively and qualitatively improve the response motivation by appealing to collegiality and by conducting the survey (for its major part) using the university's intranet in a more respondent-friendly way (e.g., without bothering people when they are at home or busy, etc.). Several positive implications of the use of the Internet are reported (Orme and King 1998). The use of the Internet significantly reduces the costs of the survey, respondents can be reached more quickly, and the response rate is higher. Finally, to further encourage participation to the survey, 25 book vouchers were divided among the respondents.

An e-mail was sent to 4,664 staff members with an e-mail address, and 891 questionnaires were completed, of which 779 were useful (i.e., completed the majority of the questions). The remaining 550 staff members without an Internet account were approached by mail. A total of 62 staff members responded, of which 55 questionnaires were useful. Hence, the response rate of our survey was 16%, i.e., double the average survey response rate in Belgium (8%). Twenty-six respondents gave inconsistent answers for the conjoint analysis (in the sense that they showed no preference for any of the eight proposed product profiles) and were eliminated from the sample. The final sample was composed of 808 respondents. 1Table 1 shows the composition of the sample. As could be expected, due to the specific university context, younger and better-educated respondents were overrepresented compared with the total Belgian or EU population. In the analysis, we verified to what extent this affected our main results.

Table 1. 
Description of the Sample
CharacteristicSample Percentage (N= 808)
 24 or younger8
 Older than 4515
 High school16
 Higher education84

The outline of the survey

As mentioned above, in order to minimize the social desirability bias when questioning issues of ethical consumption (fair trade in this case), we used a conjoint measurement of the price consumers were willing to pay when facing a coffee-buying situation in a supermarket (where most coffee types are available and which is the usual shopping place of the large majority of Belgian consumers). In a conjoint analysis, consumers are asked to indicate their preference (in this case their willingness to pay) for products with varying characteristics. By simulating real marketplace situations, conjoint analysis realistically models day-to-day consumer decisions and has a reasonable ability to predict consumer behavior. Consumers show their preferences by making trade-offs between different attributes of a product (Carroll and Green 1995; Green, Krieger, and Wind 2001; Green and Srinivasan 1978). These trade-offs can be decomposed into part-worth utilities and importance weights for each product attribute. In this way, the importance of different attributes or criteria in the consumer's evaluation of the product can be studied (Green, Rao, and Desarbo 1978).

Based on an exploratory group discussion with 12 coffee consumers of varying age, gender, and education, we determined the relevant coffee attributes and their appropriate levels as follows:

  • • Brand: manufacturer brand and private label. Manufacturer brand was presented as “Douwe Egberts” (Sara Lee), which is the market leader in Belgium, with a market share of about 50%. Private, or supermarket, label was presented as an enumeration of supermarket brands, which account in Belgium for another 25%–30% market share. The remaining market share represents smaller and more specialized coffee brands.
  • • Blending: 100% Arabica beans and a blend of Arabica and Robusta beans. A coffee is considered of high quality when the blending is 100% pure Arabica beans.
  • • Flavor: dessert, decaffeinated, and mocha. In Belgium, coffee manufacturers mainly focus on three flavors.
  • • Package: “warm” and “cold.” Consumers can be attracted to a type of coffee because of the exotic or warm appearance of the package. Therefore, a cold and a warm level of packaging were developed from the perception of the aforementioned exploratory group of consumers. A cold package consisted of a picture in blue and white. A warm package was brown and red showing a cup of freshly made coffee.
  • • The presence or absence of a fair-trade label.

Based on these attributes and their levels, 48 descriptions of coffee types were possible (2 × 2 × 3 × 2 × 2). It was evidently impossible for respondents to indicate a preference for 48 product types. Conjoint analysis instead uses a fractional design, i.e., a systematic selection of a small number out of the full set of product profiles, while maintaining the coincidence of uncorrelated levels of different attributes appearing together. This design assures that an estimate of the importance of one attribute is unaffected by the estimate of other attributes. Using Orthoplan of SPSS, a fractional (orthogonal) design of eight product profiles was generated out of the set of 48 profiles (see 2Table 2). These eight product profiles were visually presented and described to inform the respondents about the attributes. For instance, the fair-trade label was described as

Table 2. 
Coffee-Type Profiles in the Orthogonal Conjoint Design
Profile 1Fair tradePrivate label100% ArabicaWarmDessert
Profile 2No fair tradeDouwe EgbertsArabica/RobustaColdDessert
Profile 3No fair tradeDouwe Egberts100% ArabicaWarmDessert
Profile 4Fair tradeDouwe EgbertsArabica/RobustaWarmDecaffeinated
Profile 5Fair tradeDouwe Egberts100% ArabicaColdMocha
Profile 6No fair tradePrivate labelArabica/RobustaWarmMocha
Profile 7Fair tradePrivate labelArabica/RobustaColdDessert
Profile 8No fair tradePrivate label100% ArabicaColdDecaffeinated

a label on the package indicates that a fair price for the coffee harvest is guaranteed to the farmers of the South

in addition to which we mentioned as an example the name of the best-known fair-trade label in Belgium (i.e., Max Havelaar).

In the first part of the questionnaire, the respondents were asked to express their willingness to pay for the eight different types of coffee. A reference product profile was defined (and visually presented) as a coffee with dessert flavor, a warm package, a blending of Arabica and Robusta beans, without a fair-trade label, produced by Douwe Egberts, and set at a price of €1.87 (corresponding to the then-ruling shop price). Respondents were asked to express their willingness to pay by putting a price over or under the reference price of €1.87 on each of the eight coffee profiles presented. By mentioning the fair-trade label in the conjoint task, we could assume that all respondents were equally informed and that the fairly traded coffee was perceptible for everybody. The conjoint analysis method resulted in individual part-worth utilities for each level of each attribute from which the relative importance of each attribute could be derived.

In the second part of the questionnaire, the respondents were asked about their personal values and a number of sociodemographic characteristics including gender, age, and education. The Rokeach (1973) value scale was used to measure personal values. This scale contains a set of 18 terminal values that relate to end states of existence (such as the importance of material wealth for the respondent) and another set composed of 18 values that relate to modes of behavior or instrumental values (such as the importance of ambition in the life of the respondent). Originally, these values were expected to be rank ordered. In our study, the respondents were asked to rate all items separately on a 9-point Likert-type scale. In this way, nonparametric restrictions could be overcome. Literature has shown that similar results follow from the two methods (see Finegan 1994; Munson and McIntyre 1979). Munson and Posner (1980) asserted that the information about the intensity of guidance in an individual's life was more precise using a Likert-type rating scale. In addition, the rating process is quicker and is therefore more convenient for the respondent (Fritzsche 1995). We used an unbalanced scale (from irrelevant, −1, to very important, 7) in order to cope with the leniency effect (Antonides and van Raaij 1998).


Relative importance of the fair-trade attribute and segmentation of the coffee market

A conjoint analysis was carried out for each respondent (using CONJOINT of SPSS).1 It computes the part-worth utilities of the levels of the attributes and the relative importance of each attribute (i.e., the ratio of the utility range of the considered attribute and the total sum of the utility ranges of all attributes). A summary for the whole sample was obtained by averaging the part-worth utilities and relative importance of the attributes over all the respondents. In 3Table 3, we show the average relative importance of the attributes and the part-worth utilities. The Pearson correlation and Kendall rank correlation coefficients between the estimated and observed preferences give an indication of the fit of the analysis.

Table 3. 
Average Part-Worth Utilities and Relative Importance of the Coffee Attributes (Total Sample and Clusters)
 Average Part-Worth Utilities
AttributesTotal sample1234
 Fair trade3.7411.935.511.141.21
 No fair trade−3.74−11.93−5.51−1.14−1.21
 Douwe Egberts3.271.341.641.927.85
 Private brand−3.27−1.34−1.64−1.92−7.85
 100% Arabica1.04.691.531.11.37
 Pearson's R1 (.000)1 (.000).999 (.000)1 (.000)1 (.000)
 Kendall's τ1 (.0003)1 (.0003)1 (.0003)1 (.0003).982 (.0004)
 Average Relative Importance
AttributesTotal sample1234
  1. Note: p values in parentheses.

Respondents (percentage of the total sample) 11402425

As could be expected, the brand attribute had the highest relative importance for the total sample. Fair-trade label and flavor came second in the purchasing decisions of the consumers with an almost negligible difference in importance. Label is judged almost as important as flavor and somewhat less important than brand. Package and blending were of minor relative importance. However, one has to remain cautious concerning the impact of the three dominant attributes; brand, label, and flavor. Coffee is, as opposed to a washing machine for example, a relatively low-involvement good, which implies a relative low importance of any attribute. Buying a specific type of coffee is often a matter of habit. This is also shown in the conjoint analysis by the much superior part-worth utility of the constant (i.e., the part of utility of coffee not explained by the five attributes) compared with the part-worth utilities of the attribute levels.

From the results of the conjoint analysis, the consumers were assigned to different groups based on a cluster analysis of the relative importance of the attributes. First, the number of clusters was determined from a hierarchical cluster analysis using Ward's method. Next, the optimal cluster composition was determined using a K-means cluster analysis. As suggested by Hair et al. (1998), we used the percentage change of the agglomeration coefficient as a criterion to discriminate between different cluster solutions. We opted for a four-cluster solution because of the substantial acceleration of the increase of the agglomeration coefficient when passing from four to three clusters, i.e., an increase of 23% compared with the 11.6% moving average increase of the previous five cluster solutions and the 13% increase of the agglomeration coefficient when going from a five- to a four-cluster solution. A drawback of hierarchical methods is that an individual can never be removed from the cluster to which he or she has been assigned. This may result in a suboptimal clustering. Hence, once the number of clusters was determined, we performed a K-means cluster analysis to determine the optimal cluster partitioning.

Table 3 indicates that the four identified clusters differed substantially in average part-worth utilities and relative importance of the attributes. The respondents in the first cluster expressed a high and clear preference for the fair-trade label. This group was therefore denoted the fair-trade lovers cluster. Cluster 2 is characterized by a relative balance between each attribute, although the fair-trade label still came out as the most important one. We called this cluster the fair-trade likers. The flavor lovers comprise cluster 3. They prefer the flavor of their coffee and barely make a distinction between other attributes. Finally, the salient characteristic of cluster 4 is the relative importance respondents lay on the brand of their purchased coffee. They are termed brand lovers.

Willingness to pay for a fair-trade label

In order to express their preferences for a given type of coffee, respondents were asked about their willingness to pay of the eight coffee profiles, in comparison with a price of €1.87 for the reference profile. Since the orthogonal design was representative for all combinations of attribute levels, the four profiles with a label only differed from the four profiles without a label in terms of the label itself. Therefore, we could determine the willingness to pay for a fair-trade label as the difference between the average price of the coffee profiles with a label and the coffee profiles without a label (i.e., the price premium of fair-trade coffee). In 4Table 4, we show some benchmark indicators for the willingness to pay for the total sample and the four identified clusters: the average willingness to pay and the share of the respondents who were willing to pay a price premium of at least 10% and at least 27%. The latter figure represented the actual price premium of fair-trade coffee based on all types of coffee available on the market in Belgium (ACNielsen 2002).

Table 4. 
Willingness to Pay for a Fair-Trade Label
 Average Willingness to Pay for a Fair-Trade Label (Price Premium as a Percentage of No Fair-Trade Coffee)
 Total sample1234
 Share of the Respondents Willing to Pay the Price Premium (%)
Price Premium (%)Total sample1234
At least 10 (€0.19)3591651813
At least 27 (€0.50) (actual premium, all types of coffee included)10521710.7

The average willingness to pay for the total sample was 10% (€0.19) but varied substantially from 36% (€0.62) for the fair-trade lovers to less than 5% for the taste and brand lovers (€0.07 and €0.06, respectively). A total of 35% of all the respondents were prepared to pay a price premium for fair-trade coffee of at least this average; ranging from more than 90% of the fair-trade lovers to 18% (flavor lovers) and 13% (brand lovers). Flavor and brand lovers were definitely not prepared to pay the actual premium of 27%. The same applies to more than 80% of the fair-trade likers and almost half of the fair-trade lovers. Table 3 indicated that about half of the respondents considered the fair-trade label when purchasing coffee. However, when the respondents’ willingness to pay the actual price premium was taken into account, potential market penetration (the share of the consumers that can be expected to buy fair-trade at a given price premium) of fair-trade coffee dropped to about 10%. Since the amount of coffee each of the groups buys is unknown, this market penetration potential should not be interpreted as a potential market share. Furthermore, this penetration potential is only valid under the assumptions of correct information and equal availability.

The penetration potential of 10% was substantially less than the 68% in the study of Hines and Ames (2000) who claimed to have bought a product or a service because of a company's responsible reputation and the 46% of the European consumers who claimed to be willing to pay substantially more for ethical products (MORI 2000). As already mentioned, American consumers agreed with a price increase of 6.6% for green products (The Roper Organization, Inc. 1990), less than the average price premium for fair-trade coffee in our sample, but this figure may be somewhat dated. Similarly, French consumers indicate a willingness to pay 10%–25% more for apparel not made by children (CRC-Consommation 1998). Our findings for fair-trade coffee corresponded with the lower limit of this estimate.

To what extent are our results affected by the overrepresentation of young and higher-educated respondents in the sample? In order to assess this, we calculated the average part-worth utilities, average relative importance of the attributes, and average willingness to pay for the subsample of administrative and technical staff (41% of the respondents), the composition of which was more in line with the total Belgian population. From the Labour Force Survey (Eurostat 2002), the age group younger than 24 years represented 9% of the workforce in Belgium (12% in the EU, 5% of the subsample in our survey), the age group between 25 and 45 years 73% (67% in the EU, 64% of the subsample), and the age group between 45 and 64 years 17% (21% in the EU, 31% of the subsample). The higher educated remained overrepresented in the subsample of administrative and technical staff (60% of total respondents), compared with the Belgian population (40%) and EU average (33%), although less than in the full sample.

For the subsample of administrative and technical staff, the brand was the most important attribute, counting for 28% of the total utility of the five attributes, as in the full sample. Again, flavor and label are equally important, accounting on average for 26% and 25% of the total utility of all the attributes, respectively. The relative importance of package and blending was also identical for the subsample and the full sample (12% and 9%, respectively). The average willingness to pay of the subsample of the administrative and technical staff amounted to 11%, i.e., close to the 10% for the total sample. Hence, it would seem that our results are not too severely affected by the specific composition of the sample regarding age and education. Nevertheless, we have to remain cautious to extrapolate our results to the total population.

Characteristics of fair-trade consumers

From the second part of our survey, we profiled the clusters in the coffee market in demographic and personal value terms. As described above, the significance of demographic factors in ethical decisions is not clear. Based on the literature survey in the introduction of this study, only age, gender, and education were considered. The age variable was classified into three categories: 18–30 years, 31–44 years, and 45 years or older. Two levels of education were considered: high school (12 years of education or less) and higher education (more than 12 years of education).

5Table 5 gives a description of each cluster in terms of age, education, and gender. For every independent variable, the p value of the Pearson chi-square test on equal means is given in parentheses. Regarding gender, the fair-trade lovers and likers consisted of an almost equal share of men and women. This confirmed earlier studies that have concluded that ethical buying behavior was not influenced by gender (e.g., MORI 2000; Sikula and Costa 1994; Tsalikis and Ortiz-Buonafina 1990). Women represented a more than proportionate share of the brand lovers. The reason for this could be that women still do more shopping than men and therefore could be more brand aware. Men were more than proportionately flavor lovers.

Table 5. 
Cluster Demographics (in Percentage of Full Sample or Cluster)
 Total SampleFair-Trade LoversFair-Trade LikersFlavor LoversBrand Lovers
  1. Note: p values in parentheses.

Gender (.016)
Education (.037)
 High school168162115
 Higher education8492847985
Age (.000)
 45 and older15917266

People in the 31–44 year age group were dominant in the fair-trade lovers cluster. This was in line with the profile that fair-trade organizations see as their target group. It confirmed the findings of Littrell and Dickson (1999) and (to a certain extent) Roberts (1995). The results are not in line with the demographic profiles found in studies relating to other types of ethical consumer behavior, e.g., Dickson (2001), concerning the importance of the no-sweat label. The youngest age group is dominant in the brand lovers and less prevalent in the flavor lovers. Finally, the flavor lovers were older than the members of the other clusters.

The highly educated, defined as indicated above, constituted a more than proportionate share of the fair-trade lovers and a less than proportionate share of the flavor lovers. Studies tend to conclude that the ethical consumer is a person with a relatively high educational status (Carrigan and Attalla 2001; Littrell and Dickson 1999; Maignan and Ferrell 2001; Roberts 1996). This sociodemographic profile was confirmed in the context of fair-trade buying behavior in Belgium (Idea Consult 2002). However, the difference in education distribution of the clusters became insignificant if we defined those with higher and lower levels of education in alternative ways, i.e., as respondents with and without a university degree, or in three categories (university degree, college nonuniversity degree, and high school). It would seem that the relevant education difference is between high-school-only consumers and higher-education consumers.

To define the basic dimensions that underlie the Rokeach scale measurements, an exploratory factor analysis was carried out on the 36 scale items.2 We distinguished five factors: a conventionalism factor, a competence factor, a sincere and social factor, an idealistic factor, and a personal gratification factor. Their reliability was determined, using Cronbach's alpha. This is a coefficient of reliability that measures how consistent a set of items (or variables) measures a single dimensional latent construct. Alphas should be higher than .7 but may decrease to .6 in exploratory research (Hair et al. 1998). Some of the five factors that result from this analysis are also found in previous studies (e.g., Crosby, Bitner, and Gill 1990; Vinson, Munson, and Nakanishi 1977), but other studies have found less factors (e.g., Dickson 2000; Gibbins and Walker 1993) or concluded that the number of dimensions found are dependent upon moderating variables such as cross-cultural differences (see, for instance, the overview by Meglino and Ravlin 1998).

An analysis of variance for every factor was conducted to check for significant differences between the four previously identified clusters using the Bonferroni post hoc test. 6Table 6 provides the differences in means between the clusters for each factor. The means ranged between 3 and almost 5 on the scale between −1 and 7 and were mostly higher than the midpoint scale values.

Table 6. 
Bonferroni Post Hoc Tests for the Five Underlying Factors
 Fair-Trade LoversFair-Trade LikersFlavor LoversBrand Lovers
  1. Note: The figures in parentheses indicate significant differences between cluster pairs of the mean factor scores at the .05 level.

Conventionalism3.02 (1,3); (1,4)3.35 (2,4)3.65 (3,1)3.78 (4,1); (4,2)
Sincere and social4.914.794.704.72
Idealistic4.53 (1,3); (1,4)4.47 (2,3); (2,4)4.06 (3,1); (3,2)4.00 (4,1); (4,2)
Personal gratification3.47 (1,4)3.60 (2,4)3.59 (3,4)3.96 (4,1); (4,2); (4,3)

The competence factor and the sincere and social factor were not significantly different across the clusters. Fair-trade lovers were less conventional than flavor lovers and brand lovers. Next, fair-trade lovers and fair-trade likers were significantly more idealistic than the other two clusters. Roozen, De Pelsmacker, and Bostyn (2001) also found idealism as a determining factor for ethical behavior. Dickson (2000) found that societal-centered values predicted attitudes toward socially responsible businesses. Finally, the brand lovers were significantly more motivated by personal gratification than any other group of respondents.

Overall, the four clusters could be described as follows. Fair-trade lovers accounted for 11% of the sample. For this group, a fair-trade coffee label represented the dominant attribute when buying coffee. Members of the group were largely between 31 and 44 years of age and higher educated than the other clusters. They were, together with the flavor lovers, predominantly male, more so than the other two clusters. This cluster also tended to be more idealistic and less conventional. The fair-trade likers formed the largest group, with 40% of the sample. They tended to choose a fair-trade label on coffee but also had the highest preference for blending and the second highest preference for any other attribute. Demographics were not notably different from other clusters, and they adhered to the same values as the fair-trade lovers. The flavor lovers consisted of approximately one-quarter of the total sample (24%). There were more men in this cluster. They were older and less educated (although this could be intrinsic to the fact of being older, i.e., a generation effect). They were more conventional than the two former clusters and less idealistic. The brand lovers accounted for 25% of the sample and were mainly women with an education profile similar to the fair-trade likers. Brand lovers were significantly younger than any other group. Considering personal gratification as a way of life made the difference between this and the other clusters.


In this paper, we investigated to what extent consumers consider a fair-trade label when purchasing coffee, using a sample of Belgian consumers. A substantial number of surveys showed that consumers value the ethical aspect in a product. However, consumers’ behavior in the marketplace is apparently not consistent with their reported attitude toward products with an ethical dimension. In this study, we tried to avoid the misleading general attitude indications by capturing the hypothetical purchase intention for fair-trade coffee. In presenting a choice situation to consumers in a close-to-reality setting, we tried to determine the value of a fair-trade label, and hence the importance of ethics, by including simultaneously all the relevant dimensions of coffee-buying intentions. The brand was the most important attribute of coffee, closely followed by flavor and fair-trade label in third. The willingness to pay for a fair-trade label on coffee of the respondents indicated that about 10% of the sample wanted to pay the current price premium of 27% in Belgium.

Clusters based on differences in preference were defined to estimate market opportunities for fairly traded coffee and to profile potential consumers. Profiling was done by means of demographic features: age, gender, and education level, as well as underlying factors of the Rokeach personal values scale. Four clusters were identified. The fair-trade lovers accounted for 11% of the sample and were predominantly aged 31–45 years. They were more idealistic and less conventional compared with other groups. The fair-trade likers represented the largest group. They did not differ significantly from the rest of the sample in terms of demographic characteristics, but they were relatively more idealistic. The flavor lovers and the brand lovers each accounted for one-quarter of the total sample and were less idealistic and more conventional. In addition, brand lovers were more likely to be women. The fair-trade lovers constituted the group that was most prepared to pay the actual price premium (slightly over 50% of them). Of the flavor lovers and the brand lovers, who account for 50% of the sample, only a small minority was prepared to pay the sample average price premium of 10% for fair-trade coffee.

The 11% fair-trade lovers cannot be considered equivalent to actual market share because the amount of coffee they buy relative to the total population is unknown. Furthermore, the results of this study were based on the assumption of extensive, equal, and correct information for all respondents and the availability of fair-trade coffee to the same extent as other brands. In reality, this is not the case. Second, the response rate of the mailing was only half that of the online survey. This adds to the explanation why higher educated and respondents younger than 45 years tended to be overrepresented in our sample (although our basic results are unaffected by the specific composition of the sample). Due to the probable, but unknown, bias in the sample toward fair-trade issues, and because of the overrepresentation of the higher educated, we consider the market penetration potential of 10% as an upper limit. Nevertheless, the gap between this and the actual market share of fair-trade coffee of 1% on the Belgian market suggests that even at the actual price premium, there is an unexploited market potential for fair-trade coffee; however, this is somewhat more modest than that suggested by some other studies.

Although the fair-trade lovers are a considerable niche, the size of the fair-trade likers segment indicated an even larger market potential of fair-trade coffee. Fair-trade lovers and likers covered 50% of the consumers. These two groups could be convinced to buy fair-trade coffee if better informed and the right marketing efforts are pursued. Fair-trade likers also attach a lot of importance to attributes such as brand and flavor. To appeal to them, the quality of the fair-trade coffee should match that of regular brands. The creation of a genuine fair-trade brand, instead of labeling other brands with fair-trade, may be a more efficient and credible alternative to promote the fair-trade idea to a broader audience. Finally, although the willingness to pay the actual price premium for fair-trade coffee was relatively strong in the fair-trade lovers group, it was on the contrary relatively weak in the fair-trade likers segment. Apparently, the appreciation for the fair-trade attribute was not strong enough to support the actual price premium. Maybe the most efficient way to expand the market would be to reduce the price premium of fair-trade coffee to a level more acceptable by larger parts of the population, e.g., by giving tax incentives such as lower value-added tax, similar to the tax reductions for environmentally friendly products in some countries.

There are at least two directions in which this research could be expanded. One direction for future research is related to the increasing competition on the “good-cause label” market. A growing number of products carry “green,”“bio,”“social,” or fair-trade labels. Ethically oriented consumers are increasingly faced with the choice between these labels. A similar study to ours is currently being conducted in which the type of label (and not just the presence or the absence of a fair-trade label) is incorporated as a product attribute. This will provide more insight into the willingness to pay for different types of ethical products in terms of what or whom they benefit (the environment versus people, the immediate context of the consumer versus geographically remote contexts, etc.). The willingness to buy and to pay for fair-trade products may also be contingent upon the availability of these products in shops, the credibility of the issuer of the label, and the amount and quality of the fair-trade information. A follow-up study is currently under way in which these aspects are also incorporated as product attributes. This will enable fair-trade marketeers and governments to optimize their marketing and information efforts.

A second direction for further research may focus on the increasing use of fair-trade labels on other types of products besides coffee, such as bananas, honey, chocolate, and clothing. A similar study as for fair-trade coffee could be set up to investigate the relative importance of fair-trade and other ethical labels across product categories, to assess the willingness to pay a price premium for the ethical label attribute, to estimate the market potential of fair-trade and other ethically labeled products, and to define promising market segments of consumers in different product categories and markets.


  • 1

    All the computations were performed using SPSS.

  • 2

    For the sake of brevity, factor loadings and other detailed figures are not reported. They are available from the authors upon request.