Fairtrade coffee consumption in Spain: Employing dual attitudes and construal level theory to draw insights on the ethical purchasing gap

As a contribution to the debate about Fair Trade contributions to the United Nations Sustainability Development Goals, this article investigates Spanish shoppers’ behaviour towards Fairtrade coffee. Although consumers generally state that they purchase fairly traded products, the market shares of most of them remain low, a phenomenon known as the ethical purchasing gap. Our review identifies a gap in extant literature to draw insights on the ethical purchasing gap, utilising two existing theories: attitudes and construal level as appropriate theoretical framework. The first theory highlights the duality of individuals’ attitudes towards an object: explicit attitudes are accessible to the consumers, whereas implicit attitudes are the ones they cannot recall, but nonetheless affect behaviour. The second theory examines the influence of low-level construal (concrete, specific) or high-level construal (general) information on decision-making. A three-stage experiment took place in two sessions in a large university in Madrid in order to apply these two theories. It was based on an online survey on explicit attitudes and purchase intention, and an Implicit Association Test (IAT) to identify implicit attitudes. It was run two weeks apart to capture three points of time effects. The results reveal that, despite exposure to different stimuli, implicit attitudes remain stable along three points of time. The average difference in purchase intentions was positive for low-level construal and negative for high-level construal. Explicit attitudes were not influenced by the exposure to the stimuli. No correlation was found between purchase intentions and implicit or explicit attitudes. These findings have useful managerial implications for both Fair Trade practitioners and academics.


Introduction
The earth has reached its limits (Biermann, 2012;Dao, Peduzzi & Friot, 2018;Wijkman, Rockström & Rockström, 2013) and it is now high time to rethink our consumption and production practices.Climate change, fresh-water depletion, deforestation, over-fishing, as well as pollution and increased rates of hunger, have helped push the sustainability debate to the forefront of scientific discussions across many disciplines.The continued pressure to lower agricultural prices has been putting immense pressure on those at the weakest end of the supply chain, that is, producers who work hard to produce the food we put on our tables.
Humanity continues to face serious humanitarian issues, such as hunger, migration, war, while access to water, education and medical care are still challenging in many parts of the world.Often migration is the result of falling agricultural prices, which motivates people to leave the countryside and seek employment in the city.
Alarming is also the fact that the income gap between the rich and the poor was in 2015 at its highest in 30 years in many countries (OECD, 2015, pp. 3, 15).An impressive 85 per cent of the world´s poor live in rural areas (Alkire et al., 2014).Income inequality results in decreased access to education, which in turn leads to wasted potential and lower social mobility (OECD, 2015, pp. 3, 15).This is known to slow GDP growth (OECD, 2015, p. 3).
For social change to happen, humanity needs to find new ways out of the current vicious circle.This implies changing production, as well as consumption patterns.In this paper, we will discuss the case of fairly traded coffee in the realm of sustainable or ethical consumption.Fair Trade has the mission to secure humane minimum prices for producers, to enable them, their families and their communities to thrive, away from the frequently exploitative practices in world trade.It therefore contributes to the achievement of several of the United Nations Sustainability Development Goals.Fairtrade International identifies Goal 1 'End poverty in all its forms everywhere' as central to Fairtrade´s mission.Moreover, Goal 2,5,8,12,13,16 and 17 are also addressed through the work of Fairtrade (Fairtrade.net, n.d.-b), as Fairtrade does not only concentrate on monetary benefits for the farmers and plantation workers but also aspects of social and environmental sustainability.
Fairtrade addresses issues such as gender equality, inclusive, sustainable economic growth, combating climate change, ensuring sustainable consumption and production patterns, promotion of peaceful and inclusive societies and strengthening the global partnership for sustainable development.
Despite the global community's commitment to the SDGs and the relevance of Fairtrade to those, Spain is one of the least developed Fairtrade markets (see the section on Fairtrade Consumption in Spain).With the present study, the authors aim to contribute to a deeper understanding of the decision-making around the best known and selling Fairtrade product, coffee, through the lens of dual attitudes and construal level theory in Spain, with the aim of offering insights that will increase its consumption.To date, there has been very little research on the influence of implicit attitudes on the consumption of Fairtrade products (Kossmann & Gómez-Suárez, 2018, 2019).The objective of the research is to identify if there are shifts in explicit and implicit attitudes, and purchase intentions over time, following exposure to two stimuli: highlevel and low-level construal.
Therefore, the main research questions are: 1. Will the exposure to high (low) level construal information lead to increased (decreased) purchase intention for Fairtrade coffee? 2. Will the effect be larger in the short than in the long run?
3. What will be the effect of this exposure to high (low) level construal information on implicit attitudes?4. Will the effect differ between the short and the long run?
5. What will be the effect of this exposure to high (low) level construal information on explicit attitudes?6.Will the effect differ between the short and the long run? 7. Will implicit or explicit attitudes be better predictors for purchase intentions?
To be able to assess those shifts a longitudinal, experimental research design was deployed.This study implemented a one-factor, two-level (information: concrete, abstract) between-subjects, repeated measures design with explicit attitudes, implicit attitudes and purchase intentions as dependent variables.Through a structured questionnaire, participants were asked about their explicit attitudes and intention to purchase FT coffee and finally about their demographics.For the implicit attitudes part, the standard Implicit Attitudes Test (IAT) was modified and adapted to the research context (Greenwald, McGhee & Schwartz, 1998).For the explicit attitudes (EA) and purchase intentions (PI) part, Qualtrics was used for the data collection.

The case for Fair Trade coffee
Coffee is the second most valuable commodity after petroleum in the global market (Loureiro & Lotade, 2005).
Around the world, 25 million small-scale coffee farmers produce 70-80 per cent of the world´s coffee (Fairtrade.net, n.d.-a), and depend on it as their main income source.
Fair trade developed as a movement in the post-Second-World War era as an attempt to reverse global income inequality (Bakker, van Dijk & Wicherts, 2012) in the post-colonial era.As farmers face increased income volatility, the adverse effects of climate change and an immense push on prices, several certification schemes emerged to remedy this.Through certification schemes, such as Fairtrade, farmers have access to a more stable income, pre-financing to secure the smooth running of their operations, or also as a help to their communities, receive training and other support to organise themselves democratically, but also as businesses, or increase their yields and thus their productivity.Currently, there is a proliferation of sustainability labels, some of them addressing issues of fair trade, for example, World Fair Trade Organisation (WFTO), BanaFair, Gepa or their own initiatives from firms in the organic food market, such as Rapunzel's Hand in Hand initiative.
Fairtrade consumption can be defined as part of sustainable consumption.Luchs et al. (2011, p. 1) define sustainable consumption as 'consumption that simultaneously optimizes the environmental, social, and economic consequences of consumption to meet the needs of both current and future generations'.Consumers mindful of the impact of their consumption choices on the environment, animals or other human beings (Barnett, Cafaro & Newholm, 2005) are the main shoppers for fairly traded products.
Fairtrade certifies small producer organisations as well as plantations.The latter intend to demonstrate their adherence to workers' rights through the certification of their produce.
Despite the growth of Fairtrade globally, but also in Spain, the market shares for coffee, the oldest, most widely distributed and most recognised product, remain low, at just over 3 per cent in 2017 (Statista, n.d.;Fairtrade International, 2019).With such low market shares in the global market, it can hardly be argued that the majority of coffee farmers are selling their produce on Fairtrade terms and thus that they have escaped the poverty circle.We hope that our research is but a small contribution in better understanding the consumer decision-making behind purchases of Fairtrade coffee and thus to offer relevant insights for practitioners and academia, with the ultimate goal to increase sales of fairly traded coffee and market access for coffee farmers.

Fairtrade consumption in Spain
The Fairtrade mark is less developed in Spain in comparison to other West and North European countries.In the Iberian peninsula sales of Fairtrade products were estimated at €94 million (Fairtrade Ibérica, n.d.) in 2018.
In comparison to the global sales of Fairtrade at €9.8 billion in 2018, this makes Spain one of the smallest Fairtrade markets, with around 1% of the global Fairtrade sales.In a country with a population of approximately 47 million (World Bank, n.d.-a) the aforementioned sales amount to €2 per capita, much behind the world leader, Switzerland, where per capita Fairtrade sales amount to €87 (Max Havelaar, n.d.; World Bank, n.d.-b).Nevertheless, Fairtrade is but one of the many Spanish initiatives in the area of ethical consumerism.
Oxfam, World Fairtrade Organisation, REAS and IDEAS are further actors in the ethical consumerism movement.Moreover, the solidarity economy in Spain adds up to 10% of the gross national product with 190,000 jobs, 80 per cent of which are under permanent contracts (Ballesteros, 2017).

Research questions
Although research generally concentrated on values and attitudes, further to a recent systematic literature review (Kossmann & Gómez-Suárez, 2019) only one paper was identified that used the Implicit Association Test (IAT) to identify implicit attitudes relating to the purchase of fairly traded products (Govind et al., 2019).
Practitioners would welcome such a piece of research (Kossmann & Gómez-Suárez, 2018).This is a promising method that can contribute to a better understanding of marketing measures that will increase the consumption of fairly traded products.Moreover, although the construal-level theory has been applied with relative success in social psychology, to the best of our knowledge this theory has not been applied to fair trade consumerism.
The development of the research questions was based on the above two theories.We investigate the effect of the exposure to high-or low-level construal information and its impact on implicit, explicit attitudes and purchase intentions both in the short term, that is directly after exposure to the stimulus, and in the long term, two weeks later.

Fair trade consumerism
According to a literature review published in 2012, most pieces of research on fair trade consumerism apply social psychology approaches, while fewer concentrate on economic approaches (Andorfer & Liebe, 2012).The number of papers devoted to this topic has grown, especially since 2015.As in the period prior to 2012, most papers concentrate on social psychology approaches.These mainly cover the following thematic areas: values, guilt and pride, willingness-to-pay, labels and certification, theory of reasoned action and theory of planned behaviour (Kossmann & Gómez-Suárez, 2019).Values are considered to be desirable end-states which influence people's decision-making and as a result action.Therefore, values and their influence on Fairtrade consumption have been researched by several academics (Bratanova et al., 2015;Doran, 2010;Kim, Lee & Park, 2010;Ladhari & Tchetgna, 2015).Different sets of values have been tested using diverse experimental methods.Power and social status values have been identified as irrelevant to fair trade consumption, whereas universalism, a world of beauty and benevolence have positive effect on fair trade consumption (Kossmann & Gómez-Suárez, 2019).
A smaller number of researchers have approached the subject from the perspective of willingness-to-pay, that is, the price premium that consumers are willing to pay for fair trade products (Bosbach & Maietta, 2019;De Pelsmacker, Driesen, & Rayp, 2005;Konuk, 2019;Lappeman et al., 2019;Schenk, 2019;Vecchio & Annunziata, 2015;Yoganathan, Osburg, & Akhtar, 2019;Zerbini, Vergura, & Latusi, 2019).This is an especially interesting aspect for marketers, as price is the only marketing mix element that directly generates income (Breidert, 2006).The price premium consumers are willing to pay varies per country and category, highlighting the need to conduct local market research before pricing decisions are made.
Antonetti has researched in two papers the impact of guilt on fair trade consumption (Antonetti, Baines & Jain, 2018;Antonetti & Maklan, 2014).Guilt helps an individual see herself as the cause of an action, while the increased transportation caused by guilt impacts positively on affective, cognitive and behavioural levels.In another piece of research anticipated consumer guilt is found to mediate the effects of its antecedents on fair trade buying intention.Anticipated consumer guilt is composed of two components: negative affect and selfdirected ethical judgement.Finally, self-efficacy was proven to have 'direct positive effects on anticipated consumer guilt and fair-trade buying behaviour' (Lindenmeier et al., 2017, p. 9).

Attitudes
An attitude is defined as a lasting, general evaluation of the self and others, objects, advertisements or issues (Baron, 1984) and is directed to an attitude object.It may serve distinct functions: utilitarian, value-expressive, ego-defensive and knowledge.According to the utilitarian functions, individuals develop some attitudes based on whether something causes pleasure or pain.The value-expressive function expresses the consumer's core values, while the ego-defensive attitudes are formed to protect the person from external threats or internal feelings.Finally, the knowledge function is the result of a need for order, structure or meaning (Solomon et al., 2006, pp. 138-139).An attitude may serve more than one function, often one being dominant.In the case of FT consumption, it is expected that attitudes can mainly be placed under the first three categories: utilitarian, value-expressive and ego-defensive.

Theory of Reasoned Action and Theory of Planned Behaviour
The Theory of Reasoned Action (TRA), put forward by Fishbein andAjzen in the 1970s (1975), and the Theory of Planned Behaviour (TPB), proposed by Ajzen in 1985 and updated in 1991, have been extremely influential in highlighting the relationship between attitude and behaviour by introducing the mediating role of intention and have been used in many types of 'ethical purchasing' (Hassan et al. 2016, p. 220).
According to TRA, actual behaviour is determined by behavioural intention, which is in turn dependent upon two variables: subjective norms and attitude towards the behaviour (Hassan et al., 2016).Subjective norms are defined as 'the perceived social pressure to perform or not to perform the behaviour' (Ajzen, 1991).
Subjective norms are based on an individual's perception of how most people who are important to her would consider his actions (Trafimow et al., 2002).
TPB extends TRA by including a measure of perceived behavioural control.This measure accounts for behaviours that are outside or within an individual's volitional control (Hassan et al., 2016).Following Ajzen (1991, p. 188) the above three dimensions are 'three conceptually independent determinants of intention', as such a person may be more or less influenced in her decision-making by the perceptions of her social group.
Research has demonstrated that the strength of an attitude is conducive to its ability to predict behaviour.
Strong attitudes, that is 'attitudinal positions held with confidence, vested interest in the topic, or attitudes highly accessible in memory' are better predictors of behaviour than weak attitudes (Ajzen, Czasch, & Flood, 2009).Some researchers claim that attitude is the first step to product consideration (Setyobudi et al., 2015) and thus has a central role in consumer decision-making.Furthermore, research by Yamoah et al. (2016) concluded that there is a significant effect of attitude on purchase intention (PI) for fair trade food products in the UK

Dual-process theories
Although TPB has been a popular theory, it has also been the object of debate and criticism, with some researchers even denying its adequacy to explain human social behaviour (Ajzen, 2011(Ajzen, , p. 1113)).Greenwald and Banaji (1995) posited that behaviour is also driven by implicit attitudes.If attitudes are to be used as estimates of behaviour, then both implicit and explicit attitudes should be considered, as well as other factors (Setyobudi et al., 2015).
The theory of dual attitudes is based on the notion that behaviour is not always based on conscious control (Greenwald & Banaji, 1995).Implicit cognition suggests that past experience influences the actor in ways that are unknown to him (Belletier et al., 2018, p. 61).Other terms that have been used to denote this distinction are conscious/unconscious, automatic/controlled, direct/indirect, etc.In this sense, attitudes are influenced by past experience that the actor cannot recall or report (Greenwald & Banaji, 1995, pp. 4-5).An explicit response is controllable, intended, made with awareness, and requires cognitive resources (Nosek, 2007, p. 65).Implicit attitudes are attitudes that are automatically activated upon exposure to the object, contrasted with explicit attitudes that require deliberate, conscious thought in their forming process (Setyobudi et al., 2015, p. 316).Wilson, Lindsey and Schooler (2000, p. 102) posited that attitudes are not 'stored evaluations of objects and issues' but rather that at times 'people construct on-the-spot attitudes based on information that happens to be accessible at that point in time'.Both the former proposal of 'Attitudes-as-Stored-Evaluations', as well as the latter 'Attitudes-as-Constructions' are subject to limitations.However, both approaches coexist.Wilson et al. (2000, p. 102) suggest that one attitude will not replace the other if one moves from one attitude A 0 to A 1 , but that the two attitudes will coexist.Thus, the attitude people choose at a point in time will depend on their capacity to retrieve the explicit attitude and whether this will, in the end, replace the implicit attitude.In the case of FT consumption, it is expected that, due in part to the social desirability bias, the explicit attitudes will differ significantly from the implicit ones.Research conducted by Govind et al. (2019) proved that, in the case of ethical consumerism, explicit attitudes had no impact on the choice of consumers, with implicit attitudes guiding behaviour.

Construal level theory
Construal level theory examines the effect of psychological distance on the mental interpretation of inputs and the importance attached to high and low-level criteria (Lynch & Zauberman, 2007), contributing greatly to the understanding of consumer psychology and consumer decision-making.Psychological distance is defined as the distance between the self in the here and now and the object and is a subjective experience (Trope & Liberman, 2010).The psychological distance can otherwise be defined as the 'divergence from direct experience' (Liberman & Förster, 2009, p. 203).Psychological distance may take any of the following dimensions: temporal -how far in the future or past is the object; spatial -how near or far is the object; socialhow close or far from me is the object on social terms; or hypothetical, in other words, the probability that an event takes place.
These distances are cognitively related to one another, similarly influence and are influenced by the level of the mental construal and in turn affect prediction, preference and action (Trope & Liberman, 2010).It is important to highlight the bi-directional relationship between the level of construal and the perception of psychological distance.
Construal level theory describes a process that involves a mental interpretative approach (Trope & Liberman, 2010, p. 440).High-level construals are those related to essential, central, abstract and global features and low-level construals are related to peripheral, incidental, concrete, specific and local features (Wakslak et al., 2006).The perceived distance from objects or events is associated with how abstractly those things are construed or represented.Higher-level representations are abstract and consist of 'schematic summaries that capture the essence or gist of something, whereas concrete, lower-level representations incorporate more contextual, readily observable features' (Henderson & Wakslak, 2010, p. 390).According to Bar-Anan et al. (2006, p. 609) perceivers are more likely to form higher-level construals, the greater the distance of objects and events.The more an event becomes removed in time, the more people's decisions are driven by superordinate concerns (ends) and central features, rather than subordinate concerns (means to the ends) and peripheral features.(Lynch & Zauberman, 2007).
Literature highlights that dual-process theory and construal level theory (Amit & Greene, 2012;Körner & Volk, 2014) can complement each other, in that dual-process theory acknowledges that one individual may have two distinct and conflicting attitudes towards one object.Through manipulating the psychological distance in experiments, scientists have highlighted how different types and sizes of psychological distance can influence moral judgements and decision-making.

Hypothesis Development
Participants in a series of four experiments placed events described in abstract terms (higher-level construal) further in the future (large temporal distance) than those described in more concrete terms (lower-level construal) (Liberman et al., 2007, p. 148).Given the bi-directional relationship between construal level and psychological distance (Trope, Liberman, & Wakslak, 2007), it is expected that exposure to general (high-level construal) or concrete (low-level construal) information on Fairtrade will influence the development of concrete plans to purchase Fairtrade coffee.Purchase intentions can be considered as manifestations of a plan.
According to TRA and TPB intention precedes behaviour (de Leeuw et al., 2015;Shaw et al., 2000).Therefore, in this experiment purchase intentions are tested between subjects further to an exposure to two different conditions, one being high-and the other low-level construal.Participants were asked in this study if and when they plan to purchase Fairtrade coffee, in order to test the effect of the exposure to the high-or low-level construal on purchase intentions.
H1a: Individuals who are exposed to high-level construal information will have less concrete plans to purchase Fairtrade coffee in the near future than those exposed to low-level construal information.Govind and colleagues (2019) proved that explicit attitudes were influenced by the nature of the stimuli.Thus, it is expected that there will be variation in the groups exposed to high-and low-level construal.It is assumed that there will be a higher effect directly post-exposure (short term) as situational cues (stimuli) affect attitude formation (Ajzen & Fishbein, 2000, p. 16;Wilson et al., 2000;Zanna, 1990, p. 98).The level at which situational cues affect attitudes and behaviour differs for each individual and attitude, however, the expectation is that the effect of the situational cues will be higher short-term rather than long-term.As mentioned above, the exposure to low-level construal information leads to the formation of concrete plans.
According to TRA and TPB attitudes precede intention and behaviour.Therefore, it is assumed that the effect of the situational cues on purchase intentions will be higher directly post-exposure than in the long run.
Moreover, it is expected that the exposure to concrete or specific (low-level construal) information will positively influence explicit attitudes towards Fairtrade coffee, more than in the case of exposure to abstract or general (high-level construal) information.
H1b: The effect of the exposure to stimuli on purchase intentions will be higher in the short run than in the long run.
H2a: A single incidence of low-level construal information will affect EA towards Fairtrade coffee more than in the case of high-level construal.
H2b: This influence will be larger in the short than in the long run.
In contrast to explicit attitudes, implicit attitudes are activated outside of the conscious control of an individual.They are largely based on past experience and influence judgements in ways that are known to the individual (Greenwald & Banaji, 1995, p. 4).Therefore self-reported measures are inappropriate to test implicit attitudes, as these suggest the ability to consciously retrieve those attitudes.For implicit attitudes indirect measures are necessary (Greenwald & Banaji, 1995, p. 5).As previously mentioned, implicit and explicit attitudes can coexist.The attitude that guides behaviour is dependent on the actor´s cognitive capacity.
Implicit attitudes are widely considered to be stable and more resistant to situational influences.Nevertheless, in a recent paper Gawronski et al. (2017) proved that explicit attitudes exhibit more temporal stability than implicit attitudes.However, Gawronski et al. tested the change over a period of one to two months, rather than two weeks, which was the time period employed by Govind et al. (2019) and in the survey carried out in this research.
H3: For a single incidence of exposure to high (low) level construal information, implicit attitudes will remain unchanged both in the short run and in the long run.Govind and colleagues (2019) found a significant effect between implicit and explicit attitudes and brand preference post-exposure to positive or negative information about the brand.Implicit attitudes had a significant effect on brand choice, whereas explicit attitudes did not.Similarly, in this study, the impact of highand low-level construal information on purchase intention has been tested.It is expected that across conditions implicit attitudes and not explicit attitudes have a higher correlation with PI.
H4: Across conditions, implicit attitudes (but not explicit attitudes) will predict the purchase intention.

Research Design
The survey instrument for the explicit attitudes was a questionnaire with four parts based on Yamoah et al. (2016), comprising an affective and a cognitive component.A five-point Likert scale was employed (1 = strongly disagree, 5 = strongly agree) for the responses.To assess the purchase intention of the subjects, the subjects had to indicate how probable it was they would purchase Fairtrade coffee in the next two weeks.A five-point Likert scale (1 = very unlikely, 5 = very likely) was deployed to assess the responses.A scale with a larger number of possible responses was avoided, as this often goes beyond the 'discrimination ability' of many subjects and may lead to satisficing behaviours as participants seek to reduce the cognitive burden (Xu & Leung, 2018, p. 119).
The IA part on the standard IAT (Greenwald et al., 1998) with seven stages was adapted to accommodate the needs of this research.According to Greenwald and colleagues, IAT evaluates 'implicit attitudes by measuring their underlying automatic evaluation ' (1998, p. 1464) and measures the association between a 'target-concept discrimination and an attribute dimension ' (1998, p. 1465).In this case, the implicit attitudes were evaluated for Fairtrade coffee and possible shifts post-exposure in the short and long run were also valued.The strength of IAT is that it can uncover attitudes which may otherwise not be expressed by the subjects.IAT was run on the online version of Millisecond.
The statistical analysis of the results of the EA and IA parts of the experiment was run in IBM SPSS Statistics Version 25.
Participants were 54 university students and staff from a large university in Madrid (Spain), as well as through convenience sampling people from the wider social circle of the researchers.The research presented in this paper was divided into three stages, which took places in two sessions (Table 1).Session one had a longer duration, as the participants had to undertake the same tasks twice and fill in their demographic information.
To control for social desirability bias, participants were assured by the research assistant that their answers would be handled with confidentiality.Moreover, a confidentiality agreement and a consent form were handed out to each participant individually. 1Participants were informed that they could leave the experiment at any time.Enough space was devoted to each participant that their responses could not be viewable by other participants.Finally, participants were assigned participant numbers to be able to match their responses in the three stages of the experiment.
To start with, the participants took part in IAT.Here it was tested how they perceive fairly traded coffee brand 'Tierra Madre' and mainstream brand 'Marcilla' (see Appendix 1).The participants were shown subsequently packaging of the two products followed by seven negative and seven positive adjectives.The adjectives list provided and tested by Govind and colleagues (2019) were used and double-blinded translated into Spanish (see Appendix 2).Fairtrade coffee consumption in Spain Elena Kossmann, Mónica Veloso and Mónica Gómez-Suárez The Oxfam coffee Tierra Madre was chosen, as it is a non-conventional brand, sold under the name of a large, established non-governmental organisation (NGO), rather than a Fairtrade variant of a large, conventional brand.As a conventional brand, the leading player in coffee, Marcilla (Statista, 2018) was chosen.
Following the IAT, the participants were asked to proceed to the Explicit Attitudes (EA) test, followed by a question on the Purchase Intention for Fairtrade coffee.At the end of the session, the participants were randomly assigned to one of the two condition groups.One group received a text about Fairtrade with highlevel construal (abstract) information and the other one with low-level construal (concrete) information.The high-level construal text describes the Fairtrade system more generally, whereas the low-level construal text described the impact of Fairtrade on a specific community.Both texts were sourced from different Fairtrade websites originally written in Spanish, and were proof-read by native speaker researchers in the research group (see Appendix 2).The level of construal in the text was expected to be a moderator to the psychological distance and therefore to impact the decision-making, but also the development of plans to buy or not Fairtrade coffee in the future.At the end of this part, a filler task took place with participants reading a short passage about the weather in Madrid (see Appendix 3).
After reading the text about the weather in Madrid, participants were asked to proceed immediately to Session 1b.Session 1b had the same structure as Session 1a, however, at the end of this session, participants were asked to fill in their demographic information.The participants were then thanked and informed of the next appointment.Session 2 took place two weeks later.The participants had to follow the same structure and flow as in session 1a: IAT, EA and PI.

Methodology
For the descriptive statistics of the sample, frequency tables were used due to the categorical nature of the current data.To evaluate the reliability of the instrument for the explicit attitudes Cronbach's α was calculated.Reliability analysis tests an instrument's ability to consistently reflect what it is measuring (Field, 2018(Field, , p. 1033)).In general, cases with values greater than 0.8 are considered to show an adequate internal consistency.Cronbach's α may be higher when a scale consists of more items, thus leading to false conclusions about the internal consistency of a scale (Field, 2018, pp. 821-823).In this case the scale consisted of four items, thus this is an appropriate use.Since the Cronbach's α values met the above criterion, a factor analysis was conducted on the explicit attitudes' variables and three factor scores were computed, one for each phase of the experiment.
All quantitative variables were tested for normality with the Kolmogorov-Smirnov test.The Kolmogorov-Smirnov test checks if 'a distribution of scores is significantly different from a normal distribution' (Field, 2018(Field, , p. 1022)).The Kolmogorov-Smirnov test with the Liliefors significance correction was run to check for normality for the explicit attitudes factors, the D-Score resulting from IAT and the purchase intention.At PI there is no normality.
Paired sample and independent sample t-tests were run in order to compare the means of the two groups exposed to the two stimuli (Kim, 2015).Independent sample t-tests are appropriate when the two groups in question are independent or consisting of different entities (Field, 2018, p. 445).Paired sample t-tests are used when the two means come from 'conditions consisting of the same or related entities' (Field, 2018, p. 445).In this case, paired sample tests were used when analysing the results of the participants within each condition at different points in time.Independent t-tests were used when the means of the different scores between conditions were compared.
For the overall evaluation of the 'condition' and 'phase' factors on the IA factor, a mixed-design ANOVA was utilised where 'phase' was set as a within-subjects factor and 'condition' as a between-subjects factor.
To test for the correlation between PI and EA or IA, non-parametric correlation coefficients were used as it is not assumed that there are specific data distributions (Herzog, Francis & Clarke, 2019, p. 52).Spearman's rho is an appropriate test, especially with a small sample (Field, 2018, p. 344).Moreover, the normality assumption was not met for PI, so only non-parametric tests were possible.
In all tests, a 5 per cent level of significance was used.

Instrument reliability
To check for instrument reliability for the questionnaire on explicit attitudes, Cronbach's α for the four dimensions of the explicit attitudes for Fairtrade coffee was calculated.The score proved to be high for the three phases 0.803, 0.901 and 0.866 (>0.8) and appropriate for cognitive tests, especially when a scale has a lower number of items (Field, 2018, p. 823).Therefore, three factor-scores were extracted to denote the values of EA for the three phases of the experiment.
In line with extant literature on the theories of planned behaviour and reasoned action, attitude precedes intention and behaviour (de Leeuw et al., 2015;Shaw et al., 2000).In this analysis, adding PI reduces Cronbach's α to 0.651, proving the hypothesis that explicit attitudes and purchase intentions are not unidimensional.

Hypothesis 1a and 1b
Hypothesis 1b states that exposure to high (low) level construal will have a greater effect on purchase intention (PI) in the short run than in the long run.Moreover, in hypothesis 1a, it is expected that the exposure to high-level information will lead to the formation of less concrete plans to purchase Fairtrade coffee in the near future than in the case of exposure to low-level construal. 2To check for the baseline PI score for both groups, independent samples t-tests were run and exhibited equal variances.Therefore, it can be assumed that the baseline score has no statistically significant differences (t = 1.504, df = 22, p = 0.147).
The differences between the two conditions3 were tested through independent samples t-tests.Factor condition has a statistically significant effect on the magnitude of the difference between PI 2 and PI 1 (t = -2.686,df = 13.725,p = 0.018).For exposure to condition a (PIdif 2_1a ) the difference on average decreases (mean = -0.8462),whereas for exposure to condition b (PIdif 2_1b ) the mean increases slightly (mean = 0.0909).As expected, the average difference is higher for PI a than PI b .
For hypothesis 1b to hold paired samples t-tests were ran.For subjects exposed to condition a, the null hypothesis of equality of means was supported (p = 1.000), as well as for those exposed to condition b (0.588).
Therefore, the alternative hypothesis cannot be confirmed.

Hypothesis 2
The second hypothesis states that a single incident of exposure to low-level construal information will positively affect explicit attitudes more than the exposure to high-level construal.This effect will be larger in the short run than in the long run.Baseline (t 1 ) scores between groups were found to have equality of means in the independent samples t-tests (t = -1.024,df = 22, p = 0.3217), proving condition 1.4 The independent t-tests confirmed the null hypothesis of equality of means for EA (t = 0.555, df = 22, p = 0.584).Therefore condition 3 does not hold.
For the fifth condition paired samples t-tests were run.In the paired difference both condition a and b had p-values greater than 0.05 (p a = 0.97, p b = 0.346), therefore the null hypothesis of equality of means is supported and condition 5 does not hold either for condition a or for condition b.

Hypothesis 3
Hypothesis 3 states that implicit attitudes will remain relatively stable in the three phases of the experiment, regardless of the exposure to condition a or b.5 To test this, a mixed-design ANOVA was run for IA with time (phases 1, 2, 3) as within-subjects factor and condition as the between-subjects factor.Factor time showed a non-significant effect on IA (F = 0.688, df1 = 2, df2 = 44, p = 0.508).Factor Condition demonstrated a statistically non-significant effect on IA (F = 1.727, df1 = 1, df2 = 22, p = 0.202) at a 5 per cent significance level.The interaction between time and condition does not have a statistically significant effect on IA (F = 2.983, df1 = 2, df2 = 44, p = 0.061).The results of this test confirm the hypothesis that IA remain stable over time, despite exposure to different conditions.

Hypothesis 4
Hypothesis 4 states that across conditions implicit attitudes, but not explicit attitudes, will predict the purchase intention.To test this hypothesis, non-parametric correlations were run.In all combinations of interest, the correlation coefficients do not exhibit statistically significant results.As expected, across conditions purchase intentions do not exhibit a correlation with explicit attitudes.However, within the current sample, neither do implicit attitudes correlate with purchase intentions.(See Table 3.)

Conclusions and Discussion
Implicit associations influence attitudes in ways that are inaccessible to the actor (Greenwald & Banaji, 1995).They are automatic and occur outside the cognition of an individual (Miles, Charron-Chénier & Schleifer, 2019).As such they are less susceptible to social desirability bias, which is especially important in cases of virtuous or ethical consumption.The IAT allowed to test participants' implicit attitudes towards Fairtrade coffee.IAT helps deliver results that are not influenced by social desirability, contrary to self-claimed behaviour.In this study, the assumption that implicit attitudes remained unchanged over time was confirmed, despite exposure to different stimuli, higher and lower-level construal.
In this paper, purchase intentions were considered the closest step to behaviour.As mentioned, attitudes precede intention, which in turn precedes behaviour.For those participants exposed to condition a, on average Gracias a los recursos provistos por Fairtrade y generados con la venta de plátanos certificados, en el departamento de La Guajira (Colombia), la Fundación de Trabajadores Unidos de Bananera Don Marce (FUTUBAN) entregó el proyecto de construcción de un restaurante escolar, la adecuación de un aula de clase y un tanque para almacenamiento de agua en la granja Marbasella.Marbasella, es un asentamiento en el que viven un promedio de 200 personas indígenas pertenecientes a la etnia Wayúu, que se encuentra ubicada aproximadamente a 30 kilómetros de la plantación de plátanos.50 de sus habitantes son trabajadores(as) en dicha plantación.Este proyecto nació gracias a la preocupación existente ante las condiciones de calidad sanitaria y de la infraestructura del aula.A través de la entrega del mismo se verán beneficiados(as) 50 niños y niñas de Marballesa y de las comunidades aledañas.
Thanks to the resources provided by Fairtrade and generated from the sale of Fairtrade certified bananas in the department of La Guajira (Colombia), the Foundation for United Workers of Bananera Don Marce (FUTUBAN) was able to deliver the project to build a school restaurant, the adaptation of a classroom and a tank for water storage at the Marbasella farm.Marbasella, is a settlement where an average of 200 indigenous people belonging to the Wayúu ethnic group live.It is located approximately 30 kilometres from the banana plantation.50 of its inhabitants are workers on this plantation.This project was born thanks to the existing concern regarding the conditions of sanitary quality and the classroom infrastructure.Through this project, 50 boys and girls from Marballesa and the surrounding communities will benefit.

Table 1
Study structure

Table 2
One sample Kolmogorov-Smirnov test

Table 3
Nonparametric correlations (Spearman's rho): effect of EA and IA on PI